derek.holloway
7502bf64a4
CodeQL / Analyze (actions) (push) Canceled after 0s
CodeQL / Analyze (python) (push) Canceled after 0s
pre-commit / pre-commit (push) Canceled after 0s
Multi-Arch CI / setup (push) Canceled after 0s
Publish build_manylinux_x86_64 images / publish_build_manylinux_x86_64 (push) Canceled after 0s
Setup (Multi-Arch) / setup (push) Canceled after 0s
Multi-Arch CI / Linux::${{ fromJSON(needs.setup.outputs.linux_build_config || '{}').build_variant_label || 'skip' }} (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Publish no_rocm_image_ubi10 image / publish_no_rocm_image_ubi10 (push) Canceled after 0s
Multi-Arch CI / Windows::${{ fromJSON(needs.setup.outputs.windows_build_config || '{}').build_variant_label || 'skip' }} (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Multi-Arch CI / Manifest Diff (push) Canceled after 0s
TheRock Manifest Diff Report / Generate Manifest Diff Report (push) Canceled after 0s
Multi-Arch CI / CI Summary (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04 (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04_ocl_rt images / publish_no_rocm_image_ubuntu24_04_ocl_rt (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_rocgdb (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04_rocgdb images / publish_no_rocm_image_ubuntu24_04_rocgdb (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_ocl_rt (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Unit Tests / Unit Tests :: ubuntu-24.04 (push) Canceled after 0s
Unit Tests / Unit Tests :: windows-2022 (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_media (push) Canceled after 0s
Unit Tests / Unit Tests Summary (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_openmpi (push) Canceled after 0s
12564 lines
506 KiB
LLVM
12564 lines
506 KiB
LLVM
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
|
|
; RUN: llc -mtriple=riscv32 -mattr=+m,+d,+zvfh,+zfbfmin,+zvfbfmin,+v -target-abi=ilp32d \
|
|
; RUN: -verify-machineinstrs < %s | FileCheck %s --check-prefixes=CHECK,RV32,RV32V,RV32V-ZVFH
|
|
; RUN: llc -mtriple=riscv64 -mattr=+m,+d,+zvfh,+zfbfmin,+zvfbfmin,+v -target-abi=lp64d \
|
|
; RUN: -verify-machineinstrs < %s | FileCheck %s --check-prefixes=CHECK,RV64,RV64V,RV64V-ZVFH
|
|
; RUN: llc -mtriple=riscv32 -mattr=+m,+d,+zfhmin,+zvfhmin,+zfbfmin,+zvfbfmin,+v -target-abi=ilp32d \
|
|
; RUN: -verify-machineinstrs < %s | FileCheck %s --check-prefixes=CHECK,RV32,RV32V,RV32V-ZVFHMIN
|
|
; RUN: llc -mtriple=riscv64 -mattr=+m,+d,+zfhmin,+zvfhmin,+zfbfmin,+zvfbfmin,+v -target-abi=lp64d \
|
|
; RUN: -verify-machineinstrs < %s | FileCheck %s --check-prefixes=CHECK,RV64,RV64V,RV64V-ZVFHMIN
|
|
|
|
; RUN: llc -mtriple=riscv32 -mattr=+m,+d,+zvfh,+zfbfmin,+zvfbfmin,+zve32f,+zvl128b -target-abi=ilp32d \
|
|
; RUN: -verify-machineinstrs < %s | FileCheck %s --check-prefixes=CHECK,RV32,RV32ZVE32F,RV32ZVE32F-ZVFH
|
|
; RUN: llc -mtriple=riscv64 -mattr=+m,+d,+zvfh,+zfbfmin,+zvfbfmin,+zve32f,+zvl128b -target-abi=lp64d \
|
|
; RUN: -verify-machineinstrs < %s | FileCheck %s --check-prefixes=CHECK,RV64,RV64ZVE32F,RV64ZVE32F-ZVFH
|
|
; RUN: llc -mtriple=riscv32 -mattr=+m,+d,+zfhmin,+zvfhmin,+zfbfmin,+zvfbfmin,+zve32f,+zvl128b -target-abi=ilp32d \
|
|
; RUN: -verify-machineinstrs < %s | FileCheck %s --check-prefixes=CHECK,RV32,RV32ZVE32F,RV32ZVE32F-ZVFHMIN
|
|
; RUN: llc -mtriple=riscv64 -mattr=+m,+d,+zfhmin,+zvfhmin,+zfbfmin,+zvfbfmin,+zve32f,+zvl128b -target-abi=lp64d \
|
|
; RUN: -verify-machineinstrs < %s | FileCheck %s --check-prefixes=CHECK,RV64,RV64ZVE32F,RV64ZVE32F-ZVFHMIN
|
|
|
|
define void @mscatter_v1i8(<1 x i8> %val, <1 x ptr> %ptrs, <1 x i1> %m) {
|
|
; RV32V-LABEL: mscatter_v1i8:
|
|
; RV32V: # %bb.0:
|
|
; RV32V-NEXT: vsetivli zero, 1, e8, mf8, ta, ma
|
|
; RV32V-NEXT: vsoxei32.v v8, (zero), v9, v0.t
|
|
; RV32V-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_v1i8:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: vsetivli zero, 1, e8, mf8, ta, ma
|
|
; RV64V-NEXT: vsoxei64.v v8, (zero), v9, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV32ZVE32F-LABEL: mscatter_v1i8:
|
|
; RV32ZVE32F: # %bb.0:
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV32ZVE32F-NEXT: vsoxei32.v v8, (zero), v9, v0.t
|
|
; RV32ZVE32F-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_v1i8:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: vsetvli a1, zero, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vfirst.m a1, v0
|
|
; RV64ZVE32F-NEXT: bnez a1, .LBB0_2
|
|
; RV64ZVE32F-NEXT: # %bb.1: # %cond.store
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vse8.v v8, (a0)
|
|
; RV64ZVE32F-NEXT: .LBB0_2: # %else
|
|
; RV64ZVE32F-NEXT: ret
|
|
call void @llvm.masked.scatter.v1i8.v1p0(<1 x i8> %val, <1 x ptr> %ptrs, i32 1, <1 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_v2i8(<2 x i8> %val, <2 x ptr> %ptrs, <2 x i1> %m) {
|
|
; RV32V-LABEL: mscatter_v2i8:
|
|
; RV32V: # %bb.0:
|
|
; RV32V-NEXT: vsetivli zero, 2, e8, mf8, ta, ma
|
|
; RV32V-NEXT: vsoxei32.v v8, (zero), v9, v0.t
|
|
; RV32V-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_v2i8:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: vsetivli zero, 2, e8, mf8, ta, ma
|
|
; RV64V-NEXT: vsoxei64.v v8, (zero), v9, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV32ZVE32F-LABEL: mscatter_v2i8:
|
|
; RV32ZVE32F: # %bb.0:
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 2, e8, mf4, ta, ma
|
|
; RV32ZVE32F-NEXT: vsoxei32.v v8, (zero), v9, v0.t
|
|
; RV32ZVE32F-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_v2i8:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v0
|
|
; RV64ZVE32F-NEXT: andi a3, a2, 1
|
|
; RV64ZVE32F-NEXT: bnez a3, .LBB1_3
|
|
; RV64ZVE32F-NEXT: # %bb.1: # %else
|
|
; RV64ZVE32F-NEXT: andi a2, a2, 2
|
|
; RV64ZVE32F-NEXT: bnez a2, .LBB1_4
|
|
; RV64ZVE32F-NEXT: .LBB1_2: # %else2
|
|
; RV64ZVE32F-NEXT: ret
|
|
; RV64ZVE32F-NEXT: .LBB1_3: # %cond.store
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vse8.v v8, (a0)
|
|
; RV64ZVE32F-NEXT: andi a2, a2, 2
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB1_2
|
|
; RV64ZVE32F-NEXT: .LBB1_4: # %cond.store1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 1
|
|
; RV64ZVE32F-NEXT: vse8.v v8, (a1)
|
|
; RV64ZVE32F-NEXT: ret
|
|
call void @llvm.masked.scatter.v2i8.v2p0(<2 x i8> %val, <2 x ptr> %ptrs, i32 1, <2 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_v2i16_truncstore_v2i8(<2 x i16> %val, <2 x ptr> %ptrs, <2 x i1> %m) {
|
|
; RV32V-LABEL: mscatter_v2i16_truncstore_v2i8:
|
|
; RV32V: # %bb.0:
|
|
; RV32V-NEXT: vsetivli zero, 2, e8, mf8, ta, ma
|
|
; RV32V-NEXT: vnsrl.wi v8, v8, 0
|
|
; RV32V-NEXT: vsoxei32.v v8, (zero), v9, v0.t
|
|
; RV32V-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_v2i16_truncstore_v2i8:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: vsetivli zero, 2, e8, mf8, ta, ma
|
|
; RV64V-NEXT: vnsrl.wi v8, v8, 0
|
|
; RV64V-NEXT: vsoxei64.v v8, (zero), v9, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV32ZVE32F-LABEL: mscatter_v2i16_truncstore_v2i8:
|
|
; RV32ZVE32F: # %bb.0:
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 2, e8, mf4, ta, ma
|
|
; RV32ZVE32F-NEXT: vnsrl.wi v8, v8, 0
|
|
; RV32ZVE32F-NEXT: vsoxei32.v v8, (zero), v9, v0.t
|
|
; RV32ZVE32F-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_v2i16_truncstore_v2i8:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v0
|
|
; RV64ZVE32F-NEXT: andi a3, a2, 1
|
|
; RV64ZVE32F-NEXT: vnsrl.wi v8, v8, 0
|
|
; RV64ZVE32F-NEXT: bnez a3, .LBB2_3
|
|
; RV64ZVE32F-NEXT: # %bb.1: # %else
|
|
; RV64ZVE32F-NEXT: andi a2, a2, 2
|
|
; RV64ZVE32F-NEXT: bnez a2, .LBB2_4
|
|
; RV64ZVE32F-NEXT: .LBB2_2: # %else2
|
|
; RV64ZVE32F-NEXT: ret
|
|
; RV64ZVE32F-NEXT: .LBB2_3: # %cond.store
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vse8.v v8, (a0)
|
|
; RV64ZVE32F-NEXT: andi a2, a2, 2
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB2_2
|
|
; RV64ZVE32F-NEXT: .LBB2_4: # %cond.store1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 1
|
|
; RV64ZVE32F-NEXT: vse8.v v8, (a1)
|
|
; RV64ZVE32F-NEXT: ret
|
|
%tval = trunc <2 x i16> %val to <2 x i8>
|
|
call void @llvm.masked.scatter.v2i8.v2p0(<2 x i8> %tval, <2 x ptr> %ptrs, i32 1, <2 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_v2i32_truncstore_v2i8(<2 x i32> %val, <2 x ptr> %ptrs, <2 x i1> %m) {
|
|
; RV32V-LABEL: mscatter_v2i32_truncstore_v2i8:
|
|
; RV32V: # %bb.0:
|
|
; RV32V-NEXT: vsetivli zero, 2, e16, mf4, ta, ma
|
|
; RV32V-NEXT: vnsrl.wi v8, v8, 0
|
|
; RV32V-NEXT: vsetvli zero, zero, e8, mf8, ta, ma
|
|
; RV32V-NEXT: vnsrl.wi v8, v8, 0
|
|
; RV32V-NEXT: vsoxei32.v v8, (zero), v9, v0.t
|
|
; RV32V-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_v2i32_truncstore_v2i8:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: vsetivli zero, 2, e16, mf4, ta, ma
|
|
; RV64V-NEXT: vnsrl.wi v8, v8, 0
|
|
; RV64V-NEXT: vsetvli zero, zero, e8, mf8, ta, ma
|
|
; RV64V-NEXT: vnsrl.wi v8, v8, 0
|
|
; RV64V-NEXT: vsoxei64.v v8, (zero), v9, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV32ZVE32F-LABEL: mscatter_v2i32_truncstore_v2i8:
|
|
; RV32ZVE32F: # %bb.0:
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 2, e16, mf2, ta, ma
|
|
; RV32ZVE32F-NEXT: vnsrl.wi v8, v8, 0
|
|
; RV32ZVE32F-NEXT: vsetvli zero, zero, e8, mf4, ta, ma
|
|
; RV32ZVE32F-NEXT: vnsrl.wi v8, v8, 0
|
|
; RV32ZVE32F-NEXT: vsoxei32.v v8, (zero), v9, v0.t
|
|
; RV32ZVE32F-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_v2i32_truncstore_v2i8:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vnsrl.wi v8, v8, 0
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v0
|
|
; RV64ZVE32F-NEXT: andi a3, a2, 1
|
|
; RV64ZVE32F-NEXT: vnsrl.wi v8, v8, 0
|
|
; RV64ZVE32F-NEXT: bnez a3, .LBB3_3
|
|
; RV64ZVE32F-NEXT: # %bb.1: # %else
|
|
; RV64ZVE32F-NEXT: andi a2, a2, 2
|
|
; RV64ZVE32F-NEXT: bnez a2, .LBB3_4
|
|
; RV64ZVE32F-NEXT: .LBB3_2: # %else2
|
|
; RV64ZVE32F-NEXT: ret
|
|
; RV64ZVE32F-NEXT: .LBB3_3: # %cond.store
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vse8.v v8, (a0)
|
|
; RV64ZVE32F-NEXT: andi a2, a2, 2
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB3_2
|
|
; RV64ZVE32F-NEXT: .LBB3_4: # %cond.store1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 1
|
|
; RV64ZVE32F-NEXT: vse8.v v8, (a1)
|
|
; RV64ZVE32F-NEXT: ret
|
|
%tval = trunc <2 x i32> %val to <2 x i8>
|
|
call void @llvm.masked.scatter.v2i8.v2p0(<2 x i8> %tval, <2 x ptr> %ptrs, i32 1, <2 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_v2i64_truncstore_v2i8(<2 x i64> %val, <2 x ptr> %ptrs, <2 x i1> %m) {
|
|
; RV32V-LABEL: mscatter_v2i64_truncstore_v2i8:
|
|
; RV32V: # %bb.0:
|
|
; RV32V-NEXT: vsetivli zero, 2, e32, mf2, ta, ma
|
|
; RV32V-NEXT: vnsrl.wi v8, v8, 0
|
|
; RV32V-NEXT: vsetvli zero, zero, e16, mf4, ta, ma
|
|
; RV32V-NEXT: vnsrl.wi v8, v8, 0
|
|
; RV32V-NEXT: vsetvli zero, zero, e8, mf8, ta, ma
|
|
; RV32V-NEXT: vnsrl.wi v8, v8, 0
|
|
; RV32V-NEXT: vsoxei32.v v8, (zero), v9, v0.t
|
|
; RV32V-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_v2i64_truncstore_v2i8:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: vsetivli zero, 2, e32, mf2, ta, ma
|
|
; RV64V-NEXT: vnsrl.wi v8, v8, 0
|
|
; RV64V-NEXT: vsetvli zero, zero, e16, mf4, ta, ma
|
|
; RV64V-NEXT: vnsrl.wi v8, v8, 0
|
|
; RV64V-NEXT: vsetvli zero, zero, e8, mf8, ta, ma
|
|
; RV64V-NEXT: vnsrl.wi v8, v8, 0
|
|
; RV64V-NEXT: vsoxei64.v v8, (zero), v9, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV32ZVE32F-LABEL: mscatter_v2i64_truncstore_v2i8:
|
|
; RV32ZVE32F: # %bb.0:
|
|
; RV32ZVE32F-NEXT: lw a1, 8(a0)
|
|
; RV32ZVE32F-NEXT: lw a0, 0(a0)
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 2, e8, mf4, ta, ma
|
|
; RV32ZVE32F-NEXT: vmv.s.x v9, a1
|
|
; RV32ZVE32F-NEXT: vmv.s.x v10, a0
|
|
; RV32ZVE32F-NEXT: vslideup.vi v10, v9, 1
|
|
; RV32ZVE32F-NEXT: vsoxei32.v v10, (zero), v8, v0.t
|
|
; RV32ZVE32F-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_v2i64_truncstore_v2i8:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.s.x v9, a1
|
|
; RV64ZVE32F-NEXT: vmv.s.x v8, a0
|
|
; RV64ZVE32F-NEXT: vmv.x.s a0, v0
|
|
; RV64ZVE32F-NEXT: andi a1, a0, 1
|
|
; RV64ZVE32F-NEXT: vslideup.vi v8, v9, 1
|
|
; RV64ZVE32F-NEXT: bnez a1, .LBB4_3
|
|
; RV64ZVE32F-NEXT: # %bb.1: # %else
|
|
; RV64ZVE32F-NEXT: andi a0, a0, 2
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB4_4
|
|
; RV64ZVE32F-NEXT: .LBB4_2: # %else2
|
|
; RV64ZVE32F-NEXT: ret
|
|
; RV64ZVE32F-NEXT: .LBB4_3: # %cond.store
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vse8.v v8, (a2)
|
|
; RV64ZVE32F-NEXT: andi a0, a0, 2
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB4_2
|
|
; RV64ZVE32F-NEXT: .LBB4_4: # %cond.store1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 1
|
|
; RV64ZVE32F-NEXT: vse8.v v8, (a3)
|
|
; RV64ZVE32F-NEXT: ret
|
|
%tval = trunc <2 x i64> %val to <2 x i8>
|
|
call void @llvm.masked.scatter.v2i8.v2p0(<2 x i8> %tval, <2 x ptr> %ptrs, i32 1, <2 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_v4i8(<4 x i8> %val, <4 x ptr> %ptrs, <4 x i1> %m) {
|
|
; RV32-LABEL: mscatter_v4i8:
|
|
; RV32: # %bb.0:
|
|
; RV32-NEXT: vsetivli zero, 4, e8, mf4, ta, ma
|
|
; RV32-NEXT: vsoxei32.v v8, (zero), v9, v0.t
|
|
; RV32-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_v4i8:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: vsetivli zero, 4, e8, mf4, ta, ma
|
|
; RV64V-NEXT: vsoxei64.v v8, (zero), v10, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_v4i8:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: ld a4, 8(a0)
|
|
; RV64ZVE32F-NEXT: ld a2, 16(a0)
|
|
; RV64ZVE32F-NEXT: ld a1, 24(a0)
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a3, v0
|
|
; RV64ZVE32F-NEXT: andi a5, a3, 1
|
|
; RV64ZVE32F-NEXT: bnez a5, .LBB5_5
|
|
; RV64ZVE32F-NEXT: # %bb.1: # %else
|
|
; RV64ZVE32F-NEXT: andi a0, a3, 2
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB5_6
|
|
; RV64ZVE32F-NEXT: .LBB5_2: # %else2
|
|
; RV64ZVE32F-NEXT: andi a0, a3, 4
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB5_7
|
|
; RV64ZVE32F-NEXT: .LBB5_3: # %else4
|
|
; RV64ZVE32F-NEXT: andi a3, a3, 8
|
|
; RV64ZVE32F-NEXT: bnez a3, .LBB5_8
|
|
; RV64ZVE32F-NEXT: .LBB5_4: # %else6
|
|
; RV64ZVE32F-NEXT: ret
|
|
; RV64ZVE32F-NEXT: .LBB5_5: # %cond.store
|
|
; RV64ZVE32F-NEXT: ld a0, 0(a0)
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vse8.v v8, (a0)
|
|
; RV64ZVE32F-NEXT: andi a0, a3, 2
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB5_2
|
|
; RV64ZVE32F-NEXT: .LBB5_6: # %cond.store1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 1
|
|
; RV64ZVE32F-NEXT: vse8.v v9, (a4)
|
|
; RV64ZVE32F-NEXT: andi a0, a3, 4
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB5_3
|
|
; RV64ZVE32F-NEXT: .LBB5_7: # %cond.store3
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 2
|
|
; RV64ZVE32F-NEXT: vse8.v v9, (a2)
|
|
; RV64ZVE32F-NEXT: andi a3, a3, 8
|
|
; RV64ZVE32F-NEXT: beqz a3, .LBB5_4
|
|
; RV64ZVE32F-NEXT: .LBB5_8: # %cond.store5
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 3
|
|
; RV64ZVE32F-NEXT: vse8.v v8, (a1)
|
|
; RV64ZVE32F-NEXT: ret
|
|
call void @llvm.masked.scatter.v4i8.v4p0(<4 x i8> %val, <4 x ptr> %ptrs, i32 1, <4 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_truemask_v4i8(<4 x i8> %val, <4 x ptr> %ptrs) {
|
|
; RV32-LABEL: mscatter_truemask_v4i8:
|
|
; RV32: # %bb.0:
|
|
; RV32-NEXT: vsetivli zero, 4, e8, mf4, ta, ma
|
|
; RV32-NEXT: vsoxei32.v v8, (zero), v9
|
|
; RV32-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_truemask_v4i8:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: vsetivli zero, 4, e8, mf4, ta, ma
|
|
; RV64V-NEXT: vsoxei64.v v8, (zero), v10
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_truemask_v4i8:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: ld a1, 0(a0)
|
|
; RV64ZVE32F-NEXT: ld a2, 8(a0)
|
|
; RV64ZVE32F-NEXT: ld a3, 16(a0)
|
|
; RV64ZVE32F-NEXT: ld a0, 24(a0)
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 1
|
|
; RV64ZVE32F-NEXT: vse8.v v8, (a1)
|
|
; RV64ZVE32F-NEXT: vse8.v v9, (a2)
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 2
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 3
|
|
; RV64ZVE32F-NEXT: vse8.v v9, (a3)
|
|
; RV64ZVE32F-NEXT: vse8.v v8, (a0)
|
|
; RV64ZVE32F-NEXT: ret
|
|
call void @llvm.masked.scatter.v4i8.v4p0(<4 x i8> %val, <4 x ptr> %ptrs, i32 1, <4 x i1> splat (i1 1))
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_falsemask_v4i8(<4 x i8> %val, <4 x ptr> %ptrs) {
|
|
; CHECK-LABEL: mscatter_falsemask_v4i8:
|
|
; CHECK: # %bb.0:
|
|
; CHECK-NEXT: ret
|
|
call void @llvm.masked.scatter.v4i8.v4p0(<4 x i8> %val, <4 x ptr> %ptrs, i32 1, <4 x i1> zeroinitializer)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_v8i8(<8 x i8> %val, <8 x ptr> %ptrs, <8 x i1> %m) {
|
|
; RV32-LABEL: mscatter_v8i8:
|
|
; RV32: # %bb.0:
|
|
; RV32-NEXT: vsetivli zero, 8, e8, mf2, ta, ma
|
|
; RV32-NEXT: vsoxei32.v v8, (zero), v10, v0.t
|
|
; RV32-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_v8i8:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: vsetivli zero, 8, e8, mf2, ta, ma
|
|
; RV64V-NEXT: vsoxei64.v v8, (zero), v12, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_v8i8:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: ld a3, 40(a0)
|
|
; RV64ZVE32F-NEXT: ld a2, 48(a0)
|
|
; RV64ZVE32F-NEXT: ld a1, 56(a0)
|
|
; RV64ZVE32F-NEXT: ld t0, 8(a0)
|
|
; RV64ZVE32F-NEXT: ld a7, 16(a0)
|
|
; RV64ZVE32F-NEXT: ld a6, 24(a0)
|
|
; RV64ZVE32F-NEXT: ld a5, 32(a0)
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a4, v0
|
|
; RV64ZVE32F-NEXT: andi t1, a4, 1
|
|
; RV64ZVE32F-NEXT: bnez t1, .LBB8_9
|
|
; RV64ZVE32F-NEXT: # %bb.1: # %else
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 2
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB8_10
|
|
; RV64ZVE32F-NEXT: .LBB8_2: # %else2
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 4
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB8_11
|
|
; RV64ZVE32F-NEXT: .LBB8_3: # %else4
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 8
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB8_12
|
|
; RV64ZVE32F-NEXT: .LBB8_4: # %else6
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 16
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB8_13
|
|
; RV64ZVE32F-NEXT: .LBB8_5: # %else8
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 32
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB8_14
|
|
; RV64ZVE32F-NEXT: .LBB8_6: # %else10
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 64
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB8_15
|
|
; RV64ZVE32F-NEXT: .LBB8_7: # %else12
|
|
; RV64ZVE32F-NEXT: andi a0, a4, -128
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB8_16
|
|
; RV64ZVE32F-NEXT: .LBB8_8: # %else14
|
|
; RV64ZVE32F-NEXT: ret
|
|
; RV64ZVE32F-NEXT: .LBB8_9: # %cond.store
|
|
; RV64ZVE32F-NEXT: ld a0, 0(a0)
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vse8.v v8, (a0)
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 2
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB8_2
|
|
; RV64ZVE32F-NEXT: .LBB8_10: # %cond.store1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 1
|
|
; RV64ZVE32F-NEXT: vse8.v v9, (t0)
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 4
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB8_3
|
|
; RV64ZVE32F-NEXT: .LBB8_11: # %cond.store3
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 2
|
|
; RV64ZVE32F-NEXT: vse8.v v9, (a7)
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 8
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB8_4
|
|
; RV64ZVE32F-NEXT: .LBB8_12: # %cond.store5
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 3
|
|
; RV64ZVE32F-NEXT: vse8.v v9, (a6)
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 16
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB8_5
|
|
; RV64ZVE32F-NEXT: .LBB8_13: # %cond.store7
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 4
|
|
; RV64ZVE32F-NEXT: vse8.v v9, (a5)
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 32
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB8_6
|
|
; RV64ZVE32F-NEXT: .LBB8_14: # %cond.store9
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 5
|
|
; RV64ZVE32F-NEXT: vse8.v v9, (a3)
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 64
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB8_7
|
|
; RV64ZVE32F-NEXT: .LBB8_15: # %cond.store11
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 6
|
|
; RV64ZVE32F-NEXT: vse8.v v9, (a2)
|
|
; RV64ZVE32F-NEXT: andi a0, a4, -128
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB8_8
|
|
; RV64ZVE32F-NEXT: .LBB8_16: # %cond.store13
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 7
|
|
; RV64ZVE32F-NEXT: vse8.v v8, (a1)
|
|
; RV64ZVE32F-NEXT: ret
|
|
call void @llvm.masked.scatter.v8i8.v8p0(<8 x i8> %val, <8 x ptr> %ptrs, i32 1, <8 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_baseidx_v8i8(<8 x i8> %val, ptr %base, <8 x i8> %idxs, <8 x i1> %m) {
|
|
; RV32-LABEL: mscatter_baseidx_v8i8:
|
|
; RV32: # %bb.0:
|
|
; RV32-NEXT: vsetivli zero, 8, e32, m2, ta, ma
|
|
; RV32-NEXT: vsext.vf4 v10, v9
|
|
; RV32-NEXT: vsetvli zero, zero, e8, mf2, ta, ma
|
|
; RV32-NEXT: vsoxei32.v v8, (a0), v10, v0.t
|
|
; RV32-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_baseidx_v8i8:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: vsetivli zero, 8, e64, m4, ta, ma
|
|
; RV64V-NEXT: vsext.vf8 v12, v9
|
|
; RV64V-NEXT: vsetvli zero, zero, e8, mf2, ta, ma
|
|
; RV64V-NEXT: vsoxei64.v v8, (a0), v12, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_baseidx_v8i8:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a1, v0
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 1
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB9_2
|
|
; RV64ZVE32F-NEXT: # %bb.1: # %cond.store
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v9
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vse8.v v8, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB9_2: # %else
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 2
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB9_4
|
|
; RV64ZVE32F-NEXT: # %bb.3: # %cond.store1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v9, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v8, 1
|
|
; RV64ZVE32F-NEXT: vse8.v v10, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB9_4: # %else2
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 4
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v9, 2
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB9_6
|
|
; RV64ZVE32F-NEXT: # %bb.5: # %cond.store3
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v8, 2
|
|
; RV64ZVE32F-NEXT: vse8.v v11, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB9_6: # %else4
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 8
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB9_8
|
|
; RV64ZVE32F-NEXT: # %bb.7: # %cond.store5
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v10, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v8, 3
|
|
; RV64ZVE32F-NEXT: vse8.v v10, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB9_8: # %else6
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 16
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 4, e8, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v9, 4
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB9_10
|
|
; RV64ZVE32F-NEXT: # %bb.9: # %cond.store7
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v9
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v8, 4
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vse8.v v10, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB9_10: # %else8
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 32
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB9_12
|
|
; RV64ZVE32F-NEXT: # %bb.11: # %cond.store9
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v9, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v8, 5
|
|
; RV64ZVE32F-NEXT: vse8.v v10, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB9_12: # %else10
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 64
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v9, 2
|
|
; RV64ZVE32F-NEXT: bnez a2, .LBB9_15
|
|
; RV64ZVE32F-NEXT: # %bb.13: # %else12
|
|
; RV64ZVE32F-NEXT: andi a1, a1, -128
|
|
; RV64ZVE32F-NEXT: bnez a1, .LBB9_16
|
|
; RV64ZVE32F-NEXT: .LBB9_14: # %else14
|
|
; RV64ZVE32F-NEXT: ret
|
|
; RV64ZVE32F-NEXT: .LBB9_15: # %cond.store11
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v9
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v8, 6
|
|
; RV64ZVE32F-NEXT: vse8.v v10, (a2)
|
|
; RV64ZVE32F-NEXT: andi a1, a1, -128
|
|
; RV64ZVE32F-NEXT: beqz a1, .LBB9_14
|
|
; RV64ZVE32F-NEXT: .LBB9_16: # %cond.store13
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v9, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a1, v9
|
|
; RV64ZVE32F-NEXT: add a0, a0, a1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 7
|
|
; RV64ZVE32F-NEXT: vse8.v v8, (a0)
|
|
; RV64ZVE32F-NEXT: ret
|
|
%ptrs = getelementptr inbounds i8, ptr %base, <8 x i8> %idxs
|
|
call void @llvm.masked.scatter.v8i8.v8p0(<8 x i8> %val, <8 x ptr> %ptrs, i32 1, <8 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_v1i16(<1 x i16> %val, <1 x ptr> %ptrs, <1 x i1> %m) {
|
|
; RV32V-LABEL: mscatter_v1i16:
|
|
; RV32V: # %bb.0:
|
|
; RV32V-NEXT: vsetivli zero, 1, e16, mf4, ta, ma
|
|
; RV32V-NEXT: vsoxei32.v v8, (zero), v9, v0.t
|
|
; RV32V-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_v1i16:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: vsetivli zero, 1, e16, mf4, ta, ma
|
|
; RV64V-NEXT: vsoxei64.v v8, (zero), v9, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV32ZVE32F-LABEL: mscatter_v1i16:
|
|
; RV32ZVE32F: # %bb.0:
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
|
|
; RV32ZVE32F-NEXT: vsoxei32.v v8, (zero), v9, v0.t
|
|
; RV32ZVE32F-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_v1i16:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: vsetvli a1, zero, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vfirst.m a1, v0
|
|
; RV64ZVE32F-NEXT: bnez a1, .LBB10_2
|
|
; RV64ZVE32F-NEXT: # %bb.1: # %cond.store
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vse16.v v8, (a0)
|
|
; RV64ZVE32F-NEXT: .LBB10_2: # %else
|
|
; RV64ZVE32F-NEXT: ret
|
|
call void @llvm.masked.scatter.v1i16.v1p0(<1 x i16> %val, <1 x ptr> %ptrs, i32 2, <1 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_v2i16(<2 x i16> %val, <2 x ptr> %ptrs, <2 x i1> %m) {
|
|
; RV32V-LABEL: mscatter_v2i16:
|
|
; RV32V: # %bb.0:
|
|
; RV32V-NEXT: vsetivli zero, 2, e16, mf4, ta, ma
|
|
; RV32V-NEXT: vsoxei32.v v8, (zero), v9, v0.t
|
|
; RV32V-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_v2i16:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: vsetivli zero, 2, e16, mf4, ta, ma
|
|
; RV64V-NEXT: vsoxei64.v v8, (zero), v9, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV32ZVE32F-LABEL: mscatter_v2i16:
|
|
; RV32ZVE32F: # %bb.0:
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 2, e16, mf2, ta, ma
|
|
; RV32ZVE32F-NEXT: vsoxei32.v v8, (zero), v9, v0.t
|
|
; RV32ZVE32F-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_v2i16:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v0
|
|
; RV64ZVE32F-NEXT: andi a3, a2, 1
|
|
; RV64ZVE32F-NEXT: bnez a3, .LBB11_3
|
|
; RV64ZVE32F-NEXT: # %bb.1: # %else
|
|
; RV64ZVE32F-NEXT: andi a2, a2, 2
|
|
; RV64ZVE32F-NEXT: bnez a2, .LBB11_4
|
|
; RV64ZVE32F-NEXT: .LBB11_2: # %else2
|
|
; RV64ZVE32F-NEXT: ret
|
|
; RV64ZVE32F-NEXT: .LBB11_3: # %cond.store
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vse16.v v8, (a0)
|
|
; RV64ZVE32F-NEXT: andi a2, a2, 2
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB11_2
|
|
; RV64ZVE32F-NEXT: .LBB11_4: # %cond.store1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 1
|
|
; RV64ZVE32F-NEXT: vse16.v v8, (a1)
|
|
; RV64ZVE32F-NEXT: ret
|
|
call void @llvm.masked.scatter.v2i16.v2p0(<2 x i16> %val, <2 x ptr> %ptrs, i32 2, <2 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_v2i32_truncstore_v2i16(<2 x i32> %val, <2 x ptr> %ptrs, <2 x i1> %m) {
|
|
; RV32V-LABEL: mscatter_v2i32_truncstore_v2i16:
|
|
; RV32V: # %bb.0:
|
|
; RV32V-NEXT: vsetivli zero, 2, e16, mf4, ta, ma
|
|
; RV32V-NEXT: vnsrl.wi v8, v8, 0
|
|
; RV32V-NEXT: vsoxei32.v v8, (zero), v9, v0.t
|
|
; RV32V-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_v2i32_truncstore_v2i16:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: vsetivli zero, 2, e16, mf4, ta, ma
|
|
; RV64V-NEXT: vnsrl.wi v8, v8, 0
|
|
; RV64V-NEXT: vsoxei64.v v8, (zero), v9, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV32ZVE32F-LABEL: mscatter_v2i32_truncstore_v2i16:
|
|
; RV32ZVE32F: # %bb.0:
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 2, e16, mf2, ta, ma
|
|
; RV32ZVE32F-NEXT: vnsrl.wi v8, v8, 0
|
|
; RV32ZVE32F-NEXT: vsoxei32.v v8, (zero), v9, v0.t
|
|
; RV32ZVE32F-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_v2i32_truncstore_v2i16:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v0
|
|
; RV64ZVE32F-NEXT: andi a3, a2, 1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vnsrl.wi v8, v8, 0
|
|
; RV64ZVE32F-NEXT: bnez a3, .LBB12_3
|
|
; RV64ZVE32F-NEXT: # %bb.1: # %else
|
|
; RV64ZVE32F-NEXT: andi a2, a2, 2
|
|
; RV64ZVE32F-NEXT: bnez a2, .LBB12_4
|
|
; RV64ZVE32F-NEXT: .LBB12_2: # %else2
|
|
; RV64ZVE32F-NEXT: ret
|
|
; RV64ZVE32F-NEXT: .LBB12_3: # %cond.store
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vse16.v v8, (a0)
|
|
; RV64ZVE32F-NEXT: andi a2, a2, 2
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB12_2
|
|
; RV64ZVE32F-NEXT: .LBB12_4: # %cond.store1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 1
|
|
; RV64ZVE32F-NEXT: vse16.v v8, (a1)
|
|
; RV64ZVE32F-NEXT: ret
|
|
%tval = trunc <2 x i32> %val to <2 x i16>
|
|
call void @llvm.masked.scatter.v2i16.v2p0(<2 x i16> %tval, <2 x ptr> %ptrs, i32 2, <2 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_v2i64_truncstore_v2i16(<2 x i64> %val, <2 x ptr> %ptrs, <2 x i1> %m) {
|
|
; RV32V-LABEL: mscatter_v2i64_truncstore_v2i16:
|
|
; RV32V: # %bb.0:
|
|
; RV32V-NEXT: vsetivli zero, 2, e32, mf2, ta, ma
|
|
; RV32V-NEXT: vnsrl.wi v8, v8, 0
|
|
; RV32V-NEXT: vsetvli zero, zero, e16, mf4, ta, ma
|
|
; RV32V-NEXT: vnsrl.wi v8, v8, 0
|
|
; RV32V-NEXT: vsoxei32.v v8, (zero), v9, v0.t
|
|
; RV32V-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_v2i64_truncstore_v2i16:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: vsetivli zero, 2, e32, mf2, ta, ma
|
|
; RV64V-NEXT: vnsrl.wi v8, v8, 0
|
|
; RV64V-NEXT: vsetvli zero, zero, e16, mf4, ta, ma
|
|
; RV64V-NEXT: vnsrl.wi v8, v8, 0
|
|
; RV64V-NEXT: vsoxei64.v v8, (zero), v9, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV32ZVE32F-LABEL: mscatter_v2i64_truncstore_v2i16:
|
|
; RV32ZVE32F: # %bb.0:
|
|
; RV32ZVE32F-NEXT: lw a1, 8(a0)
|
|
; RV32ZVE32F-NEXT: lw a0, 0(a0)
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 2, e16, mf2, ta, ma
|
|
; RV32ZVE32F-NEXT: vmv.s.x v9, a1
|
|
; RV32ZVE32F-NEXT: vmv.s.x v10, a0
|
|
; RV32ZVE32F-NEXT: vslideup.vi v10, v9, 1
|
|
; RV32ZVE32F-NEXT: vsoxei32.v v10, (zero), v8, v0.t
|
|
; RV32ZVE32F-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_v2i64_truncstore_v2i16:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.s.x v9, a1
|
|
; RV64ZVE32F-NEXT: vmv.s.x v8, a0
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e8, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a0, v0
|
|
; RV64ZVE32F-NEXT: andi a1, a0, 1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslideup.vi v8, v9, 1
|
|
; RV64ZVE32F-NEXT: bnez a1, .LBB13_3
|
|
; RV64ZVE32F-NEXT: # %bb.1: # %else
|
|
; RV64ZVE32F-NEXT: andi a0, a0, 2
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB13_4
|
|
; RV64ZVE32F-NEXT: .LBB13_2: # %else2
|
|
; RV64ZVE32F-NEXT: ret
|
|
; RV64ZVE32F-NEXT: .LBB13_3: # %cond.store
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vse16.v v8, (a2)
|
|
; RV64ZVE32F-NEXT: andi a0, a0, 2
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB13_2
|
|
; RV64ZVE32F-NEXT: .LBB13_4: # %cond.store1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 1
|
|
; RV64ZVE32F-NEXT: vse16.v v8, (a3)
|
|
; RV64ZVE32F-NEXT: ret
|
|
%tval = trunc <2 x i64> %val to <2 x i16>
|
|
call void @llvm.masked.scatter.v2i16.v2p0(<2 x i16> %tval, <2 x ptr> %ptrs, i32 2, <2 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_v4i16(<4 x i16> %val, <4 x ptr> %ptrs, <4 x i1> %m) {
|
|
; RV32-LABEL: mscatter_v4i16:
|
|
; RV32: # %bb.0:
|
|
; RV32-NEXT: vsetivli zero, 4, e16, mf2, ta, ma
|
|
; RV32-NEXT: vsoxei32.v v8, (zero), v9, v0.t
|
|
; RV32-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_v4i16:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: vsetivli zero, 4, e16, mf2, ta, ma
|
|
; RV64V-NEXT: vsoxei64.v v8, (zero), v10, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_v4i16:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: ld a4, 8(a0)
|
|
; RV64ZVE32F-NEXT: ld a2, 16(a0)
|
|
; RV64ZVE32F-NEXT: ld a1, 24(a0)
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a3, v0
|
|
; RV64ZVE32F-NEXT: andi a5, a3, 1
|
|
; RV64ZVE32F-NEXT: bnez a5, .LBB14_5
|
|
; RV64ZVE32F-NEXT: # %bb.1: # %else
|
|
; RV64ZVE32F-NEXT: andi a0, a3, 2
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB14_6
|
|
; RV64ZVE32F-NEXT: .LBB14_2: # %else2
|
|
; RV64ZVE32F-NEXT: andi a0, a3, 4
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB14_7
|
|
; RV64ZVE32F-NEXT: .LBB14_3: # %else4
|
|
; RV64ZVE32F-NEXT: andi a3, a3, 8
|
|
; RV64ZVE32F-NEXT: bnez a3, .LBB14_8
|
|
; RV64ZVE32F-NEXT: .LBB14_4: # %else6
|
|
; RV64ZVE32F-NEXT: ret
|
|
; RV64ZVE32F-NEXT: .LBB14_5: # %cond.store
|
|
; RV64ZVE32F-NEXT: ld a0, 0(a0)
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vse16.v v8, (a0)
|
|
; RV64ZVE32F-NEXT: andi a0, a3, 2
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB14_2
|
|
; RV64ZVE32F-NEXT: .LBB14_6: # %cond.store1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 1
|
|
; RV64ZVE32F-NEXT: vse16.v v9, (a4)
|
|
; RV64ZVE32F-NEXT: andi a0, a3, 4
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB14_3
|
|
; RV64ZVE32F-NEXT: .LBB14_7: # %cond.store3
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 2
|
|
; RV64ZVE32F-NEXT: vse16.v v9, (a2)
|
|
; RV64ZVE32F-NEXT: andi a3, a3, 8
|
|
; RV64ZVE32F-NEXT: beqz a3, .LBB14_4
|
|
; RV64ZVE32F-NEXT: .LBB14_8: # %cond.store5
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 3
|
|
; RV64ZVE32F-NEXT: vse16.v v8, (a1)
|
|
; RV64ZVE32F-NEXT: ret
|
|
call void @llvm.masked.scatter.v4i16.v4p0(<4 x i16> %val, <4 x ptr> %ptrs, i32 2, <4 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_truemask_v4i16(<4 x i16> %val, <4 x ptr> %ptrs) {
|
|
; RV32-LABEL: mscatter_truemask_v4i16:
|
|
; RV32: # %bb.0:
|
|
; RV32-NEXT: vsetivli zero, 4, e16, mf2, ta, ma
|
|
; RV32-NEXT: vsoxei32.v v8, (zero), v9
|
|
; RV32-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_truemask_v4i16:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: vsetivli zero, 4, e16, mf2, ta, ma
|
|
; RV64V-NEXT: vsoxei64.v v8, (zero), v10
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_truemask_v4i16:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: ld a1, 0(a0)
|
|
; RV64ZVE32F-NEXT: ld a2, 8(a0)
|
|
; RV64ZVE32F-NEXT: ld a3, 16(a0)
|
|
; RV64ZVE32F-NEXT: ld a0, 24(a0)
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 1
|
|
; RV64ZVE32F-NEXT: vse16.v v8, (a1)
|
|
; RV64ZVE32F-NEXT: vse16.v v9, (a2)
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 2
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 3
|
|
; RV64ZVE32F-NEXT: vse16.v v9, (a3)
|
|
; RV64ZVE32F-NEXT: vse16.v v8, (a0)
|
|
; RV64ZVE32F-NEXT: ret
|
|
call void @llvm.masked.scatter.v4i16.v4p0(<4 x i16> %val, <4 x ptr> %ptrs, i32 2, <4 x i1> splat (i1 1))
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_falsemask_v4i16(<4 x i16> %val, <4 x ptr> %ptrs) {
|
|
; CHECK-LABEL: mscatter_falsemask_v4i16:
|
|
; CHECK: # %bb.0:
|
|
; CHECK-NEXT: ret
|
|
call void @llvm.masked.scatter.v4i16.v4p0(<4 x i16> %val, <4 x ptr> %ptrs, i32 2, <4 x i1> zeroinitializer)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_v8i16(<8 x i16> %val, <8 x ptr> %ptrs, <8 x i1> %m) {
|
|
; RV32-LABEL: mscatter_v8i16:
|
|
; RV32: # %bb.0:
|
|
; RV32-NEXT: vsetivli zero, 8, e16, m1, ta, ma
|
|
; RV32-NEXT: vsoxei32.v v8, (zero), v10, v0.t
|
|
; RV32-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_v8i16:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: vsetivli zero, 8, e16, m1, ta, ma
|
|
; RV64V-NEXT: vsoxei64.v v8, (zero), v12, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_v8i16:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: ld a3, 40(a0)
|
|
; RV64ZVE32F-NEXT: ld a2, 48(a0)
|
|
; RV64ZVE32F-NEXT: ld a1, 56(a0)
|
|
; RV64ZVE32F-NEXT: ld t0, 8(a0)
|
|
; RV64ZVE32F-NEXT: ld a7, 16(a0)
|
|
; RV64ZVE32F-NEXT: ld a6, 24(a0)
|
|
; RV64ZVE32F-NEXT: ld a5, 32(a0)
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a4, v0
|
|
; RV64ZVE32F-NEXT: andi t1, a4, 1
|
|
; RV64ZVE32F-NEXT: bnez t1, .LBB17_9
|
|
; RV64ZVE32F-NEXT: # %bb.1: # %else
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 2
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB17_10
|
|
; RV64ZVE32F-NEXT: .LBB17_2: # %else2
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 4
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB17_11
|
|
; RV64ZVE32F-NEXT: .LBB17_3: # %else4
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 8
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB17_12
|
|
; RV64ZVE32F-NEXT: .LBB17_4: # %else6
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 16
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB17_13
|
|
; RV64ZVE32F-NEXT: .LBB17_5: # %else8
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 32
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB17_14
|
|
; RV64ZVE32F-NEXT: .LBB17_6: # %else10
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 64
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB17_15
|
|
; RV64ZVE32F-NEXT: .LBB17_7: # %else12
|
|
; RV64ZVE32F-NEXT: andi a0, a4, -128
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB17_16
|
|
; RV64ZVE32F-NEXT: .LBB17_8: # %else14
|
|
; RV64ZVE32F-NEXT: ret
|
|
; RV64ZVE32F-NEXT: .LBB17_9: # %cond.store
|
|
; RV64ZVE32F-NEXT: ld a0, 0(a0)
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse16.v v8, (a0)
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 2
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB17_2
|
|
; RV64ZVE32F-NEXT: .LBB17_10: # %cond.store1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 1
|
|
; RV64ZVE32F-NEXT: vse16.v v9, (t0)
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 4
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB17_3
|
|
; RV64ZVE32F-NEXT: .LBB17_11: # %cond.store3
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 2
|
|
; RV64ZVE32F-NEXT: vse16.v v9, (a7)
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 8
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB17_4
|
|
; RV64ZVE32F-NEXT: .LBB17_12: # %cond.store5
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 3
|
|
; RV64ZVE32F-NEXT: vse16.v v9, (a6)
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 16
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB17_5
|
|
; RV64ZVE32F-NEXT: .LBB17_13: # %cond.store7
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 4
|
|
; RV64ZVE32F-NEXT: vse16.v v9, (a5)
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 32
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB17_6
|
|
; RV64ZVE32F-NEXT: .LBB17_14: # %cond.store9
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 5
|
|
; RV64ZVE32F-NEXT: vse16.v v9, (a3)
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 64
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB17_7
|
|
; RV64ZVE32F-NEXT: .LBB17_15: # %cond.store11
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 6
|
|
; RV64ZVE32F-NEXT: vse16.v v9, (a2)
|
|
; RV64ZVE32F-NEXT: andi a0, a4, -128
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB17_8
|
|
; RV64ZVE32F-NEXT: .LBB17_16: # %cond.store13
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 7
|
|
; RV64ZVE32F-NEXT: vse16.v v8, (a1)
|
|
; RV64ZVE32F-NEXT: ret
|
|
call void @llvm.masked.scatter.v8i16.v8p0(<8 x i16> %val, <8 x ptr> %ptrs, i32 2, <8 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_baseidx_v8i8_v8i16(<8 x i16> %val, ptr %base, <8 x i8> %idxs, <8 x i1> %m) {
|
|
; RV32-LABEL: mscatter_baseidx_v8i8_v8i16:
|
|
; RV32: # %bb.0:
|
|
; RV32-NEXT: vsetivli zero, 8, e32, m2, ta, ma
|
|
; RV32-NEXT: vsext.vf4 v10, v9
|
|
; RV32-NEXT: vadd.vv v10, v10, v10
|
|
; RV32-NEXT: vsetvli zero, zero, e16, m1, ta, ma
|
|
; RV32-NEXT: vsoxei32.v v8, (a0), v10, v0.t
|
|
; RV32-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_baseidx_v8i8_v8i16:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: vsetivli zero, 8, e64, m4, ta, ma
|
|
; RV64V-NEXT: vsext.vf8 v12, v9
|
|
; RV64V-NEXT: vadd.vv v12, v12, v12
|
|
; RV64V-NEXT: vsetvli zero, zero, e16, m1, ta, ma
|
|
; RV64V-NEXT: vsoxei64.v v8, (a0), v12, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_baseidx_v8i8_v8i16:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a1, v0
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 1
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB18_2
|
|
; RV64ZVE32F-NEXT: # %bb.1: # %cond.store
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v9
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse16.v v8, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB18_2: # %else
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 2
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB18_4
|
|
; RV64ZVE32F-NEXT: # %bb.3: # %cond.store1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v9, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v8, 1
|
|
; RV64ZVE32F-NEXT: vse16.v v10, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB18_4: # %else2
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 4
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v9, 2
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB18_6
|
|
; RV64ZVE32F-NEXT: # %bb.5: # %cond.store3
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v8, 2
|
|
; RV64ZVE32F-NEXT: vse16.v v11, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB18_6: # %else4
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 8
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB18_8
|
|
; RV64ZVE32F-NEXT: # %bb.7: # %cond.store5
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v10, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v8, 3
|
|
; RV64ZVE32F-NEXT: vse16.v v10, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB18_8: # %else6
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 16
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 4, e8, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v9, 4
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB18_10
|
|
; RV64ZVE32F-NEXT: # %bb.9: # %cond.store7
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v9
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v8, 4
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse16.v v10, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB18_10: # %else8
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 32
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB18_12
|
|
; RV64ZVE32F-NEXT: # %bb.11: # %cond.store9
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v9, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v8, 5
|
|
; RV64ZVE32F-NEXT: vse16.v v10, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB18_12: # %else10
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 64
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v9, 2
|
|
; RV64ZVE32F-NEXT: bnez a2, .LBB18_15
|
|
; RV64ZVE32F-NEXT: # %bb.13: # %else12
|
|
; RV64ZVE32F-NEXT: andi a1, a1, -128
|
|
; RV64ZVE32F-NEXT: bnez a1, .LBB18_16
|
|
; RV64ZVE32F-NEXT: .LBB18_14: # %else14
|
|
; RV64ZVE32F-NEXT: ret
|
|
; RV64ZVE32F-NEXT: .LBB18_15: # %cond.store11
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v9
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v8, 6
|
|
; RV64ZVE32F-NEXT: vse16.v v10, (a2)
|
|
; RV64ZVE32F-NEXT: andi a1, a1, -128
|
|
; RV64ZVE32F-NEXT: beqz a1, .LBB18_14
|
|
; RV64ZVE32F-NEXT: .LBB18_16: # %cond.store13
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v9, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a1, v9
|
|
; RV64ZVE32F-NEXT: slli a1, a1, 1
|
|
; RV64ZVE32F-NEXT: add a0, a0, a1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 7
|
|
; RV64ZVE32F-NEXT: vse16.v v8, (a0)
|
|
; RV64ZVE32F-NEXT: ret
|
|
%ptrs = getelementptr inbounds i16, ptr %base, <8 x i8> %idxs
|
|
call void @llvm.masked.scatter.v8i16.v8p0(<8 x i16> %val, <8 x ptr> %ptrs, i32 2, <8 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_baseidx_sext_v8i8_v8i16(<8 x i16> %val, ptr %base, <8 x i8> %idxs, <8 x i1> %m) {
|
|
; RV32-LABEL: mscatter_baseidx_sext_v8i8_v8i16:
|
|
; RV32: # %bb.0:
|
|
; RV32-NEXT: vsetivli zero, 8, e32, m2, ta, ma
|
|
; RV32-NEXT: vsext.vf4 v10, v9
|
|
; RV32-NEXT: vadd.vv v10, v10, v10
|
|
; RV32-NEXT: vsetvli zero, zero, e16, m1, ta, ma
|
|
; RV32-NEXT: vsoxei32.v v8, (a0), v10, v0.t
|
|
; RV32-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_baseidx_sext_v8i8_v8i16:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: vsetivli zero, 8, e64, m4, ta, ma
|
|
; RV64V-NEXT: vsext.vf8 v12, v9
|
|
; RV64V-NEXT: vadd.vv v12, v12, v12
|
|
; RV64V-NEXT: vsetvli zero, zero, e16, m1, ta, ma
|
|
; RV64V-NEXT: vsoxei64.v v8, (a0), v12, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_baseidx_sext_v8i8_v8i16:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a1, v0
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 1
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB19_2
|
|
; RV64ZVE32F-NEXT: # %bb.1: # %cond.store
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v9
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse16.v v8, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB19_2: # %else
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 2
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB19_4
|
|
; RV64ZVE32F-NEXT: # %bb.3: # %cond.store1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v9, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v8, 1
|
|
; RV64ZVE32F-NEXT: vse16.v v10, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB19_4: # %else2
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 4
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v9, 2
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB19_6
|
|
; RV64ZVE32F-NEXT: # %bb.5: # %cond.store3
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v8, 2
|
|
; RV64ZVE32F-NEXT: vse16.v v11, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB19_6: # %else4
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 8
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB19_8
|
|
; RV64ZVE32F-NEXT: # %bb.7: # %cond.store5
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v10, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v8, 3
|
|
; RV64ZVE32F-NEXT: vse16.v v10, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB19_8: # %else6
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 16
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 4, e8, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v9, 4
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB19_10
|
|
; RV64ZVE32F-NEXT: # %bb.9: # %cond.store7
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v9
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v8, 4
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse16.v v10, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB19_10: # %else8
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 32
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB19_12
|
|
; RV64ZVE32F-NEXT: # %bb.11: # %cond.store9
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v9, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v8, 5
|
|
; RV64ZVE32F-NEXT: vse16.v v10, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB19_12: # %else10
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 64
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v9, 2
|
|
; RV64ZVE32F-NEXT: bnez a2, .LBB19_15
|
|
; RV64ZVE32F-NEXT: # %bb.13: # %else12
|
|
; RV64ZVE32F-NEXT: andi a1, a1, -128
|
|
; RV64ZVE32F-NEXT: bnez a1, .LBB19_16
|
|
; RV64ZVE32F-NEXT: .LBB19_14: # %else14
|
|
; RV64ZVE32F-NEXT: ret
|
|
; RV64ZVE32F-NEXT: .LBB19_15: # %cond.store11
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v9
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v8, 6
|
|
; RV64ZVE32F-NEXT: vse16.v v10, (a2)
|
|
; RV64ZVE32F-NEXT: andi a1, a1, -128
|
|
; RV64ZVE32F-NEXT: beqz a1, .LBB19_14
|
|
; RV64ZVE32F-NEXT: .LBB19_16: # %cond.store13
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v9, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a1, v9
|
|
; RV64ZVE32F-NEXT: slli a1, a1, 1
|
|
; RV64ZVE32F-NEXT: add a0, a0, a1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 7
|
|
; RV64ZVE32F-NEXT: vse16.v v8, (a0)
|
|
; RV64ZVE32F-NEXT: ret
|
|
%eidxs = sext <8 x i8> %idxs to <8 x i16>
|
|
%ptrs = getelementptr inbounds i16, ptr %base, <8 x i16> %eidxs
|
|
call void @llvm.masked.scatter.v8i16.v8p0(<8 x i16> %val, <8 x ptr> %ptrs, i32 2, <8 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_baseidx_zext_v8i8_v8i16(<8 x i16> %val, ptr %base, <8 x i8> %idxs, <8 x i1> %m) {
|
|
; RV32-LABEL: mscatter_baseidx_zext_v8i8_v8i16:
|
|
; RV32: # %bb.0:
|
|
; RV32-NEXT: vsetivli zero, 8, e8, mf2, ta, ma
|
|
; RV32-NEXT: vwaddu.vv v10, v9, v9
|
|
; RV32-NEXT: vsetvli zero, zero, e16, m1, ta, ma
|
|
; RV32-NEXT: vsoxei16.v v8, (a0), v10, v0.t
|
|
; RV32-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_baseidx_zext_v8i8_v8i16:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: vsetivli zero, 8, e8, mf2, ta, ma
|
|
; RV64V-NEXT: vwaddu.vv v10, v9, v9
|
|
; RV64V-NEXT: vsetvli zero, zero, e16, m1, ta, ma
|
|
; RV64V-NEXT: vsoxei16.v v8, (a0), v10, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_baseidx_zext_v8i8_v8i16:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a1, v0
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 1
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB20_2
|
|
; RV64ZVE32F-NEXT: # %bb.1: # %cond.store
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v9
|
|
; RV64ZVE32F-NEXT: zext.b a2, a2
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse16.v v8, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB20_2: # %else
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 2
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB20_4
|
|
; RV64ZVE32F-NEXT: # %bb.3: # %cond.store1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v9, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: zext.b a2, a2
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v8, 1
|
|
; RV64ZVE32F-NEXT: vse16.v v10, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB20_4: # %else2
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 4
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v9, 2
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB20_6
|
|
; RV64ZVE32F-NEXT: # %bb.5: # %cond.store3
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: zext.b a2, a2
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v8, 2
|
|
; RV64ZVE32F-NEXT: vse16.v v11, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB20_6: # %else4
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 8
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB20_8
|
|
; RV64ZVE32F-NEXT: # %bb.7: # %cond.store5
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v10, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: zext.b a2, a2
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v8, 3
|
|
; RV64ZVE32F-NEXT: vse16.v v10, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB20_8: # %else6
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 16
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 4, e8, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v9, 4
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB20_10
|
|
; RV64ZVE32F-NEXT: # %bb.9: # %cond.store7
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v9
|
|
; RV64ZVE32F-NEXT: zext.b a2, a2
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v8, 4
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse16.v v10, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB20_10: # %else8
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 32
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB20_12
|
|
; RV64ZVE32F-NEXT: # %bb.11: # %cond.store9
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v9, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: zext.b a2, a2
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v8, 5
|
|
; RV64ZVE32F-NEXT: vse16.v v10, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB20_12: # %else10
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 64
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v9, 2
|
|
; RV64ZVE32F-NEXT: bnez a2, .LBB20_15
|
|
; RV64ZVE32F-NEXT: # %bb.13: # %else12
|
|
; RV64ZVE32F-NEXT: andi a1, a1, -128
|
|
; RV64ZVE32F-NEXT: bnez a1, .LBB20_16
|
|
; RV64ZVE32F-NEXT: .LBB20_14: # %else14
|
|
; RV64ZVE32F-NEXT: ret
|
|
; RV64ZVE32F-NEXT: .LBB20_15: # %cond.store11
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v9
|
|
; RV64ZVE32F-NEXT: zext.b a2, a2
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v8, 6
|
|
; RV64ZVE32F-NEXT: vse16.v v10, (a2)
|
|
; RV64ZVE32F-NEXT: andi a1, a1, -128
|
|
; RV64ZVE32F-NEXT: beqz a1, .LBB20_14
|
|
; RV64ZVE32F-NEXT: .LBB20_16: # %cond.store13
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v9, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a1, v9
|
|
; RV64ZVE32F-NEXT: zext.b a1, a1
|
|
; RV64ZVE32F-NEXT: slli a1, a1, 1
|
|
; RV64ZVE32F-NEXT: add a0, a0, a1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 7
|
|
; RV64ZVE32F-NEXT: vse16.v v8, (a0)
|
|
; RV64ZVE32F-NEXT: ret
|
|
%eidxs = zext <8 x i8> %idxs to <8 x i16>
|
|
%ptrs = getelementptr inbounds i16, ptr %base, <8 x i16> %eidxs
|
|
call void @llvm.masked.scatter.v8i16.v8p0(<8 x i16> %val, <8 x ptr> %ptrs, i32 2, <8 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_baseidx_v8i16(<8 x i16> %val, ptr %base, <8 x i16> %idxs, <8 x i1> %m) {
|
|
; RV32-LABEL: mscatter_baseidx_v8i16:
|
|
; RV32: # %bb.0:
|
|
; RV32-NEXT: vsetivli zero, 8, e16, m1, ta, ma
|
|
; RV32-NEXT: vwadd.vv v10, v9, v9
|
|
; RV32-NEXT: vsoxei32.v v8, (a0), v10, v0.t
|
|
; RV32-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_baseidx_v8i16:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: vsetivli zero, 8, e64, m4, ta, ma
|
|
; RV64V-NEXT: vsext.vf4 v12, v9
|
|
; RV64V-NEXT: vadd.vv v12, v12, v12
|
|
; RV64V-NEXT: vsetvli zero, zero, e16, m1, ta, ma
|
|
; RV64V-NEXT: vsoxei64.v v8, (a0), v12, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_baseidx_v8i16:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a1, v0
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 1
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB21_2
|
|
; RV64ZVE32F-NEXT: # %bb.1: # %cond.store
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v9
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vse16.v v8, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB21_2: # %else
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 2
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB21_4
|
|
; RV64ZVE32F-NEXT: # %bb.3: # %cond.store1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v9, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v8, 1
|
|
; RV64ZVE32F-NEXT: vse16.v v10, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB21_4: # %else2
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 4
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v9, 2
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB21_6
|
|
; RV64ZVE32F-NEXT: # %bb.5: # %cond.store3
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v8, 2
|
|
; RV64ZVE32F-NEXT: vse16.v v11, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB21_6: # %else4
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 8
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB21_8
|
|
; RV64ZVE32F-NEXT: # %bb.7: # %cond.store5
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v10, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v8, 3
|
|
; RV64ZVE32F-NEXT: vse16.v v10, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB21_8: # %else6
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 16
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 4, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v9, 4
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB21_10
|
|
; RV64ZVE32F-NEXT: # %bb.9: # %cond.store7
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v9
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v8, 4
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse16.v v10, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB21_10: # %else8
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 32
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB21_12
|
|
; RV64ZVE32F-NEXT: # %bb.11: # %cond.store9
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v9, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v8, 5
|
|
; RV64ZVE32F-NEXT: vse16.v v10, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB21_12: # %else10
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 64
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v9, 2
|
|
; RV64ZVE32F-NEXT: bnez a2, .LBB21_15
|
|
; RV64ZVE32F-NEXT: # %bb.13: # %else12
|
|
; RV64ZVE32F-NEXT: andi a1, a1, -128
|
|
; RV64ZVE32F-NEXT: bnez a1, .LBB21_16
|
|
; RV64ZVE32F-NEXT: .LBB21_14: # %else14
|
|
; RV64ZVE32F-NEXT: ret
|
|
; RV64ZVE32F-NEXT: .LBB21_15: # %cond.store11
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v9
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v8, 6
|
|
; RV64ZVE32F-NEXT: vse16.v v10, (a2)
|
|
; RV64ZVE32F-NEXT: andi a1, a1, -128
|
|
; RV64ZVE32F-NEXT: beqz a1, .LBB21_14
|
|
; RV64ZVE32F-NEXT: .LBB21_16: # %cond.store13
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v9, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a1, v9
|
|
; RV64ZVE32F-NEXT: slli a1, a1, 1
|
|
; RV64ZVE32F-NEXT: add a0, a0, a1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 7
|
|
; RV64ZVE32F-NEXT: vse16.v v8, (a0)
|
|
; RV64ZVE32F-NEXT: ret
|
|
%ptrs = getelementptr inbounds i16, ptr %base, <8 x i16> %idxs
|
|
call void @llvm.masked.scatter.v8i16.v8p0(<8 x i16> %val, <8 x ptr> %ptrs, i32 2, <8 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_v1i32(<1 x i32> %val, <1 x ptr> %ptrs, <1 x i1> %m) {
|
|
; RV32V-LABEL: mscatter_v1i32:
|
|
; RV32V: # %bb.0:
|
|
; RV32V-NEXT: vsetivli zero, 1, e32, mf2, ta, ma
|
|
; RV32V-NEXT: vsoxei32.v v8, (zero), v9, v0.t
|
|
; RV32V-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_v1i32:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: vsetivli zero, 1, e32, mf2, ta, ma
|
|
; RV64V-NEXT: vsoxei64.v v8, (zero), v9, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV32ZVE32F-LABEL: mscatter_v1i32:
|
|
; RV32ZVE32F: # %bb.0:
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vsoxei32.v v8, (zero), v9, v0.t
|
|
; RV32ZVE32F-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_v1i32:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: vsetvli a1, zero, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vfirst.m a1, v0
|
|
; RV64ZVE32F-NEXT: bnez a1, .LBB22_2
|
|
; RV64ZVE32F-NEXT: # %bb.1: # %cond.store
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v8, (a0)
|
|
; RV64ZVE32F-NEXT: .LBB22_2: # %else
|
|
; RV64ZVE32F-NEXT: ret
|
|
call void @llvm.masked.scatter.v1i32.v1p0(<1 x i32> %val, <1 x ptr> %ptrs, i32 4, <1 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_v2i32(<2 x i32> %val, <2 x ptr> %ptrs, <2 x i1> %m) {
|
|
; RV32V-LABEL: mscatter_v2i32:
|
|
; RV32V: # %bb.0:
|
|
; RV32V-NEXT: vsetivli zero, 2, e32, mf2, ta, ma
|
|
; RV32V-NEXT: vsoxei32.v v8, (zero), v9, v0.t
|
|
; RV32V-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_v2i32:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: vsetivli zero, 2, e32, mf2, ta, ma
|
|
; RV64V-NEXT: vsoxei64.v v8, (zero), v9, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV32ZVE32F-LABEL: mscatter_v2i32:
|
|
; RV32ZVE32F: # %bb.0:
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 2, e32, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vsoxei32.v v8, (zero), v9, v0.t
|
|
; RV32ZVE32F-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_v2i32:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v0
|
|
; RV64ZVE32F-NEXT: andi a3, a2, 1
|
|
; RV64ZVE32F-NEXT: bnez a3, .LBB23_3
|
|
; RV64ZVE32F-NEXT: # %bb.1: # %else
|
|
; RV64ZVE32F-NEXT: andi a2, a2, 2
|
|
; RV64ZVE32F-NEXT: bnez a2, .LBB23_4
|
|
; RV64ZVE32F-NEXT: .LBB23_2: # %else2
|
|
; RV64ZVE32F-NEXT: ret
|
|
; RV64ZVE32F-NEXT: .LBB23_3: # %cond.store
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v8, (a0)
|
|
; RV64ZVE32F-NEXT: andi a2, a2, 2
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB23_2
|
|
; RV64ZVE32F-NEXT: .LBB23_4: # %cond.store1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 1
|
|
; RV64ZVE32F-NEXT: vse32.v v8, (a1)
|
|
; RV64ZVE32F-NEXT: ret
|
|
call void @llvm.masked.scatter.v2i32.v2p0(<2 x i32> %val, <2 x ptr> %ptrs, i32 4, <2 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_v2i64_truncstore_v2i32(<2 x i64> %val, <2 x ptr> %ptrs, <2 x i1> %m) {
|
|
; RV32V-LABEL: mscatter_v2i64_truncstore_v2i32:
|
|
; RV32V: # %bb.0:
|
|
; RV32V-NEXT: vsetivli zero, 2, e32, mf2, ta, ma
|
|
; RV32V-NEXT: vnsrl.wi v8, v8, 0
|
|
; RV32V-NEXT: vsoxei32.v v8, (zero), v9, v0.t
|
|
; RV32V-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_v2i64_truncstore_v2i32:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: vsetivli zero, 2, e32, mf2, ta, ma
|
|
; RV64V-NEXT: vnsrl.wi v8, v8, 0
|
|
; RV64V-NEXT: vsoxei64.v v8, (zero), v9, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV32ZVE32F-LABEL: mscatter_v2i64_truncstore_v2i32:
|
|
; RV32ZVE32F: # %bb.0:
|
|
; RV32ZVE32F-NEXT: lw a1, 0(a0)
|
|
; RV32ZVE32F-NEXT: lw a0, 8(a0)
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 2, e32, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vmv.v.x v9, a1
|
|
; RV32ZVE32F-NEXT: vslide1down.vx v9, v9, a0
|
|
; RV32ZVE32F-NEXT: vsoxei32.v v9, (zero), v8, v0.t
|
|
; RV32ZVE32F-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_v2i64_truncstore_v2i32:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.v.x v8, a0
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a0, v0
|
|
; RV64ZVE32F-NEXT: andi a4, a0, 1
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslide1down.vx v8, v8, a1
|
|
; RV64ZVE32F-NEXT: bnez a4, .LBB24_3
|
|
; RV64ZVE32F-NEXT: # %bb.1: # %else
|
|
; RV64ZVE32F-NEXT: andi a0, a0, 2
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB24_4
|
|
; RV64ZVE32F-NEXT: .LBB24_2: # %else2
|
|
; RV64ZVE32F-NEXT: ret
|
|
; RV64ZVE32F-NEXT: .LBB24_3: # %cond.store
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v8, (a2)
|
|
; RV64ZVE32F-NEXT: andi a0, a0, 2
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB24_2
|
|
; RV64ZVE32F-NEXT: .LBB24_4: # %cond.store1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 1
|
|
; RV64ZVE32F-NEXT: vse32.v v8, (a3)
|
|
; RV64ZVE32F-NEXT: ret
|
|
%tval = trunc <2 x i64> %val to <2 x i32>
|
|
call void @llvm.masked.scatter.v2i32.v2p0(<2 x i32> %tval, <2 x ptr> %ptrs, i32 4, <2 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_v4i32(<4 x i32> %val, <4 x ptr> %ptrs, <4 x i1> %m) {
|
|
; RV32-LABEL: mscatter_v4i32:
|
|
; RV32: # %bb.0:
|
|
; RV32-NEXT: vsetivli zero, 4, e32, m1, ta, ma
|
|
; RV32-NEXT: vsoxei32.v v8, (zero), v9, v0.t
|
|
; RV32-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_v4i32:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: vsetivli zero, 4, e32, m1, ta, ma
|
|
; RV64V-NEXT: vsoxei64.v v8, (zero), v10, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_v4i32:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: ld a4, 8(a0)
|
|
; RV64ZVE32F-NEXT: ld a2, 16(a0)
|
|
; RV64ZVE32F-NEXT: ld a1, 24(a0)
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a3, v0
|
|
; RV64ZVE32F-NEXT: andi a5, a3, 1
|
|
; RV64ZVE32F-NEXT: bnez a5, .LBB25_5
|
|
; RV64ZVE32F-NEXT: # %bb.1: # %else
|
|
; RV64ZVE32F-NEXT: andi a0, a3, 2
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB25_6
|
|
; RV64ZVE32F-NEXT: .LBB25_2: # %else2
|
|
; RV64ZVE32F-NEXT: andi a0, a3, 4
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB25_7
|
|
; RV64ZVE32F-NEXT: .LBB25_3: # %else4
|
|
; RV64ZVE32F-NEXT: andi a3, a3, 8
|
|
; RV64ZVE32F-NEXT: bnez a3, .LBB25_8
|
|
; RV64ZVE32F-NEXT: .LBB25_4: # %else6
|
|
; RV64ZVE32F-NEXT: ret
|
|
; RV64ZVE32F-NEXT: .LBB25_5: # %cond.store
|
|
; RV64ZVE32F-NEXT: ld a0, 0(a0)
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v8, (a0)
|
|
; RV64ZVE32F-NEXT: andi a0, a3, 2
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB25_2
|
|
; RV64ZVE32F-NEXT: .LBB25_6: # %cond.store1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 1
|
|
; RV64ZVE32F-NEXT: vse32.v v9, (a4)
|
|
; RV64ZVE32F-NEXT: andi a0, a3, 4
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB25_3
|
|
; RV64ZVE32F-NEXT: .LBB25_7: # %cond.store3
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 2
|
|
; RV64ZVE32F-NEXT: vse32.v v9, (a2)
|
|
; RV64ZVE32F-NEXT: andi a3, a3, 8
|
|
; RV64ZVE32F-NEXT: beqz a3, .LBB25_4
|
|
; RV64ZVE32F-NEXT: .LBB25_8: # %cond.store5
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 3
|
|
; RV64ZVE32F-NEXT: vse32.v v8, (a1)
|
|
; RV64ZVE32F-NEXT: ret
|
|
call void @llvm.masked.scatter.v4i32.v4p0(<4 x i32> %val, <4 x ptr> %ptrs, i32 4, <4 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_truemask_v4i32(<4 x i32> %val, <4 x ptr> %ptrs) {
|
|
; RV32-LABEL: mscatter_truemask_v4i32:
|
|
; RV32: # %bb.0:
|
|
; RV32-NEXT: vsetivli zero, 4, e32, m1, ta, ma
|
|
; RV32-NEXT: vsoxei32.v v8, (zero), v9
|
|
; RV32-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_truemask_v4i32:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: vsetivli zero, 4, e32, m1, ta, ma
|
|
; RV64V-NEXT: vsoxei64.v v8, (zero), v10
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_truemask_v4i32:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: ld a1, 0(a0)
|
|
; RV64ZVE32F-NEXT: ld a2, 8(a0)
|
|
; RV64ZVE32F-NEXT: ld a3, 16(a0)
|
|
; RV64ZVE32F-NEXT: ld a0, 24(a0)
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 1
|
|
; RV64ZVE32F-NEXT: vse32.v v8, (a1)
|
|
; RV64ZVE32F-NEXT: vse32.v v9, (a2)
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 2
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 3
|
|
; RV64ZVE32F-NEXT: vse32.v v9, (a3)
|
|
; RV64ZVE32F-NEXT: vse32.v v8, (a0)
|
|
; RV64ZVE32F-NEXT: ret
|
|
call void @llvm.masked.scatter.v4i32.v4p0(<4 x i32> %val, <4 x ptr> %ptrs, i32 4, <4 x i1> splat (i1 1))
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_falsemask_v4i32(<4 x i32> %val, <4 x ptr> %ptrs) {
|
|
; CHECK-LABEL: mscatter_falsemask_v4i32:
|
|
; CHECK: # %bb.0:
|
|
; CHECK-NEXT: ret
|
|
call void @llvm.masked.scatter.v4i32.v4p0(<4 x i32> %val, <4 x ptr> %ptrs, i32 4, <4 x i1> zeroinitializer)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_v8i32(<8 x i32> %val, <8 x ptr> %ptrs, <8 x i1> %m) {
|
|
; RV32-LABEL: mscatter_v8i32:
|
|
; RV32: # %bb.0:
|
|
; RV32-NEXT: vsetivli zero, 8, e32, m2, ta, ma
|
|
; RV32-NEXT: vsoxei32.v v8, (zero), v10, v0.t
|
|
; RV32-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_v8i32:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: vsetivli zero, 8, e32, m2, ta, ma
|
|
; RV64V-NEXT: vsoxei64.v v8, (zero), v12, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_v8i32:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: ld a3, 40(a0)
|
|
; RV64ZVE32F-NEXT: ld a2, 48(a0)
|
|
; RV64ZVE32F-NEXT: ld a1, 56(a0)
|
|
; RV64ZVE32F-NEXT: ld t0, 8(a0)
|
|
; RV64ZVE32F-NEXT: ld a7, 16(a0)
|
|
; RV64ZVE32F-NEXT: ld a6, 24(a0)
|
|
; RV64ZVE32F-NEXT: ld a5, 32(a0)
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a4, v0
|
|
; RV64ZVE32F-NEXT: andi t1, a4, 1
|
|
; RV64ZVE32F-NEXT: bnez t1, .LBB28_9
|
|
; RV64ZVE32F-NEXT: # %bb.1: # %else
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 2
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB28_10
|
|
; RV64ZVE32F-NEXT: .LBB28_2: # %else2
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 4
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB28_11
|
|
; RV64ZVE32F-NEXT: .LBB28_3: # %else4
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 8
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB28_12
|
|
; RV64ZVE32F-NEXT: .LBB28_4: # %else6
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 16
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB28_13
|
|
; RV64ZVE32F-NEXT: .LBB28_5: # %else8
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 32
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB28_14
|
|
; RV64ZVE32F-NEXT: .LBB28_6: # %else10
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 64
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB28_15
|
|
; RV64ZVE32F-NEXT: .LBB28_7: # %else12
|
|
; RV64ZVE32F-NEXT: andi a0, a4, -128
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB28_16
|
|
; RV64ZVE32F-NEXT: .LBB28_8: # %else14
|
|
; RV64ZVE32F-NEXT: ret
|
|
; RV64ZVE32F-NEXT: .LBB28_9: # %cond.store
|
|
; RV64ZVE32F-NEXT: ld a0, 0(a0)
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v8, (a0)
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 2
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB28_2
|
|
; RV64ZVE32F-NEXT: .LBB28_10: # %cond.store1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v8, 1
|
|
; RV64ZVE32F-NEXT: vse32.v v10, (t0)
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 4
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB28_3
|
|
; RV64ZVE32F-NEXT: .LBB28_11: # %cond.store3
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v8, 2
|
|
; RV64ZVE32F-NEXT: vse32.v v10, (a7)
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 8
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB28_4
|
|
; RV64ZVE32F-NEXT: .LBB28_12: # %cond.store5
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v8, 3
|
|
; RV64ZVE32F-NEXT: vse32.v v10, (a6)
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 16
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB28_5
|
|
; RV64ZVE32F-NEXT: .LBB28_13: # %cond.store7
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v8, 4
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v10, (a5)
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 32
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB28_6
|
|
; RV64ZVE32F-NEXT: .LBB28_14: # %cond.store9
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v8, 5
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v10, (a3)
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 64
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB28_7
|
|
; RV64ZVE32F-NEXT: .LBB28_15: # %cond.store11
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v8, 6
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v10, (a2)
|
|
; RV64ZVE32F-NEXT: andi a0, a4, -128
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB28_8
|
|
; RV64ZVE32F-NEXT: .LBB28_16: # %cond.store13
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 7
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v8, (a1)
|
|
; RV64ZVE32F-NEXT: ret
|
|
call void @llvm.masked.scatter.v8i32.v8p0(<8 x i32> %val, <8 x ptr> %ptrs, i32 4, <8 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_baseidx_v8i8_v8i32(<8 x i32> %val, ptr %base, <8 x i8> %idxs, <8 x i1> %m) {
|
|
; RV32-LABEL: mscatter_baseidx_v8i8_v8i32:
|
|
; RV32: # %bb.0:
|
|
; RV32-NEXT: vsetivli zero, 8, e32, m2, ta, ma
|
|
; RV32-NEXT: vsext.vf4 v12, v10
|
|
; RV32-NEXT: vsll.vi v10, v12, 2
|
|
; RV32-NEXT: vsoxei32.v v8, (a0), v10, v0.t
|
|
; RV32-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_baseidx_v8i8_v8i32:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: vsetivli zero, 8, e64, m4, ta, ma
|
|
; RV64V-NEXT: vsext.vf8 v12, v10
|
|
; RV64V-NEXT: vsll.vi v12, v12, 2
|
|
; RV64V-NEXT: vsetvli zero, zero, e32, m2, ta, ma
|
|
; RV64V-NEXT: vsoxei64.v v8, (a0), v12, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_baseidx_v8i8_v8i32:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a1, v0
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 1
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB29_2
|
|
; RV64ZVE32F-NEXT: # %bb.1: # %cond.store
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 2
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v8, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB29_2: # %else
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 2
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB29_4
|
|
; RV64ZVE32F-NEXT: # %bb.3: # %cond.store1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v10, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v11
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 2
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v8, 1
|
|
; RV64ZVE32F-NEXT: vse32.v v11, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB29_4: # %else2
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 4
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v10, 2
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB29_6
|
|
; RV64ZVE32F-NEXT: # %bb.5: # %cond.store3
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v11
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 2
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v8, 2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v12, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB29_6: # %else4
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 8
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB29_8
|
|
; RV64ZVE32F-NEXT: # %bb.7: # %cond.store5
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v11, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v11
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 2
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v8, 3
|
|
; RV64ZVE32F-NEXT: vse32.v v11, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB29_8: # %else6
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 16
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 4, e8, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v10, 4
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB29_10
|
|
; RV64ZVE32F-NEXT: # %bb.9: # %cond.store7
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 2
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e32, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v8, 4
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v12, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB29_10: # %else8
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 32
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB29_12
|
|
; RV64ZVE32F-NEXT: # %bb.11: # %cond.store9
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v10, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v11
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 2
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v8, 5
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v12, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB29_12: # %else10
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 64
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v10, 2
|
|
; RV64ZVE32F-NEXT: bnez a2, .LBB29_15
|
|
; RV64ZVE32F-NEXT: # %bb.13: # %else12
|
|
; RV64ZVE32F-NEXT: andi a1, a1, -128
|
|
; RV64ZVE32F-NEXT: bnez a1, .LBB29_16
|
|
; RV64ZVE32F-NEXT: .LBB29_14: # %else14
|
|
; RV64ZVE32F-NEXT: ret
|
|
; RV64ZVE32F-NEXT: .LBB29_15: # %cond.store11
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 2
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v8, 6
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v12, (a2)
|
|
; RV64ZVE32F-NEXT: andi a1, a1, -128
|
|
; RV64ZVE32F-NEXT: beqz a1, .LBB29_14
|
|
; RV64ZVE32F-NEXT: .LBB29_16: # %cond.store13
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v10, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a1, v10
|
|
; RV64ZVE32F-NEXT: slli a1, a1, 2
|
|
; RV64ZVE32F-NEXT: add a0, a0, a1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 7
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v8, (a0)
|
|
; RV64ZVE32F-NEXT: ret
|
|
%ptrs = getelementptr inbounds i32, ptr %base, <8 x i8> %idxs
|
|
call void @llvm.masked.scatter.v8i32.v8p0(<8 x i32> %val, <8 x ptr> %ptrs, i32 4, <8 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_baseidx_sext_v8i8_v8i32(<8 x i32> %val, ptr %base, <8 x i8> %idxs, <8 x i1> %m) {
|
|
; RV32-LABEL: mscatter_baseidx_sext_v8i8_v8i32:
|
|
; RV32: # %bb.0:
|
|
; RV32-NEXT: vsetivli zero, 8, e32, m2, ta, ma
|
|
; RV32-NEXT: vsext.vf4 v12, v10
|
|
; RV32-NEXT: vsll.vi v10, v12, 2
|
|
; RV32-NEXT: vsoxei32.v v8, (a0), v10, v0.t
|
|
; RV32-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_baseidx_sext_v8i8_v8i32:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: vsetivli zero, 8, e64, m4, ta, ma
|
|
; RV64V-NEXT: vsext.vf8 v12, v10
|
|
; RV64V-NEXT: vsll.vi v12, v12, 2
|
|
; RV64V-NEXT: vsetvli zero, zero, e32, m2, ta, ma
|
|
; RV64V-NEXT: vsoxei64.v v8, (a0), v12, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_baseidx_sext_v8i8_v8i32:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a1, v0
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 1
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB30_2
|
|
; RV64ZVE32F-NEXT: # %bb.1: # %cond.store
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 2
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v8, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB30_2: # %else
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 2
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB30_4
|
|
; RV64ZVE32F-NEXT: # %bb.3: # %cond.store1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v10, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v11
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 2
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v8, 1
|
|
; RV64ZVE32F-NEXT: vse32.v v11, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB30_4: # %else2
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 4
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v10, 2
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB30_6
|
|
; RV64ZVE32F-NEXT: # %bb.5: # %cond.store3
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v11
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 2
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v8, 2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v12, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB30_6: # %else4
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 8
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB30_8
|
|
; RV64ZVE32F-NEXT: # %bb.7: # %cond.store5
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v11, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v11
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 2
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v8, 3
|
|
; RV64ZVE32F-NEXT: vse32.v v11, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB30_8: # %else6
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 16
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 4, e8, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v10, 4
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB30_10
|
|
; RV64ZVE32F-NEXT: # %bb.9: # %cond.store7
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 2
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e32, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v8, 4
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v12, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB30_10: # %else8
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 32
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB30_12
|
|
; RV64ZVE32F-NEXT: # %bb.11: # %cond.store9
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v10, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v11
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 2
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v8, 5
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v12, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB30_12: # %else10
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 64
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v10, 2
|
|
; RV64ZVE32F-NEXT: bnez a2, .LBB30_15
|
|
; RV64ZVE32F-NEXT: # %bb.13: # %else12
|
|
; RV64ZVE32F-NEXT: andi a1, a1, -128
|
|
; RV64ZVE32F-NEXT: bnez a1, .LBB30_16
|
|
; RV64ZVE32F-NEXT: .LBB30_14: # %else14
|
|
; RV64ZVE32F-NEXT: ret
|
|
; RV64ZVE32F-NEXT: .LBB30_15: # %cond.store11
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 2
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v8, 6
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v12, (a2)
|
|
; RV64ZVE32F-NEXT: andi a1, a1, -128
|
|
; RV64ZVE32F-NEXT: beqz a1, .LBB30_14
|
|
; RV64ZVE32F-NEXT: .LBB30_16: # %cond.store13
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v10, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a1, v10
|
|
; RV64ZVE32F-NEXT: slli a1, a1, 2
|
|
; RV64ZVE32F-NEXT: add a0, a0, a1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 7
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v8, (a0)
|
|
; RV64ZVE32F-NEXT: ret
|
|
%eidxs = sext <8 x i8> %idxs to <8 x i32>
|
|
%ptrs = getelementptr inbounds i32, ptr %base, <8 x i32> %eidxs
|
|
call void @llvm.masked.scatter.v8i32.v8p0(<8 x i32> %val, <8 x ptr> %ptrs, i32 4, <8 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_baseidx_zext_v8i8_v8i32(<8 x i32> %val, ptr %base, <8 x i8> %idxs, <8 x i1> %m) {
|
|
; RV32-LABEL: mscatter_baseidx_zext_v8i8_v8i32:
|
|
; RV32: # %bb.0:
|
|
; RV32-NEXT: li a1, 4
|
|
; RV32-NEXT: vsetivli zero, 8, e8, mf2, ta, ma
|
|
; RV32-NEXT: vwmulu.vx v11, v10, a1
|
|
; RV32-NEXT: vsetvli zero, zero, e32, m2, ta, ma
|
|
; RV32-NEXT: vsoxei16.v v8, (a0), v11, v0.t
|
|
; RV32-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_baseidx_zext_v8i8_v8i32:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: li a1, 4
|
|
; RV64V-NEXT: vsetivli zero, 8, e8, mf2, ta, ma
|
|
; RV64V-NEXT: vwmulu.vx v11, v10, a1
|
|
; RV64V-NEXT: vsetvli zero, zero, e32, m2, ta, ma
|
|
; RV64V-NEXT: vsoxei16.v v8, (a0), v11, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_baseidx_zext_v8i8_v8i32:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a1, v0
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 1
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB31_2
|
|
; RV64ZVE32F-NEXT: # %bb.1: # %cond.store
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: zext.b a2, a2
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 2
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v8, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB31_2: # %else
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 2
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB31_4
|
|
; RV64ZVE32F-NEXT: # %bb.3: # %cond.store1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v10, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v11
|
|
; RV64ZVE32F-NEXT: zext.b a2, a2
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 2
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v8, 1
|
|
; RV64ZVE32F-NEXT: vse32.v v11, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB31_4: # %else2
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 4
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v10, 2
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB31_6
|
|
; RV64ZVE32F-NEXT: # %bb.5: # %cond.store3
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v11
|
|
; RV64ZVE32F-NEXT: zext.b a2, a2
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 2
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v8, 2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v12, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB31_6: # %else4
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 8
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB31_8
|
|
; RV64ZVE32F-NEXT: # %bb.7: # %cond.store5
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v11, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v11
|
|
; RV64ZVE32F-NEXT: zext.b a2, a2
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 2
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v8, 3
|
|
; RV64ZVE32F-NEXT: vse32.v v11, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB31_8: # %else6
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 16
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 4, e8, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v10, 4
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB31_10
|
|
; RV64ZVE32F-NEXT: # %bb.9: # %cond.store7
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: zext.b a2, a2
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 2
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e32, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v8, 4
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v12, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB31_10: # %else8
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 32
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB31_12
|
|
; RV64ZVE32F-NEXT: # %bb.11: # %cond.store9
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v10, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v11
|
|
; RV64ZVE32F-NEXT: zext.b a2, a2
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 2
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v8, 5
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v12, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB31_12: # %else10
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 64
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v10, 2
|
|
; RV64ZVE32F-NEXT: bnez a2, .LBB31_15
|
|
; RV64ZVE32F-NEXT: # %bb.13: # %else12
|
|
; RV64ZVE32F-NEXT: andi a1, a1, -128
|
|
; RV64ZVE32F-NEXT: bnez a1, .LBB31_16
|
|
; RV64ZVE32F-NEXT: .LBB31_14: # %else14
|
|
; RV64ZVE32F-NEXT: ret
|
|
; RV64ZVE32F-NEXT: .LBB31_15: # %cond.store11
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: zext.b a2, a2
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 2
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v8, 6
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v12, (a2)
|
|
; RV64ZVE32F-NEXT: andi a1, a1, -128
|
|
; RV64ZVE32F-NEXT: beqz a1, .LBB31_14
|
|
; RV64ZVE32F-NEXT: .LBB31_16: # %cond.store13
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v10, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a1, v10
|
|
; RV64ZVE32F-NEXT: zext.b a1, a1
|
|
; RV64ZVE32F-NEXT: slli a1, a1, 2
|
|
; RV64ZVE32F-NEXT: add a0, a0, a1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 7
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v8, (a0)
|
|
; RV64ZVE32F-NEXT: ret
|
|
%eidxs = zext <8 x i8> %idxs to <8 x i32>
|
|
%ptrs = getelementptr inbounds i32, ptr %base, <8 x i32> %eidxs
|
|
call void @llvm.masked.scatter.v8i32.v8p0(<8 x i32> %val, <8 x ptr> %ptrs, i32 4, <8 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_baseidx_v8i16_v8i32(<8 x i32> %val, ptr %base, <8 x i16> %idxs, <8 x i1> %m) {
|
|
; RV32-LABEL: mscatter_baseidx_v8i16_v8i32:
|
|
; RV32: # %bb.0:
|
|
; RV32-NEXT: li a1, 4
|
|
; RV32-NEXT: vsetivli zero, 8, e16, m1, ta, ma
|
|
; RV32-NEXT: vwmulsu.vx v12, v10, a1
|
|
; RV32-NEXT: vsetvli zero, zero, e32, m2, ta, ma
|
|
; RV32-NEXT: vsoxei32.v v8, (a0), v12, v0.t
|
|
; RV32-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_baseidx_v8i16_v8i32:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: vsetivli zero, 8, e64, m4, ta, ma
|
|
; RV64V-NEXT: vsext.vf4 v12, v10
|
|
; RV64V-NEXT: vsll.vi v12, v12, 2
|
|
; RV64V-NEXT: vsetvli zero, zero, e32, m2, ta, ma
|
|
; RV64V-NEXT: vsoxei64.v v8, (a0), v12, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_baseidx_v8i16_v8i32:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a1, v0
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 1
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB32_2
|
|
; RV64ZVE32F-NEXT: # %bb.1: # %cond.store
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e16, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 2
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v8, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB32_2: # %else
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 2
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB32_4
|
|
; RV64ZVE32F-NEXT: # %bb.3: # %cond.store1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v10, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v11
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 2
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v8, 1
|
|
; RV64ZVE32F-NEXT: vse32.v v11, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB32_4: # %else2
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 4
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v10, 2
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB32_6
|
|
; RV64ZVE32F-NEXT: # %bb.5: # %cond.store3
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v11
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 2
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v8, 2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v12, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB32_6: # %else4
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 8
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB32_8
|
|
; RV64ZVE32F-NEXT: # %bb.7: # %cond.store5
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v11, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v11
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 2
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v8, 3
|
|
; RV64ZVE32F-NEXT: vse32.v v11, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB32_8: # %else6
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 16
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 4, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v10, 4
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB32_10
|
|
; RV64ZVE32F-NEXT: # %bb.9: # %cond.store7
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 2
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e32, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v8, 4
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v12, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB32_10: # %else8
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 32
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB32_12
|
|
; RV64ZVE32F-NEXT: # %bb.11: # %cond.store9
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v10, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v11
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 2
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v8, 5
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v12, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB32_12: # %else10
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 64
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v10, 2
|
|
; RV64ZVE32F-NEXT: bnez a2, .LBB32_15
|
|
; RV64ZVE32F-NEXT: # %bb.13: # %else12
|
|
; RV64ZVE32F-NEXT: andi a1, a1, -128
|
|
; RV64ZVE32F-NEXT: bnez a1, .LBB32_16
|
|
; RV64ZVE32F-NEXT: .LBB32_14: # %else14
|
|
; RV64ZVE32F-NEXT: ret
|
|
; RV64ZVE32F-NEXT: .LBB32_15: # %cond.store11
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 2
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v8, 6
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v12, (a2)
|
|
; RV64ZVE32F-NEXT: andi a1, a1, -128
|
|
; RV64ZVE32F-NEXT: beqz a1, .LBB32_14
|
|
; RV64ZVE32F-NEXT: .LBB32_16: # %cond.store13
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v10, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a1, v10
|
|
; RV64ZVE32F-NEXT: slli a1, a1, 2
|
|
; RV64ZVE32F-NEXT: add a0, a0, a1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 7
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v8, (a0)
|
|
; RV64ZVE32F-NEXT: ret
|
|
%ptrs = getelementptr inbounds i32, ptr %base, <8 x i16> %idxs
|
|
call void @llvm.masked.scatter.v8i32.v8p0(<8 x i32> %val, <8 x ptr> %ptrs, i32 4, <8 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_baseidx_sext_v8i16_v8i32(<8 x i32> %val, ptr %base, <8 x i16> %idxs, <8 x i1> %m) {
|
|
; RV32-LABEL: mscatter_baseidx_sext_v8i16_v8i32:
|
|
; RV32: # %bb.0:
|
|
; RV32-NEXT: li a1, 4
|
|
; RV32-NEXT: vsetivli zero, 8, e16, m1, ta, ma
|
|
; RV32-NEXT: vwmulsu.vx v12, v10, a1
|
|
; RV32-NEXT: vsetvli zero, zero, e32, m2, ta, ma
|
|
; RV32-NEXT: vsoxei32.v v8, (a0), v12, v0.t
|
|
; RV32-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_baseidx_sext_v8i16_v8i32:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: vsetivli zero, 8, e64, m4, ta, ma
|
|
; RV64V-NEXT: vsext.vf4 v12, v10
|
|
; RV64V-NEXT: vsll.vi v12, v12, 2
|
|
; RV64V-NEXT: vsetvli zero, zero, e32, m2, ta, ma
|
|
; RV64V-NEXT: vsoxei64.v v8, (a0), v12, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_baseidx_sext_v8i16_v8i32:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a1, v0
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 1
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB33_2
|
|
; RV64ZVE32F-NEXT: # %bb.1: # %cond.store
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e16, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 2
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v8, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB33_2: # %else
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 2
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB33_4
|
|
; RV64ZVE32F-NEXT: # %bb.3: # %cond.store1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v10, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v11
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 2
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v8, 1
|
|
; RV64ZVE32F-NEXT: vse32.v v11, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB33_4: # %else2
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 4
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v10, 2
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB33_6
|
|
; RV64ZVE32F-NEXT: # %bb.5: # %cond.store3
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v11
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 2
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v8, 2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v12, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB33_6: # %else4
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 8
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB33_8
|
|
; RV64ZVE32F-NEXT: # %bb.7: # %cond.store5
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v11, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v11
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 2
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v8, 3
|
|
; RV64ZVE32F-NEXT: vse32.v v11, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB33_8: # %else6
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 16
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 4, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v10, 4
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB33_10
|
|
; RV64ZVE32F-NEXT: # %bb.9: # %cond.store7
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 2
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e32, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v8, 4
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v12, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB33_10: # %else8
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 32
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB33_12
|
|
; RV64ZVE32F-NEXT: # %bb.11: # %cond.store9
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v10, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v11
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 2
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v8, 5
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v12, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB33_12: # %else10
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 64
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v10, 2
|
|
; RV64ZVE32F-NEXT: bnez a2, .LBB33_15
|
|
; RV64ZVE32F-NEXT: # %bb.13: # %else12
|
|
; RV64ZVE32F-NEXT: andi a1, a1, -128
|
|
; RV64ZVE32F-NEXT: bnez a1, .LBB33_16
|
|
; RV64ZVE32F-NEXT: .LBB33_14: # %else14
|
|
; RV64ZVE32F-NEXT: ret
|
|
; RV64ZVE32F-NEXT: .LBB33_15: # %cond.store11
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 2
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v8, 6
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v12, (a2)
|
|
; RV64ZVE32F-NEXT: andi a1, a1, -128
|
|
; RV64ZVE32F-NEXT: beqz a1, .LBB33_14
|
|
; RV64ZVE32F-NEXT: .LBB33_16: # %cond.store13
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v10, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a1, v10
|
|
; RV64ZVE32F-NEXT: slli a1, a1, 2
|
|
; RV64ZVE32F-NEXT: add a0, a0, a1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 7
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v8, (a0)
|
|
; RV64ZVE32F-NEXT: ret
|
|
%eidxs = sext <8 x i16> %idxs to <8 x i32>
|
|
%ptrs = getelementptr inbounds i32, ptr %base, <8 x i32> %eidxs
|
|
call void @llvm.masked.scatter.v8i32.v8p0(<8 x i32> %val, <8 x ptr> %ptrs, i32 4, <8 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_baseidx_zext_v8i16_v8i32(<8 x i32> %val, ptr %base, <8 x i16> %idxs, <8 x i1> %m) {
|
|
; RV32-LABEL: mscatter_baseidx_zext_v8i16_v8i32:
|
|
; RV32: # %bb.0:
|
|
; RV32-NEXT: li a1, 4
|
|
; RV32-NEXT: vsetivli zero, 8, e16, m1, ta, ma
|
|
; RV32-NEXT: vwmulu.vx v12, v10, a1
|
|
; RV32-NEXT: vsetvli zero, zero, e32, m2, ta, ma
|
|
; RV32-NEXT: vsoxei32.v v8, (a0), v12, v0.t
|
|
; RV32-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_baseidx_zext_v8i16_v8i32:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: li a1, 4
|
|
; RV64V-NEXT: vsetivli zero, 8, e16, m1, ta, ma
|
|
; RV64V-NEXT: vwmulu.vx v12, v10, a1
|
|
; RV64V-NEXT: vsetvli zero, zero, e32, m2, ta, ma
|
|
; RV64V-NEXT: vsoxei32.v v8, (a0), v12, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_baseidx_zext_v8i16_v8i32:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a1, v0
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 1
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB34_2
|
|
; RV64ZVE32F-NEXT: # %bb.1: # %cond.store
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e16, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 48
|
|
; RV64ZVE32F-NEXT: srli a2, a2, 46
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v8, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB34_2: # %else
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 2
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB34_4
|
|
; RV64ZVE32F-NEXT: # %bb.3: # %cond.store1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v10, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v11
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 48
|
|
; RV64ZVE32F-NEXT: srli a2, a2, 46
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v8, 1
|
|
; RV64ZVE32F-NEXT: vse32.v v11, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB34_4: # %else2
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 4
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v10, 2
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB34_6
|
|
; RV64ZVE32F-NEXT: # %bb.5: # %cond.store3
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v11
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 48
|
|
; RV64ZVE32F-NEXT: srli a2, a2, 46
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v8, 2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v12, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB34_6: # %else4
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 8
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB34_8
|
|
; RV64ZVE32F-NEXT: # %bb.7: # %cond.store5
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v11, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v11
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 48
|
|
; RV64ZVE32F-NEXT: srli a2, a2, 46
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v8, 3
|
|
; RV64ZVE32F-NEXT: vse32.v v11, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB34_8: # %else6
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 16
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 4, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v10, 4
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB34_10
|
|
; RV64ZVE32F-NEXT: # %bb.9: # %cond.store7
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 48
|
|
; RV64ZVE32F-NEXT: srli a2, a2, 46
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e32, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v8, 4
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v12, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB34_10: # %else8
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 32
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB34_12
|
|
; RV64ZVE32F-NEXT: # %bb.11: # %cond.store9
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v10, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v11
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 48
|
|
; RV64ZVE32F-NEXT: srli a2, a2, 46
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v8, 5
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v12, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB34_12: # %else10
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 64
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v10, 2
|
|
; RV64ZVE32F-NEXT: bnez a2, .LBB34_15
|
|
; RV64ZVE32F-NEXT: # %bb.13: # %else12
|
|
; RV64ZVE32F-NEXT: andi a1, a1, -128
|
|
; RV64ZVE32F-NEXT: bnez a1, .LBB34_16
|
|
; RV64ZVE32F-NEXT: .LBB34_14: # %else14
|
|
; RV64ZVE32F-NEXT: ret
|
|
; RV64ZVE32F-NEXT: .LBB34_15: # %cond.store11
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 48
|
|
; RV64ZVE32F-NEXT: srli a2, a2, 46
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v8, 6
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v12, (a2)
|
|
; RV64ZVE32F-NEXT: andi a1, a1, -128
|
|
; RV64ZVE32F-NEXT: beqz a1, .LBB34_14
|
|
; RV64ZVE32F-NEXT: .LBB34_16: # %cond.store13
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v10, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a1, v10
|
|
; RV64ZVE32F-NEXT: slli a1, a1, 48
|
|
; RV64ZVE32F-NEXT: srli a1, a1, 46
|
|
; RV64ZVE32F-NEXT: add a0, a0, a1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 7
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v8, (a0)
|
|
; RV64ZVE32F-NEXT: ret
|
|
%eidxs = zext <8 x i16> %idxs to <8 x i32>
|
|
%ptrs = getelementptr inbounds i32, ptr %base, <8 x i32> %eidxs
|
|
call void @llvm.masked.scatter.v8i32.v8p0(<8 x i32> %val, <8 x ptr> %ptrs, i32 4, <8 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_baseidx_v8i32(<8 x i32> %val, ptr %base, <8 x i32> %idxs, <8 x i1> %m) {
|
|
; RV32-LABEL: mscatter_baseidx_v8i32:
|
|
; RV32: # %bb.0:
|
|
; RV32-NEXT: vsetivli zero, 8, e32, m2, ta, ma
|
|
; RV32-NEXT: vsll.vi v10, v10, 2
|
|
; RV32-NEXT: vsoxei32.v v8, (a0), v10, v0.t
|
|
; RV32-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_baseidx_v8i32:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: li a1, 4
|
|
; RV64V-NEXT: vsetivli zero, 8, e32, m2, ta, ma
|
|
; RV64V-NEXT: vwmulsu.vx v12, v10, a1
|
|
; RV64V-NEXT: vsoxei64.v v8, (a0), v12, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_baseidx_v8i32:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a1, v0
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 1
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB35_2
|
|
; RV64ZVE32F-NEXT: # %bb.1: # %cond.store
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 2
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vse32.v v8, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB35_2: # %else
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 2
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB35_4
|
|
; RV64ZVE32F-NEXT: # %bb.3: # %cond.store1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v10, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v12
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 2
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v8, 1
|
|
; RV64ZVE32F-NEXT: vse32.v v12, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB35_4: # %else2
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 4
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v10, 2
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB35_6
|
|
; RV64ZVE32F-NEXT: # %bb.5: # %cond.store3
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v12
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 2
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v13, v8, 2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v13, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB35_6: # %else4
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 8
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 4, e32, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v10, 4
|
|
; RV64ZVE32F-NEXT: bnez a2, .LBB35_13
|
|
; RV64ZVE32F-NEXT: # %bb.7: # %else6
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 16
|
|
; RV64ZVE32F-NEXT: bnez a2, .LBB35_14
|
|
; RV64ZVE32F-NEXT: .LBB35_8: # %else8
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 32
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB35_10
|
|
; RV64ZVE32F-NEXT: .LBB35_9: # %cond.store9
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v10, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v11
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 2
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v8, 5
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v12, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB35_10: # %else10
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 64
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v10, 2
|
|
; RV64ZVE32F-NEXT: bnez a2, .LBB35_15
|
|
; RV64ZVE32F-NEXT: # %bb.11: # %else12
|
|
; RV64ZVE32F-NEXT: andi a1, a1, -128
|
|
; RV64ZVE32F-NEXT: bnez a1, .LBB35_16
|
|
; RV64ZVE32F-NEXT: .LBB35_12: # %else14
|
|
; RV64ZVE32F-NEXT: ret
|
|
; RV64ZVE32F-NEXT: .LBB35_13: # %cond.store5
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v12, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v11
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 2
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v8, 3
|
|
; RV64ZVE32F-NEXT: vse32.v v11, (a2)
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 16
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB35_8
|
|
; RV64ZVE32F-NEXT: .LBB35_14: # %cond.store7
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 2
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v8, 4
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v12, (a2)
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 32
|
|
; RV64ZVE32F-NEXT: bnez a2, .LBB35_9
|
|
; RV64ZVE32F-NEXT: j .LBB35_10
|
|
; RV64ZVE32F-NEXT: .LBB35_15: # %cond.store11
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 2
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v8, 6
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v12, (a2)
|
|
; RV64ZVE32F-NEXT: andi a1, a1, -128
|
|
; RV64ZVE32F-NEXT: beqz a1, .LBB35_12
|
|
; RV64ZVE32F-NEXT: .LBB35_16: # %cond.store13
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v10, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a1, v10
|
|
; RV64ZVE32F-NEXT: slli a1, a1, 2
|
|
; RV64ZVE32F-NEXT: add a0, a0, a1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 7
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v8, (a0)
|
|
; RV64ZVE32F-NEXT: ret
|
|
%ptrs = getelementptr inbounds i32, ptr %base, <8 x i32> %idxs
|
|
call void @llvm.masked.scatter.v8i32.v8p0(<8 x i32> %val, <8 x ptr> %ptrs, i32 4, <8 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_v1i64(<1 x i64> %val, <1 x ptr> %ptrs, <1 x i1> %m) {
|
|
; RV32V-LABEL: mscatter_v1i64:
|
|
; RV32V: # %bb.0:
|
|
; RV32V-NEXT: vsetivli zero, 1, e64, m1, ta, ma
|
|
; RV32V-NEXT: vsoxei32.v v8, (zero), v9, v0.t
|
|
; RV32V-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_v1i64:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: vsetivli zero, 1, e64, m1, ta, ma
|
|
; RV64V-NEXT: vsoxei64.v v8, (zero), v9, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV32ZVE32F-LABEL: mscatter_v1i64:
|
|
; RV32ZVE32F: # %bb.0:
|
|
; RV32ZVE32F-NEXT: vsetvli a2, zero, e8, mf4, ta, ma
|
|
; RV32ZVE32F-NEXT: vfirst.m a2, v0
|
|
; RV32ZVE32F-NEXT: bnez a2, .LBB36_2
|
|
; RV32ZVE32F-NEXT: # %bb.1: # %cond.store
|
|
; RV32ZVE32F-NEXT: vsetvli zero, zero, e32, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vmv.x.s a2, v8
|
|
; RV32ZVE32F-NEXT: sw a0, 0(a2)
|
|
; RV32ZVE32F-NEXT: sw a1, 4(a2)
|
|
; RV32ZVE32F-NEXT: .LBB36_2: # %else
|
|
; RV32ZVE32F-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_v1i64:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: vsetvli a2, zero, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vfirst.m a2, v0
|
|
; RV64ZVE32F-NEXT: bnez a2, .LBB36_2
|
|
; RV64ZVE32F-NEXT: # %bb.1: # %cond.store
|
|
; RV64ZVE32F-NEXT: sd a0, 0(a1)
|
|
; RV64ZVE32F-NEXT: .LBB36_2: # %else
|
|
; RV64ZVE32F-NEXT: ret
|
|
call void @llvm.masked.scatter.v1i64.v1p0(<1 x i64> %val, <1 x ptr> %ptrs, i32 8, <1 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_v2i64(<2 x i64> %val, <2 x ptr> %ptrs, <2 x i1> %m) {
|
|
; RV32V-LABEL: mscatter_v2i64:
|
|
; RV32V: # %bb.0:
|
|
; RV32V-NEXT: vsetivli zero, 2, e64, m1, ta, ma
|
|
; RV32V-NEXT: vsoxei32.v v8, (zero), v9, v0.t
|
|
; RV32V-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_v2i64:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: vsetivli zero, 2, e64, m1, ta, ma
|
|
; RV64V-NEXT: vsoxei64.v v8, (zero), v9, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV32ZVE32F-LABEL: mscatter_v2i64:
|
|
; RV32ZVE32F: # %bb.0:
|
|
; RV32ZVE32F-NEXT: lw a1, 8(a0)
|
|
; RV32ZVE32F-NEXT: lw a2, 12(a0)
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vmv.x.s a3, v0
|
|
; RV32ZVE32F-NEXT: andi a4, a3, 1
|
|
; RV32ZVE32F-NEXT: bnez a4, .LBB37_3
|
|
; RV32ZVE32F-NEXT: # %bb.1: # %else
|
|
; RV32ZVE32F-NEXT: andi a3, a3, 2
|
|
; RV32ZVE32F-NEXT: bnez a3, .LBB37_4
|
|
; RV32ZVE32F-NEXT: .LBB37_2: # %else2
|
|
; RV32ZVE32F-NEXT: ret
|
|
; RV32ZVE32F-NEXT: .LBB37_3: # %cond.store
|
|
; RV32ZVE32F-NEXT: lw a4, 0(a0)
|
|
; RV32ZVE32F-NEXT: lw a0, 4(a0)
|
|
; RV32ZVE32F-NEXT: vsetvli zero, zero, e32, m4, ta, ma
|
|
; RV32ZVE32F-NEXT: vmv.x.s a5, v8
|
|
; RV32ZVE32F-NEXT: sw a4, 0(a5)
|
|
; RV32ZVE32F-NEXT: sw a0, 4(a5)
|
|
; RV32ZVE32F-NEXT: andi a3, a3, 2
|
|
; RV32ZVE32F-NEXT: beqz a3, .LBB37_2
|
|
; RV32ZVE32F-NEXT: .LBB37_4: # %cond.store1
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v8, v8, 1
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v8
|
|
; RV32ZVE32F-NEXT: sw a1, 0(a0)
|
|
; RV32ZVE32F-NEXT: sw a2, 4(a0)
|
|
; RV32ZVE32F-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_v2i64:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a4, v0
|
|
; RV64ZVE32F-NEXT: andi a5, a4, 1
|
|
; RV64ZVE32F-NEXT: bnez a5, .LBB37_3
|
|
; RV64ZVE32F-NEXT: # %bb.1: # %else
|
|
; RV64ZVE32F-NEXT: andi a4, a4, 2
|
|
; RV64ZVE32F-NEXT: bnez a4, .LBB37_4
|
|
; RV64ZVE32F-NEXT: .LBB37_2: # %else2
|
|
; RV64ZVE32F-NEXT: ret
|
|
; RV64ZVE32F-NEXT: .LBB37_3: # %cond.store
|
|
; RV64ZVE32F-NEXT: sd a0, 0(a2)
|
|
; RV64ZVE32F-NEXT: andi a4, a4, 2
|
|
; RV64ZVE32F-NEXT: beqz a4, .LBB37_2
|
|
; RV64ZVE32F-NEXT: .LBB37_4: # %cond.store1
|
|
; RV64ZVE32F-NEXT: sd a1, 0(a3)
|
|
; RV64ZVE32F-NEXT: ret
|
|
call void @llvm.masked.scatter.v2i64.v2p0(<2 x i64> %val, <2 x ptr> %ptrs, i32 8, <2 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_v4i64(<4 x i64> %val, <4 x ptr> %ptrs, <4 x i1> %m) {
|
|
; RV32V-LABEL: mscatter_v4i64:
|
|
; RV32V: # %bb.0:
|
|
; RV32V-NEXT: vsetivli zero, 4, e64, m2, ta, ma
|
|
; RV32V-NEXT: vsoxei32.v v8, (zero), v10, v0.t
|
|
; RV32V-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_v4i64:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: vsetivli zero, 4, e64, m2, ta, ma
|
|
; RV64V-NEXT: vsoxei64.v v8, (zero), v10, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV32ZVE32F-LABEL: mscatter_v4i64:
|
|
; RV32ZVE32F: # %bb.0:
|
|
; RV32ZVE32F-NEXT: lw a1, 24(a0)
|
|
; RV32ZVE32F-NEXT: lw a2, 28(a0)
|
|
; RV32ZVE32F-NEXT: lw a6, 8(a0)
|
|
; RV32ZVE32F-NEXT: lw a7, 12(a0)
|
|
; RV32ZVE32F-NEXT: lw a3, 16(a0)
|
|
; RV32ZVE32F-NEXT: lw a4, 20(a0)
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vmv.x.s a5, v0
|
|
; RV32ZVE32F-NEXT: andi t0, a5, 1
|
|
; RV32ZVE32F-NEXT: bnez t0, .LBB38_5
|
|
; RV32ZVE32F-NEXT: # %bb.1: # %else
|
|
; RV32ZVE32F-NEXT: andi a0, a5, 2
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB38_6
|
|
; RV32ZVE32F-NEXT: .LBB38_2: # %else2
|
|
; RV32ZVE32F-NEXT: andi a0, a5, 4
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB38_7
|
|
; RV32ZVE32F-NEXT: .LBB38_3: # %else4
|
|
; RV32ZVE32F-NEXT: andi a5, a5, 8
|
|
; RV32ZVE32F-NEXT: bnez a5, .LBB38_8
|
|
; RV32ZVE32F-NEXT: .LBB38_4: # %else6
|
|
; RV32ZVE32F-NEXT: ret
|
|
; RV32ZVE32F-NEXT: .LBB38_5: # %cond.store
|
|
; RV32ZVE32F-NEXT: lw t0, 0(a0)
|
|
; RV32ZVE32F-NEXT: lw a0, 4(a0)
|
|
; RV32ZVE32F-NEXT: vsetvli zero, zero, e32, m4, ta, ma
|
|
; RV32ZVE32F-NEXT: vmv.x.s t1, v8
|
|
; RV32ZVE32F-NEXT: sw t0, 0(t1)
|
|
; RV32ZVE32F-NEXT: sw a0, 4(t1)
|
|
; RV32ZVE32F-NEXT: andi a0, a5, 2
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB38_2
|
|
; RV32ZVE32F-NEXT: .LBB38_6: # %cond.store1
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v9, v8, 1
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v9
|
|
; RV32ZVE32F-NEXT: sw a6, 0(a0)
|
|
; RV32ZVE32F-NEXT: sw a7, 4(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, a5, 4
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB38_3
|
|
; RV32ZVE32F-NEXT: .LBB38_7: # %cond.store3
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v9, v8, 2
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v9
|
|
; RV32ZVE32F-NEXT: sw a3, 0(a0)
|
|
; RV32ZVE32F-NEXT: sw a4, 4(a0)
|
|
; RV32ZVE32F-NEXT: andi a5, a5, 8
|
|
; RV32ZVE32F-NEXT: beqz a5, .LBB38_4
|
|
; RV32ZVE32F-NEXT: .LBB38_8: # %cond.store5
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v8, v8, 3
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v8
|
|
; RV32ZVE32F-NEXT: sw a1, 0(a0)
|
|
; RV32ZVE32F-NEXT: sw a2, 4(a0)
|
|
; RV32ZVE32F-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_v4i64:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: ld a6, 8(a1)
|
|
; RV64ZVE32F-NEXT: ld a4, 16(a1)
|
|
; RV64ZVE32F-NEXT: ld a2, 24(a1)
|
|
; RV64ZVE32F-NEXT: ld t0, 8(a0)
|
|
; RV64ZVE32F-NEXT: ld a5, 16(a0)
|
|
; RV64ZVE32F-NEXT: ld a3, 24(a0)
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a7, v0
|
|
; RV64ZVE32F-NEXT: andi t1, a7, 1
|
|
; RV64ZVE32F-NEXT: bnez t1, .LBB38_5
|
|
; RV64ZVE32F-NEXT: # %bb.1: # %else
|
|
; RV64ZVE32F-NEXT: andi a0, a7, 2
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB38_6
|
|
; RV64ZVE32F-NEXT: .LBB38_2: # %else2
|
|
; RV64ZVE32F-NEXT: andi a0, a7, 4
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB38_7
|
|
; RV64ZVE32F-NEXT: .LBB38_3: # %else4
|
|
; RV64ZVE32F-NEXT: andi a0, a7, 8
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB38_8
|
|
; RV64ZVE32F-NEXT: .LBB38_4: # %else6
|
|
; RV64ZVE32F-NEXT: ret
|
|
; RV64ZVE32F-NEXT: .LBB38_5: # %cond.store
|
|
; RV64ZVE32F-NEXT: ld a1, 0(a1)
|
|
; RV64ZVE32F-NEXT: ld a0, 0(a0)
|
|
; RV64ZVE32F-NEXT: sd a0, 0(a1)
|
|
; RV64ZVE32F-NEXT: andi a0, a7, 2
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB38_2
|
|
; RV64ZVE32F-NEXT: .LBB38_6: # %cond.store1
|
|
; RV64ZVE32F-NEXT: sd t0, 0(a6)
|
|
; RV64ZVE32F-NEXT: andi a0, a7, 4
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB38_3
|
|
; RV64ZVE32F-NEXT: .LBB38_7: # %cond.store3
|
|
; RV64ZVE32F-NEXT: sd a5, 0(a4)
|
|
; RV64ZVE32F-NEXT: andi a0, a7, 8
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB38_4
|
|
; RV64ZVE32F-NEXT: .LBB38_8: # %cond.store5
|
|
; RV64ZVE32F-NEXT: sd a3, 0(a2)
|
|
; RV64ZVE32F-NEXT: ret
|
|
call void @llvm.masked.scatter.v4i64.v4p0(<4 x i64> %val, <4 x ptr> %ptrs, i32 8, <4 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_truemask_v4i64(<4 x i64> %val, <4 x ptr> %ptrs) {
|
|
; RV32V-LABEL: mscatter_truemask_v4i64:
|
|
; RV32V: # %bb.0:
|
|
; RV32V-NEXT: vsetivli zero, 4, e64, m2, ta, ma
|
|
; RV32V-NEXT: vsoxei32.v v8, (zero), v10
|
|
; RV32V-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_truemask_v4i64:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: vsetivli zero, 4, e64, m2, ta, ma
|
|
; RV64V-NEXT: vsoxei64.v v8, (zero), v10
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV32ZVE32F-LABEL: mscatter_truemask_v4i64:
|
|
; RV32ZVE32F: # %bb.0:
|
|
; RV32ZVE32F-NEXT: lw a1, 16(a0)
|
|
; RV32ZVE32F-NEXT: lw a2, 20(a0)
|
|
; RV32ZVE32F-NEXT: lw a3, 24(a0)
|
|
; RV32ZVE32F-NEXT: lw a4, 28(a0)
|
|
; RV32ZVE32F-NEXT: lw a5, 0(a0)
|
|
; RV32ZVE32F-NEXT: lw a6, 4(a0)
|
|
; RV32ZVE32F-NEXT: lw a7, 8(a0)
|
|
; RV32ZVE32F-NEXT: lw a0, 12(a0)
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vmv.x.s t0, v8
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v9, v8, 1
|
|
; RV32ZVE32F-NEXT: vmv.x.s t1, v9
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v9, v8, 2
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v8, v8, 3
|
|
; RV32ZVE32F-NEXT: sw a5, 0(t0)
|
|
; RV32ZVE32F-NEXT: sw a6, 4(t0)
|
|
; RV32ZVE32F-NEXT: vmv.x.s a5, v9
|
|
; RV32ZVE32F-NEXT: vmv.x.s a6, v8
|
|
; RV32ZVE32F-NEXT: sw a7, 0(t1)
|
|
; RV32ZVE32F-NEXT: sw a0, 4(t1)
|
|
; RV32ZVE32F-NEXT: sw a1, 0(a5)
|
|
; RV32ZVE32F-NEXT: sw a2, 4(a5)
|
|
; RV32ZVE32F-NEXT: sw a3, 0(a6)
|
|
; RV32ZVE32F-NEXT: sw a4, 4(a6)
|
|
; RV32ZVE32F-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_truemask_v4i64:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: ld a2, 0(a1)
|
|
; RV64ZVE32F-NEXT: ld a3, 8(a1)
|
|
; RV64ZVE32F-NEXT: ld a4, 16(a1)
|
|
; RV64ZVE32F-NEXT: ld a1, 24(a1)
|
|
; RV64ZVE32F-NEXT: ld a5, 0(a0)
|
|
; RV64ZVE32F-NEXT: ld a6, 8(a0)
|
|
; RV64ZVE32F-NEXT: ld a7, 16(a0)
|
|
; RV64ZVE32F-NEXT: ld a0, 24(a0)
|
|
; RV64ZVE32F-NEXT: sd a5, 0(a2)
|
|
; RV64ZVE32F-NEXT: sd a6, 0(a3)
|
|
; RV64ZVE32F-NEXT: sd a7, 0(a4)
|
|
; RV64ZVE32F-NEXT: sd a0, 0(a1)
|
|
; RV64ZVE32F-NEXT: ret
|
|
call void @llvm.masked.scatter.v4i64.v4p0(<4 x i64> %val, <4 x ptr> %ptrs, i32 8, <4 x i1> splat (i1 1))
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_falsemask_v4i64(<4 x i64> %val, <4 x ptr> %ptrs) {
|
|
; CHECK-LABEL: mscatter_falsemask_v4i64:
|
|
; CHECK: # %bb.0:
|
|
; CHECK-NEXT: ret
|
|
call void @llvm.masked.scatter.v4i64.v4p0(<4 x i64> %val, <4 x ptr> %ptrs, i32 8, <4 x i1> zeroinitializer)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_v8i64(<8 x i64> %val, <8 x ptr> %ptrs, <8 x i1> %m) {
|
|
; RV32V-LABEL: mscatter_v8i64:
|
|
; RV32V: # %bb.0:
|
|
; RV32V-NEXT: vsetivli zero, 8, e64, m4, ta, ma
|
|
; RV32V-NEXT: vsoxei32.v v8, (zero), v12, v0.t
|
|
; RV32V-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_v8i64:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: vsetivli zero, 8, e64, m4, ta, ma
|
|
; RV64V-NEXT: vsoxei64.v v8, (zero), v12, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV32ZVE32F-LABEL: mscatter_v8i64:
|
|
; RV32ZVE32F: # %bb.0:
|
|
; RV32ZVE32F-NEXT: addi sp, sp, -16
|
|
; RV32ZVE32F-NEXT: .cfi_def_cfa_offset 16
|
|
; RV32ZVE32F-NEXT: sw s0, 12(sp) # 4-byte Folded Spill
|
|
; RV32ZVE32F-NEXT: sw s1, 8(sp) # 4-byte Folded Spill
|
|
; RV32ZVE32F-NEXT: sw s2, 4(sp) # 4-byte Folded Spill
|
|
; RV32ZVE32F-NEXT: .cfi_offset s0, -4
|
|
; RV32ZVE32F-NEXT: .cfi_offset s1, -8
|
|
; RV32ZVE32F-NEXT: .cfi_offset s2, -12
|
|
; RV32ZVE32F-NEXT: .cfi_remember_state
|
|
; RV32ZVE32F-NEXT: lw a1, 56(a0)
|
|
; RV32ZVE32F-NEXT: lw a2, 60(a0)
|
|
; RV32ZVE32F-NEXT: lw a5, 40(a0)
|
|
; RV32ZVE32F-NEXT: lw a6, 44(a0)
|
|
; RV32ZVE32F-NEXT: lw a3, 48(a0)
|
|
; RV32ZVE32F-NEXT: lw a4, 52(a0)
|
|
; RV32ZVE32F-NEXT: lw t2, 24(a0)
|
|
; RV32ZVE32F-NEXT: lw t3, 28(a0)
|
|
; RV32ZVE32F-NEXT: lw t0, 32(a0)
|
|
; RV32ZVE32F-NEXT: lw t1, 36(a0)
|
|
; RV32ZVE32F-NEXT: lw t6, 8(a0)
|
|
; RV32ZVE32F-NEXT: lw s0, 12(a0)
|
|
; RV32ZVE32F-NEXT: lw t4, 16(a0)
|
|
; RV32ZVE32F-NEXT: lw t5, 20(a0)
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vmv.x.s a7, v0
|
|
; RV32ZVE32F-NEXT: andi s1, a7, 1
|
|
; RV32ZVE32F-NEXT: bnez s1, .LBB41_10
|
|
; RV32ZVE32F-NEXT: # %bb.1: # %else
|
|
; RV32ZVE32F-NEXT: andi a0, a7, 2
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB41_11
|
|
; RV32ZVE32F-NEXT: .LBB41_2: # %else2
|
|
; RV32ZVE32F-NEXT: andi a0, a7, 4
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB41_12
|
|
; RV32ZVE32F-NEXT: .LBB41_3: # %else4
|
|
; RV32ZVE32F-NEXT: andi a0, a7, 8
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB41_13
|
|
; RV32ZVE32F-NEXT: .LBB41_4: # %else6
|
|
; RV32ZVE32F-NEXT: andi a0, a7, 16
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB41_14
|
|
; RV32ZVE32F-NEXT: .LBB41_5: # %else8
|
|
; RV32ZVE32F-NEXT: andi a0, a7, 32
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB41_15
|
|
; RV32ZVE32F-NEXT: .LBB41_6: # %else10
|
|
; RV32ZVE32F-NEXT: andi a0, a7, 64
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB41_16
|
|
; RV32ZVE32F-NEXT: .LBB41_7: # %else12
|
|
; RV32ZVE32F-NEXT: andi a0, a7, -128
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB41_9
|
|
; RV32ZVE32F-NEXT: .LBB41_8: # %cond.store13
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v8, v8, 7
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v8
|
|
; RV32ZVE32F-NEXT: sw a1, 0(a0)
|
|
; RV32ZVE32F-NEXT: sw a2, 4(a0)
|
|
; RV32ZVE32F-NEXT: .LBB41_9: # %else14
|
|
; RV32ZVE32F-NEXT: lw s0, 12(sp) # 4-byte Folded Reload
|
|
; RV32ZVE32F-NEXT: lw s1, 8(sp) # 4-byte Folded Reload
|
|
; RV32ZVE32F-NEXT: lw s2, 4(sp) # 4-byte Folded Reload
|
|
; RV32ZVE32F-NEXT: .cfi_restore s0
|
|
; RV32ZVE32F-NEXT: .cfi_restore s1
|
|
; RV32ZVE32F-NEXT: .cfi_restore s2
|
|
; RV32ZVE32F-NEXT: addi sp, sp, 16
|
|
; RV32ZVE32F-NEXT: .cfi_def_cfa_offset 0
|
|
; RV32ZVE32F-NEXT: ret
|
|
; RV32ZVE32F-NEXT: .LBB41_10: # %cond.store
|
|
; RV32ZVE32F-NEXT: .cfi_restore_state
|
|
; RV32ZVE32F-NEXT: lw s1, 0(a0)
|
|
; RV32ZVE32F-NEXT: lw a0, 4(a0)
|
|
; RV32ZVE32F-NEXT: vsetvli zero, zero, e32, m4, ta, ma
|
|
; RV32ZVE32F-NEXT: vmv.x.s s2, v8
|
|
; RV32ZVE32F-NEXT: sw s1, 0(s2)
|
|
; RV32ZVE32F-NEXT: sw a0, 4(s2)
|
|
; RV32ZVE32F-NEXT: andi a0, a7, 2
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB41_2
|
|
; RV32ZVE32F-NEXT: .LBB41_11: # %cond.store1
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 1
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: sw t6, 0(a0)
|
|
; RV32ZVE32F-NEXT: sw s0, 4(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, a7, 4
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB41_3
|
|
; RV32ZVE32F-NEXT: .LBB41_12: # %cond.store3
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 2
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: sw t4, 0(a0)
|
|
; RV32ZVE32F-NEXT: sw t5, 4(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, a7, 8
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB41_4
|
|
; RV32ZVE32F-NEXT: .LBB41_13: # %cond.store5
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 3
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: sw t2, 0(a0)
|
|
; RV32ZVE32F-NEXT: sw t3, 4(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, a7, 16
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB41_5
|
|
; RV32ZVE32F-NEXT: .LBB41_14: # %cond.store7
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 4
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: sw t0, 0(a0)
|
|
; RV32ZVE32F-NEXT: sw t1, 4(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, a7, 32
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB41_6
|
|
; RV32ZVE32F-NEXT: .LBB41_15: # %cond.store9
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 5
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: sw a5, 0(a0)
|
|
; RV32ZVE32F-NEXT: sw a6, 4(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, a7, 64
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB41_7
|
|
; RV32ZVE32F-NEXT: .LBB41_16: # %cond.store11
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 6
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: sw a3, 0(a0)
|
|
; RV32ZVE32F-NEXT: sw a4, 4(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, a7, -128
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB41_8
|
|
; RV32ZVE32F-NEXT: j .LBB41_9
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_v8i64:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: addi sp, sp, -32
|
|
; RV64ZVE32F-NEXT: .cfi_def_cfa_offset 32
|
|
; RV64ZVE32F-NEXT: sd s0, 24(sp) # 8-byte Folded Spill
|
|
; RV64ZVE32F-NEXT: sd s1, 16(sp) # 8-byte Folded Spill
|
|
; RV64ZVE32F-NEXT: sd s2, 8(sp) # 8-byte Folded Spill
|
|
; RV64ZVE32F-NEXT: .cfi_offset s0, -8
|
|
; RV64ZVE32F-NEXT: .cfi_offset s1, -16
|
|
; RV64ZVE32F-NEXT: .cfi_offset s2, -24
|
|
; RV64ZVE32F-NEXT: .cfi_remember_state
|
|
; RV64ZVE32F-NEXT: ld a4, 40(a1)
|
|
; RV64ZVE32F-NEXT: ld a3, 48(a1)
|
|
; RV64ZVE32F-NEXT: ld a2, 56(a1)
|
|
; RV64ZVE32F-NEXT: ld t5, 8(a1)
|
|
; RV64ZVE32F-NEXT: ld t3, 16(a1)
|
|
; RV64ZVE32F-NEXT: ld t2, 24(a1)
|
|
; RV64ZVE32F-NEXT: ld t0, 32(a1)
|
|
; RV64ZVE32F-NEXT: ld a7, 40(a0)
|
|
; RV64ZVE32F-NEXT: ld a6, 48(a0)
|
|
; RV64ZVE32F-NEXT: ld a5, 56(a0)
|
|
; RV64ZVE32F-NEXT: ld s1, 8(a0)
|
|
; RV64ZVE32F-NEXT: ld s0, 16(a0)
|
|
; RV64ZVE32F-NEXT: ld t6, 24(a0)
|
|
; RV64ZVE32F-NEXT: ld t4, 32(a0)
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s t1, v0
|
|
; RV64ZVE32F-NEXT: andi s2, t1, 1
|
|
; RV64ZVE32F-NEXT: bnez s2, .LBB41_10
|
|
; RV64ZVE32F-NEXT: # %bb.1: # %else
|
|
; RV64ZVE32F-NEXT: andi a0, t1, 2
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB41_11
|
|
; RV64ZVE32F-NEXT: .LBB41_2: # %else2
|
|
; RV64ZVE32F-NEXT: andi a0, t1, 4
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB41_12
|
|
; RV64ZVE32F-NEXT: .LBB41_3: # %else4
|
|
; RV64ZVE32F-NEXT: andi a0, t1, 8
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB41_13
|
|
; RV64ZVE32F-NEXT: .LBB41_4: # %else6
|
|
; RV64ZVE32F-NEXT: andi a0, t1, 16
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB41_14
|
|
; RV64ZVE32F-NEXT: .LBB41_5: # %else8
|
|
; RV64ZVE32F-NEXT: andi a0, t1, 32
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB41_15
|
|
; RV64ZVE32F-NEXT: .LBB41_6: # %else10
|
|
; RV64ZVE32F-NEXT: andi a0, t1, 64
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB41_16
|
|
; RV64ZVE32F-NEXT: .LBB41_7: # %else12
|
|
; RV64ZVE32F-NEXT: andi a0, t1, -128
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB41_9
|
|
; RV64ZVE32F-NEXT: .LBB41_8: # %cond.store13
|
|
; RV64ZVE32F-NEXT: sd a5, 0(a2)
|
|
; RV64ZVE32F-NEXT: .LBB41_9: # %else14
|
|
; RV64ZVE32F-NEXT: ld s0, 24(sp) # 8-byte Folded Reload
|
|
; RV64ZVE32F-NEXT: ld s1, 16(sp) # 8-byte Folded Reload
|
|
; RV64ZVE32F-NEXT: ld s2, 8(sp) # 8-byte Folded Reload
|
|
; RV64ZVE32F-NEXT: .cfi_restore s0
|
|
; RV64ZVE32F-NEXT: .cfi_restore s1
|
|
; RV64ZVE32F-NEXT: .cfi_restore s2
|
|
; RV64ZVE32F-NEXT: addi sp, sp, 32
|
|
; RV64ZVE32F-NEXT: .cfi_def_cfa_offset 0
|
|
; RV64ZVE32F-NEXT: ret
|
|
; RV64ZVE32F-NEXT: .LBB41_10: # %cond.store
|
|
; RV64ZVE32F-NEXT: .cfi_restore_state
|
|
; RV64ZVE32F-NEXT: ld a1, 0(a1)
|
|
; RV64ZVE32F-NEXT: ld a0, 0(a0)
|
|
; RV64ZVE32F-NEXT: sd a0, 0(a1)
|
|
; RV64ZVE32F-NEXT: andi a0, t1, 2
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB41_2
|
|
; RV64ZVE32F-NEXT: .LBB41_11: # %cond.store1
|
|
; RV64ZVE32F-NEXT: sd s1, 0(t5)
|
|
; RV64ZVE32F-NEXT: andi a0, t1, 4
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB41_3
|
|
; RV64ZVE32F-NEXT: .LBB41_12: # %cond.store3
|
|
; RV64ZVE32F-NEXT: sd s0, 0(t3)
|
|
; RV64ZVE32F-NEXT: andi a0, t1, 8
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB41_4
|
|
; RV64ZVE32F-NEXT: .LBB41_13: # %cond.store5
|
|
; RV64ZVE32F-NEXT: sd t6, 0(t2)
|
|
; RV64ZVE32F-NEXT: andi a0, t1, 16
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB41_5
|
|
; RV64ZVE32F-NEXT: .LBB41_14: # %cond.store7
|
|
; RV64ZVE32F-NEXT: sd t4, 0(t0)
|
|
; RV64ZVE32F-NEXT: andi a0, t1, 32
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB41_6
|
|
; RV64ZVE32F-NEXT: .LBB41_15: # %cond.store9
|
|
; RV64ZVE32F-NEXT: sd a7, 0(a4)
|
|
; RV64ZVE32F-NEXT: andi a0, t1, 64
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB41_7
|
|
; RV64ZVE32F-NEXT: .LBB41_16: # %cond.store11
|
|
; RV64ZVE32F-NEXT: sd a6, 0(a3)
|
|
; RV64ZVE32F-NEXT: andi a0, t1, -128
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB41_8
|
|
; RV64ZVE32F-NEXT: j .LBB41_9
|
|
call void @llvm.masked.scatter.v8i64.v8p0(<8 x i64> %val, <8 x ptr> %ptrs, i32 8, <8 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_baseidx_v8i8_v8i64(<8 x i64> %val, ptr %base, <8 x i8> %idxs, <8 x i1> %m) {
|
|
; RV32V-LABEL: mscatter_baseidx_v8i8_v8i64:
|
|
; RV32V: # %bb.0:
|
|
; RV32V-NEXT: vsetivli zero, 8, e32, m2, ta, ma
|
|
; RV32V-NEXT: vsext.vf4 v14, v12
|
|
; RV32V-NEXT: vsll.vi v12, v14, 3
|
|
; RV32V-NEXT: vsetvli zero, zero, e64, m4, ta, ma
|
|
; RV32V-NEXT: vsoxei32.v v8, (a0), v12, v0.t
|
|
; RV32V-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_baseidx_v8i8_v8i64:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: vsetivli zero, 8, e64, m4, ta, ma
|
|
; RV64V-NEXT: vsext.vf8 v16, v12
|
|
; RV64V-NEXT: vsll.vi v12, v16, 3
|
|
; RV64V-NEXT: vsoxei64.v v8, (a0), v12, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV32ZVE32F-LABEL: mscatter_baseidx_v8i8_v8i64:
|
|
; RV32ZVE32F: # %bb.0:
|
|
; RV32ZVE32F-NEXT: addi sp, sp, -16
|
|
; RV32ZVE32F-NEXT: .cfi_def_cfa_offset 16
|
|
; RV32ZVE32F-NEXT: sw s0, 12(sp) # 4-byte Folded Spill
|
|
; RV32ZVE32F-NEXT: sw s1, 8(sp) # 4-byte Folded Spill
|
|
; RV32ZVE32F-NEXT: sw s2, 4(sp) # 4-byte Folded Spill
|
|
; RV32ZVE32F-NEXT: .cfi_offset s0, -4
|
|
; RV32ZVE32F-NEXT: .cfi_offset s1, -8
|
|
; RV32ZVE32F-NEXT: .cfi_offset s2, -12
|
|
; RV32ZVE32F-NEXT: .cfi_remember_state
|
|
; RV32ZVE32F-NEXT: lw a2, 56(a0)
|
|
; RV32ZVE32F-NEXT: lw a3, 60(a0)
|
|
; RV32ZVE32F-NEXT: lw a6, 40(a0)
|
|
; RV32ZVE32F-NEXT: lw a7, 44(a0)
|
|
; RV32ZVE32F-NEXT: lw a4, 48(a0)
|
|
; RV32ZVE32F-NEXT: lw a5, 52(a0)
|
|
; RV32ZVE32F-NEXT: lw t3, 24(a0)
|
|
; RV32ZVE32F-NEXT: lw t4, 28(a0)
|
|
; RV32ZVE32F-NEXT: lw t1, 32(a0)
|
|
; RV32ZVE32F-NEXT: lw t2, 36(a0)
|
|
; RV32ZVE32F-NEXT: lw s0, 8(a0)
|
|
; RV32ZVE32F-NEXT: lw s1, 12(a0)
|
|
; RV32ZVE32F-NEXT: lw t5, 16(a0)
|
|
; RV32ZVE32F-NEXT: lw t6, 20(a0)
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 8, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vsext.vf4 v10, v8
|
|
; RV32ZVE32F-NEXT: vsll.vi v8, v10, 3
|
|
; RV32ZVE32F-NEXT: vsetvli zero, zero, e8, mf2, ta, ma
|
|
; RV32ZVE32F-NEXT: vmv.x.s t0, v0
|
|
; RV32ZVE32F-NEXT: andi s2, t0, 1
|
|
; RV32ZVE32F-NEXT: vsetvli zero, zero, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vadd.vx v8, v8, a1
|
|
; RV32ZVE32F-NEXT: bnez s2, .LBB42_10
|
|
; RV32ZVE32F-NEXT: # %bb.1: # %else
|
|
; RV32ZVE32F-NEXT: andi a0, t0, 2
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB42_11
|
|
; RV32ZVE32F-NEXT: .LBB42_2: # %else2
|
|
; RV32ZVE32F-NEXT: andi a0, t0, 4
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB42_12
|
|
; RV32ZVE32F-NEXT: .LBB42_3: # %else4
|
|
; RV32ZVE32F-NEXT: andi a0, t0, 8
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB42_13
|
|
; RV32ZVE32F-NEXT: .LBB42_4: # %else6
|
|
; RV32ZVE32F-NEXT: andi a0, t0, 16
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB42_14
|
|
; RV32ZVE32F-NEXT: .LBB42_5: # %else8
|
|
; RV32ZVE32F-NEXT: andi a0, t0, 32
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB42_15
|
|
; RV32ZVE32F-NEXT: .LBB42_6: # %else10
|
|
; RV32ZVE32F-NEXT: andi a0, t0, 64
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB42_16
|
|
; RV32ZVE32F-NEXT: .LBB42_7: # %else12
|
|
; RV32ZVE32F-NEXT: andi a0, t0, -128
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB42_9
|
|
; RV32ZVE32F-NEXT: .LBB42_8: # %cond.store13
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v8, v8, 7
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v8
|
|
; RV32ZVE32F-NEXT: sw a2, 0(a0)
|
|
; RV32ZVE32F-NEXT: sw a3, 4(a0)
|
|
; RV32ZVE32F-NEXT: .LBB42_9: # %else14
|
|
; RV32ZVE32F-NEXT: lw s0, 12(sp) # 4-byte Folded Reload
|
|
; RV32ZVE32F-NEXT: lw s1, 8(sp) # 4-byte Folded Reload
|
|
; RV32ZVE32F-NEXT: lw s2, 4(sp) # 4-byte Folded Reload
|
|
; RV32ZVE32F-NEXT: .cfi_restore s0
|
|
; RV32ZVE32F-NEXT: .cfi_restore s1
|
|
; RV32ZVE32F-NEXT: .cfi_restore s2
|
|
; RV32ZVE32F-NEXT: addi sp, sp, 16
|
|
; RV32ZVE32F-NEXT: .cfi_def_cfa_offset 0
|
|
; RV32ZVE32F-NEXT: ret
|
|
; RV32ZVE32F-NEXT: .LBB42_10: # %cond.store
|
|
; RV32ZVE32F-NEXT: .cfi_restore_state
|
|
; RV32ZVE32F-NEXT: lw a1, 0(a0)
|
|
; RV32ZVE32F-NEXT: lw a0, 4(a0)
|
|
; RV32ZVE32F-NEXT: vmv.x.s s2, v8
|
|
; RV32ZVE32F-NEXT: sw a1, 0(s2)
|
|
; RV32ZVE32F-NEXT: sw a0, 4(s2)
|
|
; RV32ZVE32F-NEXT: andi a0, t0, 2
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB42_2
|
|
; RV32ZVE32F-NEXT: .LBB42_11: # %cond.store1
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 1
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: sw s0, 0(a0)
|
|
; RV32ZVE32F-NEXT: sw s1, 4(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, t0, 4
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB42_3
|
|
; RV32ZVE32F-NEXT: .LBB42_12: # %cond.store3
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 2
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: sw t5, 0(a0)
|
|
; RV32ZVE32F-NEXT: sw t6, 4(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, t0, 8
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB42_4
|
|
; RV32ZVE32F-NEXT: .LBB42_13: # %cond.store5
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 3
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: sw t3, 0(a0)
|
|
; RV32ZVE32F-NEXT: sw t4, 4(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, t0, 16
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB42_5
|
|
; RV32ZVE32F-NEXT: .LBB42_14: # %cond.store7
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 4
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: sw t1, 0(a0)
|
|
; RV32ZVE32F-NEXT: sw t2, 4(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, t0, 32
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB42_6
|
|
; RV32ZVE32F-NEXT: .LBB42_15: # %cond.store9
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 5
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: sw a6, 0(a0)
|
|
; RV32ZVE32F-NEXT: sw a7, 4(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, t0, 64
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB42_7
|
|
; RV32ZVE32F-NEXT: .LBB42_16: # %cond.store11
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 6
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: sw a4, 0(a0)
|
|
; RV32ZVE32F-NEXT: sw a5, 4(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, t0, -128
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB42_8
|
|
; RV32ZVE32F-NEXT: j .LBB42_9
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_baseidx_v8i8_v8i64:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: ld a5, 40(a0)
|
|
; RV64ZVE32F-NEXT: ld a3, 48(a0)
|
|
; RV64ZVE32F-NEXT: ld a2, 56(a0)
|
|
; RV64ZVE32F-NEXT: ld t1, 8(a0)
|
|
; RV64ZVE32F-NEXT: ld t0, 16(a0)
|
|
; RV64ZVE32F-NEXT: ld a7, 24(a0)
|
|
; RV64ZVE32F-NEXT: ld a6, 32(a0)
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a4, v0
|
|
; RV64ZVE32F-NEXT: andi t2, a4, 1
|
|
; RV64ZVE32F-NEXT: beqz t2, .LBB42_2
|
|
; RV64ZVE32F-NEXT: # %bb.1: # %cond.store
|
|
; RV64ZVE32F-NEXT: ld a0, 0(a0)
|
|
; RV64ZVE32F-NEXT: vmv.x.s t2, v8
|
|
; RV64ZVE32F-NEXT: slli t2, t2, 3
|
|
; RV64ZVE32F-NEXT: add t2, a1, t2
|
|
; RV64ZVE32F-NEXT: sd a0, 0(t2)
|
|
; RV64ZVE32F-NEXT: .LBB42_2: # %else
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 2
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB42_4
|
|
; RV64ZVE32F-NEXT: # %bb.3: # %cond.store1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a0, v9
|
|
; RV64ZVE32F-NEXT: slli a0, a0, 3
|
|
; RV64ZVE32F-NEXT: add a0, a1, a0
|
|
; RV64ZVE32F-NEXT: sd t1, 0(a0)
|
|
; RV64ZVE32F-NEXT: .LBB42_4: # %else2
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 4
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 2
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB42_6
|
|
; RV64ZVE32F-NEXT: # %bb.5: # %cond.store3
|
|
; RV64ZVE32F-NEXT: vmv.x.s a0, v9
|
|
; RV64ZVE32F-NEXT: slli a0, a0, 3
|
|
; RV64ZVE32F-NEXT: add a0, a1, a0
|
|
; RV64ZVE32F-NEXT: sd t0, 0(a0)
|
|
; RV64ZVE32F-NEXT: .LBB42_6: # %else4
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 8
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB42_8
|
|
; RV64ZVE32F-NEXT: # %bb.7: # %cond.store5
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v9, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a0, v9
|
|
; RV64ZVE32F-NEXT: slli a0, a0, 3
|
|
; RV64ZVE32F-NEXT: add a0, a1, a0
|
|
; RV64ZVE32F-NEXT: sd a7, 0(a0)
|
|
; RV64ZVE32F-NEXT: .LBB42_8: # %else6
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 16
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 4, e8, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 4
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB42_10
|
|
; RV64ZVE32F-NEXT: # %bb.9: # %cond.store7
|
|
; RV64ZVE32F-NEXT: vmv.x.s a0, v8
|
|
; RV64ZVE32F-NEXT: slli a0, a0, 3
|
|
; RV64ZVE32F-NEXT: add a0, a1, a0
|
|
; RV64ZVE32F-NEXT: sd a6, 0(a0)
|
|
; RV64ZVE32F-NEXT: .LBB42_10: # %else8
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 32
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB42_12
|
|
; RV64ZVE32F-NEXT: # %bb.11: # %cond.store9
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a0, v9
|
|
; RV64ZVE32F-NEXT: slli a0, a0, 3
|
|
; RV64ZVE32F-NEXT: add a0, a1, a0
|
|
; RV64ZVE32F-NEXT: sd a5, 0(a0)
|
|
; RV64ZVE32F-NEXT: .LBB42_12: # %else10
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 64
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 2
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB42_15
|
|
; RV64ZVE32F-NEXT: # %bb.13: # %else12
|
|
; RV64ZVE32F-NEXT: andi a0, a4, -128
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB42_16
|
|
; RV64ZVE32F-NEXT: .LBB42_14: # %else14
|
|
; RV64ZVE32F-NEXT: ret
|
|
; RV64ZVE32F-NEXT: .LBB42_15: # %cond.store11
|
|
; RV64ZVE32F-NEXT: vmv.x.s a0, v8
|
|
; RV64ZVE32F-NEXT: slli a0, a0, 3
|
|
; RV64ZVE32F-NEXT: add a0, a1, a0
|
|
; RV64ZVE32F-NEXT: sd a3, 0(a0)
|
|
; RV64ZVE32F-NEXT: andi a0, a4, -128
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB42_14
|
|
; RV64ZVE32F-NEXT: .LBB42_16: # %cond.store13
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a0, v8
|
|
; RV64ZVE32F-NEXT: slli a0, a0, 3
|
|
; RV64ZVE32F-NEXT: add a0, a1, a0
|
|
; RV64ZVE32F-NEXT: sd a2, 0(a0)
|
|
; RV64ZVE32F-NEXT: ret
|
|
%ptrs = getelementptr inbounds i64, ptr %base, <8 x i8> %idxs
|
|
call void @llvm.masked.scatter.v8i64.v8p0(<8 x i64> %val, <8 x ptr> %ptrs, i32 8, <8 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_baseidx_sext_v8i8_v8i64(<8 x i64> %val, ptr %base, <8 x i8> %idxs, <8 x i1> %m) {
|
|
; RV32V-LABEL: mscatter_baseidx_sext_v8i8_v8i64:
|
|
; RV32V: # %bb.0:
|
|
; RV32V-NEXT: vsetivli zero, 8, e32, m2, ta, ma
|
|
; RV32V-NEXT: vsext.vf4 v14, v12
|
|
; RV32V-NEXT: vsll.vi v12, v14, 3
|
|
; RV32V-NEXT: vsetvli zero, zero, e64, m4, ta, ma
|
|
; RV32V-NEXT: vsoxei32.v v8, (a0), v12, v0.t
|
|
; RV32V-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_baseidx_sext_v8i8_v8i64:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: vsetivli zero, 8, e64, m4, ta, ma
|
|
; RV64V-NEXT: vsext.vf8 v16, v12
|
|
; RV64V-NEXT: vsll.vi v12, v16, 3
|
|
; RV64V-NEXT: vsoxei64.v v8, (a0), v12, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV32ZVE32F-LABEL: mscatter_baseidx_sext_v8i8_v8i64:
|
|
; RV32ZVE32F: # %bb.0:
|
|
; RV32ZVE32F-NEXT: addi sp, sp, -16
|
|
; RV32ZVE32F-NEXT: .cfi_def_cfa_offset 16
|
|
; RV32ZVE32F-NEXT: sw s0, 12(sp) # 4-byte Folded Spill
|
|
; RV32ZVE32F-NEXT: sw s1, 8(sp) # 4-byte Folded Spill
|
|
; RV32ZVE32F-NEXT: sw s2, 4(sp) # 4-byte Folded Spill
|
|
; RV32ZVE32F-NEXT: .cfi_offset s0, -4
|
|
; RV32ZVE32F-NEXT: .cfi_offset s1, -8
|
|
; RV32ZVE32F-NEXT: .cfi_offset s2, -12
|
|
; RV32ZVE32F-NEXT: .cfi_remember_state
|
|
; RV32ZVE32F-NEXT: lw a2, 56(a0)
|
|
; RV32ZVE32F-NEXT: lw a3, 60(a0)
|
|
; RV32ZVE32F-NEXT: lw a6, 40(a0)
|
|
; RV32ZVE32F-NEXT: lw a7, 44(a0)
|
|
; RV32ZVE32F-NEXT: lw a4, 48(a0)
|
|
; RV32ZVE32F-NEXT: lw a5, 52(a0)
|
|
; RV32ZVE32F-NEXT: lw t3, 24(a0)
|
|
; RV32ZVE32F-NEXT: lw t4, 28(a0)
|
|
; RV32ZVE32F-NEXT: lw t1, 32(a0)
|
|
; RV32ZVE32F-NEXT: lw t2, 36(a0)
|
|
; RV32ZVE32F-NEXT: lw s0, 8(a0)
|
|
; RV32ZVE32F-NEXT: lw s1, 12(a0)
|
|
; RV32ZVE32F-NEXT: lw t5, 16(a0)
|
|
; RV32ZVE32F-NEXT: lw t6, 20(a0)
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 8, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vsext.vf4 v10, v8
|
|
; RV32ZVE32F-NEXT: vsll.vi v8, v10, 3
|
|
; RV32ZVE32F-NEXT: vsetvli zero, zero, e8, mf2, ta, ma
|
|
; RV32ZVE32F-NEXT: vmv.x.s t0, v0
|
|
; RV32ZVE32F-NEXT: andi s2, t0, 1
|
|
; RV32ZVE32F-NEXT: vsetvli zero, zero, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vadd.vx v8, v8, a1
|
|
; RV32ZVE32F-NEXT: bnez s2, .LBB43_10
|
|
; RV32ZVE32F-NEXT: # %bb.1: # %else
|
|
; RV32ZVE32F-NEXT: andi a0, t0, 2
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB43_11
|
|
; RV32ZVE32F-NEXT: .LBB43_2: # %else2
|
|
; RV32ZVE32F-NEXT: andi a0, t0, 4
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB43_12
|
|
; RV32ZVE32F-NEXT: .LBB43_3: # %else4
|
|
; RV32ZVE32F-NEXT: andi a0, t0, 8
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB43_13
|
|
; RV32ZVE32F-NEXT: .LBB43_4: # %else6
|
|
; RV32ZVE32F-NEXT: andi a0, t0, 16
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB43_14
|
|
; RV32ZVE32F-NEXT: .LBB43_5: # %else8
|
|
; RV32ZVE32F-NEXT: andi a0, t0, 32
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB43_15
|
|
; RV32ZVE32F-NEXT: .LBB43_6: # %else10
|
|
; RV32ZVE32F-NEXT: andi a0, t0, 64
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB43_16
|
|
; RV32ZVE32F-NEXT: .LBB43_7: # %else12
|
|
; RV32ZVE32F-NEXT: andi a0, t0, -128
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB43_9
|
|
; RV32ZVE32F-NEXT: .LBB43_8: # %cond.store13
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v8, v8, 7
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v8
|
|
; RV32ZVE32F-NEXT: sw a2, 0(a0)
|
|
; RV32ZVE32F-NEXT: sw a3, 4(a0)
|
|
; RV32ZVE32F-NEXT: .LBB43_9: # %else14
|
|
; RV32ZVE32F-NEXT: lw s0, 12(sp) # 4-byte Folded Reload
|
|
; RV32ZVE32F-NEXT: lw s1, 8(sp) # 4-byte Folded Reload
|
|
; RV32ZVE32F-NEXT: lw s2, 4(sp) # 4-byte Folded Reload
|
|
; RV32ZVE32F-NEXT: .cfi_restore s0
|
|
; RV32ZVE32F-NEXT: .cfi_restore s1
|
|
; RV32ZVE32F-NEXT: .cfi_restore s2
|
|
; RV32ZVE32F-NEXT: addi sp, sp, 16
|
|
; RV32ZVE32F-NEXT: .cfi_def_cfa_offset 0
|
|
; RV32ZVE32F-NEXT: ret
|
|
; RV32ZVE32F-NEXT: .LBB43_10: # %cond.store
|
|
; RV32ZVE32F-NEXT: .cfi_restore_state
|
|
; RV32ZVE32F-NEXT: lw a1, 0(a0)
|
|
; RV32ZVE32F-NEXT: lw a0, 4(a0)
|
|
; RV32ZVE32F-NEXT: vmv.x.s s2, v8
|
|
; RV32ZVE32F-NEXT: sw a1, 0(s2)
|
|
; RV32ZVE32F-NEXT: sw a0, 4(s2)
|
|
; RV32ZVE32F-NEXT: andi a0, t0, 2
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB43_2
|
|
; RV32ZVE32F-NEXT: .LBB43_11: # %cond.store1
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 1
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: sw s0, 0(a0)
|
|
; RV32ZVE32F-NEXT: sw s1, 4(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, t0, 4
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB43_3
|
|
; RV32ZVE32F-NEXT: .LBB43_12: # %cond.store3
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 2
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: sw t5, 0(a0)
|
|
; RV32ZVE32F-NEXT: sw t6, 4(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, t0, 8
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB43_4
|
|
; RV32ZVE32F-NEXT: .LBB43_13: # %cond.store5
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 3
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: sw t3, 0(a0)
|
|
; RV32ZVE32F-NEXT: sw t4, 4(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, t0, 16
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB43_5
|
|
; RV32ZVE32F-NEXT: .LBB43_14: # %cond.store7
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 4
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: sw t1, 0(a0)
|
|
; RV32ZVE32F-NEXT: sw t2, 4(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, t0, 32
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB43_6
|
|
; RV32ZVE32F-NEXT: .LBB43_15: # %cond.store9
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 5
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: sw a6, 0(a0)
|
|
; RV32ZVE32F-NEXT: sw a7, 4(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, t0, 64
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB43_7
|
|
; RV32ZVE32F-NEXT: .LBB43_16: # %cond.store11
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 6
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: sw a4, 0(a0)
|
|
; RV32ZVE32F-NEXT: sw a5, 4(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, t0, -128
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB43_8
|
|
; RV32ZVE32F-NEXT: j .LBB43_9
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_baseidx_sext_v8i8_v8i64:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: ld a5, 40(a0)
|
|
; RV64ZVE32F-NEXT: ld a3, 48(a0)
|
|
; RV64ZVE32F-NEXT: ld a2, 56(a0)
|
|
; RV64ZVE32F-NEXT: ld t1, 8(a0)
|
|
; RV64ZVE32F-NEXT: ld t0, 16(a0)
|
|
; RV64ZVE32F-NEXT: ld a7, 24(a0)
|
|
; RV64ZVE32F-NEXT: ld a6, 32(a0)
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a4, v0
|
|
; RV64ZVE32F-NEXT: andi t2, a4, 1
|
|
; RV64ZVE32F-NEXT: beqz t2, .LBB43_2
|
|
; RV64ZVE32F-NEXT: # %bb.1: # %cond.store
|
|
; RV64ZVE32F-NEXT: ld a0, 0(a0)
|
|
; RV64ZVE32F-NEXT: vmv.x.s t2, v8
|
|
; RV64ZVE32F-NEXT: slli t2, t2, 3
|
|
; RV64ZVE32F-NEXT: add t2, a1, t2
|
|
; RV64ZVE32F-NEXT: sd a0, 0(t2)
|
|
; RV64ZVE32F-NEXT: .LBB43_2: # %else
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 2
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB43_4
|
|
; RV64ZVE32F-NEXT: # %bb.3: # %cond.store1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a0, v9
|
|
; RV64ZVE32F-NEXT: slli a0, a0, 3
|
|
; RV64ZVE32F-NEXT: add a0, a1, a0
|
|
; RV64ZVE32F-NEXT: sd t1, 0(a0)
|
|
; RV64ZVE32F-NEXT: .LBB43_4: # %else2
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 4
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 2
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB43_6
|
|
; RV64ZVE32F-NEXT: # %bb.5: # %cond.store3
|
|
; RV64ZVE32F-NEXT: vmv.x.s a0, v9
|
|
; RV64ZVE32F-NEXT: slli a0, a0, 3
|
|
; RV64ZVE32F-NEXT: add a0, a1, a0
|
|
; RV64ZVE32F-NEXT: sd t0, 0(a0)
|
|
; RV64ZVE32F-NEXT: .LBB43_6: # %else4
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 8
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB43_8
|
|
; RV64ZVE32F-NEXT: # %bb.7: # %cond.store5
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v9, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a0, v9
|
|
; RV64ZVE32F-NEXT: slli a0, a0, 3
|
|
; RV64ZVE32F-NEXT: add a0, a1, a0
|
|
; RV64ZVE32F-NEXT: sd a7, 0(a0)
|
|
; RV64ZVE32F-NEXT: .LBB43_8: # %else6
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 16
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 4, e8, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 4
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB43_10
|
|
; RV64ZVE32F-NEXT: # %bb.9: # %cond.store7
|
|
; RV64ZVE32F-NEXT: vmv.x.s a0, v8
|
|
; RV64ZVE32F-NEXT: slli a0, a0, 3
|
|
; RV64ZVE32F-NEXT: add a0, a1, a0
|
|
; RV64ZVE32F-NEXT: sd a6, 0(a0)
|
|
; RV64ZVE32F-NEXT: .LBB43_10: # %else8
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 32
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB43_12
|
|
; RV64ZVE32F-NEXT: # %bb.11: # %cond.store9
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a0, v9
|
|
; RV64ZVE32F-NEXT: slli a0, a0, 3
|
|
; RV64ZVE32F-NEXT: add a0, a1, a0
|
|
; RV64ZVE32F-NEXT: sd a5, 0(a0)
|
|
; RV64ZVE32F-NEXT: .LBB43_12: # %else10
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 64
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 2
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB43_15
|
|
; RV64ZVE32F-NEXT: # %bb.13: # %else12
|
|
; RV64ZVE32F-NEXT: andi a0, a4, -128
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB43_16
|
|
; RV64ZVE32F-NEXT: .LBB43_14: # %else14
|
|
; RV64ZVE32F-NEXT: ret
|
|
; RV64ZVE32F-NEXT: .LBB43_15: # %cond.store11
|
|
; RV64ZVE32F-NEXT: vmv.x.s a0, v8
|
|
; RV64ZVE32F-NEXT: slli a0, a0, 3
|
|
; RV64ZVE32F-NEXT: add a0, a1, a0
|
|
; RV64ZVE32F-NEXT: sd a3, 0(a0)
|
|
; RV64ZVE32F-NEXT: andi a0, a4, -128
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB43_14
|
|
; RV64ZVE32F-NEXT: .LBB43_16: # %cond.store13
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a0, v8
|
|
; RV64ZVE32F-NEXT: slli a0, a0, 3
|
|
; RV64ZVE32F-NEXT: add a0, a1, a0
|
|
; RV64ZVE32F-NEXT: sd a2, 0(a0)
|
|
; RV64ZVE32F-NEXT: ret
|
|
%eidxs = sext <8 x i8> %idxs to <8 x i64>
|
|
%ptrs = getelementptr inbounds i64, ptr %base, <8 x i64> %eidxs
|
|
call void @llvm.masked.scatter.v8i64.v8p0(<8 x i64> %val, <8 x ptr> %ptrs, i32 8, <8 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_baseidx_zext_v8i8_v8i64(<8 x i64> %val, ptr %base, <8 x i8> %idxs, <8 x i1> %m) {
|
|
; RV32V-LABEL: mscatter_baseidx_zext_v8i8_v8i64:
|
|
; RV32V: # %bb.0:
|
|
; RV32V-NEXT: li a1, 8
|
|
; RV32V-NEXT: vsetivli zero, 8, e8, mf2, ta, ma
|
|
; RV32V-NEXT: vwmulu.vx v13, v12, a1
|
|
; RV32V-NEXT: vsetvli zero, zero, e64, m4, ta, ma
|
|
; RV32V-NEXT: vsoxei16.v v8, (a0), v13, v0.t
|
|
; RV32V-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_baseidx_zext_v8i8_v8i64:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: li a1, 8
|
|
; RV64V-NEXT: vsetivli zero, 8, e8, mf2, ta, ma
|
|
; RV64V-NEXT: vwmulu.vx v13, v12, a1
|
|
; RV64V-NEXT: vsetvli zero, zero, e64, m4, ta, ma
|
|
; RV64V-NEXT: vsoxei16.v v8, (a0), v13, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV32ZVE32F-LABEL: mscatter_baseidx_zext_v8i8_v8i64:
|
|
; RV32ZVE32F: # %bb.0:
|
|
; RV32ZVE32F-NEXT: addi sp, sp, -16
|
|
; RV32ZVE32F-NEXT: .cfi_def_cfa_offset 16
|
|
; RV32ZVE32F-NEXT: sw s0, 12(sp) # 4-byte Folded Spill
|
|
; RV32ZVE32F-NEXT: sw s1, 8(sp) # 4-byte Folded Spill
|
|
; RV32ZVE32F-NEXT: sw s2, 4(sp) # 4-byte Folded Spill
|
|
; RV32ZVE32F-NEXT: .cfi_offset s0, -4
|
|
; RV32ZVE32F-NEXT: .cfi_offset s1, -8
|
|
; RV32ZVE32F-NEXT: .cfi_offset s2, -12
|
|
; RV32ZVE32F-NEXT: .cfi_remember_state
|
|
; RV32ZVE32F-NEXT: lw a2, 56(a0)
|
|
; RV32ZVE32F-NEXT: lw a3, 60(a0)
|
|
; RV32ZVE32F-NEXT: lw a6, 40(a0)
|
|
; RV32ZVE32F-NEXT: lw a7, 44(a0)
|
|
; RV32ZVE32F-NEXT: lw a4, 48(a0)
|
|
; RV32ZVE32F-NEXT: lw a5, 52(a0)
|
|
; RV32ZVE32F-NEXT: lw t3, 24(a0)
|
|
; RV32ZVE32F-NEXT: lw t4, 28(a0)
|
|
; RV32ZVE32F-NEXT: lw t1, 32(a0)
|
|
; RV32ZVE32F-NEXT: lw t2, 36(a0)
|
|
; RV32ZVE32F-NEXT: lw s0, 8(a0)
|
|
; RV32ZVE32F-NEXT: lw s1, 12(a0)
|
|
; RV32ZVE32F-NEXT: lw t5, 16(a0)
|
|
; RV32ZVE32F-NEXT: lw t6, 20(a0)
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 8, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vzext.vf4 v10, v8
|
|
; RV32ZVE32F-NEXT: vsll.vi v8, v10, 3
|
|
; RV32ZVE32F-NEXT: vsetvli zero, zero, e8, mf2, ta, ma
|
|
; RV32ZVE32F-NEXT: vmv.x.s t0, v0
|
|
; RV32ZVE32F-NEXT: andi s2, t0, 1
|
|
; RV32ZVE32F-NEXT: vsetvli zero, zero, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vadd.vx v8, v8, a1
|
|
; RV32ZVE32F-NEXT: bnez s2, .LBB44_10
|
|
; RV32ZVE32F-NEXT: # %bb.1: # %else
|
|
; RV32ZVE32F-NEXT: andi a0, t0, 2
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB44_11
|
|
; RV32ZVE32F-NEXT: .LBB44_2: # %else2
|
|
; RV32ZVE32F-NEXT: andi a0, t0, 4
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB44_12
|
|
; RV32ZVE32F-NEXT: .LBB44_3: # %else4
|
|
; RV32ZVE32F-NEXT: andi a0, t0, 8
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB44_13
|
|
; RV32ZVE32F-NEXT: .LBB44_4: # %else6
|
|
; RV32ZVE32F-NEXT: andi a0, t0, 16
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB44_14
|
|
; RV32ZVE32F-NEXT: .LBB44_5: # %else8
|
|
; RV32ZVE32F-NEXT: andi a0, t0, 32
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB44_15
|
|
; RV32ZVE32F-NEXT: .LBB44_6: # %else10
|
|
; RV32ZVE32F-NEXT: andi a0, t0, 64
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB44_16
|
|
; RV32ZVE32F-NEXT: .LBB44_7: # %else12
|
|
; RV32ZVE32F-NEXT: andi a0, t0, -128
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB44_9
|
|
; RV32ZVE32F-NEXT: .LBB44_8: # %cond.store13
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v8, v8, 7
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v8
|
|
; RV32ZVE32F-NEXT: sw a2, 0(a0)
|
|
; RV32ZVE32F-NEXT: sw a3, 4(a0)
|
|
; RV32ZVE32F-NEXT: .LBB44_9: # %else14
|
|
; RV32ZVE32F-NEXT: lw s0, 12(sp) # 4-byte Folded Reload
|
|
; RV32ZVE32F-NEXT: lw s1, 8(sp) # 4-byte Folded Reload
|
|
; RV32ZVE32F-NEXT: lw s2, 4(sp) # 4-byte Folded Reload
|
|
; RV32ZVE32F-NEXT: .cfi_restore s0
|
|
; RV32ZVE32F-NEXT: .cfi_restore s1
|
|
; RV32ZVE32F-NEXT: .cfi_restore s2
|
|
; RV32ZVE32F-NEXT: addi sp, sp, 16
|
|
; RV32ZVE32F-NEXT: .cfi_def_cfa_offset 0
|
|
; RV32ZVE32F-NEXT: ret
|
|
; RV32ZVE32F-NEXT: .LBB44_10: # %cond.store
|
|
; RV32ZVE32F-NEXT: .cfi_restore_state
|
|
; RV32ZVE32F-NEXT: lw a1, 0(a0)
|
|
; RV32ZVE32F-NEXT: lw a0, 4(a0)
|
|
; RV32ZVE32F-NEXT: vmv.x.s s2, v8
|
|
; RV32ZVE32F-NEXT: sw a1, 0(s2)
|
|
; RV32ZVE32F-NEXT: sw a0, 4(s2)
|
|
; RV32ZVE32F-NEXT: andi a0, t0, 2
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB44_2
|
|
; RV32ZVE32F-NEXT: .LBB44_11: # %cond.store1
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 1
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: sw s0, 0(a0)
|
|
; RV32ZVE32F-NEXT: sw s1, 4(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, t0, 4
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB44_3
|
|
; RV32ZVE32F-NEXT: .LBB44_12: # %cond.store3
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 2
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: sw t5, 0(a0)
|
|
; RV32ZVE32F-NEXT: sw t6, 4(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, t0, 8
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB44_4
|
|
; RV32ZVE32F-NEXT: .LBB44_13: # %cond.store5
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 3
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: sw t3, 0(a0)
|
|
; RV32ZVE32F-NEXT: sw t4, 4(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, t0, 16
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB44_5
|
|
; RV32ZVE32F-NEXT: .LBB44_14: # %cond.store7
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 4
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: sw t1, 0(a0)
|
|
; RV32ZVE32F-NEXT: sw t2, 4(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, t0, 32
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB44_6
|
|
; RV32ZVE32F-NEXT: .LBB44_15: # %cond.store9
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 5
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: sw a6, 0(a0)
|
|
; RV32ZVE32F-NEXT: sw a7, 4(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, t0, 64
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB44_7
|
|
; RV32ZVE32F-NEXT: .LBB44_16: # %cond.store11
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 6
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: sw a4, 0(a0)
|
|
; RV32ZVE32F-NEXT: sw a5, 4(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, t0, -128
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB44_8
|
|
; RV32ZVE32F-NEXT: j .LBB44_9
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_baseidx_zext_v8i8_v8i64:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: ld a5, 40(a0)
|
|
; RV64ZVE32F-NEXT: ld a3, 48(a0)
|
|
; RV64ZVE32F-NEXT: ld a2, 56(a0)
|
|
; RV64ZVE32F-NEXT: ld t1, 8(a0)
|
|
; RV64ZVE32F-NEXT: ld t0, 16(a0)
|
|
; RV64ZVE32F-NEXT: ld a7, 24(a0)
|
|
; RV64ZVE32F-NEXT: ld a6, 32(a0)
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a4, v0
|
|
; RV64ZVE32F-NEXT: andi t2, a4, 1
|
|
; RV64ZVE32F-NEXT: beqz t2, .LBB44_2
|
|
; RV64ZVE32F-NEXT: # %bb.1: # %cond.store
|
|
; RV64ZVE32F-NEXT: ld a0, 0(a0)
|
|
; RV64ZVE32F-NEXT: vmv.x.s t2, v8
|
|
; RV64ZVE32F-NEXT: zext.b t2, t2
|
|
; RV64ZVE32F-NEXT: slli t2, t2, 3
|
|
; RV64ZVE32F-NEXT: add t2, a1, t2
|
|
; RV64ZVE32F-NEXT: sd a0, 0(t2)
|
|
; RV64ZVE32F-NEXT: .LBB44_2: # %else
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 2
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB44_4
|
|
; RV64ZVE32F-NEXT: # %bb.3: # %cond.store1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a0, v9
|
|
; RV64ZVE32F-NEXT: zext.b a0, a0
|
|
; RV64ZVE32F-NEXT: slli a0, a0, 3
|
|
; RV64ZVE32F-NEXT: add a0, a1, a0
|
|
; RV64ZVE32F-NEXT: sd t1, 0(a0)
|
|
; RV64ZVE32F-NEXT: .LBB44_4: # %else2
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 4
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 2
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB44_6
|
|
; RV64ZVE32F-NEXT: # %bb.5: # %cond.store3
|
|
; RV64ZVE32F-NEXT: vmv.x.s a0, v9
|
|
; RV64ZVE32F-NEXT: zext.b a0, a0
|
|
; RV64ZVE32F-NEXT: slli a0, a0, 3
|
|
; RV64ZVE32F-NEXT: add a0, a1, a0
|
|
; RV64ZVE32F-NEXT: sd t0, 0(a0)
|
|
; RV64ZVE32F-NEXT: .LBB44_6: # %else4
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 8
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB44_8
|
|
; RV64ZVE32F-NEXT: # %bb.7: # %cond.store5
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v9, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a0, v9
|
|
; RV64ZVE32F-NEXT: zext.b a0, a0
|
|
; RV64ZVE32F-NEXT: slli a0, a0, 3
|
|
; RV64ZVE32F-NEXT: add a0, a1, a0
|
|
; RV64ZVE32F-NEXT: sd a7, 0(a0)
|
|
; RV64ZVE32F-NEXT: .LBB44_8: # %else6
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 16
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 4, e8, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 4
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB44_10
|
|
; RV64ZVE32F-NEXT: # %bb.9: # %cond.store7
|
|
; RV64ZVE32F-NEXT: vmv.x.s a0, v8
|
|
; RV64ZVE32F-NEXT: zext.b a0, a0
|
|
; RV64ZVE32F-NEXT: slli a0, a0, 3
|
|
; RV64ZVE32F-NEXT: add a0, a1, a0
|
|
; RV64ZVE32F-NEXT: sd a6, 0(a0)
|
|
; RV64ZVE32F-NEXT: .LBB44_10: # %else8
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 32
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB44_12
|
|
; RV64ZVE32F-NEXT: # %bb.11: # %cond.store9
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a0, v9
|
|
; RV64ZVE32F-NEXT: zext.b a0, a0
|
|
; RV64ZVE32F-NEXT: slli a0, a0, 3
|
|
; RV64ZVE32F-NEXT: add a0, a1, a0
|
|
; RV64ZVE32F-NEXT: sd a5, 0(a0)
|
|
; RV64ZVE32F-NEXT: .LBB44_12: # %else10
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 64
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 2
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB44_15
|
|
; RV64ZVE32F-NEXT: # %bb.13: # %else12
|
|
; RV64ZVE32F-NEXT: andi a0, a4, -128
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB44_16
|
|
; RV64ZVE32F-NEXT: .LBB44_14: # %else14
|
|
; RV64ZVE32F-NEXT: ret
|
|
; RV64ZVE32F-NEXT: .LBB44_15: # %cond.store11
|
|
; RV64ZVE32F-NEXT: vmv.x.s a0, v8
|
|
; RV64ZVE32F-NEXT: zext.b a0, a0
|
|
; RV64ZVE32F-NEXT: slli a0, a0, 3
|
|
; RV64ZVE32F-NEXT: add a0, a1, a0
|
|
; RV64ZVE32F-NEXT: sd a3, 0(a0)
|
|
; RV64ZVE32F-NEXT: andi a0, a4, -128
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB44_14
|
|
; RV64ZVE32F-NEXT: .LBB44_16: # %cond.store13
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a0, v8
|
|
; RV64ZVE32F-NEXT: zext.b a0, a0
|
|
; RV64ZVE32F-NEXT: slli a0, a0, 3
|
|
; RV64ZVE32F-NEXT: add a0, a1, a0
|
|
; RV64ZVE32F-NEXT: sd a2, 0(a0)
|
|
; RV64ZVE32F-NEXT: ret
|
|
%eidxs = zext <8 x i8> %idxs to <8 x i64>
|
|
%ptrs = getelementptr inbounds i64, ptr %base, <8 x i64> %eidxs
|
|
call void @llvm.masked.scatter.v8i64.v8p0(<8 x i64> %val, <8 x ptr> %ptrs, i32 8, <8 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_baseidx_v8i16_v8i64(<8 x i64> %val, ptr %base, <8 x i16> %idxs, <8 x i1> %m) {
|
|
; RV32V-LABEL: mscatter_baseidx_v8i16_v8i64:
|
|
; RV32V: # %bb.0:
|
|
; RV32V-NEXT: li a1, 8
|
|
; RV32V-NEXT: vsetivli zero, 8, e16, m1, ta, ma
|
|
; RV32V-NEXT: vwmulsu.vx v14, v12, a1
|
|
; RV32V-NEXT: vsetvli zero, zero, e64, m4, ta, ma
|
|
; RV32V-NEXT: vsoxei32.v v8, (a0), v14, v0.t
|
|
; RV32V-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_baseidx_v8i16_v8i64:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: vsetivli zero, 8, e64, m4, ta, ma
|
|
; RV64V-NEXT: vsext.vf4 v16, v12
|
|
; RV64V-NEXT: vsll.vi v12, v16, 3
|
|
; RV64V-NEXT: vsoxei64.v v8, (a0), v12, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV32ZVE32F-LABEL: mscatter_baseidx_v8i16_v8i64:
|
|
; RV32ZVE32F: # %bb.0:
|
|
; RV32ZVE32F-NEXT: addi sp, sp, -16
|
|
; RV32ZVE32F-NEXT: .cfi_def_cfa_offset 16
|
|
; RV32ZVE32F-NEXT: sw s0, 12(sp) # 4-byte Folded Spill
|
|
; RV32ZVE32F-NEXT: sw s1, 8(sp) # 4-byte Folded Spill
|
|
; RV32ZVE32F-NEXT: sw s2, 4(sp) # 4-byte Folded Spill
|
|
; RV32ZVE32F-NEXT: .cfi_offset s0, -4
|
|
; RV32ZVE32F-NEXT: .cfi_offset s1, -8
|
|
; RV32ZVE32F-NEXT: .cfi_offset s2, -12
|
|
; RV32ZVE32F-NEXT: .cfi_remember_state
|
|
; RV32ZVE32F-NEXT: lw a2, 56(a0)
|
|
; RV32ZVE32F-NEXT: lw a3, 60(a0)
|
|
; RV32ZVE32F-NEXT: lw a6, 40(a0)
|
|
; RV32ZVE32F-NEXT: lw a7, 44(a0)
|
|
; RV32ZVE32F-NEXT: lw a4, 48(a0)
|
|
; RV32ZVE32F-NEXT: lw a5, 52(a0)
|
|
; RV32ZVE32F-NEXT: lw t2, 24(a0)
|
|
; RV32ZVE32F-NEXT: lw t3, 28(a0)
|
|
; RV32ZVE32F-NEXT: lw t0, 32(a0)
|
|
; RV32ZVE32F-NEXT: lw t1, 36(a0)
|
|
; RV32ZVE32F-NEXT: lw t6, 8(a0)
|
|
; RV32ZVE32F-NEXT: lw s0, 12(a0)
|
|
; RV32ZVE32F-NEXT: lw t4, 16(a0)
|
|
; RV32ZVE32F-NEXT: lw t5, 20(a0)
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 8, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vmv.v.x v10, a1
|
|
; RV32ZVE32F-NEXT: li s1, 8
|
|
; RV32ZVE32F-NEXT: vsetvli zero, zero, e8, mf2, ta, ma
|
|
; RV32ZVE32F-NEXT: vmv.x.s a1, v0
|
|
; RV32ZVE32F-NEXT: andi s2, a1, 1
|
|
; RV32ZVE32F-NEXT: vsetvli zero, zero, e16, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vwmaccus.vx v10, s1, v8
|
|
; RV32ZVE32F-NEXT: bnez s2, .LBB45_10
|
|
; RV32ZVE32F-NEXT: # %bb.1: # %else
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 2
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB45_11
|
|
; RV32ZVE32F-NEXT: .LBB45_2: # %else2
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 4
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB45_12
|
|
; RV32ZVE32F-NEXT: .LBB45_3: # %else4
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 8
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB45_13
|
|
; RV32ZVE32F-NEXT: .LBB45_4: # %else6
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 16
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB45_14
|
|
; RV32ZVE32F-NEXT: .LBB45_5: # %else8
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 32
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB45_15
|
|
; RV32ZVE32F-NEXT: .LBB45_6: # %else10
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 64
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB45_16
|
|
; RV32ZVE32F-NEXT: .LBB45_7: # %else12
|
|
; RV32ZVE32F-NEXT: andi a0, a1, -128
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB45_9
|
|
; RV32ZVE32F-NEXT: .LBB45_8: # %cond.store13
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v8, v10, 7
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v8
|
|
; RV32ZVE32F-NEXT: sw a2, 0(a0)
|
|
; RV32ZVE32F-NEXT: sw a3, 4(a0)
|
|
; RV32ZVE32F-NEXT: .LBB45_9: # %else14
|
|
; RV32ZVE32F-NEXT: lw s0, 12(sp) # 4-byte Folded Reload
|
|
; RV32ZVE32F-NEXT: lw s1, 8(sp) # 4-byte Folded Reload
|
|
; RV32ZVE32F-NEXT: lw s2, 4(sp) # 4-byte Folded Reload
|
|
; RV32ZVE32F-NEXT: .cfi_restore s0
|
|
; RV32ZVE32F-NEXT: .cfi_restore s1
|
|
; RV32ZVE32F-NEXT: .cfi_restore s2
|
|
; RV32ZVE32F-NEXT: addi sp, sp, 16
|
|
; RV32ZVE32F-NEXT: .cfi_def_cfa_offset 0
|
|
; RV32ZVE32F-NEXT: ret
|
|
; RV32ZVE32F-NEXT: .LBB45_10: # %cond.store
|
|
; RV32ZVE32F-NEXT: .cfi_restore_state
|
|
; RV32ZVE32F-NEXT: lw s1, 0(a0)
|
|
; RV32ZVE32F-NEXT: lw a0, 4(a0)
|
|
; RV32ZVE32F-NEXT: vsetvli zero, zero, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vmv.x.s s2, v10
|
|
; RV32ZVE32F-NEXT: sw s1, 0(s2)
|
|
; RV32ZVE32F-NEXT: sw a0, 4(s2)
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 2
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB45_2
|
|
; RV32ZVE32F-NEXT: .LBB45_11: # %cond.store1
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v8, v10, 1
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v8
|
|
; RV32ZVE32F-NEXT: sw t6, 0(a0)
|
|
; RV32ZVE32F-NEXT: sw s0, 4(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 4
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB45_3
|
|
; RV32ZVE32F-NEXT: .LBB45_12: # %cond.store3
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v8, v10, 2
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v8
|
|
; RV32ZVE32F-NEXT: sw t4, 0(a0)
|
|
; RV32ZVE32F-NEXT: sw t5, 4(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 8
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB45_4
|
|
; RV32ZVE32F-NEXT: .LBB45_13: # %cond.store5
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v8, v10, 3
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v8
|
|
; RV32ZVE32F-NEXT: sw t2, 0(a0)
|
|
; RV32ZVE32F-NEXT: sw t3, 4(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 16
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB45_5
|
|
; RV32ZVE32F-NEXT: .LBB45_14: # %cond.store7
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v8, v10, 4
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v8
|
|
; RV32ZVE32F-NEXT: sw t0, 0(a0)
|
|
; RV32ZVE32F-NEXT: sw t1, 4(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 32
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB45_6
|
|
; RV32ZVE32F-NEXT: .LBB45_15: # %cond.store9
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v8, v10, 5
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v8
|
|
; RV32ZVE32F-NEXT: sw a6, 0(a0)
|
|
; RV32ZVE32F-NEXT: sw a7, 4(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 64
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB45_7
|
|
; RV32ZVE32F-NEXT: .LBB45_16: # %cond.store11
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v8, v10, 6
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v8
|
|
; RV32ZVE32F-NEXT: sw a4, 0(a0)
|
|
; RV32ZVE32F-NEXT: sw a5, 4(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, a1, -128
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB45_8
|
|
; RV32ZVE32F-NEXT: j .LBB45_9
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_baseidx_v8i16_v8i64:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: ld a5, 40(a0)
|
|
; RV64ZVE32F-NEXT: ld a3, 48(a0)
|
|
; RV64ZVE32F-NEXT: ld a2, 56(a0)
|
|
; RV64ZVE32F-NEXT: ld t1, 8(a0)
|
|
; RV64ZVE32F-NEXT: ld t0, 16(a0)
|
|
; RV64ZVE32F-NEXT: ld a7, 24(a0)
|
|
; RV64ZVE32F-NEXT: ld a6, 32(a0)
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a4, v0
|
|
; RV64ZVE32F-NEXT: andi t2, a4, 1
|
|
; RV64ZVE32F-NEXT: beqz t2, .LBB45_2
|
|
; RV64ZVE32F-NEXT: # %bb.1: # %cond.store
|
|
; RV64ZVE32F-NEXT: ld a0, 0(a0)
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e16, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s t2, v8
|
|
; RV64ZVE32F-NEXT: slli t2, t2, 3
|
|
; RV64ZVE32F-NEXT: add t2, a1, t2
|
|
; RV64ZVE32F-NEXT: sd a0, 0(t2)
|
|
; RV64ZVE32F-NEXT: .LBB45_2: # %else
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 2
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB45_4
|
|
; RV64ZVE32F-NEXT: # %bb.3: # %cond.store1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a0, v9
|
|
; RV64ZVE32F-NEXT: slli a0, a0, 3
|
|
; RV64ZVE32F-NEXT: add a0, a1, a0
|
|
; RV64ZVE32F-NEXT: sd t1, 0(a0)
|
|
; RV64ZVE32F-NEXT: .LBB45_4: # %else2
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 4
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 2
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB45_6
|
|
; RV64ZVE32F-NEXT: # %bb.5: # %cond.store3
|
|
; RV64ZVE32F-NEXT: vmv.x.s a0, v9
|
|
; RV64ZVE32F-NEXT: slli a0, a0, 3
|
|
; RV64ZVE32F-NEXT: add a0, a1, a0
|
|
; RV64ZVE32F-NEXT: sd t0, 0(a0)
|
|
; RV64ZVE32F-NEXT: .LBB45_6: # %else4
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 8
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB45_8
|
|
; RV64ZVE32F-NEXT: # %bb.7: # %cond.store5
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v9, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a0, v9
|
|
; RV64ZVE32F-NEXT: slli a0, a0, 3
|
|
; RV64ZVE32F-NEXT: add a0, a1, a0
|
|
; RV64ZVE32F-NEXT: sd a7, 0(a0)
|
|
; RV64ZVE32F-NEXT: .LBB45_8: # %else6
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 16
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 4, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 4
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB45_10
|
|
; RV64ZVE32F-NEXT: # %bb.9: # %cond.store7
|
|
; RV64ZVE32F-NEXT: vmv.x.s a0, v8
|
|
; RV64ZVE32F-NEXT: slli a0, a0, 3
|
|
; RV64ZVE32F-NEXT: add a0, a1, a0
|
|
; RV64ZVE32F-NEXT: sd a6, 0(a0)
|
|
; RV64ZVE32F-NEXT: .LBB45_10: # %else8
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 32
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB45_12
|
|
; RV64ZVE32F-NEXT: # %bb.11: # %cond.store9
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a0, v9
|
|
; RV64ZVE32F-NEXT: slli a0, a0, 3
|
|
; RV64ZVE32F-NEXT: add a0, a1, a0
|
|
; RV64ZVE32F-NEXT: sd a5, 0(a0)
|
|
; RV64ZVE32F-NEXT: .LBB45_12: # %else10
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 64
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 2
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB45_15
|
|
; RV64ZVE32F-NEXT: # %bb.13: # %else12
|
|
; RV64ZVE32F-NEXT: andi a0, a4, -128
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB45_16
|
|
; RV64ZVE32F-NEXT: .LBB45_14: # %else14
|
|
; RV64ZVE32F-NEXT: ret
|
|
; RV64ZVE32F-NEXT: .LBB45_15: # %cond.store11
|
|
; RV64ZVE32F-NEXT: vmv.x.s a0, v8
|
|
; RV64ZVE32F-NEXT: slli a0, a0, 3
|
|
; RV64ZVE32F-NEXT: add a0, a1, a0
|
|
; RV64ZVE32F-NEXT: sd a3, 0(a0)
|
|
; RV64ZVE32F-NEXT: andi a0, a4, -128
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB45_14
|
|
; RV64ZVE32F-NEXT: .LBB45_16: # %cond.store13
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a0, v8
|
|
; RV64ZVE32F-NEXT: slli a0, a0, 3
|
|
; RV64ZVE32F-NEXT: add a0, a1, a0
|
|
; RV64ZVE32F-NEXT: sd a2, 0(a0)
|
|
; RV64ZVE32F-NEXT: ret
|
|
%ptrs = getelementptr inbounds i64, ptr %base, <8 x i16> %idxs
|
|
call void @llvm.masked.scatter.v8i64.v8p0(<8 x i64> %val, <8 x ptr> %ptrs, i32 8, <8 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_baseidx_sext_v8i16_v8i64(<8 x i64> %val, ptr %base, <8 x i16> %idxs, <8 x i1> %m) {
|
|
; RV32V-LABEL: mscatter_baseidx_sext_v8i16_v8i64:
|
|
; RV32V: # %bb.0:
|
|
; RV32V-NEXT: li a1, 8
|
|
; RV32V-NEXT: vsetivli zero, 8, e16, m1, ta, ma
|
|
; RV32V-NEXT: vwmulsu.vx v14, v12, a1
|
|
; RV32V-NEXT: vsetvli zero, zero, e64, m4, ta, ma
|
|
; RV32V-NEXT: vsoxei32.v v8, (a0), v14, v0.t
|
|
; RV32V-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_baseidx_sext_v8i16_v8i64:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: vsetivli zero, 8, e64, m4, ta, ma
|
|
; RV64V-NEXT: vsext.vf4 v16, v12
|
|
; RV64V-NEXT: vsll.vi v12, v16, 3
|
|
; RV64V-NEXT: vsoxei64.v v8, (a0), v12, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV32ZVE32F-LABEL: mscatter_baseidx_sext_v8i16_v8i64:
|
|
; RV32ZVE32F: # %bb.0:
|
|
; RV32ZVE32F-NEXT: addi sp, sp, -16
|
|
; RV32ZVE32F-NEXT: .cfi_def_cfa_offset 16
|
|
; RV32ZVE32F-NEXT: sw s0, 12(sp) # 4-byte Folded Spill
|
|
; RV32ZVE32F-NEXT: sw s1, 8(sp) # 4-byte Folded Spill
|
|
; RV32ZVE32F-NEXT: sw s2, 4(sp) # 4-byte Folded Spill
|
|
; RV32ZVE32F-NEXT: .cfi_offset s0, -4
|
|
; RV32ZVE32F-NEXT: .cfi_offset s1, -8
|
|
; RV32ZVE32F-NEXT: .cfi_offset s2, -12
|
|
; RV32ZVE32F-NEXT: .cfi_remember_state
|
|
; RV32ZVE32F-NEXT: lw a2, 56(a0)
|
|
; RV32ZVE32F-NEXT: lw a3, 60(a0)
|
|
; RV32ZVE32F-NEXT: lw a6, 40(a0)
|
|
; RV32ZVE32F-NEXT: lw a7, 44(a0)
|
|
; RV32ZVE32F-NEXT: lw a4, 48(a0)
|
|
; RV32ZVE32F-NEXT: lw a5, 52(a0)
|
|
; RV32ZVE32F-NEXT: lw t2, 24(a0)
|
|
; RV32ZVE32F-NEXT: lw t3, 28(a0)
|
|
; RV32ZVE32F-NEXT: lw t0, 32(a0)
|
|
; RV32ZVE32F-NEXT: lw t1, 36(a0)
|
|
; RV32ZVE32F-NEXT: lw t6, 8(a0)
|
|
; RV32ZVE32F-NEXT: lw s0, 12(a0)
|
|
; RV32ZVE32F-NEXT: lw t4, 16(a0)
|
|
; RV32ZVE32F-NEXT: lw t5, 20(a0)
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 8, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vmv.v.x v10, a1
|
|
; RV32ZVE32F-NEXT: li s1, 8
|
|
; RV32ZVE32F-NEXT: vsetvli zero, zero, e8, mf2, ta, ma
|
|
; RV32ZVE32F-NEXT: vmv.x.s a1, v0
|
|
; RV32ZVE32F-NEXT: andi s2, a1, 1
|
|
; RV32ZVE32F-NEXT: vsetvli zero, zero, e16, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vwmaccus.vx v10, s1, v8
|
|
; RV32ZVE32F-NEXT: bnez s2, .LBB46_10
|
|
; RV32ZVE32F-NEXT: # %bb.1: # %else
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 2
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB46_11
|
|
; RV32ZVE32F-NEXT: .LBB46_2: # %else2
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 4
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB46_12
|
|
; RV32ZVE32F-NEXT: .LBB46_3: # %else4
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 8
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB46_13
|
|
; RV32ZVE32F-NEXT: .LBB46_4: # %else6
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 16
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB46_14
|
|
; RV32ZVE32F-NEXT: .LBB46_5: # %else8
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 32
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB46_15
|
|
; RV32ZVE32F-NEXT: .LBB46_6: # %else10
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 64
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB46_16
|
|
; RV32ZVE32F-NEXT: .LBB46_7: # %else12
|
|
; RV32ZVE32F-NEXT: andi a0, a1, -128
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB46_9
|
|
; RV32ZVE32F-NEXT: .LBB46_8: # %cond.store13
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v8, v10, 7
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v8
|
|
; RV32ZVE32F-NEXT: sw a2, 0(a0)
|
|
; RV32ZVE32F-NEXT: sw a3, 4(a0)
|
|
; RV32ZVE32F-NEXT: .LBB46_9: # %else14
|
|
; RV32ZVE32F-NEXT: lw s0, 12(sp) # 4-byte Folded Reload
|
|
; RV32ZVE32F-NEXT: lw s1, 8(sp) # 4-byte Folded Reload
|
|
; RV32ZVE32F-NEXT: lw s2, 4(sp) # 4-byte Folded Reload
|
|
; RV32ZVE32F-NEXT: .cfi_restore s0
|
|
; RV32ZVE32F-NEXT: .cfi_restore s1
|
|
; RV32ZVE32F-NEXT: .cfi_restore s2
|
|
; RV32ZVE32F-NEXT: addi sp, sp, 16
|
|
; RV32ZVE32F-NEXT: .cfi_def_cfa_offset 0
|
|
; RV32ZVE32F-NEXT: ret
|
|
; RV32ZVE32F-NEXT: .LBB46_10: # %cond.store
|
|
; RV32ZVE32F-NEXT: .cfi_restore_state
|
|
; RV32ZVE32F-NEXT: lw s1, 0(a0)
|
|
; RV32ZVE32F-NEXT: lw a0, 4(a0)
|
|
; RV32ZVE32F-NEXT: vsetvli zero, zero, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vmv.x.s s2, v10
|
|
; RV32ZVE32F-NEXT: sw s1, 0(s2)
|
|
; RV32ZVE32F-NEXT: sw a0, 4(s2)
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 2
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB46_2
|
|
; RV32ZVE32F-NEXT: .LBB46_11: # %cond.store1
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v8, v10, 1
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v8
|
|
; RV32ZVE32F-NEXT: sw t6, 0(a0)
|
|
; RV32ZVE32F-NEXT: sw s0, 4(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 4
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB46_3
|
|
; RV32ZVE32F-NEXT: .LBB46_12: # %cond.store3
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v8, v10, 2
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v8
|
|
; RV32ZVE32F-NEXT: sw t4, 0(a0)
|
|
; RV32ZVE32F-NEXT: sw t5, 4(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 8
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB46_4
|
|
; RV32ZVE32F-NEXT: .LBB46_13: # %cond.store5
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v8, v10, 3
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v8
|
|
; RV32ZVE32F-NEXT: sw t2, 0(a0)
|
|
; RV32ZVE32F-NEXT: sw t3, 4(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 16
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB46_5
|
|
; RV32ZVE32F-NEXT: .LBB46_14: # %cond.store7
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v8, v10, 4
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v8
|
|
; RV32ZVE32F-NEXT: sw t0, 0(a0)
|
|
; RV32ZVE32F-NEXT: sw t1, 4(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 32
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB46_6
|
|
; RV32ZVE32F-NEXT: .LBB46_15: # %cond.store9
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v8, v10, 5
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v8
|
|
; RV32ZVE32F-NEXT: sw a6, 0(a0)
|
|
; RV32ZVE32F-NEXT: sw a7, 4(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 64
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB46_7
|
|
; RV32ZVE32F-NEXT: .LBB46_16: # %cond.store11
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v8, v10, 6
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v8
|
|
; RV32ZVE32F-NEXT: sw a4, 0(a0)
|
|
; RV32ZVE32F-NEXT: sw a5, 4(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, a1, -128
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB46_8
|
|
; RV32ZVE32F-NEXT: j .LBB46_9
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_baseidx_sext_v8i16_v8i64:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: ld a5, 40(a0)
|
|
; RV64ZVE32F-NEXT: ld a3, 48(a0)
|
|
; RV64ZVE32F-NEXT: ld a2, 56(a0)
|
|
; RV64ZVE32F-NEXT: ld t1, 8(a0)
|
|
; RV64ZVE32F-NEXT: ld t0, 16(a0)
|
|
; RV64ZVE32F-NEXT: ld a7, 24(a0)
|
|
; RV64ZVE32F-NEXT: ld a6, 32(a0)
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a4, v0
|
|
; RV64ZVE32F-NEXT: andi t2, a4, 1
|
|
; RV64ZVE32F-NEXT: beqz t2, .LBB46_2
|
|
; RV64ZVE32F-NEXT: # %bb.1: # %cond.store
|
|
; RV64ZVE32F-NEXT: ld a0, 0(a0)
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e16, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s t2, v8
|
|
; RV64ZVE32F-NEXT: slli t2, t2, 3
|
|
; RV64ZVE32F-NEXT: add t2, a1, t2
|
|
; RV64ZVE32F-NEXT: sd a0, 0(t2)
|
|
; RV64ZVE32F-NEXT: .LBB46_2: # %else
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 2
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB46_4
|
|
; RV64ZVE32F-NEXT: # %bb.3: # %cond.store1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a0, v9
|
|
; RV64ZVE32F-NEXT: slli a0, a0, 3
|
|
; RV64ZVE32F-NEXT: add a0, a1, a0
|
|
; RV64ZVE32F-NEXT: sd t1, 0(a0)
|
|
; RV64ZVE32F-NEXT: .LBB46_4: # %else2
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 4
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 2
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB46_6
|
|
; RV64ZVE32F-NEXT: # %bb.5: # %cond.store3
|
|
; RV64ZVE32F-NEXT: vmv.x.s a0, v9
|
|
; RV64ZVE32F-NEXT: slli a0, a0, 3
|
|
; RV64ZVE32F-NEXT: add a0, a1, a0
|
|
; RV64ZVE32F-NEXT: sd t0, 0(a0)
|
|
; RV64ZVE32F-NEXT: .LBB46_6: # %else4
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 8
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB46_8
|
|
; RV64ZVE32F-NEXT: # %bb.7: # %cond.store5
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v9, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a0, v9
|
|
; RV64ZVE32F-NEXT: slli a0, a0, 3
|
|
; RV64ZVE32F-NEXT: add a0, a1, a0
|
|
; RV64ZVE32F-NEXT: sd a7, 0(a0)
|
|
; RV64ZVE32F-NEXT: .LBB46_8: # %else6
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 16
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 4, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 4
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB46_10
|
|
; RV64ZVE32F-NEXT: # %bb.9: # %cond.store7
|
|
; RV64ZVE32F-NEXT: vmv.x.s a0, v8
|
|
; RV64ZVE32F-NEXT: slli a0, a0, 3
|
|
; RV64ZVE32F-NEXT: add a0, a1, a0
|
|
; RV64ZVE32F-NEXT: sd a6, 0(a0)
|
|
; RV64ZVE32F-NEXT: .LBB46_10: # %else8
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 32
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB46_12
|
|
; RV64ZVE32F-NEXT: # %bb.11: # %cond.store9
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a0, v9
|
|
; RV64ZVE32F-NEXT: slli a0, a0, 3
|
|
; RV64ZVE32F-NEXT: add a0, a1, a0
|
|
; RV64ZVE32F-NEXT: sd a5, 0(a0)
|
|
; RV64ZVE32F-NEXT: .LBB46_12: # %else10
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 64
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 2
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB46_15
|
|
; RV64ZVE32F-NEXT: # %bb.13: # %else12
|
|
; RV64ZVE32F-NEXT: andi a0, a4, -128
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB46_16
|
|
; RV64ZVE32F-NEXT: .LBB46_14: # %else14
|
|
; RV64ZVE32F-NEXT: ret
|
|
; RV64ZVE32F-NEXT: .LBB46_15: # %cond.store11
|
|
; RV64ZVE32F-NEXT: vmv.x.s a0, v8
|
|
; RV64ZVE32F-NEXT: slli a0, a0, 3
|
|
; RV64ZVE32F-NEXT: add a0, a1, a0
|
|
; RV64ZVE32F-NEXT: sd a3, 0(a0)
|
|
; RV64ZVE32F-NEXT: andi a0, a4, -128
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB46_14
|
|
; RV64ZVE32F-NEXT: .LBB46_16: # %cond.store13
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a0, v8
|
|
; RV64ZVE32F-NEXT: slli a0, a0, 3
|
|
; RV64ZVE32F-NEXT: add a0, a1, a0
|
|
; RV64ZVE32F-NEXT: sd a2, 0(a0)
|
|
; RV64ZVE32F-NEXT: ret
|
|
%eidxs = sext <8 x i16> %idxs to <8 x i64>
|
|
%ptrs = getelementptr inbounds i64, ptr %base, <8 x i64> %eidxs
|
|
call void @llvm.masked.scatter.v8i64.v8p0(<8 x i64> %val, <8 x ptr> %ptrs, i32 8, <8 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_baseidx_zext_v8i16_v8i64(<8 x i64> %val, ptr %base, <8 x i16> %idxs, <8 x i1> %m) {
|
|
; RV32V-LABEL: mscatter_baseidx_zext_v8i16_v8i64:
|
|
; RV32V: # %bb.0:
|
|
; RV32V-NEXT: li a1, 8
|
|
; RV32V-NEXT: vsetivli zero, 8, e16, m1, ta, ma
|
|
; RV32V-NEXT: vwmulu.vx v14, v12, a1
|
|
; RV32V-NEXT: vsetvli zero, zero, e64, m4, ta, ma
|
|
; RV32V-NEXT: vsoxei32.v v8, (a0), v14, v0.t
|
|
; RV32V-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_baseidx_zext_v8i16_v8i64:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: li a1, 8
|
|
; RV64V-NEXT: vsetivli zero, 8, e16, m1, ta, ma
|
|
; RV64V-NEXT: vwmulu.vx v14, v12, a1
|
|
; RV64V-NEXT: vsetvli zero, zero, e64, m4, ta, ma
|
|
; RV64V-NEXT: vsoxei32.v v8, (a0), v14, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV32ZVE32F-LABEL: mscatter_baseidx_zext_v8i16_v8i64:
|
|
; RV32ZVE32F: # %bb.0:
|
|
; RV32ZVE32F-NEXT: addi sp, sp, -16
|
|
; RV32ZVE32F-NEXT: .cfi_def_cfa_offset 16
|
|
; RV32ZVE32F-NEXT: sw s0, 12(sp) # 4-byte Folded Spill
|
|
; RV32ZVE32F-NEXT: sw s1, 8(sp) # 4-byte Folded Spill
|
|
; RV32ZVE32F-NEXT: sw s2, 4(sp) # 4-byte Folded Spill
|
|
; RV32ZVE32F-NEXT: .cfi_offset s0, -4
|
|
; RV32ZVE32F-NEXT: .cfi_offset s1, -8
|
|
; RV32ZVE32F-NEXT: .cfi_offset s2, -12
|
|
; RV32ZVE32F-NEXT: .cfi_remember_state
|
|
; RV32ZVE32F-NEXT: lw a2, 56(a0)
|
|
; RV32ZVE32F-NEXT: lw a3, 60(a0)
|
|
; RV32ZVE32F-NEXT: lw a6, 40(a0)
|
|
; RV32ZVE32F-NEXT: lw a7, 44(a0)
|
|
; RV32ZVE32F-NEXT: lw a4, 48(a0)
|
|
; RV32ZVE32F-NEXT: lw a5, 52(a0)
|
|
; RV32ZVE32F-NEXT: lw t2, 24(a0)
|
|
; RV32ZVE32F-NEXT: lw t3, 28(a0)
|
|
; RV32ZVE32F-NEXT: lw t0, 32(a0)
|
|
; RV32ZVE32F-NEXT: lw t1, 36(a0)
|
|
; RV32ZVE32F-NEXT: lw t6, 8(a0)
|
|
; RV32ZVE32F-NEXT: lw s0, 12(a0)
|
|
; RV32ZVE32F-NEXT: lw t4, 16(a0)
|
|
; RV32ZVE32F-NEXT: lw t5, 20(a0)
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 8, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vmv.v.x v10, a1
|
|
; RV32ZVE32F-NEXT: li s1, 8
|
|
; RV32ZVE32F-NEXT: vsetvli zero, zero, e8, mf2, ta, ma
|
|
; RV32ZVE32F-NEXT: vmv.x.s a1, v0
|
|
; RV32ZVE32F-NEXT: andi s2, a1, 1
|
|
; RV32ZVE32F-NEXT: vsetvli zero, zero, e16, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vwmaccu.vx v10, s1, v8
|
|
; RV32ZVE32F-NEXT: bnez s2, .LBB47_10
|
|
; RV32ZVE32F-NEXT: # %bb.1: # %else
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 2
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB47_11
|
|
; RV32ZVE32F-NEXT: .LBB47_2: # %else2
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 4
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB47_12
|
|
; RV32ZVE32F-NEXT: .LBB47_3: # %else4
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 8
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB47_13
|
|
; RV32ZVE32F-NEXT: .LBB47_4: # %else6
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 16
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB47_14
|
|
; RV32ZVE32F-NEXT: .LBB47_5: # %else8
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 32
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB47_15
|
|
; RV32ZVE32F-NEXT: .LBB47_6: # %else10
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 64
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB47_16
|
|
; RV32ZVE32F-NEXT: .LBB47_7: # %else12
|
|
; RV32ZVE32F-NEXT: andi a0, a1, -128
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB47_9
|
|
; RV32ZVE32F-NEXT: .LBB47_8: # %cond.store13
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v8, v10, 7
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v8
|
|
; RV32ZVE32F-NEXT: sw a2, 0(a0)
|
|
; RV32ZVE32F-NEXT: sw a3, 4(a0)
|
|
; RV32ZVE32F-NEXT: .LBB47_9: # %else14
|
|
; RV32ZVE32F-NEXT: lw s0, 12(sp) # 4-byte Folded Reload
|
|
; RV32ZVE32F-NEXT: lw s1, 8(sp) # 4-byte Folded Reload
|
|
; RV32ZVE32F-NEXT: lw s2, 4(sp) # 4-byte Folded Reload
|
|
; RV32ZVE32F-NEXT: .cfi_restore s0
|
|
; RV32ZVE32F-NEXT: .cfi_restore s1
|
|
; RV32ZVE32F-NEXT: .cfi_restore s2
|
|
; RV32ZVE32F-NEXT: addi sp, sp, 16
|
|
; RV32ZVE32F-NEXT: .cfi_def_cfa_offset 0
|
|
; RV32ZVE32F-NEXT: ret
|
|
; RV32ZVE32F-NEXT: .LBB47_10: # %cond.store
|
|
; RV32ZVE32F-NEXT: .cfi_restore_state
|
|
; RV32ZVE32F-NEXT: lw s1, 0(a0)
|
|
; RV32ZVE32F-NEXT: lw a0, 4(a0)
|
|
; RV32ZVE32F-NEXT: vsetvli zero, zero, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vmv.x.s s2, v10
|
|
; RV32ZVE32F-NEXT: sw s1, 0(s2)
|
|
; RV32ZVE32F-NEXT: sw a0, 4(s2)
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 2
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB47_2
|
|
; RV32ZVE32F-NEXT: .LBB47_11: # %cond.store1
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v8, v10, 1
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v8
|
|
; RV32ZVE32F-NEXT: sw t6, 0(a0)
|
|
; RV32ZVE32F-NEXT: sw s0, 4(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 4
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB47_3
|
|
; RV32ZVE32F-NEXT: .LBB47_12: # %cond.store3
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v8, v10, 2
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v8
|
|
; RV32ZVE32F-NEXT: sw t4, 0(a0)
|
|
; RV32ZVE32F-NEXT: sw t5, 4(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 8
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB47_4
|
|
; RV32ZVE32F-NEXT: .LBB47_13: # %cond.store5
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v8, v10, 3
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v8
|
|
; RV32ZVE32F-NEXT: sw t2, 0(a0)
|
|
; RV32ZVE32F-NEXT: sw t3, 4(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 16
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB47_5
|
|
; RV32ZVE32F-NEXT: .LBB47_14: # %cond.store7
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v8, v10, 4
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v8
|
|
; RV32ZVE32F-NEXT: sw t0, 0(a0)
|
|
; RV32ZVE32F-NEXT: sw t1, 4(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 32
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB47_6
|
|
; RV32ZVE32F-NEXT: .LBB47_15: # %cond.store9
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v8, v10, 5
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v8
|
|
; RV32ZVE32F-NEXT: sw a6, 0(a0)
|
|
; RV32ZVE32F-NEXT: sw a7, 4(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 64
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB47_7
|
|
; RV32ZVE32F-NEXT: .LBB47_16: # %cond.store11
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v8, v10, 6
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v8
|
|
; RV32ZVE32F-NEXT: sw a4, 0(a0)
|
|
; RV32ZVE32F-NEXT: sw a5, 4(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, a1, -128
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB47_8
|
|
; RV32ZVE32F-NEXT: j .LBB47_9
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_baseidx_zext_v8i16_v8i64:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: ld a5, 40(a0)
|
|
; RV64ZVE32F-NEXT: ld a3, 48(a0)
|
|
; RV64ZVE32F-NEXT: ld a2, 56(a0)
|
|
; RV64ZVE32F-NEXT: ld t1, 8(a0)
|
|
; RV64ZVE32F-NEXT: ld t0, 16(a0)
|
|
; RV64ZVE32F-NEXT: ld a7, 24(a0)
|
|
; RV64ZVE32F-NEXT: ld a6, 32(a0)
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a4, v0
|
|
; RV64ZVE32F-NEXT: andi t2, a4, 1
|
|
; RV64ZVE32F-NEXT: beqz t2, .LBB47_2
|
|
; RV64ZVE32F-NEXT: # %bb.1: # %cond.store
|
|
; RV64ZVE32F-NEXT: ld a0, 0(a0)
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e16, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s t2, v8
|
|
; RV64ZVE32F-NEXT: slli t2, t2, 48
|
|
; RV64ZVE32F-NEXT: srli t2, t2, 45
|
|
; RV64ZVE32F-NEXT: add t2, a1, t2
|
|
; RV64ZVE32F-NEXT: sd a0, 0(t2)
|
|
; RV64ZVE32F-NEXT: .LBB47_2: # %else
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 2
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB47_4
|
|
; RV64ZVE32F-NEXT: # %bb.3: # %cond.store1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a0, v9
|
|
; RV64ZVE32F-NEXT: slli a0, a0, 48
|
|
; RV64ZVE32F-NEXT: srli a0, a0, 45
|
|
; RV64ZVE32F-NEXT: add a0, a1, a0
|
|
; RV64ZVE32F-NEXT: sd t1, 0(a0)
|
|
; RV64ZVE32F-NEXT: .LBB47_4: # %else2
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 4
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 2
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB47_6
|
|
; RV64ZVE32F-NEXT: # %bb.5: # %cond.store3
|
|
; RV64ZVE32F-NEXT: vmv.x.s a0, v9
|
|
; RV64ZVE32F-NEXT: slli a0, a0, 48
|
|
; RV64ZVE32F-NEXT: srli a0, a0, 45
|
|
; RV64ZVE32F-NEXT: add a0, a1, a0
|
|
; RV64ZVE32F-NEXT: sd t0, 0(a0)
|
|
; RV64ZVE32F-NEXT: .LBB47_6: # %else4
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 8
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB47_8
|
|
; RV64ZVE32F-NEXT: # %bb.7: # %cond.store5
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v9, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a0, v9
|
|
; RV64ZVE32F-NEXT: slli a0, a0, 48
|
|
; RV64ZVE32F-NEXT: srli a0, a0, 45
|
|
; RV64ZVE32F-NEXT: add a0, a1, a0
|
|
; RV64ZVE32F-NEXT: sd a7, 0(a0)
|
|
; RV64ZVE32F-NEXT: .LBB47_8: # %else6
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 16
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 4, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 4
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB47_10
|
|
; RV64ZVE32F-NEXT: # %bb.9: # %cond.store7
|
|
; RV64ZVE32F-NEXT: vmv.x.s a0, v8
|
|
; RV64ZVE32F-NEXT: slli a0, a0, 48
|
|
; RV64ZVE32F-NEXT: srli a0, a0, 45
|
|
; RV64ZVE32F-NEXT: add a0, a1, a0
|
|
; RV64ZVE32F-NEXT: sd a6, 0(a0)
|
|
; RV64ZVE32F-NEXT: .LBB47_10: # %else8
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 32
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB47_12
|
|
; RV64ZVE32F-NEXT: # %bb.11: # %cond.store9
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a0, v9
|
|
; RV64ZVE32F-NEXT: slli a0, a0, 48
|
|
; RV64ZVE32F-NEXT: srli a0, a0, 45
|
|
; RV64ZVE32F-NEXT: add a0, a1, a0
|
|
; RV64ZVE32F-NEXT: sd a5, 0(a0)
|
|
; RV64ZVE32F-NEXT: .LBB47_12: # %else10
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 64
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 2
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB47_15
|
|
; RV64ZVE32F-NEXT: # %bb.13: # %else12
|
|
; RV64ZVE32F-NEXT: andi a0, a4, -128
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB47_16
|
|
; RV64ZVE32F-NEXT: .LBB47_14: # %else14
|
|
; RV64ZVE32F-NEXT: ret
|
|
; RV64ZVE32F-NEXT: .LBB47_15: # %cond.store11
|
|
; RV64ZVE32F-NEXT: vmv.x.s a0, v8
|
|
; RV64ZVE32F-NEXT: slli a0, a0, 48
|
|
; RV64ZVE32F-NEXT: srli a0, a0, 45
|
|
; RV64ZVE32F-NEXT: add a0, a1, a0
|
|
; RV64ZVE32F-NEXT: sd a3, 0(a0)
|
|
; RV64ZVE32F-NEXT: andi a0, a4, -128
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB47_14
|
|
; RV64ZVE32F-NEXT: .LBB47_16: # %cond.store13
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a0, v8
|
|
; RV64ZVE32F-NEXT: slli a0, a0, 48
|
|
; RV64ZVE32F-NEXT: srli a0, a0, 45
|
|
; RV64ZVE32F-NEXT: add a0, a1, a0
|
|
; RV64ZVE32F-NEXT: sd a2, 0(a0)
|
|
; RV64ZVE32F-NEXT: ret
|
|
%eidxs = zext <8 x i16> %idxs to <8 x i64>
|
|
%ptrs = getelementptr inbounds i64, ptr %base, <8 x i64> %eidxs
|
|
call void @llvm.masked.scatter.v8i64.v8p0(<8 x i64> %val, <8 x ptr> %ptrs, i32 8, <8 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_baseidx_v8i32_v8i64(<8 x i64> %val, ptr %base, <8 x i32> %idxs, <8 x i1> %m) {
|
|
; RV32V-LABEL: mscatter_baseidx_v8i32_v8i64:
|
|
; RV32V: # %bb.0:
|
|
; RV32V-NEXT: vsetivli zero, 8, e32, m2, ta, ma
|
|
; RV32V-NEXT: vsll.vi v12, v12, 3
|
|
; RV32V-NEXT: vsetvli zero, zero, e64, m4, ta, ma
|
|
; RV32V-NEXT: vsoxei32.v v8, (a0), v12, v0.t
|
|
; RV32V-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_baseidx_v8i32_v8i64:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: li a1, 8
|
|
; RV64V-NEXT: vsetivli zero, 8, e32, m2, ta, ma
|
|
; RV64V-NEXT: vwmulsu.vx v16, v12, a1
|
|
; RV64V-NEXT: vsetvli zero, zero, e64, m4, ta, ma
|
|
; RV64V-NEXT: vsoxei64.v v8, (a0), v16, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV32ZVE32F-LABEL: mscatter_baseidx_v8i32_v8i64:
|
|
; RV32ZVE32F: # %bb.0:
|
|
; RV32ZVE32F-NEXT: addi sp, sp, -16
|
|
; RV32ZVE32F-NEXT: .cfi_def_cfa_offset 16
|
|
; RV32ZVE32F-NEXT: sw s0, 12(sp) # 4-byte Folded Spill
|
|
; RV32ZVE32F-NEXT: sw s1, 8(sp) # 4-byte Folded Spill
|
|
; RV32ZVE32F-NEXT: sw s2, 4(sp) # 4-byte Folded Spill
|
|
; RV32ZVE32F-NEXT: .cfi_offset s0, -4
|
|
; RV32ZVE32F-NEXT: .cfi_offset s1, -8
|
|
; RV32ZVE32F-NEXT: .cfi_offset s2, -12
|
|
; RV32ZVE32F-NEXT: .cfi_remember_state
|
|
; RV32ZVE32F-NEXT: lw a2, 56(a0)
|
|
; RV32ZVE32F-NEXT: lw a3, 60(a0)
|
|
; RV32ZVE32F-NEXT: lw a6, 40(a0)
|
|
; RV32ZVE32F-NEXT: lw a7, 44(a0)
|
|
; RV32ZVE32F-NEXT: lw a4, 48(a0)
|
|
; RV32ZVE32F-NEXT: lw a5, 52(a0)
|
|
; RV32ZVE32F-NEXT: lw t3, 24(a0)
|
|
; RV32ZVE32F-NEXT: lw t4, 28(a0)
|
|
; RV32ZVE32F-NEXT: lw t1, 32(a0)
|
|
; RV32ZVE32F-NEXT: lw t2, 36(a0)
|
|
; RV32ZVE32F-NEXT: lw s0, 8(a0)
|
|
; RV32ZVE32F-NEXT: lw s1, 12(a0)
|
|
; RV32ZVE32F-NEXT: lw t5, 16(a0)
|
|
; RV32ZVE32F-NEXT: lw t6, 20(a0)
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 8, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vsll.vi v8, v8, 3
|
|
; RV32ZVE32F-NEXT: vsetvli zero, zero, e8, mf2, ta, ma
|
|
; RV32ZVE32F-NEXT: vmv.x.s t0, v0
|
|
; RV32ZVE32F-NEXT: andi s2, t0, 1
|
|
; RV32ZVE32F-NEXT: vsetvli zero, zero, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vadd.vx v8, v8, a1
|
|
; RV32ZVE32F-NEXT: bnez s2, .LBB48_10
|
|
; RV32ZVE32F-NEXT: # %bb.1: # %else
|
|
; RV32ZVE32F-NEXT: andi a0, t0, 2
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB48_11
|
|
; RV32ZVE32F-NEXT: .LBB48_2: # %else2
|
|
; RV32ZVE32F-NEXT: andi a0, t0, 4
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB48_12
|
|
; RV32ZVE32F-NEXT: .LBB48_3: # %else4
|
|
; RV32ZVE32F-NEXT: andi a0, t0, 8
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB48_13
|
|
; RV32ZVE32F-NEXT: .LBB48_4: # %else6
|
|
; RV32ZVE32F-NEXT: andi a0, t0, 16
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB48_14
|
|
; RV32ZVE32F-NEXT: .LBB48_5: # %else8
|
|
; RV32ZVE32F-NEXT: andi a0, t0, 32
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB48_15
|
|
; RV32ZVE32F-NEXT: .LBB48_6: # %else10
|
|
; RV32ZVE32F-NEXT: andi a0, t0, 64
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB48_16
|
|
; RV32ZVE32F-NEXT: .LBB48_7: # %else12
|
|
; RV32ZVE32F-NEXT: andi a0, t0, -128
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB48_9
|
|
; RV32ZVE32F-NEXT: .LBB48_8: # %cond.store13
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v8, v8, 7
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v8
|
|
; RV32ZVE32F-NEXT: sw a2, 0(a0)
|
|
; RV32ZVE32F-NEXT: sw a3, 4(a0)
|
|
; RV32ZVE32F-NEXT: .LBB48_9: # %else14
|
|
; RV32ZVE32F-NEXT: lw s0, 12(sp) # 4-byte Folded Reload
|
|
; RV32ZVE32F-NEXT: lw s1, 8(sp) # 4-byte Folded Reload
|
|
; RV32ZVE32F-NEXT: lw s2, 4(sp) # 4-byte Folded Reload
|
|
; RV32ZVE32F-NEXT: .cfi_restore s0
|
|
; RV32ZVE32F-NEXT: .cfi_restore s1
|
|
; RV32ZVE32F-NEXT: .cfi_restore s2
|
|
; RV32ZVE32F-NEXT: addi sp, sp, 16
|
|
; RV32ZVE32F-NEXT: .cfi_def_cfa_offset 0
|
|
; RV32ZVE32F-NEXT: ret
|
|
; RV32ZVE32F-NEXT: .LBB48_10: # %cond.store
|
|
; RV32ZVE32F-NEXT: .cfi_restore_state
|
|
; RV32ZVE32F-NEXT: lw a1, 0(a0)
|
|
; RV32ZVE32F-NEXT: lw a0, 4(a0)
|
|
; RV32ZVE32F-NEXT: vmv.x.s s2, v8
|
|
; RV32ZVE32F-NEXT: sw a1, 0(s2)
|
|
; RV32ZVE32F-NEXT: sw a0, 4(s2)
|
|
; RV32ZVE32F-NEXT: andi a0, t0, 2
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB48_2
|
|
; RV32ZVE32F-NEXT: .LBB48_11: # %cond.store1
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 1
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: sw s0, 0(a0)
|
|
; RV32ZVE32F-NEXT: sw s1, 4(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, t0, 4
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB48_3
|
|
; RV32ZVE32F-NEXT: .LBB48_12: # %cond.store3
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 2
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: sw t5, 0(a0)
|
|
; RV32ZVE32F-NEXT: sw t6, 4(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, t0, 8
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB48_4
|
|
; RV32ZVE32F-NEXT: .LBB48_13: # %cond.store5
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 3
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: sw t3, 0(a0)
|
|
; RV32ZVE32F-NEXT: sw t4, 4(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, t0, 16
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB48_5
|
|
; RV32ZVE32F-NEXT: .LBB48_14: # %cond.store7
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 4
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: sw t1, 0(a0)
|
|
; RV32ZVE32F-NEXT: sw t2, 4(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, t0, 32
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB48_6
|
|
; RV32ZVE32F-NEXT: .LBB48_15: # %cond.store9
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 5
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: sw a6, 0(a0)
|
|
; RV32ZVE32F-NEXT: sw a7, 4(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, t0, 64
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB48_7
|
|
; RV32ZVE32F-NEXT: .LBB48_16: # %cond.store11
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 6
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: sw a4, 0(a0)
|
|
; RV32ZVE32F-NEXT: sw a5, 4(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, t0, -128
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB48_8
|
|
; RV32ZVE32F-NEXT: j .LBB48_9
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_baseidx_v8i32_v8i64:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: ld a5, 40(a0)
|
|
; RV64ZVE32F-NEXT: ld a3, 48(a0)
|
|
; RV64ZVE32F-NEXT: ld a2, 56(a0)
|
|
; RV64ZVE32F-NEXT: ld t1, 8(a0)
|
|
; RV64ZVE32F-NEXT: ld t0, 16(a0)
|
|
; RV64ZVE32F-NEXT: ld a7, 24(a0)
|
|
; RV64ZVE32F-NEXT: ld a6, 32(a0)
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a4, v0
|
|
; RV64ZVE32F-NEXT: andi t2, a4, 1
|
|
; RV64ZVE32F-NEXT: beqz t2, .LBB48_2
|
|
; RV64ZVE32F-NEXT: # %bb.1: # %cond.store
|
|
; RV64ZVE32F-NEXT: ld a0, 0(a0)
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e32, m4, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s t2, v8
|
|
; RV64ZVE32F-NEXT: slli t2, t2, 3
|
|
; RV64ZVE32F-NEXT: add t2, a1, t2
|
|
; RV64ZVE32F-NEXT: sd a0, 0(t2)
|
|
; RV64ZVE32F-NEXT: .LBB48_2: # %else
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 2
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB48_4
|
|
; RV64ZVE32F-NEXT: # %bb.3: # %cond.store1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v8, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV64ZVE32F-NEXT: slli a0, a0, 3
|
|
; RV64ZVE32F-NEXT: add a0, a1, a0
|
|
; RV64ZVE32F-NEXT: sd t1, 0(a0)
|
|
; RV64ZVE32F-NEXT: .LBB48_4: # %else2
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 4
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v8, 2
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB48_6
|
|
; RV64ZVE32F-NEXT: # %bb.5: # %cond.store3
|
|
; RV64ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV64ZVE32F-NEXT: slli a0, a0, 3
|
|
; RV64ZVE32F-NEXT: add a0, a1, a0
|
|
; RV64ZVE32F-NEXT: sd t0, 0(a0)
|
|
; RV64ZVE32F-NEXT: .LBB48_6: # %else4
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 8
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 4, e32, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 4
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB48_13
|
|
; RV64ZVE32F-NEXT: # %bb.7: # %else6
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 16
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB48_14
|
|
; RV64ZVE32F-NEXT: .LBB48_8: # %else8
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 32
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB48_10
|
|
; RV64ZVE32F-NEXT: .LBB48_9: # %cond.store9
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a0, v9
|
|
; RV64ZVE32F-NEXT: slli a0, a0, 3
|
|
; RV64ZVE32F-NEXT: add a0, a1, a0
|
|
; RV64ZVE32F-NEXT: sd a5, 0(a0)
|
|
; RV64ZVE32F-NEXT: .LBB48_10: # %else10
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 64
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 2
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB48_15
|
|
; RV64ZVE32F-NEXT: # %bb.11: # %else12
|
|
; RV64ZVE32F-NEXT: andi a0, a4, -128
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB48_16
|
|
; RV64ZVE32F-NEXT: .LBB48_12: # %else14
|
|
; RV64ZVE32F-NEXT: ret
|
|
; RV64ZVE32F-NEXT: .LBB48_13: # %cond.store5
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v10, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a0, v9
|
|
; RV64ZVE32F-NEXT: slli a0, a0, 3
|
|
; RV64ZVE32F-NEXT: add a0, a1, a0
|
|
; RV64ZVE32F-NEXT: sd a7, 0(a0)
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 16
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB48_8
|
|
; RV64ZVE32F-NEXT: .LBB48_14: # %cond.store7
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a0, v8
|
|
; RV64ZVE32F-NEXT: slli a0, a0, 3
|
|
; RV64ZVE32F-NEXT: add a0, a1, a0
|
|
; RV64ZVE32F-NEXT: sd a6, 0(a0)
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 32
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB48_9
|
|
; RV64ZVE32F-NEXT: j .LBB48_10
|
|
; RV64ZVE32F-NEXT: .LBB48_15: # %cond.store11
|
|
; RV64ZVE32F-NEXT: vmv.x.s a0, v8
|
|
; RV64ZVE32F-NEXT: slli a0, a0, 3
|
|
; RV64ZVE32F-NEXT: add a0, a1, a0
|
|
; RV64ZVE32F-NEXT: sd a3, 0(a0)
|
|
; RV64ZVE32F-NEXT: andi a0, a4, -128
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB48_12
|
|
; RV64ZVE32F-NEXT: .LBB48_16: # %cond.store13
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a0, v8
|
|
; RV64ZVE32F-NEXT: slli a0, a0, 3
|
|
; RV64ZVE32F-NEXT: add a0, a1, a0
|
|
; RV64ZVE32F-NEXT: sd a2, 0(a0)
|
|
; RV64ZVE32F-NEXT: ret
|
|
%ptrs = getelementptr inbounds i64, ptr %base, <8 x i32> %idxs
|
|
call void @llvm.masked.scatter.v8i64.v8p0(<8 x i64> %val, <8 x ptr> %ptrs, i32 8, <8 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_baseidx_sext_v8i32_v8i64(<8 x i64> %val, ptr %base, <8 x i32> %idxs, <8 x i1> %m) {
|
|
; RV32V-LABEL: mscatter_baseidx_sext_v8i32_v8i64:
|
|
; RV32V: # %bb.0:
|
|
; RV32V-NEXT: vsetivli zero, 8, e32, m2, ta, ma
|
|
; RV32V-NEXT: vsll.vi v12, v12, 3
|
|
; RV32V-NEXT: vsetvli zero, zero, e64, m4, ta, ma
|
|
; RV32V-NEXT: vsoxei32.v v8, (a0), v12, v0.t
|
|
; RV32V-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_baseidx_sext_v8i32_v8i64:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: li a1, 8
|
|
; RV64V-NEXT: vsetivli zero, 8, e32, m2, ta, ma
|
|
; RV64V-NEXT: vwmulsu.vx v16, v12, a1
|
|
; RV64V-NEXT: vsetvli zero, zero, e64, m4, ta, ma
|
|
; RV64V-NEXT: vsoxei64.v v8, (a0), v16, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV32ZVE32F-LABEL: mscatter_baseidx_sext_v8i32_v8i64:
|
|
; RV32ZVE32F: # %bb.0:
|
|
; RV32ZVE32F-NEXT: addi sp, sp, -16
|
|
; RV32ZVE32F-NEXT: .cfi_def_cfa_offset 16
|
|
; RV32ZVE32F-NEXT: sw s0, 12(sp) # 4-byte Folded Spill
|
|
; RV32ZVE32F-NEXT: sw s1, 8(sp) # 4-byte Folded Spill
|
|
; RV32ZVE32F-NEXT: sw s2, 4(sp) # 4-byte Folded Spill
|
|
; RV32ZVE32F-NEXT: .cfi_offset s0, -4
|
|
; RV32ZVE32F-NEXT: .cfi_offset s1, -8
|
|
; RV32ZVE32F-NEXT: .cfi_offset s2, -12
|
|
; RV32ZVE32F-NEXT: .cfi_remember_state
|
|
; RV32ZVE32F-NEXT: lw a2, 56(a0)
|
|
; RV32ZVE32F-NEXT: lw a3, 60(a0)
|
|
; RV32ZVE32F-NEXT: lw a6, 40(a0)
|
|
; RV32ZVE32F-NEXT: lw a7, 44(a0)
|
|
; RV32ZVE32F-NEXT: lw a4, 48(a0)
|
|
; RV32ZVE32F-NEXT: lw a5, 52(a0)
|
|
; RV32ZVE32F-NEXT: lw t3, 24(a0)
|
|
; RV32ZVE32F-NEXT: lw t4, 28(a0)
|
|
; RV32ZVE32F-NEXT: lw t1, 32(a0)
|
|
; RV32ZVE32F-NEXT: lw t2, 36(a0)
|
|
; RV32ZVE32F-NEXT: lw s0, 8(a0)
|
|
; RV32ZVE32F-NEXT: lw s1, 12(a0)
|
|
; RV32ZVE32F-NEXT: lw t5, 16(a0)
|
|
; RV32ZVE32F-NEXT: lw t6, 20(a0)
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 8, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vsll.vi v8, v8, 3
|
|
; RV32ZVE32F-NEXT: vsetvli zero, zero, e8, mf2, ta, ma
|
|
; RV32ZVE32F-NEXT: vmv.x.s t0, v0
|
|
; RV32ZVE32F-NEXT: andi s2, t0, 1
|
|
; RV32ZVE32F-NEXT: vsetvli zero, zero, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vadd.vx v8, v8, a1
|
|
; RV32ZVE32F-NEXT: bnez s2, .LBB49_10
|
|
; RV32ZVE32F-NEXT: # %bb.1: # %else
|
|
; RV32ZVE32F-NEXT: andi a0, t0, 2
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB49_11
|
|
; RV32ZVE32F-NEXT: .LBB49_2: # %else2
|
|
; RV32ZVE32F-NEXT: andi a0, t0, 4
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB49_12
|
|
; RV32ZVE32F-NEXT: .LBB49_3: # %else4
|
|
; RV32ZVE32F-NEXT: andi a0, t0, 8
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB49_13
|
|
; RV32ZVE32F-NEXT: .LBB49_4: # %else6
|
|
; RV32ZVE32F-NEXT: andi a0, t0, 16
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB49_14
|
|
; RV32ZVE32F-NEXT: .LBB49_5: # %else8
|
|
; RV32ZVE32F-NEXT: andi a0, t0, 32
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB49_15
|
|
; RV32ZVE32F-NEXT: .LBB49_6: # %else10
|
|
; RV32ZVE32F-NEXT: andi a0, t0, 64
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB49_16
|
|
; RV32ZVE32F-NEXT: .LBB49_7: # %else12
|
|
; RV32ZVE32F-NEXT: andi a0, t0, -128
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB49_9
|
|
; RV32ZVE32F-NEXT: .LBB49_8: # %cond.store13
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v8, v8, 7
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v8
|
|
; RV32ZVE32F-NEXT: sw a2, 0(a0)
|
|
; RV32ZVE32F-NEXT: sw a3, 4(a0)
|
|
; RV32ZVE32F-NEXT: .LBB49_9: # %else14
|
|
; RV32ZVE32F-NEXT: lw s0, 12(sp) # 4-byte Folded Reload
|
|
; RV32ZVE32F-NEXT: lw s1, 8(sp) # 4-byte Folded Reload
|
|
; RV32ZVE32F-NEXT: lw s2, 4(sp) # 4-byte Folded Reload
|
|
; RV32ZVE32F-NEXT: .cfi_restore s0
|
|
; RV32ZVE32F-NEXT: .cfi_restore s1
|
|
; RV32ZVE32F-NEXT: .cfi_restore s2
|
|
; RV32ZVE32F-NEXT: addi sp, sp, 16
|
|
; RV32ZVE32F-NEXT: .cfi_def_cfa_offset 0
|
|
; RV32ZVE32F-NEXT: ret
|
|
; RV32ZVE32F-NEXT: .LBB49_10: # %cond.store
|
|
; RV32ZVE32F-NEXT: .cfi_restore_state
|
|
; RV32ZVE32F-NEXT: lw a1, 0(a0)
|
|
; RV32ZVE32F-NEXT: lw a0, 4(a0)
|
|
; RV32ZVE32F-NEXT: vmv.x.s s2, v8
|
|
; RV32ZVE32F-NEXT: sw a1, 0(s2)
|
|
; RV32ZVE32F-NEXT: sw a0, 4(s2)
|
|
; RV32ZVE32F-NEXT: andi a0, t0, 2
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB49_2
|
|
; RV32ZVE32F-NEXT: .LBB49_11: # %cond.store1
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 1
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: sw s0, 0(a0)
|
|
; RV32ZVE32F-NEXT: sw s1, 4(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, t0, 4
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB49_3
|
|
; RV32ZVE32F-NEXT: .LBB49_12: # %cond.store3
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 2
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: sw t5, 0(a0)
|
|
; RV32ZVE32F-NEXT: sw t6, 4(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, t0, 8
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB49_4
|
|
; RV32ZVE32F-NEXT: .LBB49_13: # %cond.store5
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 3
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: sw t3, 0(a0)
|
|
; RV32ZVE32F-NEXT: sw t4, 4(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, t0, 16
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB49_5
|
|
; RV32ZVE32F-NEXT: .LBB49_14: # %cond.store7
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 4
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: sw t1, 0(a0)
|
|
; RV32ZVE32F-NEXT: sw t2, 4(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, t0, 32
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB49_6
|
|
; RV32ZVE32F-NEXT: .LBB49_15: # %cond.store9
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 5
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: sw a6, 0(a0)
|
|
; RV32ZVE32F-NEXT: sw a7, 4(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, t0, 64
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB49_7
|
|
; RV32ZVE32F-NEXT: .LBB49_16: # %cond.store11
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 6
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: sw a4, 0(a0)
|
|
; RV32ZVE32F-NEXT: sw a5, 4(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, t0, -128
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB49_8
|
|
; RV32ZVE32F-NEXT: j .LBB49_9
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_baseidx_sext_v8i32_v8i64:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: ld a5, 40(a0)
|
|
; RV64ZVE32F-NEXT: ld a3, 48(a0)
|
|
; RV64ZVE32F-NEXT: ld a2, 56(a0)
|
|
; RV64ZVE32F-NEXT: ld t1, 8(a0)
|
|
; RV64ZVE32F-NEXT: ld t0, 16(a0)
|
|
; RV64ZVE32F-NEXT: ld a7, 24(a0)
|
|
; RV64ZVE32F-NEXT: ld a6, 32(a0)
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a4, v0
|
|
; RV64ZVE32F-NEXT: andi t2, a4, 1
|
|
; RV64ZVE32F-NEXT: beqz t2, .LBB49_2
|
|
; RV64ZVE32F-NEXT: # %bb.1: # %cond.store
|
|
; RV64ZVE32F-NEXT: ld a0, 0(a0)
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e32, m4, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s t2, v8
|
|
; RV64ZVE32F-NEXT: slli t2, t2, 3
|
|
; RV64ZVE32F-NEXT: add t2, a1, t2
|
|
; RV64ZVE32F-NEXT: sd a0, 0(t2)
|
|
; RV64ZVE32F-NEXT: .LBB49_2: # %else
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 2
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB49_4
|
|
; RV64ZVE32F-NEXT: # %bb.3: # %cond.store1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v8, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV64ZVE32F-NEXT: slli a0, a0, 3
|
|
; RV64ZVE32F-NEXT: add a0, a1, a0
|
|
; RV64ZVE32F-NEXT: sd t1, 0(a0)
|
|
; RV64ZVE32F-NEXT: .LBB49_4: # %else2
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 4
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v8, 2
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB49_6
|
|
; RV64ZVE32F-NEXT: # %bb.5: # %cond.store3
|
|
; RV64ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV64ZVE32F-NEXT: slli a0, a0, 3
|
|
; RV64ZVE32F-NEXT: add a0, a1, a0
|
|
; RV64ZVE32F-NEXT: sd t0, 0(a0)
|
|
; RV64ZVE32F-NEXT: .LBB49_6: # %else4
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 8
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 4, e32, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 4
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB49_13
|
|
; RV64ZVE32F-NEXT: # %bb.7: # %else6
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 16
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB49_14
|
|
; RV64ZVE32F-NEXT: .LBB49_8: # %else8
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 32
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB49_10
|
|
; RV64ZVE32F-NEXT: .LBB49_9: # %cond.store9
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a0, v9
|
|
; RV64ZVE32F-NEXT: slli a0, a0, 3
|
|
; RV64ZVE32F-NEXT: add a0, a1, a0
|
|
; RV64ZVE32F-NEXT: sd a5, 0(a0)
|
|
; RV64ZVE32F-NEXT: .LBB49_10: # %else10
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 64
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 2
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB49_15
|
|
; RV64ZVE32F-NEXT: # %bb.11: # %else12
|
|
; RV64ZVE32F-NEXT: andi a0, a4, -128
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB49_16
|
|
; RV64ZVE32F-NEXT: .LBB49_12: # %else14
|
|
; RV64ZVE32F-NEXT: ret
|
|
; RV64ZVE32F-NEXT: .LBB49_13: # %cond.store5
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v10, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a0, v9
|
|
; RV64ZVE32F-NEXT: slli a0, a0, 3
|
|
; RV64ZVE32F-NEXT: add a0, a1, a0
|
|
; RV64ZVE32F-NEXT: sd a7, 0(a0)
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 16
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB49_8
|
|
; RV64ZVE32F-NEXT: .LBB49_14: # %cond.store7
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a0, v8
|
|
; RV64ZVE32F-NEXT: slli a0, a0, 3
|
|
; RV64ZVE32F-NEXT: add a0, a1, a0
|
|
; RV64ZVE32F-NEXT: sd a6, 0(a0)
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 32
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB49_9
|
|
; RV64ZVE32F-NEXT: j .LBB49_10
|
|
; RV64ZVE32F-NEXT: .LBB49_15: # %cond.store11
|
|
; RV64ZVE32F-NEXT: vmv.x.s a0, v8
|
|
; RV64ZVE32F-NEXT: slli a0, a0, 3
|
|
; RV64ZVE32F-NEXT: add a0, a1, a0
|
|
; RV64ZVE32F-NEXT: sd a3, 0(a0)
|
|
; RV64ZVE32F-NEXT: andi a0, a4, -128
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB49_12
|
|
; RV64ZVE32F-NEXT: .LBB49_16: # %cond.store13
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a0, v8
|
|
; RV64ZVE32F-NEXT: slli a0, a0, 3
|
|
; RV64ZVE32F-NEXT: add a0, a1, a0
|
|
; RV64ZVE32F-NEXT: sd a2, 0(a0)
|
|
; RV64ZVE32F-NEXT: ret
|
|
%eidxs = sext <8 x i32> %idxs to <8 x i64>
|
|
%ptrs = getelementptr inbounds i64, ptr %base, <8 x i64> %eidxs
|
|
call void @llvm.masked.scatter.v8i64.v8p0(<8 x i64> %val, <8 x ptr> %ptrs, i32 8, <8 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_baseidx_zext_v8i32_v8i64(<8 x i64> %val, ptr %base, <8 x i32> %idxs, <8 x i1> %m) {
|
|
; RV32V-LABEL: mscatter_baseidx_zext_v8i32_v8i64:
|
|
; RV32V: # %bb.0:
|
|
; RV32V-NEXT: vsetivli zero, 8, e32, m2, ta, ma
|
|
; RV32V-NEXT: vsll.vi v12, v12, 3
|
|
; RV32V-NEXT: vsetvli zero, zero, e64, m4, ta, ma
|
|
; RV32V-NEXT: vsoxei32.v v8, (a0), v12, v0.t
|
|
; RV32V-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_baseidx_zext_v8i32_v8i64:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: li a1, 8
|
|
; RV64V-NEXT: vsetivli zero, 8, e32, m2, ta, ma
|
|
; RV64V-NEXT: vwmulu.vx v16, v12, a1
|
|
; RV64V-NEXT: vsetvli zero, zero, e64, m4, ta, ma
|
|
; RV64V-NEXT: vsoxei64.v v8, (a0), v16, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV32ZVE32F-LABEL: mscatter_baseidx_zext_v8i32_v8i64:
|
|
; RV32ZVE32F: # %bb.0:
|
|
; RV32ZVE32F-NEXT: addi sp, sp, -16
|
|
; RV32ZVE32F-NEXT: .cfi_def_cfa_offset 16
|
|
; RV32ZVE32F-NEXT: sw s0, 12(sp) # 4-byte Folded Spill
|
|
; RV32ZVE32F-NEXT: sw s1, 8(sp) # 4-byte Folded Spill
|
|
; RV32ZVE32F-NEXT: sw s2, 4(sp) # 4-byte Folded Spill
|
|
; RV32ZVE32F-NEXT: .cfi_offset s0, -4
|
|
; RV32ZVE32F-NEXT: .cfi_offset s1, -8
|
|
; RV32ZVE32F-NEXT: .cfi_offset s2, -12
|
|
; RV32ZVE32F-NEXT: .cfi_remember_state
|
|
; RV32ZVE32F-NEXT: lw a2, 56(a0)
|
|
; RV32ZVE32F-NEXT: lw a3, 60(a0)
|
|
; RV32ZVE32F-NEXT: lw a6, 40(a0)
|
|
; RV32ZVE32F-NEXT: lw a7, 44(a0)
|
|
; RV32ZVE32F-NEXT: lw a4, 48(a0)
|
|
; RV32ZVE32F-NEXT: lw a5, 52(a0)
|
|
; RV32ZVE32F-NEXT: lw t3, 24(a0)
|
|
; RV32ZVE32F-NEXT: lw t4, 28(a0)
|
|
; RV32ZVE32F-NEXT: lw t1, 32(a0)
|
|
; RV32ZVE32F-NEXT: lw t2, 36(a0)
|
|
; RV32ZVE32F-NEXT: lw s0, 8(a0)
|
|
; RV32ZVE32F-NEXT: lw s1, 12(a0)
|
|
; RV32ZVE32F-NEXT: lw t5, 16(a0)
|
|
; RV32ZVE32F-NEXT: lw t6, 20(a0)
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 8, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vsll.vi v8, v8, 3
|
|
; RV32ZVE32F-NEXT: vsetvli zero, zero, e8, mf2, ta, ma
|
|
; RV32ZVE32F-NEXT: vmv.x.s t0, v0
|
|
; RV32ZVE32F-NEXT: andi s2, t0, 1
|
|
; RV32ZVE32F-NEXT: vsetvli zero, zero, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vadd.vx v8, v8, a1
|
|
; RV32ZVE32F-NEXT: bnez s2, .LBB50_10
|
|
; RV32ZVE32F-NEXT: # %bb.1: # %else
|
|
; RV32ZVE32F-NEXT: andi a0, t0, 2
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB50_11
|
|
; RV32ZVE32F-NEXT: .LBB50_2: # %else2
|
|
; RV32ZVE32F-NEXT: andi a0, t0, 4
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB50_12
|
|
; RV32ZVE32F-NEXT: .LBB50_3: # %else4
|
|
; RV32ZVE32F-NEXT: andi a0, t0, 8
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB50_13
|
|
; RV32ZVE32F-NEXT: .LBB50_4: # %else6
|
|
; RV32ZVE32F-NEXT: andi a0, t0, 16
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB50_14
|
|
; RV32ZVE32F-NEXT: .LBB50_5: # %else8
|
|
; RV32ZVE32F-NEXT: andi a0, t0, 32
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB50_15
|
|
; RV32ZVE32F-NEXT: .LBB50_6: # %else10
|
|
; RV32ZVE32F-NEXT: andi a0, t0, 64
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB50_16
|
|
; RV32ZVE32F-NEXT: .LBB50_7: # %else12
|
|
; RV32ZVE32F-NEXT: andi a0, t0, -128
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB50_9
|
|
; RV32ZVE32F-NEXT: .LBB50_8: # %cond.store13
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v8, v8, 7
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v8
|
|
; RV32ZVE32F-NEXT: sw a2, 0(a0)
|
|
; RV32ZVE32F-NEXT: sw a3, 4(a0)
|
|
; RV32ZVE32F-NEXT: .LBB50_9: # %else14
|
|
; RV32ZVE32F-NEXT: lw s0, 12(sp) # 4-byte Folded Reload
|
|
; RV32ZVE32F-NEXT: lw s1, 8(sp) # 4-byte Folded Reload
|
|
; RV32ZVE32F-NEXT: lw s2, 4(sp) # 4-byte Folded Reload
|
|
; RV32ZVE32F-NEXT: .cfi_restore s0
|
|
; RV32ZVE32F-NEXT: .cfi_restore s1
|
|
; RV32ZVE32F-NEXT: .cfi_restore s2
|
|
; RV32ZVE32F-NEXT: addi sp, sp, 16
|
|
; RV32ZVE32F-NEXT: .cfi_def_cfa_offset 0
|
|
; RV32ZVE32F-NEXT: ret
|
|
; RV32ZVE32F-NEXT: .LBB50_10: # %cond.store
|
|
; RV32ZVE32F-NEXT: .cfi_restore_state
|
|
; RV32ZVE32F-NEXT: lw a1, 0(a0)
|
|
; RV32ZVE32F-NEXT: lw a0, 4(a0)
|
|
; RV32ZVE32F-NEXT: vmv.x.s s2, v8
|
|
; RV32ZVE32F-NEXT: sw a1, 0(s2)
|
|
; RV32ZVE32F-NEXT: sw a0, 4(s2)
|
|
; RV32ZVE32F-NEXT: andi a0, t0, 2
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB50_2
|
|
; RV32ZVE32F-NEXT: .LBB50_11: # %cond.store1
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 1
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: sw s0, 0(a0)
|
|
; RV32ZVE32F-NEXT: sw s1, 4(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, t0, 4
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB50_3
|
|
; RV32ZVE32F-NEXT: .LBB50_12: # %cond.store3
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 2
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: sw t5, 0(a0)
|
|
; RV32ZVE32F-NEXT: sw t6, 4(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, t0, 8
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB50_4
|
|
; RV32ZVE32F-NEXT: .LBB50_13: # %cond.store5
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 3
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: sw t3, 0(a0)
|
|
; RV32ZVE32F-NEXT: sw t4, 4(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, t0, 16
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB50_5
|
|
; RV32ZVE32F-NEXT: .LBB50_14: # %cond.store7
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 4
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: sw t1, 0(a0)
|
|
; RV32ZVE32F-NEXT: sw t2, 4(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, t0, 32
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB50_6
|
|
; RV32ZVE32F-NEXT: .LBB50_15: # %cond.store9
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 5
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: sw a6, 0(a0)
|
|
; RV32ZVE32F-NEXT: sw a7, 4(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, t0, 64
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB50_7
|
|
; RV32ZVE32F-NEXT: .LBB50_16: # %cond.store11
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 6
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: sw a4, 0(a0)
|
|
; RV32ZVE32F-NEXT: sw a5, 4(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, t0, -128
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB50_8
|
|
; RV32ZVE32F-NEXT: j .LBB50_9
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_baseidx_zext_v8i32_v8i64:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: ld a5, 40(a0)
|
|
; RV64ZVE32F-NEXT: ld a3, 48(a0)
|
|
; RV64ZVE32F-NEXT: ld a2, 56(a0)
|
|
; RV64ZVE32F-NEXT: ld t1, 8(a0)
|
|
; RV64ZVE32F-NEXT: ld t0, 16(a0)
|
|
; RV64ZVE32F-NEXT: ld a7, 24(a0)
|
|
; RV64ZVE32F-NEXT: ld a6, 32(a0)
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a4, v0
|
|
; RV64ZVE32F-NEXT: andi t2, a4, 1
|
|
; RV64ZVE32F-NEXT: beqz t2, .LBB50_2
|
|
; RV64ZVE32F-NEXT: # %bb.1: # %cond.store
|
|
; RV64ZVE32F-NEXT: ld a0, 0(a0)
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e32, m4, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s t2, v8
|
|
; RV64ZVE32F-NEXT: slli t2, t2, 32
|
|
; RV64ZVE32F-NEXT: srli t2, t2, 29
|
|
; RV64ZVE32F-NEXT: add t2, a1, t2
|
|
; RV64ZVE32F-NEXT: sd a0, 0(t2)
|
|
; RV64ZVE32F-NEXT: .LBB50_2: # %else
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 2
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB50_4
|
|
; RV64ZVE32F-NEXT: # %bb.3: # %cond.store1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v8, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV64ZVE32F-NEXT: slli a0, a0, 32
|
|
; RV64ZVE32F-NEXT: srli a0, a0, 29
|
|
; RV64ZVE32F-NEXT: add a0, a1, a0
|
|
; RV64ZVE32F-NEXT: sd t1, 0(a0)
|
|
; RV64ZVE32F-NEXT: .LBB50_4: # %else2
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 4
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v8, 2
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB50_6
|
|
; RV64ZVE32F-NEXT: # %bb.5: # %cond.store3
|
|
; RV64ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV64ZVE32F-NEXT: slli a0, a0, 32
|
|
; RV64ZVE32F-NEXT: srli a0, a0, 29
|
|
; RV64ZVE32F-NEXT: add a0, a1, a0
|
|
; RV64ZVE32F-NEXT: sd t0, 0(a0)
|
|
; RV64ZVE32F-NEXT: .LBB50_6: # %else4
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 8
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 4, e32, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 4
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB50_13
|
|
; RV64ZVE32F-NEXT: # %bb.7: # %else6
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 16
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB50_14
|
|
; RV64ZVE32F-NEXT: .LBB50_8: # %else8
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 32
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB50_10
|
|
; RV64ZVE32F-NEXT: .LBB50_9: # %cond.store9
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a0, v9
|
|
; RV64ZVE32F-NEXT: slli a0, a0, 32
|
|
; RV64ZVE32F-NEXT: srli a0, a0, 29
|
|
; RV64ZVE32F-NEXT: add a0, a1, a0
|
|
; RV64ZVE32F-NEXT: sd a5, 0(a0)
|
|
; RV64ZVE32F-NEXT: .LBB50_10: # %else10
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 64
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 2
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB50_15
|
|
; RV64ZVE32F-NEXT: # %bb.11: # %else12
|
|
; RV64ZVE32F-NEXT: andi a0, a4, -128
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB50_16
|
|
; RV64ZVE32F-NEXT: .LBB50_12: # %else14
|
|
; RV64ZVE32F-NEXT: ret
|
|
; RV64ZVE32F-NEXT: .LBB50_13: # %cond.store5
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v10, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a0, v9
|
|
; RV64ZVE32F-NEXT: slli a0, a0, 32
|
|
; RV64ZVE32F-NEXT: srli a0, a0, 29
|
|
; RV64ZVE32F-NEXT: add a0, a1, a0
|
|
; RV64ZVE32F-NEXT: sd a7, 0(a0)
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 16
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB50_8
|
|
; RV64ZVE32F-NEXT: .LBB50_14: # %cond.store7
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a0, v8
|
|
; RV64ZVE32F-NEXT: slli a0, a0, 32
|
|
; RV64ZVE32F-NEXT: srli a0, a0, 29
|
|
; RV64ZVE32F-NEXT: add a0, a1, a0
|
|
; RV64ZVE32F-NEXT: sd a6, 0(a0)
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 32
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB50_9
|
|
; RV64ZVE32F-NEXT: j .LBB50_10
|
|
; RV64ZVE32F-NEXT: .LBB50_15: # %cond.store11
|
|
; RV64ZVE32F-NEXT: vmv.x.s a0, v8
|
|
; RV64ZVE32F-NEXT: slli a0, a0, 32
|
|
; RV64ZVE32F-NEXT: srli a0, a0, 29
|
|
; RV64ZVE32F-NEXT: add a0, a1, a0
|
|
; RV64ZVE32F-NEXT: sd a3, 0(a0)
|
|
; RV64ZVE32F-NEXT: andi a0, a4, -128
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB50_12
|
|
; RV64ZVE32F-NEXT: .LBB50_16: # %cond.store13
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a0, v8
|
|
; RV64ZVE32F-NEXT: slli a0, a0, 32
|
|
; RV64ZVE32F-NEXT: srli a0, a0, 29
|
|
; RV64ZVE32F-NEXT: add a0, a1, a0
|
|
; RV64ZVE32F-NEXT: sd a2, 0(a0)
|
|
; RV64ZVE32F-NEXT: ret
|
|
%eidxs = zext <8 x i32> %idxs to <8 x i64>
|
|
%ptrs = getelementptr inbounds i64, ptr %base, <8 x i64> %eidxs
|
|
call void @llvm.masked.scatter.v8i64.v8p0(<8 x i64> %val, <8 x ptr> %ptrs, i32 8, <8 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_baseidx_v8i64(<8 x i64> %val, ptr %base, <8 x i64> %idxs, <8 x i1> %m) {
|
|
; RV32V-LABEL: mscatter_baseidx_v8i64:
|
|
; RV32V: # %bb.0:
|
|
; RV32V-NEXT: vsetivli zero, 8, e32, m2, ta, ma
|
|
; RV32V-NEXT: vnsrl.wi v16, v12, 0
|
|
; RV32V-NEXT: vsll.vi v12, v16, 3
|
|
; RV32V-NEXT: vsetvli zero, zero, e64, m4, ta, ma
|
|
; RV32V-NEXT: vsoxei32.v v8, (a0), v12, v0.t
|
|
; RV32V-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_baseidx_v8i64:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: vsetivli zero, 8, e64, m4, ta, ma
|
|
; RV64V-NEXT: vsll.vi v12, v12, 3
|
|
; RV64V-NEXT: vsoxei64.v v8, (a0), v12, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV32ZVE32F-LABEL: mscatter_baseidx_v8i64:
|
|
; RV32ZVE32F: # %bb.0:
|
|
; RV32ZVE32F-NEXT: addi sp, sp, -48
|
|
; RV32ZVE32F-NEXT: .cfi_def_cfa_offset 48
|
|
; RV32ZVE32F-NEXT: sw s0, 44(sp) # 4-byte Folded Spill
|
|
; RV32ZVE32F-NEXT: sw s1, 40(sp) # 4-byte Folded Spill
|
|
; RV32ZVE32F-NEXT: sw s2, 36(sp) # 4-byte Folded Spill
|
|
; RV32ZVE32F-NEXT: sw s3, 32(sp) # 4-byte Folded Spill
|
|
; RV32ZVE32F-NEXT: sw s4, 28(sp) # 4-byte Folded Spill
|
|
; RV32ZVE32F-NEXT: sw s5, 24(sp) # 4-byte Folded Spill
|
|
; RV32ZVE32F-NEXT: sw s6, 20(sp) # 4-byte Folded Spill
|
|
; RV32ZVE32F-NEXT: sw s7, 16(sp) # 4-byte Folded Spill
|
|
; RV32ZVE32F-NEXT: sw s8, 12(sp) # 4-byte Folded Spill
|
|
; RV32ZVE32F-NEXT: sw s9, 8(sp) # 4-byte Folded Spill
|
|
; RV32ZVE32F-NEXT: .cfi_offset s0, -4
|
|
; RV32ZVE32F-NEXT: .cfi_offset s1, -8
|
|
; RV32ZVE32F-NEXT: .cfi_offset s2, -12
|
|
; RV32ZVE32F-NEXT: .cfi_offset s3, -16
|
|
; RV32ZVE32F-NEXT: .cfi_offset s4, -20
|
|
; RV32ZVE32F-NEXT: .cfi_offset s5, -24
|
|
; RV32ZVE32F-NEXT: .cfi_offset s6, -28
|
|
; RV32ZVE32F-NEXT: .cfi_offset s7, -32
|
|
; RV32ZVE32F-NEXT: .cfi_offset s8, -36
|
|
; RV32ZVE32F-NEXT: .cfi_offset s9, -40
|
|
; RV32ZVE32F-NEXT: .cfi_remember_state
|
|
; RV32ZVE32F-NEXT: lw a3, 56(a0)
|
|
; RV32ZVE32F-NEXT: lw a4, 60(a0)
|
|
; RV32ZVE32F-NEXT: lw a7, 40(a0)
|
|
; RV32ZVE32F-NEXT: lw t0, 44(a0)
|
|
; RV32ZVE32F-NEXT: lw a5, 48(a0)
|
|
; RV32ZVE32F-NEXT: lw a6, 52(a0)
|
|
; RV32ZVE32F-NEXT: lw t3, 24(a0)
|
|
; RV32ZVE32F-NEXT: lw t4, 28(a0)
|
|
; RV32ZVE32F-NEXT: lw t1, 32(a0)
|
|
; RV32ZVE32F-NEXT: lw t2, 36(a0)
|
|
; RV32ZVE32F-NEXT: lw s0, 8(a0)
|
|
; RV32ZVE32F-NEXT: lw s1, 12(a0)
|
|
; RV32ZVE32F-NEXT: lw t5, 16(a0)
|
|
; RV32ZVE32F-NEXT: lw t6, 20(a0)
|
|
; RV32ZVE32F-NEXT: lw s2, 32(a2)
|
|
; RV32ZVE32F-NEXT: lw s3, 40(a2)
|
|
; RV32ZVE32F-NEXT: lw s4, 48(a2)
|
|
; RV32ZVE32F-NEXT: lw s5, 56(a2)
|
|
; RV32ZVE32F-NEXT: lw s6, 0(a2)
|
|
; RV32ZVE32F-NEXT: lw s7, 8(a2)
|
|
; RV32ZVE32F-NEXT: lw s8, 16(a2)
|
|
; RV32ZVE32F-NEXT: lw s9, 24(a2)
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 8, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vmv.v.x v8, s6
|
|
; RV32ZVE32F-NEXT: vsetvli zero, zero, e8, mf2, ta, ma
|
|
; RV32ZVE32F-NEXT: vmv.x.s a2, v0
|
|
; RV32ZVE32F-NEXT: vsetvli zero, zero, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslide1down.vx v8, v8, s7
|
|
; RV32ZVE32F-NEXT: vslide1down.vx v8, v8, s8
|
|
; RV32ZVE32F-NEXT: vslide1down.vx v8, v8, s9
|
|
; RV32ZVE32F-NEXT: vslide1down.vx v8, v8, s2
|
|
; RV32ZVE32F-NEXT: vslide1down.vx v8, v8, s3
|
|
; RV32ZVE32F-NEXT: vslide1down.vx v8, v8, s4
|
|
; RV32ZVE32F-NEXT: vslide1down.vx v8, v8, s5
|
|
; RV32ZVE32F-NEXT: vsll.vi v8, v8, 3
|
|
; RV32ZVE32F-NEXT: andi s2, a2, 1
|
|
; RV32ZVE32F-NEXT: vadd.vx v8, v8, a1
|
|
; RV32ZVE32F-NEXT: bnez s2, .LBB51_10
|
|
; RV32ZVE32F-NEXT: # %bb.1: # %else
|
|
; RV32ZVE32F-NEXT: andi a0, a2, 2
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB51_11
|
|
; RV32ZVE32F-NEXT: .LBB51_2: # %else2
|
|
; RV32ZVE32F-NEXT: andi a0, a2, 4
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB51_12
|
|
; RV32ZVE32F-NEXT: .LBB51_3: # %else4
|
|
; RV32ZVE32F-NEXT: andi a0, a2, 8
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB51_13
|
|
; RV32ZVE32F-NEXT: .LBB51_4: # %else6
|
|
; RV32ZVE32F-NEXT: andi a0, a2, 16
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB51_14
|
|
; RV32ZVE32F-NEXT: .LBB51_5: # %else8
|
|
; RV32ZVE32F-NEXT: andi a0, a2, 32
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB51_15
|
|
; RV32ZVE32F-NEXT: .LBB51_6: # %else10
|
|
; RV32ZVE32F-NEXT: andi a0, a2, 64
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB51_16
|
|
; RV32ZVE32F-NEXT: .LBB51_7: # %else12
|
|
; RV32ZVE32F-NEXT: andi a0, a2, -128
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB51_9
|
|
; RV32ZVE32F-NEXT: .LBB51_8: # %cond.store13
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v8, v8, 7
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v8
|
|
; RV32ZVE32F-NEXT: sw a3, 0(a0)
|
|
; RV32ZVE32F-NEXT: sw a4, 4(a0)
|
|
; RV32ZVE32F-NEXT: .LBB51_9: # %else14
|
|
; RV32ZVE32F-NEXT: lw s0, 44(sp) # 4-byte Folded Reload
|
|
; RV32ZVE32F-NEXT: lw s1, 40(sp) # 4-byte Folded Reload
|
|
; RV32ZVE32F-NEXT: lw s2, 36(sp) # 4-byte Folded Reload
|
|
; RV32ZVE32F-NEXT: lw s3, 32(sp) # 4-byte Folded Reload
|
|
; RV32ZVE32F-NEXT: lw s4, 28(sp) # 4-byte Folded Reload
|
|
; RV32ZVE32F-NEXT: lw s5, 24(sp) # 4-byte Folded Reload
|
|
; RV32ZVE32F-NEXT: lw s6, 20(sp) # 4-byte Folded Reload
|
|
; RV32ZVE32F-NEXT: lw s7, 16(sp) # 4-byte Folded Reload
|
|
; RV32ZVE32F-NEXT: lw s8, 12(sp) # 4-byte Folded Reload
|
|
; RV32ZVE32F-NEXT: lw s9, 8(sp) # 4-byte Folded Reload
|
|
; RV32ZVE32F-NEXT: .cfi_restore s0
|
|
; RV32ZVE32F-NEXT: .cfi_restore s1
|
|
; RV32ZVE32F-NEXT: .cfi_restore s2
|
|
; RV32ZVE32F-NEXT: .cfi_restore s3
|
|
; RV32ZVE32F-NEXT: .cfi_restore s4
|
|
; RV32ZVE32F-NEXT: .cfi_restore s5
|
|
; RV32ZVE32F-NEXT: .cfi_restore s6
|
|
; RV32ZVE32F-NEXT: .cfi_restore s7
|
|
; RV32ZVE32F-NEXT: .cfi_restore s8
|
|
; RV32ZVE32F-NEXT: .cfi_restore s9
|
|
; RV32ZVE32F-NEXT: addi sp, sp, 48
|
|
; RV32ZVE32F-NEXT: .cfi_def_cfa_offset 0
|
|
; RV32ZVE32F-NEXT: ret
|
|
; RV32ZVE32F-NEXT: .LBB51_10: # %cond.store
|
|
; RV32ZVE32F-NEXT: .cfi_restore_state
|
|
; RV32ZVE32F-NEXT: lw a1, 0(a0)
|
|
; RV32ZVE32F-NEXT: lw a0, 4(a0)
|
|
; RV32ZVE32F-NEXT: vmv.x.s s2, v8
|
|
; RV32ZVE32F-NEXT: sw a1, 0(s2)
|
|
; RV32ZVE32F-NEXT: sw a0, 4(s2)
|
|
; RV32ZVE32F-NEXT: andi a0, a2, 2
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB51_2
|
|
; RV32ZVE32F-NEXT: .LBB51_11: # %cond.store1
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 1
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: sw s0, 0(a0)
|
|
; RV32ZVE32F-NEXT: sw s1, 4(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, a2, 4
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB51_3
|
|
; RV32ZVE32F-NEXT: .LBB51_12: # %cond.store3
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 2
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: sw t5, 0(a0)
|
|
; RV32ZVE32F-NEXT: sw t6, 4(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, a2, 8
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB51_4
|
|
; RV32ZVE32F-NEXT: .LBB51_13: # %cond.store5
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 3
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: sw t3, 0(a0)
|
|
; RV32ZVE32F-NEXT: sw t4, 4(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, a2, 16
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB51_5
|
|
; RV32ZVE32F-NEXT: .LBB51_14: # %cond.store7
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 4
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: sw t1, 0(a0)
|
|
; RV32ZVE32F-NEXT: sw t2, 4(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, a2, 32
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB51_6
|
|
; RV32ZVE32F-NEXT: .LBB51_15: # %cond.store9
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 5
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: sw a7, 0(a0)
|
|
; RV32ZVE32F-NEXT: sw t0, 4(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, a2, 64
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB51_7
|
|
; RV32ZVE32F-NEXT: .LBB51_16: # %cond.store11
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 6
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: sw a5, 0(a0)
|
|
; RV32ZVE32F-NEXT: sw a6, 4(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, a2, -128
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB51_8
|
|
; RV32ZVE32F-NEXT: j .LBB51_9
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_baseidx_v8i64:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: addi sp, sp, -32
|
|
; RV64ZVE32F-NEXT: .cfi_def_cfa_offset 32
|
|
; RV64ZVE32F-NEXT: sd s0, 24(sp) # 8-byte Folded Spill
|
|
; RV64ZVE32F-NEXT: sd s1, 16(sp) # 8-byte Folded Spill
|
|
; RV64ZVE32F-NEXT: sd s2, 8(sp) # 8-byte Folded Spill
|
|
; RV64ZVE32F-NEXT: sd s3, 0(sp) # 8-byte Folded Spill
|
|
; RV64ZVE32F-NEXT: .cfi_offset s0, -8
|
|
; RV64ZVE32F-NEXT: .cfi_offset s1, -16
|
|
; RV64ZVE32F-NEXT: .cfi_offset s2, -24
|
|
; RV64ZVE32F-NEXT: .cfi_offset s3, -32
|
|
; RV64ZVE32F-NEXT: .cfi_remember_state
|
|
; RV64ZVE32F-NEXT: ld a5, 40(a0)
|
|
; RV64ZVE32F-NEXT: ld a4, 48(a0)
|
|
; RV64ZVE32F-NEXT: ld a3, 56(a0)
|
|
; RV64ZVE32F-NEXT: ld s0, 8(a0)
|
|
; RV64ZVE32F-NEXT: ld t5, 16(a0)
|
|
; RV64ZVE32F-NEXT: ld t3, 24(a0)
|
|
; RV64ZVE32F-NEXT: ld t1, 32(a0)
|
|
; RV64ZVE32F-NEXT: ld s2, 8(a2)
|
|
; RV64ZVE32F-NEXT: ld s1, 16(a2)
|
|
; RV64ZVE32F-NEXT: ld t6, 24(a2)
|
|
; RV64ZVE32F-NEXT: ld t4, 32(a2)
|
|
; RV64ZVE32F-NEXT: ld t2, 40(a2)
|
|
; RV64ZVE32F-NEXT: ld t0, 48(a2)
|
|
; RV64ZVE32F-NEXT: ld a6, 56(a2)
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a7, v0
|
|
; RV64ZVE32F-NEXT: andi s3, a7, 1
|
|
; RV64ZVE32F-NEXT: bnez s3, .LBB51_10
|
|
; RV64ZVE32F-NEXT: # %bb.1: # %else
|
|
; RV64ZVE32F-NEXT: andi a0, a7, 2
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB51_11
|
|
; RV64ZVE32F-NEXT: .LBB51_2: # %else2
|
|
; RV64ZVE32F-NEXT: andi a0, a7, 4
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB51_12
|
|
; RV64ZVE32F-NEXT: .LBB51_3: # %else4
|
|
; RV64ZVE32F-NEXT: andi a0, a7, 8
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB51_13
|
|
; RV64ZVE32F-NEXT: .LBB51_4: # %else6
|
|
; RV64ZVE32F-NEXT: andi a0, a7, 16
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB51_14
|
|
; RV64ZVE32F-NEXT: .LBB51_5: # %else8
|
|
; RV64ZVE32F-NEXT: andi a0, a7, 32
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB51_15
|
|
; RV64ZVE32F-NEXT: .LBB51_6: # %else10
|
|
; RV64ZVE32F-NEXT: andi a0, a7, 64
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB51_16
|
|
; RV64ZVE32F-NEXT: .LBB51_7: # %else12
|
|
; RV64ZVE32F-NEXT: andi a0, a7, -128
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB51_9
|
|
; RV64ZVE32F-NEXT: .LBB51_8: # %cond.store13
|
|
; RV64ZVE32F-NEXT: slli a6, a6, 3
|
|
; RV64ZVE32F-NEXT: add a1, a1, a6
|
|
; RV64ZVE32F-NEXT: sd a3, 0(a1)
|
|
; RV64ZVE32F-NEXT: .LBB51_9: # %else14
|
|
; RV64ZVE32F-NEXT: ld s0, 24(sp) # 8-byte Folded Reload
|
|
; RV64ZVE32F-NEXT: ld s1, 16(sp) # 8-byte Folded Reload
|
|
; RV64ZVE32F-NEXT: ld s2, 8(sp) # 8-byte Folded Reload
|
|
; RV64ZVE32F-NEXT: ld s3, 0(sp) # 8-byte Folded Reload
|
|
; RV64ZVE32F-NEXT: .cfi_restore s0
|
|
; RV64ZVE32F-NEXT: .cfi_restore s1
|
|
; RV64ZVE32F-NEXT: .cfi_restore s2
|
|
; RV64ZVE32F-NEXT: .cfi_restore s3
|
|
; RV64ZVE32F-NEXT: addi sp, sp, 32
|
|
; RV64ZVE32F-NEXT: .cfi_def_cfa_offset 0
|
|
; RV64ZVE32F-NEXT: ret
|
|
; RV64ZVE32F-NEXT: .LBB51_10: # %cond.store
|
|
; RV64ZVE32F-NEXT: .cfi_restore_state
|
|
; RV64ZVE32F-NEXT: ld a2, 0(a2)
|
|
; RV64ZVE32F-NEXT: ld a0, 0(a0)
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 3
|
|
; RV64ZVE32F-NEXT: add a2, a1, a2
|
|
; RV64ZVE32F-NEXT: sd a0, 0(a2)
|
|
; RV64ZVE32F-NEXT: andi a0, a7, 2
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB51_2
|
|
; RV64ZVE32F-NEXT: .LBB51_11: # %cond.store1
|
|
; RV64ZVE32F-NEXT: slli s2, s2, 3
|
|
; RV64ZVE32F-NEXT: add s2, a1, s2
|
|
; RV64ZVE32F-NEXT: sd s0, 0(s2)
|
|
; RV64ZVE32F-NEXT: andi a0, a7, 4
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB51_3
|
|
; RV64ZVE32F-NEXT: .LBB51_12: # %cond.store3
|
|
; RV64ZVE32F-NEXT: slli s1, s1, 3
|
|
; RV64ZVE32F-NEXT: add s1, a1, s1
|
|
; RV64ZVE32F-NEXT: sd t5, 0(s1)
|
|
; RV64ZVE32F-NEXT: andi a0, a7, 8
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB51_4
|
|
; RV64ZVE32F-NEXT: .LBB51_13: # %cond.store5
|
|
; RV64ZVE32F-NEXT: slli t6, t6, 3
|
|
; RV64ZVE32F-NEXT: add t6, a1, t6
|
|
; RV64ZVE32F-NEXT: sd t3, 0(t6)
|
|
; RV64ZVE32F-NEXT: andi a0, a7, 16
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB51_5
|
|
; RV64ZVE32F-NEXT: .LBB51_14: # %cond.store7
|
|
; RV64ZVE32F-NEXT: slli t4, t4, 3
|
|
; RV64ZVE32F-NEXT: add t4, a1, t4
|
|
; RV64ZVE32F-NEXT: sd t1, 0(t4)
|
|
; RV64ZVE32F-NEXT: andi a0, a7, 32
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB51_6
|
|
; RV64ZVE32F-NEXT: .LBB51_15: # %cond.store9
|
|
; RV64ZVE32F-NEXT: slli t2, t2, 3
|
|
; RV64ZVE32F-NEXT: add t2, a1, t2
|
|
; RV64ZVE32F-NEXT: sd a5, 0(t2)
|
|
; RV64ZVE32F-NEXT: andi a0, a7, 64
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB51_7
|
|
; RV64ZVE32F-NEXT: .LBB51_16: # %cond.store11
|
|
; RV64ZVE32F-NEXT: slli t0, t0, 3
|
|
; RV64ZVE32F-NEXT: add t0, a1, t0
|
|
; RV64ZVE32F-NEXT: sd a4, 0(t0)
|
|
; RV64ZVE32F-NEXT: andi a0, a7, -128
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB51_8
|
|
; RV64ZVE32F-NEXT: j .LBB51_9
|
|
%ptrs = getelementptr inbounds i64, ptr %base, <8 x i64> %idxs
|
|
call void @llvm.masked.scatter.v8i64.v8p0(<8 x i64> %val, <8 x ptr> %ptrs, i32 8, <8 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_v1bf16(<1 x bfloat> %val, <1 x ptr> %ptrs, <1 x i1> %m) {
|
|
; RV32V-LABEL: mscatter_v1bf16:
|
|
; RV32V: # %bb.0:
|
|
; RV32V-NEXT: vsetivli zero, 1, e16, mf4, ta, ma
|
|
; RV32V-NEXT: vsoxei32.v v8, (zero), v9, v0.t
|
|
; RV32V-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_v1bf16:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: vsetivli zero, 1, e16, mf4, ta, ma
|
|
; RV64V-NEXT: vsoxei64.v v8, (zero), v9, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV32ZVE32F-LABEL: mscatter_v1bf16:
|
|
; RV32ZVE32F: # %bb.0:
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
|
|
; RV32ZVE32F-NEXT: vsoxei32.v v8, (zero), v9, v0.t
|
|
; RV32ZVE32F-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_v1bf16:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: vsetvli a1, zero, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vfirst.m a1, v0
|
|
; RV64ZVE32F-NEXT: bnez a1, .LBB52_2
|
|
; RV64ZVE32F-NEXT: # %bb.1: # %cond.store
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a1, v8
|
|
; RV64ZVE32F-NEXT: fmv.h.x fa5, a1
|
|
; RV64ZVE32F-NEXT: fsh fa5, 0(a0)
|
|
; RV64ZVE32F-NEXT: .LBB52_2: # %else
|
|
; RV64ZVE32F-NEXT: ret
|
|
call void @llvm.masked.scatter.v1bf16.v1p0(<1 x bfloat> %val, <1 x ptr> %ptrs, i32 2, <1 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_v2bf16(<2 x bfloat> %val, <2 x ptr> %ptrs, <2 x i1> %m) {
|
|
; RV32V-LABEL: mscatter_v2bf16:
|
|
; RV32V: # %bb.0:
|
|
; RV32V-NEXT: vsetivli zero, 2, e16, mf4, ta, ma
|
|
; RV32V-NEXT: vsoxei32.v v8, (zero), v9, v0.t
|
|
; RV32V-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_v2bf16:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: vsetivli zero, 2, e16, mf4, ta, ma
|
|
; RV64V-NEXT: vsoxei64.v v8, (zero), v9, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV32ZVE32F-LABEL: mscatter_v2bf16:
|
|
; RV32ZVE32F: # %bb.0:
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 2, e16, mf2, ta, ma
|
|
; RV32ZVE32F-NEXT: vsoxei32.v v8, (zero), v9, v0.t
|
|
; RV32ZVE32F-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_v2bf16:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v0
|
|
; RV64ZVE32F-NEXT: andi a3, a2, 1
|
|
; RV64ZVE32F-NEXT: bnez a3, .LBB53_3
|
|
; RV64ZVE32F-NEXT: # %bb.1: # %else
|
|
; RV64ZVE32F-NEXT: andi a2, a2, 2
|
|
; RV64ZVE32F-NEXT: bnez a2, .LBB53_4
|
|
; RV64ZVE32F-NEXT: .LBB53_2: # %else2
|
|
; RV64ZVE32F-NEXT: ret
|
|
; RV64ZVE32F-NEXT: .LBB53_3: # %cond.store
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e16, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a3, v8
|
|
; RV64ZVE32F-NEXT: fmv.h.x fa5, a3
|
|
; RV64ZVE32F-NEXT: fsh fa5, 0(a0)
|
|
; RV64ZVE32F-NEXT: andi a2, a2, 2
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB53_2
|
|
; RV64ZVE32F-NEXT: .LBB53_4: # %cond.store1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a0, v8
|
|
; RV64ZVE32F-NEXT: fmv.h.x fa5, a0
|
|
; RV64ZVE32F-NEXT: fsh fa5, 0(a1)
|
|
; RV64ZVE32F-NEXT: ret
|
|
call void @llvm.masked.scatter.v2bf16.v2p0(<2 x bfloat> %val, <2 x ptr> %ptrs, i32 2, <2 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_v4bf16(<4 x bfloat> %val, <4 x ptr> %ptrs, <4 x i1> %m) {
|
|
; RV32-LABEL: mscatter_v4bf16:
|
|
; RV32: # %bb.0:
|
|
; RV32-NEXT: vsetivli zero, 4, e16, mf2, ta, ma
|
|
; RV32-NEXT: vsoxei32.v v8, (zero), v9, v0.t
|
|
; RV32-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_v4bf16:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: vsetivli zero, 4, e16, mf2, ta, ma
|
|
; RV64V-NEXT: vsoxei64.v v8, (zero), v10, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_v4bf16:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: ld a4, 8(a0)
|
|
; RV64ZVE32F-NEXT: ld a2, 16(a0)
|
|
; RV64ZVE32F-NEXT: ld a1, 24(a0)
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a3, v0
|
|
; RV64ZVE32F-NEXT: andi a5, a3, 1
|
|
; RV64ZVE32F-NEXT: bnez a5, .LBB54_5
|
|
; RV64ZVE32F-NEXT: # %bb.1: # %else
|
|
; RV64ZVE32F-NEXT: andi a0, a3, 2
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB54_6
|
|
; RV64ZVE32F-NEXT: .LBB54_2: # %else2
|
|
; RV64ZVE32F-NEXT: andi a0, a3, 4
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB54_7
|
|
; RV64ZVE32F-NEXT: .LBB54_3: # %else4
|
|
; RV64ZVE32F-NEXT: andi a3, a3, 8
|
|
; RV64ZVE32F-NEXT: bnez a3, .LBB54_8
|
|
; RV64ZVE32F-NEXT: .LBB54_4: # %else6
|
|
; RV64ZVE32F-NEXT: ret
|
|
; RV64ZVE32F-NEXT: .LBB54_5: # %cond.store
|
|
; RV64ZVE32F-NEXT: ld a0, 0(a0)
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e16, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a5, v8
|
|
; RV64ZVE32F-NEXT: fmv.h.x fa5, a5
|
|
; RV64ZVE32F-NEXT: fsh fa5, 0(a0)
|
|
; RV64ZVE32F-NEXT: andi a0, a3, 2
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB54_2
|
|
; RV64ZVE32F-NEXT: .LBB54_6: # %cond.store1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a0, v9
|
|
; RV64ZVE32F-NEXT: fmv.h.x fa5, a0
|
|
; RV64ZVE32F-NEXT: fsh fa5, 0(a4)
|
|
; RV64ZVE32F-NEXT: andi a0, a3, 4
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB54_3
|
|
; RV64ZVE32F-NEXT: .LBB54_7: # %cond.store3
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 2
|
|
; RV64ZVE32F-NEXT: vmv.x.s a0, v9
|
|
; RV64ZVE32F-NEXT: fmv.h.x fa5, a0
|
|
; RV64ZVE32F-NEXT: fsh fa5, 0(a2)
|
|
; RV64ZVE32F-NEXT: andi a3, a3, 8
|
|
; RV64ZVE32F-NEXT: beqz a3, .LBB54_4
|
|
; RV64ZVE32F-NEXT: .LBB54_8: # %cond.store5
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 3
|
|
; RV64ZVE32F-NEXT: vmv.x.s a0, v8
|
|
; RV64ZVE32F-NEXT: fmv.h.x fa5, a0
|
|
; RV64ZVE32F-NEXT: fsh fa5, 0(a1)
|
|
; RV64ZVE32F-NEXT: ret
|
|
call void @llvm.masked.scatter.v4bf16.v4p0(<4 x bfloat> %val, <4 x ptr> %ptrs, i32 2, <4 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_truemask_v4bf16(<4 x bfloat> %val, <4 x ptr> %ptrs) {
|
|
; RV32-LABEL: mscatter_truemask_v4bf16:
|
|
; RV32: # %bb.0:
|
|
; RV32-NEXT: vsetivli zero, 4, e16, mf2, ta, ma
|
|
; RV32-NEXT: vsoxei32.v v8, (zero), v9
|
|
; RV32-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_truemask_v4bf16:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: vsetivli zero, 4, e16, mf2, ta, ma
|
|
; RV64V-NEXT: vsoxei64.v v8, (zero), v10
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_truemask_v4bf16:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: ld a1, 0(a0)
|
|
; RV64ZVE32F-NEXT: ld a2, 8(a0)
|
|
; RV64ZVE32F-NEXT: ld a3, 16(a0)
|
|
; RV64ZVE32F-NEXT: ld a0, 24(a0)
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a4, v8
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 1
|
|
; RV64ZVE32F-NEXT: fmv.h.x fa5, a4
|
|
; RV64ZVE32F-NEXT: vmv.x.s a4, v9
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 2
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 3
|
|
; RV64ZVE32F-NEXT: fsh fa5, 0(a1)
|
|
; RV64ZVE32F-NEXT: vmv.x.s a1, v9
|
|
; RV64ZVE32F-NEXT: fmv.h.x fa5, a4
|
|
; RV64ZVE32F-NEXT: vmv.x.s a4, v8
|
|
; RV64ZVE32F-NEXT: fsh fa5, 0(a2)
|
|
; RV64ZVE32F-NEXT: fmv.h.x fa5, a1
|
|
; RV64ZVE32F-NEXT: fsh fa5, 0(a3)
|
|
; RV64ZVE32F-NEXT: fmv.h.x fa5, a4
|
|
; RV64ZVE32F-NEXT: fsh fa5, 0(a0)
|
|
; RV64ZVE32F-NEXT: ret
|
|
call void @llvm.masked.scatter.v4bf16.v4p0(<4 x bfloat> %val, <4 x ptr> %ptrs, i32 2, <4 x i1> splat (i1 1))
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_falsemask_v4bf16(<4 x bfloat> %val, <4 x ptr> %ptrs) {
|
|
; CHECK-LABEL: mscatter_falsemask_v4bf16:
|
|
; CHECK: # %bb.0:
|
|
; CHECK-NEXT: ret
|
|
call void @llvm.masked.scatter.v4bf16.v4p0(<4 x bfloat> %val, <4 x ptr> %ptrs, i32 2, <4 x i1> zeroinitializer)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_v8bf16(<8 x bfloat> %val, <8 x ptr> %ptrs, <8 x i1> %m) {
|
|
; RV32-LABEL: mscatter_v8bf16:
|
|
; RV32: # %bb.0:
|
|
; RV32-NEXT: vsetivli zero, 8, e16, m1, ta, ma
|
|
; RV32-NEXT: vsoxei32.v v8, (zero), v10, v0.t
|
|
; RV32-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_v8bf16:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: vsetivli zero, 8, e16, m1, ta, ma
|
|
; RV64V-NEXT: vsoxei64.v v8, (zero), v12, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_v8bf16:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: ld a3, 40(a0)
|
|
; RV64ZVE32F-NEXT: ld a2, 48(a0)
|
|
; RV64ZVE32F-NEXT: ld a1, 56(a0)
|
|
; RV64ZVE32F-NEXT: ld t0, 8(a0)
|
|
; RV64ZVE32F-NEXT: ld a7, 16(a0)
|
|
; RV64ZVE32F-NEXT: ld a6, 24(a0)
|
|
; RV64ZVE32F-NEXT: ld a5, 32(a0)
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a4, v0
|
|
; RV64ZVE32F-NEXT: andi t1, a4, 1
|
|
; RV64ZVE32F-NEXT: bnez t1, .LBB57_9
|
|
; RV64ZVE32F-NEXT: # %bb.1: # %else
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 2
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB57_10
|
|
; RV64ZVE32F-NEXT: .LBB57_2: # %else2
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 4
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB57_11
|
|
; RV64ZVE32F-NEXT: .LBB57_3: # %else4
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 8
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB57_12
|
|
; RV64ZVE32F-NEXT: .LBB57_4: # %else6
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 16
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB57_13
|
|
; RV64ZVE32F-NEXT: .LBB57_5: # %else8
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 32
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB57_14
|
|
; RV64ZVE32F-NEXT: .LBB57_6: # %else10
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 64
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB57_15
|
|
; RV64ZVE32F-NEXT: .LBB57_7: # %else12
|
|
; RV64ZVE32F-NEXT: andi a0, a4, -128
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB57_16
|
|
; RV64ZVE32F-NEXT: .LBB57_8: # %else14
|
|
; RV64ZVE32F-NEXT: ret
|
|
; RV64ZVE32F-NEXT: .LBB57_9: # %cond.store
|
|
; RV64ZVE32F-NEXT: ld a0, 0(a0)
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e16, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s t1, v8
|
|
; RV64ZVE32F-NEXT: fmv.h.x fa5, t1
|
|
; RV64ZVE32F-NEXT: fsh fa5, 0(a0)
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 2
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB57_2
|
|
; RV64ZVE32F-NEXT: .LBB57_10: # %cond.store1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a0, v9
|
|
; RV64ZVE32F-NEXT: fmv.h.x fa5, a0
|
|
; RV64ZVE32F-NEXT: fsh fa5, 0(t0)
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 4
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB57_3
|
|
; RV64ZVE32F-NEXT: .LBB57_11: # %cond.store3
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 2
|
|
; RV64ZVE32F-NEXT: vmv.x.s a0, v9
|
|
; RV64ZVE32F-NEXT: fmv.h.x fa5, a0
|
|
; RV64ZVE32F-NEXT: fsh fa5, 0(a7)
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 8
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB57_4
|
|
; RV64ZVE32F-NEXT: .LBB57_12: # %cond.store5
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 3
|
|
; RV64ZVE32F-NEXT: vmv.x.s a0, v9
|
|
; RV64ZVE32F-NEXT: fmv.h.x fa5, a0
|
|
; RV64ZVE32F-NEXT: fsh fa5, 0(a6)
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 16
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB57_5
|
|
; RV64ZVE32F-NEXT: .LBB57_13: # %cond.store7
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 4
|
|
; RV64ZVE32F-NEXT: vmv.x.s a0, v9
|
|
; RV64ZVE32F-NEXT: fmv.h.x fa5, a0
|
|
; RV64ZVE32F-NEXT: fsh fa5, 0(a5)
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 32
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB57_6
|
|
; RV64ZVE32F-NEXT: .LBB57_14: # %cond.store9
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 5
|
|
; RV64ZVE32F-NEXT: vmv.x.s a0, v9
|
|
; RV64ZVE32F-NEXT: fmv.h.x fa5, a0
|
|
; RV64ZVE32F-NEXT: fsh fa5, 0(a3)
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 64
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB57_7
|
|
; RV64ZVE32F-NEXT: .LBB57_15: # %cond.store11
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 6
|
|
; RV64ZVE32F-NEXT: vmv.x.s a0, v9
|
|
; RV64ZVE32F-NEXT: fmv.h.x fa5, a0
|
|
; RV64ZVE32F-NEXT: fsh fa5, 0(a2)
|
|
; RV64ZVE32F-NEXT: andi a0, a4, -128
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB57_8
|
|
; RV64ZVE32F-NEXT: .LBB57_16: # %cond.store13
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 7
|
|
; RV64ZVE32F-NEXT: vmv.x.s a0, v8
|
|
; RV64ZVE32F-NEXT: fmv.h.x fa5, a0
|
|
; RV64ZVE32F-NEXT: fsh fa5, 0(a1)
|
|
; RV64ZVE32F-NEXT: ret
|
|
call void @llvm.masked.scatter.v8bf16.v8p0(<8 x bfloat> %val, <8 x ptr> %ptrs, i32 2, <8 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_baseidx_v8i8_v8bf16(<8 x bfloat> %val, ptr %base, <8 x i8> %idxs, <8 x i1> %m) {
|
|
; RV32-LABEL: mscatter_baseidx_v8i8_v8bf16:
|
|
; RV32: # %bb.0:
|
|
; RV32-NEXT: vsetivli zero, 8, e32, m2, ta, ma
|
|
; RV32-NEXT: vsext.vf4 v10, v9
|
|
; RV32-NEXT: vadd.vv v10, v10, v10
|
|
; RV32-NEXT: vsetvli zero, zero, e16, m1, ta, ma
|
|
; RV32-NEXT: vsoxei32.v v8, (a0), v10, v0.t
|
|
; RV32-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_baseidx_v8i8_v8bf16:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: vsetivli zero, 8, e64, m4, ta, ma
|
|
; RV64V-NEXT: vsext.vf8 v12, v9
|
|
; RV64V-NEXT: vadd.vv v12, v12, v12
|
|
; RV64V-NEXT: vsetvli zero, zero, e16, m1, ta, ma
|
|
; RV64V-NEXT: vsoxei64.v v8, (a0), v12, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_baseidx_v8i8_v8bf16:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a1, v0
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 1
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB58_2
|
|
; RV64ZVE32F-NEXT: # %bb.1: # %cond.store
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v9
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e16, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a3, v8
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fmv.h.x fa5, a3
|
|
; RV64ZVE32F-NEXT: fsh fa5, 0(a2)
|
|
; RV64ZVE32F-NEXT: .LBB58_2: # %else
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 2
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB58_4
|
|
; RV64ZVE32F-NEXT: # %bb.3: # %cond.store1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v9, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v8, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a3, v10
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fmv.h.x fa5, a3
|
|
; RV64ZVE32F-NEXT: fsh fa5, 0(a2)
|
|
; RV64ZVE32F-NEXT: .LBB58_4: # %else2
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 4
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v9, 2
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB58_6
|
|
; RV64ZVE32F-NEXT: # %bb.5: # %cond.store3
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v8, 2
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a3, v11
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fmv.h.x fa5, a3
|
|
; RV64ZVE32F-NEXT: fsh fa5, 0(a2)
|
|
; RV64ZVE32F-NEXT: .LBB58_6: # %else4
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 8
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB58_8
|
|
; RV64ZVE32F-NEXT: # %bb.7: # %cond.store5
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v10, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v8, 3
|
|
; RV64ZVE32F-NEXT: vmv.x.s a3, v10
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fmv.h.x fa5, a3
|
|
; RV64ZVE32F-NEXT: fsh fa5, 0(a2)
|
|
; RV64ZVE32F-NEXT: .LBB58_8: # %else6
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 16
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 4, e8, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v9, 4
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB58_10
|
|
; RV64ZVE32F-NEXT: # %bb.9: # %cond.store7
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v9
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v8, 4
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a3, v10
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fmv.h.x fa5, a3
|
|
; RV64ZVE32F-NEXT: fsh fa5, 0(a2)
|
|
; RV64ZVE32F-NEXT: .LBB58_10: # %else8
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 32
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB58_12
|
|
; RV64ZVE32F-NEXT: # %bb.11: # %cond.store9
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v9, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v8, 5
|
|
; RV64ZVE32F-NEXT: vmv.x.s a3, v10
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fmv.h.x fa5, a3
|
|
; RV64ZVE32F-NEXT: fsh fa5, 0(a2)
|
|
; RV64ZVE32F-NEXT: .LBB58_12: # %else10
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 64
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v9, 2
|
|
; RV64ZVE32F-NEXT: bnez a2, .LBB58_15
|
|
; RV64ZVE32F-NEXT: # %bb.13: # %else12
|
|
; RV64ZVE32F-NEXT: andi a1, a1, -128
|
|
; RV64ZVE32F-NEXT: bnez a1, .LBB58_16
|
|
; RV64ZVE32F-NEXT: .LBB58_14: # %else14
|
|
; RV64ZVE32F-NEXT: ret
|
|
; RV64ZVE32F-NEXT: .LBB58_15: # %cond.store11
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v9
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v8, 6
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a3, v10
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fmv.h.x fa5, a3
|
|
; RV64ZVE32F-NEXT: fsh fa5, 0(a2)
|
|
; RV64ZVE32F-NEXT: andi a1, a1, -128
|
|
; RV64ZVE32F-NEXT: beqz a1, .LBB58_14
|
|
; RV64ZVE32F-NEXT: .LBB58_16: # %cond.store13
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v9, 1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 7
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e8, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a1, v9
|
|
; RV64ZVE32F-NEXT: slli a1, a1, 1
|
|
; RV64ZVE32F-NEXT: add a0, a0, a1
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a1, v8
|
|
; RV64ZVE32F-NEXT: fmv.h.x fa5, a1
|
|
; RV64ZVE32F-NEXT: fsh fa5, 0(a0)
|
|
; RV64ZVE32F-NEXT: ret
|
|
%ptrs = getelementptr inbounds bfloat, ptr %base, <8 x i8> %idxs
|
|
call void @llvm.masked.scatter.v8bf16.v8p0(<8 x bfloat> %val, <8 x ptr> %ptrs, i32 2, <8 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_baseidx_sext_v8i8_v8bf16(<8 x bfloat> %val, ptr %base, <8 x i8> %idxs, <8 x i1> %m) {
|
|
; RV32-LABEL: mscatter_baseidx_sext_v8i8_v8bf16:
|
|
; RV32: # %bb.0:
|
|
; RV32-NEXT: vsetivli zero, 8, e32, m2, ta, ma
|
|
; RV32-NEXT: vsext.vf4 v10, v9
|
|
; RV32-NEXT: vadd.vv v10, v10, v10
|
|
; RV32-NEXT: vsetvli zero, zero, e16, m1, ta, ma
|
|
; RV32-NEXT: vsoxei32.v v8, (a0), v10, v0.t
|
|
; RV32-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_baseidx_sext_v8i8_v8bf16:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: vsetivli zero, 8, e64, m4, ta, ma
|
|
; RV64V-NEXT: vsext.vf8 v12, v9
|
|
; RV64V-NEXT: vadd.vv v12, v12, v12
|
|
; RV64V-NEXT: vsetvli zero, zero, e16, m1, ta, ma
|
|
; RV64V-NEXT: vsoxei64.v v8, (a0), v12, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_baseidx_sext_v8i8_v8bf16:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a1, v0
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 1
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB59_2
|
|
; RV64ZVE32F-NEXT: # %bb.1: # %cond.store
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v9
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e16, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a3, v8
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fmv.h.x fa5, a3
|
|
; RV64ZVE32F-NEXT: fsh fa5, 0(a2)
|
|
; RV64ZVE32F-NEXT: .LBB59_2: # %else
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 2
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB59_4
|
|
; RV64ZVE32F-NEXT: # %bb.3: # %cond.store1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v9, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v8, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a3, v10
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fmv.h.x fa5, a3
|
|
; RV64ZVE32F-NEXT: fsh fa5, 0(a2)
|
|
; RV64ZVE32F-NEXT: .LBB59_4: # %else2
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 4
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v9, 2
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB59_6
|
|
; RV64ZVE32F-NEXT: # %bb.5: # %cond.store3
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v8, 2
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a3, v11
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fmv.h.x fa5, a3
|
|
; RV64ZVE32F-NEXT: fsh fa5, 0(a2)
|
|
; RV64ZVE32F-NEXT: .LBB59_6: # %else4
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 8
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB59_8
|
|
; RV64ZVE32F-NEXT: # %bb.7: # %cond.store5
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v10, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v8, 3
|
|
; RV64ZVE32F-NEXT: vmv.x.s a3, v10
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fmv.h.x fa5, a3
|
|
; RV64ZVE32F-NEXT: fsh fa5, 0(a2)
|
|
; RV64ZVE32F-NEXT: .LBB59_8: # %else6
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 16
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 4, e8, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v9, 4
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB59_10
|
|
; RV64ZVE32F-NEXT: # %bb.9: # %cond.store7
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v9
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v8, 4
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a3, v10
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fmv.h.x fa5, a3
|
|
; RV64ZVE32F-NEXT: fsh fa5, 0(a2)
|
|
; RV64ZVE32F-NEXT: .LBB59_10: # %else8
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 32
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB59_12
|
|
; RV64ZVE32F-NEXT: # %bb.11: # %cond.store9
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v9, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v8, 5
|
|
; RV64ZVE32F-NEXT: vmv.x.s a3, v10
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fmv.h.x fa5, a3
|
|
; RV64ZVE32F-NEXT: fsh fa5, 0(a2)
|
|
; RV64ZVE32F-NEXT: .LBB59_12: # %else10
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 64
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v9, 2
|
|
; RV64ZVE32F-NEXT: bnez a2, .LBB59_15
|
|
; RV64ZVE32F-NEXT: # %bb.13: # %else12
|
|
; RV64ZVE32F-NEXT: andi a1, a1, -128
|
|
; RV64ZVE32F-NEXT: bnez a1, .LBB59_16
|
|
; RV64ZVE32F-NEXT: .LBB59_14: # %else14
|
|
; RV64ZVE32F-NEXT: ret
|
|
; RV64ZVE32F-NEXT: .LBB59_15: # %cond.store11
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v9
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v8, 6
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a3, v10
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fmv.h.x fa5, a3
|
|
; RV64ZVE32F-NEXT: fsh fa5, 0(a2)
|
|
; RV64ZVE32F-NEXT: andi a1, a1, -128
|
|
; RV64ZVE32F-NEXT: beqz a1, .LBB59_14
|
|
; RV64ZVE32F-NEXT: .LBB59_16: # %cond.store13
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v9, 1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 7
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e8, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a1, v9
|
|
; RV64ZVE32F-NEXT: slli a1, a1, 1
|
|
; RV64ZVE32F-NEXT: add a0, a0, a1
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a1, v8
|
|
; RV64ZVE32F-NEXT: fmv.h.x fa5, a1
|
|
; RV64ZVE32F-NEXT: fsh fa5, 0(a0)
|
|
; RV64ZVE32F-NEXT: ret
|
|
%eidxs = sext <8 x i8> %idxs to <8 x i16>
|
|
%ptrs = getelementptr inbounds bfloat, ptr %base, <8 x i16> %eidxs
|
|
call void @llvm.masked.scatter.v8bf16.v8p0(<8 x bfloat> %val, <8 x ptr> %ptrs, i32 2, <8 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_baseidx_zext_v8i8_v8bf16(<8 x bfloat> %val, ptr %base, <8 x i8> %idxs, <8 x i1> %m) {
|
|
; RV32-LABEL: mscatter_baseidx_zext_v8i8_v8bf16:
|
|
; RV32: # %bb.0:
|
|
; RV32-NEXT: vsetivli zero, 8, e8, mf2, ta, ma
|
|
; RV32-NEXT: vwaddu.vv v10, v9, v9
|
|
; RV32-NEXT: vsetvli zero, zero, e16, m1, ta, ma
|
|
; RV32-NEXT: vsoxei16.v v8, (a0), v10, v0.t
|
|
; RV32-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_baseidx_zext_v8i8_v8bf16:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: vsetivli zero, 8, e8, mf2, ta, ma
|
|
; RV64V-NEXT: vwaddu.vv v10, v9, v9
|
|
; RV64V-NEXT: vsetvli zero, zero, e16, m1, ta, ma
|
|
; RV64V-NEXT: vsoxei16.v v8, (a0), v10, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_baseidx_zext_v8i8_v8bf16:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a1, v0
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 1
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB60_2
|
|
; RV64ZVE32F-NEXT: # %bb.1: # %cond.store
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v9
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e16, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a3, v8
|
|
; RV64ZVE32F-NEXT: zext.b a2, a2
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fmv.h.x fa5, a3
|
|
; RV64ZVE32F-NEXT: fsh fa5, 0(a2)
|
|
; RV64ZVE32F-NEXT: .LBB60_2: # %else
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 2
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB60_4
|
|
; RV64ZVE32F-NEXT: # %bb.3: # %cond.store1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v9, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v8, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a3, v10
|
|
; RV64ZVE32F-NEXT: zext.b a2, a2
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fmv.h.x fa5, a3
|
|
; RV64ZVE32F-NEXT: fsh fa5, 0(a2)
|
|
; RV64ZVE32F-NEXT: .LBB60_4: # %else2
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 4
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v9, 2
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB60_6
|
|
; RV64ZVE32F-NEXT: # %bb.5: # %cond.store3
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v8, 2
|
|
; RV64ZVE32F-NEXT: zext.b a2, a2
|
|
; RV64ZVE32F-NEXT: vmv.x.s a3, v11
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fmv.h.x fa5, a3
|
|
; RV64ZVE32F-NEXT: fsh fa5, 0(a2)
|
|
; RV64ZVE32F-NEXT: .LBB60_6: # %else4
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 8
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB60_8
|
|
; RV64ZVE32F-NEXT: # %bb.7: # %cond.store5
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v10, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v8, 3
|
|
; RV64ZVE32F-NEXT: vmv.x.s a3, v10
|
|
; RV64ZVE32F-NEXT: zext.b a2, a2
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fmv.h.x fa5, a3
|
|
; RV64ZVE32F-NEXT: fsh fa5, 0(a2)
|
|
; RV64ZVE32F-NEXT: .LBB60_8: # %else6
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 16
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 4, e8, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v9, 4
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB60_10
|
|
; RV64ZVE32F-NEXT: # %bb.9: # %cond.store7
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v9
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v8, 4
|
|
; RV64ZVE32F-NEXT: zext.b a2, a2
|
|
; RV64ZVE32F-NEXT: vmv.x.s a3, v10
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fmv.h.x fa5, a3
|
|
; RV64ZVE32F-NEXT: fsh fa5, 0(a2)
|
|
; RV64ZVE32F-NEXT: .LBB60_10: # %else8
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 32
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB60_12
|
|
; RV64ZVE32F-NEXT: # %bb.11: # %cond.store9
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v9, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v8, 5
|
|
; RV64ZVE32F-NEXT: vmv.x.s a3, v10
|
|
; RV64ZVE32F-NEXT: zext.b a2, a2
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fmv.h.x fa5, a3
|
|
; RV64ZVE32F-NEXT: fsh fa5, 0(a2)
|
|
; RV64ZVE32F-NEXT: .LBB60_12: # %else10
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 64
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v9, 2
|
|
; RV64ZVE32F-NEXT: bnez a2, .LBB60_15
|
|
; RV64ZVE32F-NEXT: # %bb.13: # %else12
|
|
; RV64ZVE32F-NEXT: andi a1, a1, -128
|
|
; RV64ZVE32F-NEXT: bnez a1, .LBB60_16
|
|
; RV64ZVE32F-NEXT: .LBB60_14: # %else14
|
|
; RV64ZVE32F-NEXT: ret
|
|
; RV64ZVE32F-NEXT: .LBB60_15: # %cond.store11
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v9
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v8, 6
|
|
; RV64ZVE32F-NEXT: zext.b a2, a2
|
|
; RV64ZVE32F-NEXT: vmv.x.s a3, v10
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fmv.h.x fa5, a3
|
|
; RV64ZVE32F-NEXT: fsh fa5, 0(a2)
|
|
; RV64ZVE32F-NEXT: andi a1, a1, -128
|
|
; RV64ZVE32F-NEXT: beqz a1, .LBB60_14
|
|
; RV64ZVE32F-NEXT: .LBB60_16: # %cond.store13
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v9, 1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 7
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e8, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a1, v9
|
|
; RV64ZVE32F-NEXT: zext.b a1, a1
|
|
; RV64ZVE32F-NEXT: slli a1, a1, 1
|
|
; RV64ZVE32F-NEXT: add a0, a0, a1
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a1, v8
|
|
; RV64ZVE32F-NEXT: fmv.h.x fa5, a1
|
|
; RV64ZVE32F-NEXT: fsh fa5, 0(a0)
|
|
; RV64ZVE32F-NEXT: ret
|
|
%eidxs = zext <8 x i8> %idxs to <8 x i16>
|
|
%ptrs = getelementptr inbounds bfloat, ptr %base, <8 x i16> %eidxs
|
|
call void @llvm.masked.scatter.v8bf16.v8p0(<8 x bfloat> %val, <8 x ptr> %ptrs, i32 2, <8 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_baseidx_v8bf16(<8 x bfloat> %val, ptr %base, <8 x i16> %idxs, <8 x i1> %m) {
|
|
; RV32-LABEL: mscatter_baseidx_v8bf16:
|
|
; RV32: # %bb.0:
|
|
; RV32-NEXT: vsetivli zero, 8, e16, m1, ta, ma
|
|
; RV32-NEXT: vwadd.vv v10, v9, v9
|
|
; RV32-NEXT: vsoxei32.v v8, (a0), v10, v0.t
|
|
; RV32-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_baseidx_v8bf16:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: vsetivli zero, 8, e64, m4, ta, ma
|
|
; RV64V-NEXT: vsext.vf4 v12, v9
|
|
; RV64V-NEXT: vadd.vv v12, v12, v12
|
|
; RV64V-NEXT: vsetvli zero, zero, e16, m1, ta, ma
|
|
; RV64V-NEXT: vsoxei64.v v8, (a0), v12, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_baseidx_v8bf16:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a1, v0
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 1
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB61_2
|
|
; RV64ZVE32F-NEXT: # %bb.1: # %cond.store
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e16, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v9
|
|
; RV64ZVE32F-NEXT: vmv.x.s a3, v8
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fmv.h.x fa5, a3
|
|
; RV64ZVE32F-NEXT: fsh fa5, 0(a2)
|
|
; RV64ZVE32F-NEXT: .LBB61_2: # %else
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 2
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB61_4
|
|
; RV64ZVE32F-NEXT: # %bb.3: # %cond.store1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v9, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v8, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a3, v10
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fmv.h.x fa5, a3
|
|
; RV64ZVE32F-NEXT: fsh fa5, 0(a2)
|
|
; RV64ZVE32F-NEXT: .LBB61_4: # %else2
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 4
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v9, 2
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB61_6
|
|
; RV64ZVE32F-NEXT: # %bb.5: # %cond.store3
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v8, 2
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a3, v11
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fmv.h.x fa5, a3
|
|
; RV64ZVE32F-NEXT: fsh fa5, 0(a2)
|
|
; RV64ZVE32F-NEXT: .LBB61_6: # %else4
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 8
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB61_8
|
|
; RV64ZVE32F-NEXT: # %bb.7: # %cond.store5
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v10, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v8, 3
|
|
; RV64ZVE32F-NEXT: vmv.x.s a3, v10
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fmv.h.x fa5, a3
|
|
; RV64ZVE32F-NEXT: fsh fa5, 0(a2)
|
|
; RV64ZVE32F-NEXT: .LBB61_8: # %else6
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 16
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 4, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v9, 4
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB61_10
|
|
; RV64ZVE32F-NEXT: # %bb.9: # %cond.store7
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v9
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v8, 4
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a3, v10
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fmv.h.x fa5, a3
|
|
; RV64ZVE32F-NEXT: fsh fa5, 0(a2)
|
|
; RV64ZVE32F-NEXT: .LBB61_10: # %else8
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 32
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB61_12
|
|
; RV64ZVE32F-NEXT: # %bb.11: # %cond.store9
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v9, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v8, 5
|
|
; RV64ZVE32F-NEXT: vmv.x.s a3, v10
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fmv.h.x fa5, a3
|
|
; RV64ZVE32F-NEXT: fsh fa5, 0(a2)
|
|
; RV64ZVE32F-NEXT: .LBB61_12: # %else10
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 64
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v9, 2
|
|
; RV64ZVE32F-NEXT: bnez a2, .LBB61_15
|
|
; RV64ZVE32F-NEXT: # %bb.13: # %else12
|
|
; RV64ZVE32F-NEXT: andi a1, a1, -128
|
|
; RV64ZVE32F-NEXT: bnez a1, .LBB61_16
|
|
; RV64ZVE32F-NEXT: .LBB61_14: # %else14
|
|
; RV64ZVE32F-NEXT: ret
|
|
; RV64ZVE32F-NEXT: .LBB61_15: # %cond.store11
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v9
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v8, 6
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a3, v10
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fmv.h.x fa5, a3
|
|
; RV64ZVE32F-NEXT: fsh fa5, 0(a2)
|
|
; RV64ZVE32F-NEXT: andi a1, a1, -128
|
|
; RV64ZVE32F-NEXT: beqz a1, .LBB61_14
|
|
; RV64ZVE32F-NEXT: .LBB61_16: # %cond.store13
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v9, 1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 7
|
|
; RV64ZVE32F-NEXT: vmv.x.s a1, v9
|
|
; RV64ZVE32F-NEXT: slli a1, a1, 1
|
|
; RV64ZVE32F-NEXT: add a0, a0, a1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a1, v8
|
|
; RV64ZVE32F-NEXT: fmv.h.x fa5, a1
|
|
; RV64ZVE32F-NEXT: fsh fa5, 0(a0)
|
|
; RV64ZVE32F-NEXT: ret
|
|
%ptrs = getelementptr inbounds bfloat, ptr %base, <8 x i16> %idxs
|
|
call void @llvm.masked.scatter.v8bf16.v8p0(<8 x bfloat> %val, <8 x ptr> %ptrs, i32 2, <8 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_v1f16(<1 x half> %val, <1 x ptr> %ptrs, <1 x i1> %m) {
|
|
; RV32V-LABEL: mscatter_v1f16:
|
|
; RV32V: # %bb.0:
|
|
; RV32V-NEXT: vsetivli zero, 1, e16, mf4, ta, ma
|
|
; RV32V-NEXT: vsoxei32.v v8, (zero), v9, v0.t
|
|
; RV32V-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_v1f16:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: vsetivli zero, 1, e16, mf4, ta, ma
|
|
; RV64V-NEXT: vsoxei64.v v8, (zero), v9, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV32ZVE32F-LABEL: mscatter_v1f16:
|
|
; RV32ZVE32F: # %bb.0:
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
|
|
; RV32ZVE32F-NEXT: vsoxei32.v v8, (zero), v9, v0.t
|
|
; RV32ZVE32F-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-ZVFH-LABEL: mscatter_v1f16:
|
|
; RV64ZVE32F-ZVFH: # %bb.0:
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetvli a1, zero, e8, mf4, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vfirst.m a1, v0
|
|
; RV64ZVE32F-ZVFH-NEXT: bnez a1, .LBB62_2
|
|
; RV64ZVE32F-ZVFH-NEXT: # %bb.1: # %cond.store
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vse16.v v8, (a0)
|
|
; RV64ZVE32F-ZVFH-NEXT: .LBB62_2: # %else
|
|
; RV64ZVE32F-ZVFH-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-ZVFHMIN-LABEL: mscatter_v1f16:
|
|
; RV64ZVE32F-ZVFHMIN: # %bb.0:
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetvli a1, zero, e8, mf4, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vfirst.m a1, v0
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: bnez a1, .LBB62_2
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: # %bb.1: # %cond.store
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetvli zero, zero, e16, mf2, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a1, v8
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fmv.h.x fa5, a1
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fsh fa5, 0(a0)
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: .LBB62_2: # %else
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: ret
|
|
call void @llvm.masked.scatter.v1f16.v1p0(<1 x half> %val, <1 x ptr> %ptrs, i32 2, <1 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_v2f16(<2 x half> %val, <2 x ptr> %ptrs, <2 x i1> %m) {
|
|
; RV32V-LABEL: mscatter_v2f16:
|
|
; RV32V: # %bb.0:
|
|
; RV32V-NEXT: vsetivli zero, 2, e16, mf4, ta, ma
|
|
; RV32V-NEXT: vsoxei32.v v8, (zero), v9, v0.t
|
|
; RV32V-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_v2f16:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: vsetivli zero, 2, e16, mf4, ta, ma
|
|
; RV64V-NEXT: vsoxei64.v v8, (zero), v9, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV32ZVE32F-LABEL: mscatter_v2f16:
|
|
; RV32ZVE32F: # %bb.0:
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 2, e16, mf2, ta, ma
|
|
; RV32ZVE32F-NEXT: vsoxei32.v v8, (zero), v9, v0.t
|
|
; RV32ZVE32F-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-ZVFH-LABEL: mscatter_v2f16:
|
|
; RV64ZVE32F-ZVFH: # %bb.0:
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vmv.x.s a2, v0
|
|
; RV64ZVE32F-ZVFH-NEXT: andi a3, a2, 1
|
|
; RV64ZVE32F-ZVFH-NEXT: bnez a3, .LBB63_3
|
|
; RV64ZVE32F-ZVFH-NEXT: # %bb.1: # %else
|
|
; RV64ZVE32F-ZVFH-NEXT: andi a2, a2, 2
|
|
; RV64ZVE32F-ZVFH-NEXT: bnez a2, .LBB63_4
|
|
; RV64ZVE32F-ZVFH-NEXT: .LBB63_2: # %else2
|
|
; RV64ZVE32F-ZVFH-NEXT: ret
|
|
; RV64ZVE32F-ZVFH-NEXT: .LBB63_3: # %cond.store
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vse16.v v8, (a0)
|
|
; RV64ZVE32F-ZVFH-NEXT: andi a2, a2, 2
|
|
; RV64ZVE32F-ZVFH-NEXT: beqz a2, .LBB63_2
|
|
; RV64ZVE32F-ZVFH-NEXT: .LBB63_4: # %cond.store1
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vslidedown.vi v8, v8, 1
|
|
; RV64ZVE32F-ZVFH-NEXT: vse16.v v8, (a1)
|
|
; RV64ZVE32F-ZVFH-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-ZVFHMIN-LABEL: mscatter_v2f16:
|
|
; RV64ZVE32F-ZVFHMIN: # %bb.0:
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a2, v0
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: andi a3, a2, 1
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: bnez a3, .LBB63_3
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: # %bb.1: # %else
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: andi a2, a2, 2
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: bnez a2, .LBB63_4
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: .LBB63_2: # %else2
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: ret
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: .LBB63_3: # %cond.store
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetvli zero, zero, e16, m2, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a3, v8
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fmv.h.x fa5, a3
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fsh fa5, 0(a0)
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: andi a2, a2, 2
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: beqz a2, .LBB63_2
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: .LBB63_4: # %cond.store1
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vslidedown.vi v8, v8, 1
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a0, v8
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fmv.h.x fa5, a0
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fsh fa5, 0(a1)
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: ret
|
|
call void @llvm.masked.scatter.v2f16.v2p0(<2 x half> %val, <2 x ptr> %ptrs, i32 2, <2 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_v4f16(<4 x half> %val, <4 x ptr> %ptrs, <4 x i1> %m) {
|
|
; RV32-LABEL: mscatter_v4f16:
|
|
; RV32: # %bb.0:
|
|
; RV32-NEXT: vsetivli zero, 4, e16, mf2, ta, ma
|
|
; RV32-NEXT: vsoxei32.v v8, (zero), v9, v0.t
|
|
; RV32-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_v4f16:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: vsetivli zero, 4, e16, mf2, ta, ma
|
|
; RV64V-NEXT: vsoxei64.v v8, (zero), v10, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-ZVFH-LABEL: mscatter_v4f16:
|
|
; RV64ZVE32F-ZVFH: # %bb.0:
|
|
; RV64ZVE32F-ZVFH-NEXT: ld a4, 8(a0)
|
|
; RV64ZVE32F-ZVFH-NEXT: ld a2, 16(a0)
|
|
; RV64ZVE32F-ZVFH-NEXT: ld a1, 24(a0)
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vmv.x.s a3, v0
|
|
; RV64ZVE32F-ZVFH-NEXT: andi a5, a3, 1
|
|
; RV64ZVE32F-ZVFH-NEXT: bnez a5, .LBB64_5
|
|
; RV64ZVE32F-ZVFH-NEXT: # %bb.1: # %else
|
|
; RV64ZVE32F-ZVFH-NEXT: andi a0, a3, 2
|
|
; RV64ZVE32F-ZVFH-NEXT: bnez a0, .LBB64_6
|
|
; RV64ZVE32F-ZVFH-NEXT: .LBB64_2: # %else2
|
|
; RV64ZVE32F-ZVFH-NEXT: andi a0, a3, 4
|
|
; RV64ZVE32F-ZVFH-NEXT: bnez a0, .LBB64_7
|
|
; RV64ZVE32F-ZVFH-NEXT: .LBB64_3: # %else4
|
|
; RV64ZVE32F-ZVFH-NEXT: andi a3, a3, 8
|
|
; RV64ZVE32F-ZVFH-NEXT: bnez a3, .LBB64_8
|
|
; RV64ZVE32F-ZVFH-NEXT: .LBB64_4: # %else6
|
|
; RV64ZVE32F-ZVFH-NEXT: ret
|
|
; RV64ZVE32F-ZVFH-NEXT: .LBB64_5: # %cond.store
|
|
; RV64ZVE32F-ZVFH-NEXT: ld a0, 0(a0)
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vse16.v v8, (a0)
|
|
; RV64ZVE32F-ZVFH-NEXT: andi a0, a3, 2
|
|
; RV64ZVE32F-ZVFH-NEXT: beqz a0, .LBB64_2
|
|
; RV64ZVE32F-ZVFH-NEXT: .LBB64_6: # %cond.store1
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vslidedown.vi v9, v8, 1
|
|
; RV64ZVE32F-ZVFH-NEXT: vse16.v v9, (a4)
|
|
; RV64ZVE32F-ZVFH-NEXT: andi a0, a3, 4
|
|
; RV64ZVE32F-ZVFH-NEXT: beqz a0, .LBB64_3
|
|
; RV64ZVE32F-ZVFH-NEXT: .LBB64_7: # %cond.store3
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vslidedown.vi v9, v8, 2
|
|
; RV64ZVE32F-ZVFH-NEXT: vse16.v v9, (a2)
|
|
; RV64ZVE32F-ZVFH-NEXT: andi a3, a3, 8
|
|
; RV64ZVE32F-ZVFH-NEXT: beqz a3, .LBB64_4
|
|
; RV64ZVE32F-ZVFH-NEXT: .LBB64_8: # %cond.store5
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vslidedown.vi v8, v8, 3
|
|
; RV64ZVE32F-ZVFH-NEXT: vse16.v v8, (a1)
|
|
; RV64ZVE32F-ZVFH-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-ZVFHMIN-LABEL: mscatter_v4f16:
|
|
; RV64ZVE32F-ZVFHMIN: # %bb.0:
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: ld a4, 8(a0)
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: ld a2, 16(a0)
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: ld a1, 24(a0)
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a3, v0
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: andi a5, a3, 1
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: bnez a5, .LBB64_5
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: # %bb.1: # %else
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: andi a0, a3, 2
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: bnez a0, .LBB64_6
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: .LBB64_2: # %else2
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: andi a0, a3, 4
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: bnez a0, .LBB64_7
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: .LBB64_3: # %else4
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: andi a3, a3, 8
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: bnez a3, .LBB64_8
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: .LBB64_4: # %else6
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: ret
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: .LBB64_5: # %cond.store
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: ld a0, 0(a0)
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetvli zero, zero, e16, m2, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a5, v8
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fmv.h.x fa5, a5
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fsh fa5, 0(a0)
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: andi a0, a3, 2
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: beqz a0, .LBB64_2
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: .LBB64_6: # %cond.store1
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vslidedown.vi v9, v8, 1
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a0, v9
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fmv.h.x fa5, a0
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fsh fa5, 0(a4)
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: andi a0, a3, 4
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: beqz a0, .LBB64_3
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: .LBB64_7: # %cond.store3
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vslidedown.vi v9, v8, 2
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a0, v9
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fmv.h.x fa5, a0
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fsh fa5, 0(a2)
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: andi a3, a3, 8
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: beqz a3, .LBB64_4
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: .LBB64_8: # %cond.store5
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vslidedown.vi v8, v8, 3
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a0, v8
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fmv.h.x fa5, a0
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fsh fa5, 0(a1)
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: ret
|
|
call void @llvm.masked.scatter.v4f16.v4p0(<4 x half> %val, <4 x ptr> %ptrs, i32 2, <4 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_truemask_v4f16(<4 x half> %val, <4 x ptr> %ptrs) {
|
|
; RV32-LABEL: mscatter_truemask_v4f16:
|
|
; RV32: # %bb.0:
|
|
; RV32-NEXT: vsetivli zero, 4, e16, mf2, ta, ma
|
|
; RV32-NEXT: vsoxei32.v v8, (zero), v9
|
|
; RV32-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_truemask_v4f16:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: vsetivli zero, 4, e16, mf2, ta, ma
|
|
; RV64V-NEXT: vsoxei64.v v8, (zero), v10
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-ZVFH-LABEL: mscatter_truemask_v4f16:
|
|
; RV64ZVE32F-ZVFH: # %bb.0:
|
|
; RV64ZVE32F-ZVFH-NEXT: ld a1, 0(a0)
|
|
; RV64ZVE32F-ZVFH-NEXT: ld a2, 8(a0)
|
|
; RV64ZVE32F-ZVFH-NEXT: ld a3, 16(a0)
|
|
; RV64ZVE32F-ZVFH-NEXT: ld a0, 24(a0)
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vslidedown.vi v9, v8, 1
|
|
; RV64ZVE32F-ZVFH-NEXT: vse16.v v8, (a1)
|
|
; RV64ZVE32F-ZVFH-NEXT: vse16.v v9, (a2)
|
|
; RV64ZVE32F-ZVFH-NEXT: vslidedown.vi v9, v8, 2
|
|
; RV64ZVE32F-ZVFH-NEXT: vslidedown.vi v8, v8, 3
|
|
; RV64ZVE32F-ZVFH-NEXT: vse16.v v9, (a3)
|
|
; RV64ZVE32F-ZVFH-NEXT: vse16.v v8, (a0)
|
|
; RV64ZVE32F-ZVFH-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-ZVFHMIN-LABEL: mscatter_truemask_v4f16:
|
|
; RV64ZVE32F-ZVFHMIN: # %bb.0:
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: ld a1, 0(a0)
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: ld a2, 8(a0)
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: ld a3, 16(a0)
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: ld a0, 24(a0)
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a4, v8
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vslidedown.vi v9, v8, 1
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fmv.h.x fa5, a4
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a4, v9
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vslidedown.vi v9, v8, 2
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vslidedown.vi v8, v8, 3
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fsh fa5, 0(a1)
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a1, v9
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fmv.h.x fa5, a4
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a4, v8
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fsh fa5, 0(a2)
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fmv.h.x fa5, a1
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fsh fa5, 0(a3)
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fmv.h.x fa5, a4
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fsh fa5, 0(a0)
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: ret
|
|
call void @llvm.masked.scatter.v4f16.v4p0(<4 x half> %val, <4 x ptr> %ptrs, i32 2, <4 x i1> splat (i1 1))
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_falsemask_v4f16(<4 x half> %val, <4 x ptr> %ptrs) {
|
|
; CHECK-LABEL: mscatter_falsemask_v4f16:
|
|
; CHECK: # %bb.0:
|
|
; CHECK-NEXT: ret
|
|
call void @llvm.masked.scatter.v4f16.v4p0(<4 x half> %val, <4 x ptr> %ptrs, i32 2, <4 x i1> zeroinitializer)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_v8f16(<8 x half> %val, <8 x ptr> %ptrs, <8 x i1> %m) {
|
|
; RV32-LABEL: mscatter_v8f16:
|
|
; RV32: # %bb.0:
|
|
; RV32-NEXT: vsetivli zero, 8, e16, m1, ta, ma
|
|
; RV32-NEXT: vsoxei32.v v8, (zero), v10, v0.t
|
|
; RV32-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_v8f16:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: vsetivli zero, 8, e16, m1, ta, ma
|
|
; RV64V-NEXT: vsoxei64.v v8, (zero), v12, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-ZVFH-LABEL: mscatter_v8f16:
|
|
; RV64ZVE32F-ZVFH: # %bb.0:
|
|
; RV64ZVE32F-ZVFH-NEXT: ld a3, 40(a0)
|
|
; RV64ZVE32F-ZVFH-NEXT: ld a2, 48(a0)
|
|
; RV64ZVE32F-ZVFH-NEXT: ld a1, 56(a0)
|
|
; RV64ZVE32F-ZVFH-NEXT: ld t0, 8(a0)
|
|
; RV64ZVE32F-ZVFH-NEXT: ld a7, 16(a0)
|
|
; RV64ZVE32F-ZVFH-NEXT: ld a6, 24(a0)
|
|
; RV64ZVE32F-ZVFH-NEXT: ld a5, 32(a0)
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vmv.x.s a4, v0
|
|
; RV64ZVE32F-ZVFH-NEXT: andi t1, a4, 1
|
|
; RV64ZVE32F-ZVFH-NEXT: bnez t1, .LBB67_9
|
|
; RV64ZVE32F-ZVFH-NEXT: # %bb.1: # %else
|
|
; RV64ZVE32F-ZVFH-NEXT: andi a0, a4, 2
|
|
; RV64ZVE32F-ZVFH-NEXT: bnez a0, .LBB67_10
|
|
; RV64ZVE32F-ZVFH-NEXT: .LBB67_2: # %else2
|
|
; RV64ZVE32F-ZVFH-NEXT: andi a0, a4, 4
|
|
; RV64ZVE32F-ZVFH-NEXT: bnez a0, .LBB67_11
|
|
; RV64ZVE32F-ZVFH-NEXT: .LBB67_3: # %else4
|
|
; RV64ZVE32F-ZVFH-NEXT: andi a0, a4, 8
|
|
; RV64ZVE32F-ZVFH-NEXT: bnez a0, .LBB67_12
|
|
; RV64ZVE32F-ZVFH-NEXT: .LBB67_4: # %else6
|
|
; RV64ZVE32F-ZVFH-NEXT: andi a0, a4, 16
|
|
; RV64ZVE32F-ZVFH-NEXT: bnez a0, .LBB67_13
|
|
; RV64ZVE32F-ZVFH-NEXT: .LBB67_5: # %else8
|
|
; RV64ZVE32F-ZVFH-NEXT: andi a0, a4, 32
|
|
; RV64ZVE32F-ZVFH-NEXT: bnez a0, .LBB67_14
|
|
; RV64ZVE32F-ZVFH-NEXT: .LBB67_6: # %else10
|
|
; RV64ZVE32F-ZVFH-NEXT: andi a0, a4, 64
|
|
; RV64ZVE32F-ZVFH-NEXT: bnez a0, .LBB67_15
|
|
; RV64ZVE32F-ZVFH-NEXT: .LBB67_7: # %else12
|
|
; RV64ZVE32F-ZVFH-NEXT: andi a0, a4, -128
|
|
; RV64ZVE32F-ZVFH-NEXT: bnez a0, .LBB67_16
|
|
; RV64ZVE32F-ZVFH-NEXT: .LBB67_8: # %else14
|
|
; RV64ZVE32F-ZVFH-NEXT: ret
|
|
; RV64ZVE32F-ZVFH-NEXT: .LBB67_9: # %cond.store
|
|
; RV64ZVE32F-ZVFH-NEXT: ld a0, 0(a0)
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vse16.v v8, (a0)
|
|
; RV64ZVE32F-ZVFH-NEXT: andi a0, a4, 2
|
|
; RV64ZVE32F-ZVFH-NEXT: beqz a0, .LBB67_2
|
|
; RV64ZVE32F-ZVFH-NEXT: .LBB67_10: # %cond.store1
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vslidedown.vi v9, v8, 1
|
|
; RV64ZVE32F-ZVFH-NEXT: vse16.v v9, (t0)
|
|
; RV64ZVE32F-ZVFH-NEXT: andi a0, a4, 4
|
|
; RV64ZVE32F-ZVFH-NEXT: beqz a0, .LBB67_3
|
|
; RV64ZVE32F-ZVFH-NEXT: .LBB67_11: # %cond.store3
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vslidedown.vi v9, v8, 2
|
|
; RV64ZVE32F-ZVFH-NEXT: vse16.v v9, (a7)
|
|
; RV64ZVE32F-ZVFH-NEXT: andi a0, a4, 8
|
|
; RV64ZVE32F-ZVFH-NEXT: beqz a0, .LBB67_4
|
|
; RV64ZVE32F-ZVFH-NEXT: .LBB67_12: # %cond.store5
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vslidedown.vi v9, v8, 3
|
|
; RV64ZVE32F-ZVFH-NEXT: vse16.v v9, (a6)
|
|
; RV64ZVE32F-ZVFH-NEXT: andi a0, a4, 16
|
|
; RV64ZVE32F-ZVFH-NEXT: beqz a0, .LBB67_5
|
|
; RV64ZVE32F-ZVFH-NEXT: .LBB67_13: # %cond.store7
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vslidedown.vi v9, v8, 4
|
|
; RV64ZVE32F-ZVFH-NEXT: vse16.v v9, (a5)
|
|
; RV64ZVE32F-ZVFH-NEXT: andi a0, a4, 32
|
|
; RV64ZVE32F-ZVFH-NEXT: beqz a0, .LBB67_6
|
|
; RV64ZVE32F-ZVFH-NEXT: .LBB67_14: # %cond.store9
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vslidedown.vi v9, v8, 5
|
|
; RV64ZVE32F-ZVFH-NEXT: vse16.v v9, (a3)
|
|
; RV64ZVE32F-ZVFH-NEXT: andi a0, a4, 64
|
|
; RV64ZVE32F-ZVFH-NEXT: beqz a0, .LBB67_7
|
|
; RV64ZVE32F-ZVFH-NEXT: .LBB67_15: # %cond.store11
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vslidedown.vi v9, v8, 6
|
|
; RV64ZVE32F-ZVFH-NEXT: vse16.v v9, (a2)
|
|
; RV64ZVE32F-ZVFH-NEXT: andi a0, a4, -128
|
|
; RV64ZVE32F-ZVFH-NEXT: beqz a0, .LBB67_8
|
|
; RV64ZVE32F-ZVFH-NEXT: .LBB67_16: # %cond.store13
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vslidedown.vi v8, v8, 7
|
|
; RV64ZVE32F-ZVFH-NEXT: vse16.v v8, (a1)
|
|
; RV64ZVE32F-ZVFH-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-ZVFHMIN-LABEL: mscatter_v8f16:
|
|
; RV64ZVE32F-ZVFHMIN: # %bb.0:
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: ld a3, 40(a0)
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: ld a2, 48(a0)
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: ld a1, 56(a0)
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: ld t0, 8(a0)
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: ld a7, 16(a0)
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: ld a6, 24(a0)
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: ld a5, 32(a0)
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a4, v0
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: andi t1, a4, 1
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: bnez t1, .LBB67_9
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: # %bb.1: # %else
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: andi a0, a4, 2
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: bnez a0, .LBB67_10
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: .LBB67_2: # %else2
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: andi a0, a4, 4
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: bnez a0, .LBB67_11
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: .LBB67_3: # %else4
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: andi a0, a4, 8
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: bnez a0, .LBB67_12
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: .LBB67_4: # %else6
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: andi a0, a4, 16
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: bnez a0, .LBB67_13
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: .LBB67_5: # %else8
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: andi a0, a4, 32
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: bnez a0, .LBB67_14
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: .LBB67_6: # %else10
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: andi a0, a4, 64
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: bnez a0, .LBB67_15
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: .LBB67_7: # %else12
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: andi a0, a4, -128
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: bnez a0, .LBB67_16
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: .LBB67_8: # %else14
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: ret
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: .LBB67_9: # %cond.store
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: ld a0, 0(a0)
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetvli zero, zero, e16, m2, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s t1, v8
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fmv.h.x fa5, t1
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fsh fa5, 0(a0)
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: andi a0, a4, 2
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: beqz a0, .LBB67_2
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: .LBB67_10: # %cond.store1
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vslidedown.vi v9, v8, 1
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a0, v9
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fmv.h.x fa5, a0
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fsh fa5, 0(t0)
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: andi a0, a4, 4
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: beqz a0, .LBB67_3
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: .LBB67_11: # %cond.store3
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vslidedown.vi v9, v8, 2
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a0, v9
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fmv.h.x fa5, a0
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fsh fa5, 0(a7)
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: andi a0, a4, 8
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: beqz a0, .LBB67_4
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: .LBB67_12: # %cond.store5
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vslidedown.vi v9, v8, 3
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a0, v9
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fmv.h.x fa5, a0
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fsh fa5, 0(a6)
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: andi a0, a4, 16
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: beqz a0, .LBB67_5
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: .LBB67_13: # %cond.store7
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vslidedown.vi v9, v8, 4
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a0, v9
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fmv.h.x fa5, a0
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fsh fa5, 0(a5)
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: andi a0, a4, 32
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: beqz a0, .LBB67_6
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: .LBB67_14: # %cond.store9
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vslidedown.vi v9, v8, 5
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a0, v9
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fmv.h.x fa5, a0
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fsh fa5, 0(a3)
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: andi a0, a4, 64
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: beqz a0, .LBB67_7
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: .LBB67_15: # %cond.store11
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vslidedown.vi v9, v8, 6
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a0, v9
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fmv.h.x fa5, a0
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fsh fa5, 0(a2)
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: andi a0, a4, -128
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: beqz a0, .LBB67_8
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: .LBB67_16: # %cond.store13
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vslidedown.vi v8, v8, 7
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a0, v8
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fmv.h.x fa5, a0
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fsh fa5, 0(a1)
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: ret
|
|
call void @llvm.masked.scatter.v8f16.v8p0(<8 x half> %val, <8 x ptr> %ptrs, i32 2, <8 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_baseidx_v8i8_v8f16(<8 x half> %val, ptr %base, <8 x i8> %idxs, <8 x i1> %m) {
|
|
; RV32-LABEL: mscatter_baseidx_v8i8_v8f16:
|
|
; RV32: # %bb.0:
|
|
; RV32-NEXT: vsetivli zero, 8, e32, m2, ta, ma
|
|
; RV32-NEXT: vsext.vf4 v10, v9
|
|
; RV32-NEXT: vadd.vv v10, v10, v10
|
|
; RV32-NEXT: vsetvli zero, zero, e16, m1, ta, ma
|
|
; RV32-NEXT: vsoxei32.v v8, (a0), v10, v0.t
|
|
; RV32-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_baseidx_v8i8_v8f16:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: vsetivli zero, 8, e64, m4, ta, ma
|
|
; RV64V-NEXT: vsext.vf8 v12, v9
|
|
; RV64V-NEXT: vadd.vv v12, v12, v12
|
|
; RV64V-NEXT: vsetvli zero, zero, e16, m1, ta, ma
|
|
; RV64V-NEXT: vsoxei64.v v8, (a0), v12, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-ZVFH-LABEL: mscatter_baseidx_v8i8_v8f16:
|
|
; RV64ZVE32F-ZVFH: # %bb.0:
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vmv.x.s a1, v0
|
|
; RV64ZVE32F-ZVFH-NEXT: andi a2, a1, 1
|
|
; RV64ZVE32F-ZVFH-NEXT: beqz a2, .LBB68_2
|
|
; RV64ZVE32F-ZVFH-NEXT: # %bb.1: # %cond.store
|
|
; RV64ZVE32F-ZVFH-NEXT: vmv.x.s a2, v9
|
|
; RV64ZVE32F-ZVFH-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-ZVFH-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vse16.v v8, (a2)
|
|
; RV64ZVE32F-ZVFH-NEXT: .LBB68_2: # %else
|
|
; RV64ZVE32F-ZVFH-NEXT: andi a2, a1, 2
|
|
; RV64ZVE32F-ZVFH-NEXT: beqz a2, .LBB68_4
|
|
; RV64ZVE32F-ZVFH-NEXT: # %bb.3: # %cond.store1
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vslidedown.vi v10, v9, 1
|
|
; RV64ZVE32F-ZVFH-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-ZVFH-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-ZVFH-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vslidedown.vi v10, v8, 1
|
|
; RV64ZVE32F-ZVFH-NEXT: vse16.v v10, (a2)
|
|
; RV64ZVE32F-ZVFH-NEXT: .LBB68_4: # %else2
|
|
; RV64ZVE32F-ZVFH-NEXT: andi a2, a1, 4
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetivli zero, 2, e8, mf4, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vslidedown.vi v10, v9, 2
|
|
; RV64ZVE32F-ZVFH-NEXT: beqz a2, .LBB68_6
|
|
; RV64ZVE32F-ZVFH-NEXT: # %bb.5: # %cond.store3
|
|
; RV64ZVE32F-ZVFH-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-ZVFH-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-ZVFH-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vslidedown.vi v11, v8, 2
|
|
; RV64ZVE32F-ZVFH-NEXT: vse16.v v11, (a2)
|
|
; RV64ZVE32F-ZVFH-NEXT: .LBB68_6: # %else4
|
|
; RV64ZVE32F-ZVFH-NEXT: andi a2, a1, 8
|
|
; RV64ZVE32F-ZVFH-NEXT: beqz a2, .LBB68_8
|
|
; RV64ZVE32F-ZVFH-NEXT: # %bb.7: # %cond.store5
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vslidedown.vi v10, v10, 1
|
|
; RV64ZVE32F-ZVFH-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-ZVFH-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-ZVFH-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vslidedown.vi v10, v8, 3
|
|
; RV64ZVE32F-ZVFH-NEXT: vse16.v v10, (a2)
|
|
; RV64ZVE32F-ZVFH-NEXT: .LBB68_8: # %else6
|
|
; RV64ZVE32F-ZVFH-NEXT: andi a2, a1, 16
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetivli zero, 4, e8, mf2, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vslidedown.vi v9, v9, 4
|
|
; RV64ZVE32F-ZVFH-NEXT: beqz a2, .LBB68_10
|
|
; RV64ZVE32F-ZVFH-NEXT: # %bb.9: # %cond.store7
|
|
; RV64ZVE32F-ZVFH-NEXT: vmv.x.s a2, v9
|
|
; RV64ZVE32F-ZVFH-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-ZVFH-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetvli zero, zero, e16, m1, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vslidedown.vi v10, v8, 4
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vse16.v v10, (a2)
|
|
; RV64ZVE32F-ZVFH-NEXT: .LBB68_10: # %else8
|
|
; RV64ZVE32F-ZVFH-NEXT: andi a2, a1, 32
|
|
; RV64ZVE32F-ZVFH-NEXT: beqz a2, .LBB68_12
|
|
; RV64ZVE32F-ZVFH-NEXT: # %bb.11: # %cond.store9
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vslidedown.vi v10, v9, 1
|
|
; RV64ZVE32F-ZVFH-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-ZVFH-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-ZVFH-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vslidedown.vi v10, v8, 5
|
|
; RV64ZVE32F-ZVFH-NEXT: vse16.v v10, (a2)
|
|
; RV64ZVE32F-ZVFH-NEXT: .LBB68_12: # %else10
|
|
; RV64ZVE32F-ZVFH-NEXT: andi a2, a1, 64
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetivli zero, 2, e8, mf4, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vslidedown.vi v9, v9, 2
|
|
; RV64ZVE32F-ZVFH-NEXT: bnez a2, .LBB68_15
|
|
; RV64ZVE32F-ZVFH-NEXT: # %bb.13: # %else12
|
|
; RV64ZVE32F-ZVFH-NEXT: andi a1, a1, -128
|
|
; RV64ZVE32F-ZVFH-NEXT: bnez a1, .LBB68_16
|
|
; RV64ZVE32F-ZVFH-NEXT: .LBB68_14: # %else14
|
|
; RV64ZVE32F-ZVFH-NEXT: ret
|
|
; RV64ZVE32F-ZVFH-NEXT: .LBB68_15: # %cond.store11
|
|
; RV64ZVE32F-ZVFH-NEXT: vmv.x.s a2, v9
|
|
; RV64ZVE32F-ZVFH-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-ZVFH-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vslidedown.vi v10, v8, 6
|
|
; RV64ZVE32F-ZVFH-NEXT: vse16.v v10, (a2)
|
|
; RV64ZVE32F-ZVFH-NEXT: andi a1, a1, -128
|
|
; RV64ZVE32F-ZVFH-NEXT: beqz a1, .LBB68_14
|
|
; RV64ZVE32F-ZVFH-NEXT: .LBB68_16: # %cond.store13
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vslidedown.vi v9, v9, 1
|
|
; RV64ZVE32F-ZVFH-NEXT: vmv.x.s a1, v9
|
|
; RV64ZVE32F-ZVFH-NEXT: slli a1, a1, 1
|
|
; RV64ZVE32F-ZVFH-NEXT: add a0, a0, a1
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vslidedown.vi v8, v8, 7
|
|
; RV64ZVE32F-ZVFH-NEXT: vse16.v v8, (a0)
|
|
; RV64ZVE32F-ZVFH-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-ZVFHMIN-LABEL: mscatter_baseidx_v8i8_v8f16:
|
|
; RV64ZVE32F-ZVFHMIN: # %bb.0:
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a1, v0
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: andi a2, a1, 1
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: beqz a2, .LBB68_2
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: # %bb.1: # %cond.store
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a2, v9
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetvli zero, zero, e16, m2, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a3, v8
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fmv.h.x fa5, a3
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fsh fa5, 0(a2)
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: .LBB68_2: # %else
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: andi a2, a1, 2
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: beqz a2, .LBB68_4
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: # %bb.3: # %cond.store1
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vslidedown.vi v10, v9, 1
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vslidedown.vi v10, v8, 1
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a3, v10
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fmv.h.x fa5, a3
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fsh fa5, 0(a2)
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: .LBB68_4: # %else2
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: andi a2, a1, 4
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetivli zero, 2, e8, mf4, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vslidedown.vi v10, v9, 2
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: beqz a2, .LBB68_6
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: # %bb.5: # %cond.store3
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vslidedown.vi v11, v8, 2
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a3, v11
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fmv.h.x fa5, a3
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fsh fa5, 0(a2)
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: .LBB68_6: # %else4
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: andi a2, a1, 8
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: beqz a2, .LBB68_8
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: # %bb.7: # %cond.store5
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vslidedown.vi v10, v10, 1
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vslidedown.vi v10, v8, 3
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a3, v10
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fmv.h.x fa5, a3
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fsh fa5, 0(a2)
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: .LBB68_8: # %else6
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: andi a2, a1, 16
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetivli zero, 4, e8, mf2, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vslidedown.vi v9, v9, 4
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: beqz a2, .LBB68_10
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: # %bb.9: # %cond.store7
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a2, v9
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetvli zero, zero, e16, m1, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vslidedown.vi v10, v8, 4
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a3, v10
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fmv.h.x fa5, a3
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fsh fa5, 0(a2)
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: .LBB68_10: # %else8
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: andi a2, a1, 32
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: beqz a2, .LBB68_12
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: # %bb.11: # %cond.store9
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vslidedown.vi v10, v9, 1
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vslidedown.vi v10, v8, 5
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a3, v10
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fmv.h.x fa5, a3
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fsh fa5, 0(a2)
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: .LBB68_12: # %else10
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: andi a2, a1, 64
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetivli zero, 2, e8, mf4, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vslidedown.vi v9, v9, 2
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: bnez a2, .LBB68_15
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: # %bb.13: # %else12
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: andi a1, a1, -128
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: bnez a1, .LBB68_16
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: .LBB68_14: # %else14
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: ret
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: .LBB68_15: # %cond.store11
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a2, v9
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vslidedown.vi v10, v8, 6
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a3, v10
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fmv.h.x fa5, a3
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fsh fa5, 0(a2)
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: andi a1, a1, -128
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: beqz a1, .LBB68_14
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: .LBB68_16: # %cond.store13
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vslidedown.vi v9, v9, 1
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vslidedown.vi v8, v8, 7
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetvli zero, zero, e8, mf2, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a1, v9
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: slli a1, a1, 1
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: add a0, a0, a1
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetvli zero, zero, e16, m1, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a1, v8
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fmv.h.x fa5, a1
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fsh fa5, 0(a0)
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: ret
|
|
%ptrs = getelementptr inbounds half, ptr %base, <8 x i8> %idxs
|
|
call void @llvm.masked.scatter.v8f16.v8p0(<8 x half> %val, <8 x ptr> %ptrs, i32 2, <8 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_baseidx_sext_v8i8_v8f16(<8 x half> %val, ptr %base, <8 x i8> %idxs, <8 x i1> %m) {
|
|
; RV32-LABEL: mscatter_baseidx_sext_v8i8_v8f16:
|
|
; RV32: # %bb.0:
|
|
; RV32-NEXT: vsetivli zero, 8, e32, m2, ta, ma
|
|
; RV32-NEXT: vsext.vf4 v10, v9
|
|
; RV32-NEXT: vadd.vv v10, v10, v10
|
|
; RV32-NEXT: vsetvli zero, zero, e16, m1, ta, ma
|
|
; RV32-NEXT: vsoxei32.v v8, (a0), v10, v0.t
|
|
; RV32-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_baseidx_sext_v8i8_v8f16:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: vsetivli zero, 8, e64, m4, ta, ma
|
|
; RV64V-NEXT: vsext.vf8 v12, v9
|
|
; RV64V-NEXT: vadd.vv v12, v12, v12
|
|
; RV64V-NEXT: vsetvli zero, zero, e16, m1, ta, ma
|
|
; RV64V-NEXT: vsoxei64.v v8, (a0), v12, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-ZVFH-LABEL: mscatter_baseidx_sext_v8i8_v8f16:
|
|
; RV64ZVE32F-ZVFH: # %bb.0:
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vmv.x.s a1, v0
|
|
; RV64ZVE32F-ZVFH-NEXT: andi a2, a1, 1
|
|
; RV64ZVE32F-ZVFH-NEXT: beqz a2, .LBB69_2
|
|
; RV64ZVE32F-ZVFH-NEXT: # %bb.1: # %cond.store
|
|
; RV64ZVE32F-ZVFH-NEXT: vmv.x.s a2, v9
|
|
; RV64ZVE32F-ZVFH-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-ZVFH-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vse16.v v8, (a2)
|
|
; RV64ZVE32F-ZVFH-NEXT: .LBB69_2: # %else
|
|
; RV64ZVE32F-ZVFH-NEXT: andi a2, a1, 2
|
|
; RV64ZVE32F-ZVFH-NEXT: beqz a2, .LBB69_4
|
|
; RV64ZVE32F-ZVFH-NEXT: # %bb.3: # %cond.store1
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vslidedown.vi v10, v9, 1
|
|
; RV64ZVE32F-ZVFH-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-ZVFH-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-ZVFH-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vslidedown.vi v10, v8, 1
|
|
; RV64ZVE32F-ZVFH-NEXT: vse16.v v10, (a2)
|
|
; RV64ZVE32F-ZVFH-NEXT: .LBB69_4: # %else2
|
|
; RV64ZVE32F-ZVFH-NEXT: andi a2, a1, 4
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetivli zero, 2, e8, mf4, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vslidedown.vi v10, v9, 2
|
|
; RV64ZVE32F-ZVFH-NEXT: beqz a2, .LBB69_6
|
|
; RV64ZVE32F-ZVFH-NEXT: # %bb.5: # %cond.store3
|
|
; RV64ZVE32F-ZVFH-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-ZVFH-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-ZVFH-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vslidedown.vi v11, v8, 2
|
|
; RV64ZVE32F-ZVFH-NEXT: vse16.v v11, (a2)
|
|
; RV64ZVE32F-ZVFH-NEXT: .LBB69_6: # %else4
|
|
; RV64ZVE32F-ZVFH-NEXT: andi a2, a1, 8
|
|
; RV64ZVE32F-ZVFH-NEXT: beqz a2, .LBB69_8
|
|
; RV64ZVE32F-ZVFH-NEXT: # %bb.7: # %cond.store5
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vslidedown.vi v10, v10, 1
|
|
; RV64ZVE32F-ZVFH-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-ZVFH-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-ZVFH-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vslidedown.vi v10, v8, 3
|
|
; RV64ZVE32F-ZVFH-NEXT: vse16.v v10, (a2)
|
|
; RV64ZVE32F-ZVFH-NEXT: .LBB69_8: # %else6
|
|
; RV64ZVE32F-ZVFH-NEXT: andi a2, a1, 16
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetivli zero, 4, e8, mf2, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vslidedown.vi v9, v9, 4
|
|
; RV64ZVE32F-ZVFH-NEXT: beqz a2, .LBB69_10
|
|
; RV64ZVE32F-ZVFH-NEXT: # %bb.9: # %cond.store7
|
|
; RV64ZVE32F-ZVFH-NEXT: vmv.x.s a2, v9
|
|
; RV64ZVE32F-ZVFH-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-ZVFH-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetvli zero, zero, e16, m1, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vslidedown.vi v10, v8, 4
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vse16.v v10, (a2)
|
|
; RV64ZVE32F-ZVFH-NEXT: .LBB69_10: # %else8
|
|
; RV64ZVE32F-ZVFH-NEXT: andi a2, a1, 32
|
|
; RV64ZVE32F-ZVFH-NEXT: beqz a2, .LBB69_12
|
|
; RV64ZVE32F-ZVFH-NEXT: # %bb.11: # %cond.store9
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vslidedown.vi v10, v9, 1
|
|
; RV64ZVE32F-ZVFH-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-ZVFH-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-ZVFH-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vslidedown.vi v10, v8, 5
|
|
; RV64ZVE32F-ZVFH-NEXT: vse16.v v10, (a2)
|
|
; RV64ZVE32F-ZVFH-NEXT: .LBB69_12: # %else10
|
|
; RV64ZVE32F-ZVFH-NEXT: andi a2, a1, 64
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetivli zero, 2, e8, mf4, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vslidedown.vi v9, v9, 2
|
|
; RV64ZVE32F-ZVFH-NEXT: bnez a2, .LBB69_15
|
|
; RV64ZVE32F-ZVFH-NEXT: # %bb.13: # %else12
|
|
; RV64ZVE32F-ZVFH-NEXT: andi a1, a1, -128
|
|
; RV64ZVE32F-ZVFH-NEXT: bnez a1, .LBB69_16
|
|
; RV64ZVE32F-ZVFH-NEXT: .LBB69_14: # %else14
|
|
; RV64ZVE32F-ZVFH-NEXT: ret
|
|
; RV64ZVE32F-ZVFH-NEXT: .LBB69_15: # %cond.store11
|
|
; RV64ZVE32F-ZVFH-NEXT: vmv.x.s a2, v9
|
|
; RV64ZVE32F-ZVFH-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-ZVFH-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vslidedown.vi v10, v8, 6
|
|
; RV64ZVE32F-ZVFH-NEXT: vse16.v v10, (a2)
|
|
; RV64ZVE32F-ZVFH-NEXT: andi a1, a1, -128
|
|
; RV64ZVE32F-ZVFH-NEXT: beqz a1, .LBB69_14
|
|
; RV64ZVE32F-ZVFH-NEXT: .LBB69_16: # %cond.store13
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vslidedown.vi v9, v9, 1
|
|
; RV64ZVE32F-ZVFH-NEXT: vmv.x.s a1, v9
|
|
; RV64ZVE32F-ZVFH-NEXT: slli a1, a1, 1
|
|
; RV64ZVE32F-ZVFH-NEXT: add a0, a0, a1
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vslidedown.vi v8, v8, 7
|
|
; RV64ZVE32F-ZVFH-NEXT: vse16.v v8, (a0)
|
|
; RV64ZVE32F-ZVFH-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-ZVFHMIN-LABEL: mscatter_baseidx_sext_v8i8_v8f16:
|
|
; RV64ZVE32F-ZVFHMIN: # %bb.0:
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a1, v0
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: andi a2, a1, 1
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: beqz a2, .LBB69_2
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: # %bb.1: # %cond.store
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a2, v9
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetvli zero, zero, e16, m2, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a3, v8
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fmv.h.x fa5, a3
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fsh fa5, 0(a2)
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: .LBB69_2: # %else
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: andi a2, a1, 2
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: beqz a2, .LBB69_4
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: # %bb.3: # %cond.store1
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vslidedown.vi v10, v9, 1
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vslidedown.vi v10, v8, 1
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a3, v10
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fmv.h.x fa5, a3
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fsh fa5, 0(a2)
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: .LBB69_4: # %else2
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: andi a2, a1, 4
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetivli zero, 2, e8, mf4, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vslidedown.vi v10, v9, 2
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: beqz a2, .LBB69_6
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: # %bb.5: # %cond.store3
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vslidedown.vi v11, v8, 2
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a3, v11
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fmv.h.x fa5, a3
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fsh fa5, 0(a2)
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: .LBB69_6: # %else4
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: andi a2, a1, 8
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: beqz a2, .LBB69_8
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: # %bb.7: # %cond.store5
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vslidedown.vi v10, v10, 1
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vslidedown.vi v10, v8, 3
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a3, v10
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fmv.h.x fa5, a3
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fsh fa5, 0(a2)
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: .LBB69_8: # %else6
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: andi a2, a1, 16
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetivli zero, 4, e8, mf2, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vslidedown.vi v9, v9, 4
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: beqz a2, .LBB69_10
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: # %bb.9: # %cond.store7
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a2, v9
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetvli zero, zero, e16, m1, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vslidedown.vi v10, v8, 4
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a3, v10
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fmv.h.x fa5, a3
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fsh fa5, 0(a2)
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: .LBB69_10: # %else8
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: andi a2, a1, 32
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: beqz a2, .LBB69_12
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: # %bb.11: # %cond.store9
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vslidedown.vi v10, v9, 1
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vslidedown.vi v10, v8, 5
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a3, v10
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fmv.h.x fa5, a3
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fsh fa5, 0(a2)
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: .LBB69_12: # %else10
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: andi a2, a1, 64
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetivli zero, 2, e8, mf4, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vslidedown.vi v9, v9, 2
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: bnez a2, .LBB69_15
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: # %bb.13: # %else12
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: andi a1, a1, -128
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: bnez a1, .LBB69_16
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: .LBB69_14: # %else14
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: ret
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: .LBB69_15: # %cond.store11
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a2, v9
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vslidedown.vi v10, v8, 6
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a3, v10
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fmv.h.x fa5, a3
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fsh fa5, 0(a2)
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: andi a1, a1, -128
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: beqz a1, .LBB69_14
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: .LBB69_16: # %cond.store13
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vslidedown.vi v9, v9, 1
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vslidedown.vi v8, v8, 7
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetvli zero, zero, e8, mf2, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a1, v9
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: slli a1, a1, 1
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: add a0, a0, a1
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetvli zero, zero, e16, m1, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a1, v8
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fmv.h.x fa5, a1
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fsh fa5, 0(a0)
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: ret
|
|
%eidxs = sext <8 x i8> %idxs to <8 x i16>
|
|
%ptrs = getelementptr inbounds half, ptr %base, <8 x i16> %eidxs
|
|
call void @llvm.masked.scatter.v8f16.v8p0(<8 x half> %val, <8 x ptr> %ptrs, i32 2, <8 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_baseidx_zext_v8i8_v8f16(<8 x half> %val, ptr %base, <8 x i8> %idxs, <8 x i1> %m) {
|
|
; RV32-LABEL: mscatter_baseidx_zext_v8i8_v8f16:
|
|
; RV32: # %bb.0:
|
|
; RV32-NEXT: vsetivli zero, 8, e8, mf2, ta, ma
|
|
; RV32-NEXT: vwaddu.vv v10, v9, v9
|
|
; RV32-NEXT: vsetvli zero, zero, e16, m1, ta, ma
|
|
; RV32-NEXT: vsoxei16.v v8, (a0), v10, v0.t
|
|
; RV32-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_baseidx_zext_v8i8_v8f16:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: vsetivli zero, 8, e8, mf2, ta, ma
|
|
; RV64V-NEXT: vwaddu.vv v10, v9, v9
|
|
; RV64V-NEXT: vsetvli zero, zero, e16, m1, ta, ma
|
|
; RV64V-NEXT: vsoxei16.v v8, (a0), v10, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-ZVFH-LABEL: mscatter_baseidx_zext_v8i8_v8f16:
|
|
; RV64ZVE32F-ZVFH: # %bb.0:
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vmv.x.s a1, v0
|
|
; RV64ZVE32F-ZVFH-NEXT: andi a2, a1, 1
|
|
; RV64ZVE32F-ZVFH-NEXT: beqz a2, .LBB70_2
|
|
; RV64ZVE32F-ZVFH-NEXT: # %bb.1: # %cond.store
|
|
; RV64ZVE32F-ZVFH-NEXT: vmv.x.s a2, v9
|
|
; RV64ZVE32F-ZVFH-NEXT: zext.b a2, a2
|
|
; RV64ZVE32F-ZVFH-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-ZVFH-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vse16.v v8, (a2)
|
|
; RV64ZVE32F-ZVFH-NEXT: .LBB70_2: # %else
|
|
; RV64ZVE32F-ZVFH-NEXT: andi a2, a1, 2
|
|
; RV64ZVE32F-ZVFH-NEXT: beqz a2, .LBB70_4
|
|
; RV64ZVE32F-ZVFH-NEXT: # %bb.3: # %cond.store1
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vslidedown.vi v10, v9, 1
|
|
; RV64ZVE32F-ZVFH-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-ZVFH-NEXT: zext.b a2, a2
|
|
; RV64ZVE32F-ZVFH-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-ZVFH-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vslidedown.vi v10, v8, 1
|
|
; RV64ZVE32F-ZVFH-NEXT: vse16.v v10, (a2)
|
|
; RV64ZVE32F-ZVFH-NEXT: .LBB70_4: # %else2
|
|
; RV64ZVE32F-ZVFH-NEXT: andi a2, a1, 4
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetivli zero, 2, e8, mf4, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vslidedown.vi v10, v9, 2
|
|
; RV64ZVE32F-ZVFH-NEXT: beqz a2, .LBB70_6
|
|
; RV64ZVE32F-ZVFH-NEXT: # %bb.5: # %cond.store3
|
|
; RV64ZVE32F-ZVFH-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-ZVFH-NEXT: zext.b a2, a2
|
|
; RV64ZVE32F-ZVFH-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-ZVFH-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vslidedown.vi v11, v8, 2
|
|
; RV64ZVE32F-ZVFH-NEXT: vse16.v v11, (a2)
|
|
; RV64ZVE32F-ZVFH-NEXT: .LBB70_6: # %else4
|
|
; RV64ZVE32F-ZVFH-NEXT: andi a2, a1, 8
|
|
; RV64ZVE32F-ZVFH-NEXT: beqz a2, .LBB70_8
|
|
; RV64ZVE32F-ZVFH-NEXT: # %bb.7: # %cond.store5
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vslidedown.vi v10, v10, 1
|
|
; RV64ZVE32F-ZVFH-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-ZVFH-NEXT: zext.b a2, a2
|
|
; RV64ZVE32F-ZVFH-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-ZVFH-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vslidedown.vi v10, v8, 3
|
|
; RV64ZVE32F-ZVFH-NEXT: vse16.v v10, (a2)
|
|
; RV64ZVE32F-ZVFH-NEXT: .LBB70_8: # %else6
|
|
; RV64ZVE32F-ZVFH-NEXT: andi a2, a1, 16
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetivli zero, 4, e8, mf2, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vslidedown.vi v9, v9, 4
|
|
; RV64ZVE32F-ZVFH-NEXT: beqz a2, .LBB70_10
|
|
; RV64ZVE32F-ZVFH-NEXT: # %bb.9: # %cond.store7
|
|
; RV64ZVE32F-ZVFH-NEXT: vmv.x.s a2, v9
|
|
; RV64ZVE32F-ZVFH-NEXT: zext.b a2, a2
|
|
; RV64ZVE32F-ZVFH-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-ZVFH-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetvli zero, zero, e16, m1, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vslidedown.vi v10, v8, 4
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vse16.v v10, (a2)
|
|
; RV64ZVE32F-ZVFH-NEXT: .LBB70_10: # %else8
|
|
; RV64ZVE32F-ZVFH-NEXT: andi a2, a1, 32
|
|
; RV64ZVE32F-ZVFH-NEXT: beqz a2, .LBB70_12
|
|
; RV64ZVE32F-ZVFH-NEXT: # %bb.11: # %cond.store9
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vslidedown.vi v10, v9, 1
|
|
; RV64ZVE32F-ZVFH-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-ZVFH-NEXT: zext.b a2, a2
|
|
; RV64ZVE32F-ZVFH-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-ZVFH-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vslidedown.vi v10, v8, 5
|
|
; RV64ZVE32F-ZVFH-NEXT: vse16.v v10, (a2)
|
|
; RV64ZVE32F-ZVFH-NEXT: .LBB70_12: # %else10
|
|
; RV64ZVE32F-ZVFH-NEXT: andi a2, a1, 64
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetivli zero, 2, e8, mf4, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vslidedown.vi v9, v9, 2
|
|
; RV64ZVE32F-ZVFH-NEXT: bnez a2, .LBB70_15
|
|
; RV64ZVE32F-ZVFH-NEXT: # %bb.13: # %else12
|
|
; RV64ZVE32F-ZVFH-NEXT: andi a1, a1, -128
|
|
; RV64ZVE32F-ZVFH-NEXT: bnez a1, .LBB70_16
|
|
; RV64ZVE32F-ZVFH-NEXT: .LBB70_14: # %else14
|
|
; RV64ZVE32F-ZVFH-NEXT: ret
|
|
; RV64ZVE32F-ZVFH-NEXT: .LBB70_15: # %cond.store11
|
|
; RV64ZVE32F-ZVFH-NEXT: vmv.x.s a2, v9
|
|
; RV64ZVE32F-ZVFH-NEXT: zext.b a2, a2
|
|
; RV64ZVE32F-ZVFH-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-ZVFH-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vslidedown.vi v10, v8, 6
|
|
; RV64ZVE32F-ZVFH-NEXT: vse16.v v10, (a2)
|
|
; RV64ZVE32F-ZVFH-NEXT: andi a1, a1, -128
|
|
; RV64ZVE32F-ZVFH-NEXT: beqz a1, .LBB70_14
|
|
; RV64ZVE32F-ZVFH-NEXT: .LBB70_16: # %cond.store13
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vslidedown.vi v9, v9, 1
|
|
; RV64ZVE32F-ZVFH-NEXT: vmv.x.s a1, v9
|
|
; RV64ZVE32F-ZVFH-NEXT: zext.b a1, a1
|
|
; RV64ZVE32F-ZVFH-NEXT: slli a1, a1, 1
|
|
; RV64ZVE32F-ZVFH-NEXT: add a0, a0, a1
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vslidedown.vi v8, v8, 7
|
|
; RV64ZVE32F-ZVFH-NEXT: vse16.v v8, (a0)
|
|
; RV64ZVE32F-ZVFH-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-ZVFHMIN-LABEL: mscatter_baseidx_zext_v8i8_v8f16:
|
|
; RV64ZVE32F-ZVFHMIN: # %bb.0:
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a1, v0
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: andi a2, a1, 1
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: beqz a2, .LBB70_2
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: # %bb.1: # %cond.store
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a2, v9
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetvli zero, zero, e16, m2, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a3, v8
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: zext.b a2, a2
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fmv.h.x fa5, a3
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fsh fa5, 0(a2)
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: .LBB70_2: # %else
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: andi a2, a1, 2
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: beqz a2, .LBB70_4
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: # %bb.3: # %cond.store1
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vslidedown.vi v10, v9, 1
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vslidedown.vi v10, v8, 1
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a3, v10
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: zext.b a2, a2
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fmv.h.x fa5, a3
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fsh fa5, 0(a2)
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: .LBB70_4: # %else2
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: andi a2, a1, 4
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetivli zero, 2, e8, mf4, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vslidedown.vi v10, v9, 2
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: beqz a2, .LBB70_6
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: # %bb.5: # %cond.store3
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vslidedown.vi v11, v8, 2
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: zext.b a2, a2
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a3, v11
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fmv.h.x fa5, a3
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fsh fa5, 0(a2)
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: .LBB70_6: # %else4
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: andi a2, a1, 8
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: beqz a2, .LBB70_8
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: # %bb.7: # %cond.store5
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vslidedown.vi v10, v10, 1
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vslidedown.vi v10, v8, 3
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a3, v10
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: zext.b a2, a2
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fmv.h.x fa5, a3
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fsh fa5, 0(a2)
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: .LBB70_8: # %else6
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: andi a2, a1, 16
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetivli zero, 4, e8, mf2, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vslidedown.vi v9, v9, 4
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: beqz a2, .LBB70_10
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: # %bb.9: # %cond.store7
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a2, v9
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetvli zero, zero, e16, m1, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vslidedown.vi v10, v8, 4
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: zext.b a2, a2
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a3, v10
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fmv.h.x fa5, a3
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fsh fa5, 0(a2)
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: .LBB70_10: # %else8
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: andi a2, a1, 32
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: beqz a2, .LBB70_12
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: # %bb.11: # %cond.store9
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vslidedown.vi v10, v9, 1
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vslidedown.vi v10, v8, 5
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a3, v10
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: zext.b a2, a2
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fmv.h.x fa5, a3
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fsh fa5, 0(a2)
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: .LBB70_12: # %else10
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: andi a2, a1, 64
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetivli zero, 2, e8, mf4, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vslidedown.vi v9, v9, 2
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: bnez a2, .LBB70_15
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: # %bb.13: # %else12
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: andi a1, a1, -128
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: bnez a1, .LBB70_16
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: .LBB70_14: # %else14
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: ret
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: .LBB70_15: # %cond.store11
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a2, v9
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vslidedown.vi v10, v8, 6
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: zext.b a2, a2
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a3, v10
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fmv.h.x fa5, a3
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fsh fa5, 0(a2)
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: andi a1, a1, -128
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: beqz a1, .LBB70_14
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: .LBB70_16: # %cond.store13
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vslidedown.vi v9, v9, 1
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vslidedown.vi v8, v8, 7
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetvli zero, zero, e8, mf2, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a1, v9
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: zext.b a1, a1
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: slli a1, a1, 1
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: add a0, a0, a1
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetvli zero, zero, e16, m1, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a1, v8
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fmv.h.x fa5, a1
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fsh fa5, 0(a0)
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: ret
|
|
%eidxs = zext <8 x i8> %idxs to <8 x i16>
|
|
%ptrs = getelementptr inbounds half, ptr %base, <8 x i16> %eidxs
|
|
call void @llvm.masked.scatter.v8f16.v8p0(<8 x half> %val, <8 x ptr> %ptrs, i32 2, <8 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_baseidx_v8f16(<8 x half> %val, ptr %base, <8 x i16> %idxs, <8 x i1> %m) {
|
|
; RV32-LABEL: mscatter_baseidx_v8f16:
|
|
; RV32: # %bb.0:
|
|
; RV32-NEXT: vsetivli zero, 8, e16, m1, ta, ma
|
|
; RV32-NEXT: vwadd.vv v10, v9, v9
|
|
; RV32-NEXT: vsoxei32.v v8, (a0), v10, v0.t
|
|
; RV32-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_baseidx_v8f16:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: vsetivli zero, 8, e64, m4, ta, ma
|
|
; RV64V-NEXT: vsext.vf4 v12, v9
|
|
; RV64V-NEXT: vadd.vv v12, v12, v12
|
|
; RV64V-NEXT: vsetvli zero, zero, e16, m1, ta, ma
|
|
; RV64V-NEXT: vsoxei64.v v8, (a0), v12, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-ZVFH-LABEL: mscatter_baseidx_v8f16:
|
|
; RV64ZVE32F-ZVFH: # %bb.0:
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vmv.x.s a1, v0
|
|
; RV64ZVE32F-ZVFH-NEXT: andi a2, a1, 1
|
|
; RV64ZVE32F-ZVFH-NEXT: beqz a2, .LBB71_2
|
|
; RV64ZVE32F-ZVFH-NEXT: # %bb.1: # %cond.store
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vmv.x.s a2, v9
|
|
; RV64ZVE32F-ZVFH-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-ZVFH-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-ZVFH-NEXT: vse16.v v8, (a2)
|
|
; RV64ZVE32F-ZVFH-NEXT: .LBB71_2: # %else
|
|
; RV64ZVE32F-ZVFH-NEXT: andi a2, a1, 2
|
|
; RV64ZVE32F-ZVFH-NEXT: beqz a2, .LBB71_4
|
|
; RV64ZVE32F-ZVFH-NEXT: # %bb.3: # %cond.store1
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vslidedown.vi v10, v9, 1
|
|
; RV64ZVE32F-ZVFH-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-ZVFH-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-ZVFH-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vslidedown.vi v10, v8, 1
|
|
; RV64ZVE32F-ZVFH-NEXT: vse16.v v10, (a2)
|
|
; RV64ZVE32F-ZVFH-NEXT: .LBB71_4: # %else2
|
|
; RV64ZVE32F-ZVFH-NEXT: andi a2, a1, 4
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetivli zero, 2, e16, mf2, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vslidedown.vi v10, v9, 2
|
|
; RV64ZVE32F-ZVFH-NEXT: beqz a2, .LBB71_6
|
|
; RV64ZVE32F-ZVFH-NEXT: # %bb.5: # %cond.store3
|
|
; RV64ZVE32F-ZVFH-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-ZVFH-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-ZVFH-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vslidedown.vi v11, v8, 2
|
|
; RV64ZVE32F-ZVFH-NEXT: vse16.v v11, (a2)
|
|
; RV64ZVE32F-ZVFH-NEXT: .LBB71_6: # %else4
|
|
; RV64ZVE32F-ZVFH-NEXT: andi a2, a1, 8
|
|
; RV64ZVE32F-ZVFH-NEXT: beqz a2, .LBB71_8
|
|
; RV64ZVE32F-ZVFH-NEXT: # %bb.7: # %cond.store5
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vslidedown.vi v10, v10, 1
|
|
; RV64ZVE32F-ZVFH-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-ZVFH-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-ZVFH-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vslidedown.vi v10, v8, 3
|
|
; RV64ZVE32F-ZVFH-NEXT: vse16.v v10, (a2)
|
|
; RV64ZVE32F-ZVFH-NEXT: .LBB71_8: # %else6
|
|
; RV64ZVE32F-ZVFH-NEXT: andi a2, a1, 16
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetivli zero, 4, e16, m1, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vslidedown.vi v9, v9, 4
|
|
; RV64ZVE32F-ZVFH-NEXT: beqz a2, .LBB71_10
|
|
; RV64ZVE32F-ZVFH-NEXT: # %bb.9: # %cond.store7
|
|
; RV64ZVE32F-ZVFH-NEXT: vmv.x.s a2, v9
|
|
; RV64ZVE32F-ZVFH-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-ZVFH-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-ZVFH-NEXT: vslidedown.vi v10, v8, 4
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vse16.v v10, (a2)
|
|
; RV64ZVE32F-ZVFH-NEXT: .LBB71_10: # %else8
|
|
; RV64ZVE32F-ZVFH-NEXT: andi a2, a1, 32
|
|
; RV64ZVE32F-ZVFH-NEXT: beqz a2, .LBB71_12
|
|
; RV64ZVE32F-ZVFH-NEXT: # %bb.11: # %cond.store9
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vslidedown.vi v10, v9, 1
|
|
; RV64ZVE32F-ZVFH-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-ZVFH-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-ZVFH-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vslidedown.vi v10, v8, 5
|
|
; RV64ZVE32F-ZVFH-NEXT: vse16.v v10, (a2)
|
|
; RV64ZVE32F-ZVFH-NEXT: .LBB71_12: # %else10
|
|
; RV64ZVE32F-ZVFH-NEXT: andi a2, a1, 64
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetivli zero, 2, e16, mf2, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vslidedown.vi v9, v9, 2
|
|
; RV64ZVE32F-ZVFH-NEXT: bnez a2, .LBB71_15
|
|
; RV64ZVE32F-ZVFH-NEXT: # %bb.13: # %else12
|
|
; RV64ZVE32F-ZVFH-NEXT: andi a1, a1, -128
|
|
; RV64ZVE32F-ZVFH-NEXT: bnez a1, .LBB71_16
|
|
; RV64ZVE32F-ZVFH-NEXT: .LBB71_14: # %else14
|
|
; RV64ZVE32F-ZVFH-NEXT: ret
|
|
; RV64ZVE32F-ZVFH-NEXT: .LBB71_15: # %cond.store11
|
|
; RV64ZVE32F-ZVFH-NEXT: vmv.x.s a2, v9
|
|
; RV64ZVE32F-ZVFH-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-ZVFH-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vslidedown.vi v10, v8, 6
|
|
; RV64ZVE32F-ZVFH-NEXT: vse16.v v10, (a2)
|
|
; RV64ZVE32F-ZVFH-NEXT: andi a1, a1, -128
|
|
; RV64ZVE32F-ZVFH-NEXT: beqz a1, .LBB71_14
|
|
; RV64ZVE32F-ZVFH-NEXT: .LBB71_16: # %cond.store13
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vslidedown.vi v9, v9, 1
|
|
; RV64ZVE32F-ZVFH-NEXT: vmv.x.s a1, v9
|
|
; RV64ZVE32F-ZVFH-NEXT: slli a1, a1, 1
|
|
; RV64ZVE32F-ZVFH-NEXT: add a0, a0, a1
|
|
; RV64ZVE32F-ZVFH-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-ZVFH-NEXT: vslidedown.vi v8, v8, 7
|
|
; RV64ZVE32F-ZVFH-NEXT: vse16.v v8, (a0)
|
|
; RV64ZVE32F-ZVFH-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-ZVFHMIN-LABEL: mscatter_baseidx_v8f16:
|
|
; RV64ZVE32F-ZVFHMIN: # %bb.0:
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a1, v0
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: andi a2, a1, 1
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: beqz a2, .LBB71_2
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: # %bb.1: # %cond.store
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetvli zero, zero, e16, m2, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a2, v9
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a3, v8
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fmv.h.x fa5, a3
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fsh fa5, 0(a2)
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: .LBB71_2: # %else
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: andi a2, a1, 2
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: beqz a2, .LBB71_4
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: # %bb.3: # %cond.store1
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vslidedown.vi v10, v9, 1
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vslidedown.vi v10, v8, 1
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a3, v10
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fmv.h.x fa5, a3
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fsh fa5, 0(a2)
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: .LBB71_4: # %else2
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: andi a2, a1, 4
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetivli zero, 2, e16, mf2, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vslidedown.vi v10, v9, 2
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: beqz a2, .LBB71_6
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: # %bb.5: # %cond.store3
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vslidedown.vi v11, v8, 2
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a3, v11
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fmv.h.x fa5, a3
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fsh fa5, 0(a2)
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: .LBB71_6: # %else4
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: andi a2, a1, 8
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: beqz a2, .LBB71_8
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: # %bb.7: # %cond.store5
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vslidedown.vi v10, v10, 1
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vslidedown.vi v10, v8, 3
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a3, v10
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fmv.h.x fa5, a3
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fsh fa5, 0(a2)
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: .LBB71_8: # %else6
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: andi a2, a1, 16
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetivli zero, 4, e16, m1, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vslidedown.vi v9, v9, 4
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: beqz a2, .LBB71_10
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: # %bb.9: # %cond.store7
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a2, v9
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vslidedown.vi v10, v8, 4
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a3, v10
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fmv.h.x fa5, a3
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fsh fa5, 0(a2)
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: .LBB71_10: # %else8
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: andi a2, a1, 32
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: beqz a2, .LBB71_12
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: # %bb.11: # %cond.store9
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vslidedown.vi v10, v9, 1
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vslidedown.vi v10, v8, 5
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a3, v10
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fmv.h.x fa5, a3
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fsh fa5, 0(a2)
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: .LBB71_12: # %else10
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: andi a2, a1, 64
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetivli zero, 2, e16, mf2, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vslidedown.vi v9, v9, 2
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: bnez a2, .LBB71_15
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: # %bb.13: # %else12
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: andi a1, a1, -128
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: bnez a1, .LBB71_16
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: .LBB71_14: # %else14
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: ret
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: .LBB71_15: # %cond.store11
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a2, v9
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vslidedown.vi v10, v8, 6
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: slli a2, a2, 1
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a3, v10
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fmv.h.x fa5, a3
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fsh fa5, 0(a2)
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: andi a1, a1, -128
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: beqz a1, .LBB71_14
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: .LBB71_16: # %cond.store13
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vslidedown.vi v9, v9, 1
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vslidedown.vi v8, v8, 7
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a1, v9
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: slli a1, a1, 1
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: add a0, a0, a1
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: vmv.x.s a1, v8
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fmv.h.x fa5, a1
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: fsh fa5, 0(a0)
|
|
; RV64ZVE32F-ZVFHMIN-NEXT: ret
|
|
%ptrs = getelementptr inbounds half, ptr %base, <8 x i16> %idxs
|
|
call void @llvm.masked.scatter.v8f16.v8p0(<8 x half> %val, <8 x ptr> %ptrs, i32 2, <8 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_v1f32(<1 x float> %val, <1 x ptr> %ptrs, <1 x i1> %m) {
|
|
; RV32V-LABEL: mscatter_v1f32:
|
|
; RV32V: # %bb.0:
|
|
; RV32V-NEXT: vsetivli zero, 1, e32, mf2, ta, ma
|
|
; RV32V-NEXT: vsoxei32.v v8, (zero), v9, v0.t
|
|
; RV32V-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_v1f32:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: vsetivli zero, 1, e32, mf2, ta, ma
|
|
; RV64V-NEXT: vsoxei64.v v8, (zero), v9, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV32ZVE32F-LABEL: mscatter_v1f32:
|
|
; RV32ZVE32F: # %bb.0:
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vsoxei32.v v8, (zero), v9, v0.t
|
|
; RV32ZVE32F-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_v1f32:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: vsetvli a1, zero, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vfirst.m a1, v0
|
|
; RV64ZVE32F-NEXT: bnez a1, .LBB72_2
|
|
; RV64ZVE32F-NEXT: # %bb.1: # %cond.store
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v8, (a0)
|
|
; RV64ZVE32F-NEXT: .LBB72_2: # %else
|
|
; RV64ZVE32F-NEXT: ret
|
|
call void @llvm.masked.scatter.v1f32.v1p0(<1 x float> %val, <1 x ptr> %ptrs, i32 4, <1 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_v2f32(<2 x float> %val, <2 x ptr> %ptrs, <2 x i1> %m) {
|
|
; RV32V-LABEL: mscatter_v2f32:
|
|
; RV32V: # %bb.0:
|
|
; RV32V-NEXT: vsetivli zero, 2, e32, mf2, ta, ma
|
|
; RV32V-NEXT: vsoxei32.v v8, (zero), v9, v0.t
|
|
; RV32V-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_v2f32:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: vsetivli zero, 2, e32, mf2, ta, ma
|
|
; RV64V-NEXT: vsoxei64.v v8, (zero), v9, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV32ZVE32F-LABEL: mscatter_v2f32:
|
|
; RV32ZVE32F: # %bb.0:
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 2, e32, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vsoxei32.v v8, (zero), v9, v0.t
|
|
; RV32ZVE32F-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_v2f32:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v0
|
|
; RV64ZVE32F-NEXT: andi a3, a2, 1
|
|
; RV64ZVE32F-NEXT: bnez a3, .LBB73_3
|
|
; RV64ZVE32F-NEXT: # %bb.1: # %else
|
|
; RV64ZVE32F-NEXT: andi a2, a2, 2
|
|
; RV64ZVE32F-NEXT: bnez a2, .LBB73_4
|
|
; RV64ZVE32F-NEXT: .LBB73_2: # %else2
|
|
; RV64ZVE32F-NEXT: ret
|
|
; RV64ZVE32F-NEXT: .LBB73_3: # %cond.store
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v8, (a0)
|
|
; RV64ZVE32F-NEXT: andi a2, a2, 2
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB73_2
|
|
; RV64ZVE32F-NEXT: .LBB73_4: # %cond.store1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 1
|
|
; RV64ZVE32F-NEXT: vse32.v v8, (a1)
|
|
; RV64ZVE32F-NEXT: ret
|
|
call void @llvm.masked.scatter.v2f32.v2p0(<2 x float> %val, <2 x ptr> %ptrs, i32 4, <2 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_v4f32(<4 x float> %val, <4 x ptr> %ptrs, <4 x i1> %m) {
|
|
; RV32-LABEL: mscatter_v4f32:
|
|
; RV32: # %bb.0:
|
|
; RV32-NEXT: vsetivli zero, 4, e32, m1, ta, ma
|
|
; RV32-NEXT: vsoxei32.v v8, (zero), v9, v0.t
|
|
; RV32-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_v4f32:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: vsetivli zero, 4, e32, m1, ta, ma
|
|
; RV64V-NEXT: vsoxei64.v v8, (zero), v10, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_v4f32:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: ld a4, 8(a0)
|
|
; RV64ZVE32F-NEXT: ld a2, 16(a0)
|
|
; RV64ZVE32F-NEXT: ld a1, 24(a0)
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a3, v0
|
|
; RV64ZVE32F-NEXT: andi a5, a3, 1
|
|
; RV64ZVE32F-NEXT: bnez a5, .LBB74_5
|
|
; RV64ZVE32F-NEXT: # %bb.1: # %else
|
|
; RV64ZVE32F-NEXT: andi a0, a3, 2
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB74_6
|
|
; RV64ZVE32F-NEXT: .LBB74_2: # %else2
|
|
; RV64ZVE32F-NEXT: andi a0, a3, 4
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB74_7
|
|
; RV64ZVE32F-NEXT: .LBB74_3: # %else4
|
|
; RV64ZVE32F-NEXT: andi a3, a3, 8
|
|
; RV64ZVE32F-NEXT: bnez a3, .LBB74_8
|
|
; RV64ZVE32F-NEXT: .LBB74_4: # %else6
|
|
; RV64ZVE32F-NEXT: ret
|
|
; RV64ZVE32F-NEXT: .LBB74_5: # %cond.store
|
|
; RV64ZVE32F-NEXT: ld a0, 0(a0)
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v8, (a0)
|
|
; RV64ZVE32F-NEXT: andi a0, a3, 2
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB74_2
|
|
; RV64ZVE32F-NEXT: .LBB74_6: # %cond.store1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 1
|
|
; RV64ZVE32F-NEXT: vse32.v v9, (a4)
|
|
; RV64ZVE32F-NEXT: andi a0, a3, 4
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB74_3
|
|
; RV64ZVE32F-NEXT: .LBB74_7: # %cond.store3
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 2
|
|
; RV64ZVE32F-NEXT: vse32.v v9, (a2)
|
|
; RV64ZVE32F-NEXT: andi a3, a3, 8
|
|
; RV64ZVE32F-NEXT: beqz a3, .LBB74_4
|
|
; RV64ZVE32F-NEXT: .LBB74_8: # %cond.store5
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 3
|
|
; RV64ZVE32F-NEXT: vse32.v v8, (a1)
|
|
; RV64ZVE32F-NEXT: ret
|
|
call void @llvm.masked.scatter.v4f32.v4p0(<4 x float> %val, <4 x ptr> %ptrs, i32 4, <4 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_truemask_v4f32(<4 x float> %val, <4 x ptr> %ptrs) {
|
|
; RV32-LABEL: mscatter_truemask_v4f32:
|
|
; RV32: # %bb.0:
|
|
; RV32-NEXT: vsetivli zero, 4, e32, m1, ta, ma
|
|
; RV32-NEXT: vsoxei32.v v8, (zero), v9
|
|
; RV32-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_truemask_v4f32:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: vsetivli zero, 4, e32, m1, ta, ma
|
|
; RV64V-NEXT: vsoxei64.v v8, (zero), v10
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_truemask_v4f32:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: ld a1, 0(a0)
|
|
; RV64ZVE32F-NEXT: ld a2, 8(a0)
|
|
; RV64ZVE32F-NEXT: ld a3, 16(a0)
|
|
; RV64ZVE32F-NEXT: ld a0, 24(a0)
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 1
|
|
; RV64ZVE32F-NEXT: vse32.v v8, (a1)
|
|
; RV64ZVE32F-NEXT: vse32.v v9, (a2)
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 2
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 3
|
|
; RV64ZVE32F-NEXT: vse32.v v9, (a3)
|
|
; RV64ZVE32F-NEXT: vse32.v v8, (a0)
|
|
; RV64ZVE32F-NEXT: ret
|
|
call void @llvm.masked.scatter.v4f32.v4p0(<4 x float> %val, <4 x ptr> %ptrs, i32 4, <4 x i1> splat (i1 1))
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_falsemask_v4f32(<4 x float> %val, <4 x ptr> %ptrs) {
|
|
; CHECK-LABEL: mscatter_falsemask_v4f32:
|
|
; CHECK: # %bb.0:
|
|
; CHECK-NEXT: ret
|
|
call void @llvm.masked.scatter.v4f32.v4p0(<4 x float> %val, <4 x ptr> %ptrs, i32 4, <4 x i1> zeroinitializer)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_v8f32(<8 x float> %val, <8 x ptr> %ptrs, <8 x i1> %m) {
|
|
; RV32-LABEL: mscatter_v8f32:
|
|
; RV32: # %bb.0:
|
|
; RV32-NEXT: vsetivli zero, 8, e32, m2, ta, ma
|
|
; RV32-NEXT: vsoxei32.v v8, (zero), v10, v0.t
|
|
; RV32-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_v8f32:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: vsetivli zero, 8, e32, m2, ta, ma
|
|
; RV64V-NEXT: vsoxei64.v v8, (zero), v12, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_v8f32:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: ld a3, 40(a0)
|
|
; RV64ZVE32F-NEXT: ld a2, 48(a0)
|
|
; RV64ZVE32F-NEXT: ld a1, 56(a0)
|
|
; RV64ZVE32F-NEXT: ld t0, 8(a0)
|
|
; RV64ZVE32F-NEXT: ld a7, 16(a0)
|
|
; RV64ZVE32F-NEXT: ld a6, 24(a0)
|
|
; RV64ZVE32F-NEXT: ld a5, 32(a0)
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a4, v0
|
|
; RV64ZVE32F-NEXT: andi t1, a4, 1
|
|
; RV64ZVE32F-NEXT: bnez t1, .LBB77_9
|
|
; RV64ZVE32F-NEXT: # %bb.1: # %else
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 2
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB77_10
|
|
; RV64ZVE32F-NEXT: .LBB77_2: # %else2
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 4
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB77_11
|
|
; RV64ZVE32F-NEXT: .LBB77_3: # %else4
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 8
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB77_12
|
|
; RV64ZVE32F-NEXT: .LBB77_4: # %else6
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 16
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB77_13
|
|
; RV64ZVE32F-NEXT: .LBB77_5: # %else8
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 32
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB77_14
|
|
; RV64ZVE32F-NEXT: .LBB77_6: # %else10
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 64
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB77_15
|
|
; RV64ZVE32F-NEXT: .LBB77_7: # %else12
|
|
; RV64ZVE32F-NEXT: andi a0, a4, -128
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB77_16
|
|
; RV64ZVE32F-NEXT: .LBB77_8: # %else14
|
|
; RV64ZVE32F-NEXT: ret
|
|
; RV64ZVE32F-NEXT: .LBB77_9: # %cond.store
|
|
; RV64ZVE32F-NEXT: ld a0, 0(a0)
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v8, (a0)
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 2
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB77_2
|
|
; RV64ZVE32F-NEXT: .LBB77_10: # %cond.store1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v8, 1
|
|
; RV64ZVE32F-NEXT: vse32.v v10, (t0)
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 4
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB77_3
|
|
; RV64ZVE32F-NEXT: .LBB77_11: # %cond.store3
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v8, 2
|
|
; RV64ZVE32F-NEXT: vse32.v v10, (a7)
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 8
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB77_4
|
|
; RV64ZVE32F-NEXT: .LBB77_12: # %cond.store5
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v8, 3
|
|
; RV64ZVE32F-NEXT: vse32.v v10, (a6)
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 16
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB77_5
|
|
; RV64ZVE32F-NEXT: .LBB77_13: # %cond.store7
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v8, 4
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v10, (a5)
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 32
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB77_6
|
|
; RV64ZVE32F-NEXT: .LBB77_14: # %cond.store9
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v8, 5
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v10, (a3)
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 64
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB77_7
|
|
; RV64ZVE32F-NEXT: .LBB77_15: # %cond.store11
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v8, 6
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v10, (a2)
|
|
; RV64ZVE32F-NEXT: andi a0, a4, -128
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB77_8
|
|
; RV64ZVE32F-NEXT: .LBB77_16: # %cond.store13
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 7
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v8, (a1)
|
|
; RV64ZVE32F-NEXT: ret
|
|
call void @llvm.masked.scatter.v8f32.v8p0(<8 x float> %val, <8 x ptr> %ptrs, i32 4, <8 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_baseidx_v8i8_v8f32(<8 x float> %val, ptr %base, <8 x i8> %idxs, <8 x i1> %m) {
|
|
; RV32-LABEL: mscatter_baseidx_v8i8_v8f32:
|
|
; RV32: # %bb.0:
|
|
; RV32-NEXT: vsetivli zero, 8, e32, m2, ta, ma
|
|
; RV32-NEXT: vsext.vf4 v12, v10
|
|
; RV32-NEXT: vsll.vi v10, v12, 2
|
|
; RV32-NEXT: vsoxei32.v v8, (a0), v10, v0.t
|
|
; RV32-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_baseidx_v8i8_v8f32:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: vsetivli zero, 8, e64, m4, ta, ma
|
|
; RV64V-NEXT: vsext.vf8 v12, v10
|
|
; RV64V-NEXT: vsll.vi v12, v12, 2
|
|
; RV64V-NEXT: vsetvli zero, zero, e32, m2, ta, ma
|
|
; RV64V-NEXT: vsoxei64.v v8, (a0), v12, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_baseidx_v8i8_v8f32:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a1, v0
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 1
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB78_2
|
|
; RV64ZVE32F-NEXT: # %bb.1: # %cond.store
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 2
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v8, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB78_2: # %else
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 2
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB78_4
|
|
; RV64ZVE32F-NEXT: # %bb.3: # %cond.store1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v10, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v11
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 2
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v8, 1
|
|
; RV64ZVE32F-NEXT: vse32.v v11, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB78_4: # %else2
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 4
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v10, 2
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB78_6
|
|
; RV64ZVE32F-NEXT: # %bb.5: # %cond.store3
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v11
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 2
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v8, 2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v12, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB78_6: # %else4
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 8
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB78_8
|
|
; RV64ZVE32F-NEXT: # %bb.7: # %cond.store5
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v11, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v11
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 2
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v8, 3
|
|
; RV64ZVE32F-NEXT: vse32.v v11, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB78_8: # %else6
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 16
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 4, e8, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v10, 4
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB78_10
|
|
; RV64ZVE32F-NEXT: # %bb.9: # %cond.store7
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 2
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e32, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v8, 4
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v12, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB78_10: # %else8
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 32
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB78_12
|
|
; RV64ZVE32F-NEXT: # %bb.11: # %cond.store9
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v10, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v11
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 2
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v8, 5
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v12, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB78_12: # %else10
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 64
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v10, 2
|
|
; RV64ZVE32F-NEXT: bnez a2, .LBB78_15
|
|
; RV64ZVE32F-NEXT: # %bb.13: # %else12
|
|
; RV64ZVE32F-NEXT: andi a1, a1, -128
|
|
; RV64ZVE32F-NEXT: bnez a1, .LBB78_16
|
|
; RV64ZVE32F-NEXT: .LBB78_14: # %else14
|
|
; RV64ZVE32F-NEXT: ret
|
|
; RV64ZVE32F-NEXT: .LBB78_15: # %cond.store11
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 2
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v8, 6
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v12, (a2)
|
|
; RV64ZVE32F-NEXT: andi a1, a1, -128
|
|
; RV64ZVE32F-NEXT: beqz a1, .LBB78_14
|
|
; RV64ZVE32F-NEXT: .LBB78_16: # %cond.store13
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v10, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a1, v10
|
|
; RV64ZVE32F-NEXT: slli a1, a1, 2
|
|
; RV64ZVE32F-NEXT: add a0, a0, a1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 7
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v8, (a0)
|
|
; RV64ZVE32F-NEXT: ret
|
|
%ptrs = getelementptr inbounds float, ptr %base, <8 x i8> %idxs
|
|
call void @llvm.masked.scatter.v8f32.v8p0(<8 x float> %val, <8 x ptr> %ptrs, i32 4, <8 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_baseidx_sext_v8i8_v8f32(<8 x float> %val, ptr %base, <8 x i8> %idxs, <8 x i1> %m) {
|
|
; RV32-LABEL: mscatter_baseidx_sext_v8i8_v8f32:
|
|
; RV32: # %bb.0:
|
|
; RV32-NEXT: vsetivli zero, 8, e32, m2, ta, ma
|
|
; RV32-NEXT: vsext.vf4 v12, v10
|
|
; RV32-NEXT: vsll.vi v10, v12, 2
|
|
; RV32-NEXT: vsoxei32.v v8, (a0), v10, v0.t
|
|
; RV32-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_baseidx_sext_v8i8_v8f32:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: vsetivli zero, 8, e64, m4, ta, ma
|
|
; RV64V-NEXT: vsext.vf8 v12, v10
|
|
; RV64V-NEXT: vsll.vi v12, v12, 2
|
|
; RV64V-NEXT: vsetvli zero, zero, e32, m2, ta, ma
|
|
; RV64V-NEXT: vsoxei64.v v8, (a0), v12, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_baseidx_sext_v8i8_v8f32:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a1, v0
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 1
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB79_2
|
|
; RV64ZVE32F-NEXT: # %bb.1: # %cond.store
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 2
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v8, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB79_2: # %else
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 2
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB79_4
|
|
; RV64ZVE32F-NEXT: # %bb.3: # %cond.store1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v10, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v11
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 2
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v8, 1
|
|
; RV64ZVE32F-NEXT: vse32.v v11, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB79_4: # %else2
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 4
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v10, 2
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB79_6
|
|
; RV64ZVE32F-NEXT: # %bb.5: # %cond.store3
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v11
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 2
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v8, 2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v12, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB79_6: # %else4
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 8
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB79_8
|
|
; RV64ZVE32F-NEXT: # %bb.7: # %cond.store5
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v11, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v11
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 2
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v8, 3
|
|
; RV64ZVE32F-NEXT: vse32.v v11, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB79_8: # %else6
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 16
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 4, e8, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v10, 4
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB79_10
|
|
; RV64ZVE32F-NEXT: # %bb.9: # %cond.store7
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 2
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e32, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v8, 4
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v12, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB79_10: # %else8
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 32
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB79_12
|
|
; RV64ZVE32F-NEXT: # %bb.11: # %cond.store9
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v10, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v11
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 2
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v8, 5
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v12, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB79_12: # %else10
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 64
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v10, 2
|
|
; RV64ZVE32F-NEXT: bnez a2, .LBB79_15
|
|
; RV64ZVE32F-NEXT: # %bb.13: # %else12
|
|
; RV64ZVE32F-NEXT: andi a1, a1, -128
|
|
; RV64ZVE32F-NEXT: bnez a1, .LBB79_16
|
|
; RV64ZVE32F-NEXT: .LBB79_14: # %else14
|
|
; RV64ZVE32F-NEXT: ret
|
|
; RV64ZVE32F-NEXT: .LBB79_15: # %cond.store11
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 2
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v8, 6
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v12, (a2)
|
|
; RV64ZVE32F-NEXT: andi a1, a1, -128
|
|
; RV64ZVE32F-NEXT: beqz a1, .LBB79_14
|
|
; RV64ZVE32F-NEXT: .LBB79_16: # %cond.store13
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v10, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a1, v10
|
|
; RV64ZVE32F-NEXT: slli a1, a1, 2
|
|
; RV64ZVE32F-NEXT: add a0, a0, a1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 7
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v8, (a0)
|
|
; RV64ZVE32F-NEXT: ret
|
|
%eidxs = sext <8 x i8> %idxs to <8 x i32>
|
|
%ptrs = getelementptr inbounds float, ptr %base, <8 x i32> %eidxs
|
|
call void @llvm.masked.scatter.v8f32.v8p0(<8 x float> %val, <8 x ptr> %ptrs, i32 4, <8 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_baseidx_zext_v8i8_v8f32(<8 x float> %val, ptr %base, <8 x i8> %idxs, <8 x i1> %m) {
|
|
; RV32-LABEL: mscatter_baseidx_zext_v8i8_v8f32:
|
|
; RV32: # %bb.0:
|
|
; RV32-NEXT: li a1, 4
|
|
; RV32-NEXT: vsetivli zero, 8, e8, mf2, ta, ma
|
|
; RV32-NEXT: vwmulu.vx v11, v10, a1
|
|
; RV32-NEXT: vsetvli zero, zero, e32, m2, ta, ma
|
|
; RV32-NEXT: vsoxei16.v v8, (a0), v11, v0.t
|
|
; RV32-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_baseidx_zext_v8i8_v8f32:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: li a1, 4
|
|
; RV64V-NEXT: vsetivli zero, 8, e8, mf2, ta, ma
|
|
; RV64V-NEXT: vwmulu.vx v11, v10, a1
|
|
; RV64V-NEXT: vsetvli zero, zero, e32, m2, ta, ma
|
|
; RV64V-NEXT: vsoxei16.v v8, (a0), v11, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_baseidx_zext_v8i8_v8f32:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a1, v0
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 1
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB80_2
|
|
; RV64ZVE32F-NEXT: # %bb.1: # %cond.store
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: zext.b a2, a2
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 2
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v8, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB80_2: # %else
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 2
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB80_4
|
|
; RV64ZVE32F-NEXT: # %bb.3: # %cond.store1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v10, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v11
|
|
; RV64ZVE32F-NEXT: zext.b a2, a2
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 2
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v8, 1
|
|
; RV64ZVE32F-NEXT: vse32.v v11, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB80_4: # %else2
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 4
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v10, 2
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB80_6
|
|
; RV64ZVE32F-NEXT: # %bb.5: # %cond.store3
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v11
|
|
; RV64ZVE32F-NEXT: zext.b a2, a2
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 2
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v8, 2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v12, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB80_6: # %else4
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 8
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB80_8
|
|
; RV64ZVE32F-NEXT: # %bb.7: # %cond.store5
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v11, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v11
|
|
; RV64ZVE32F-NEXT: zext.b a2, a2
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 2
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v8, 3
|
|
; RV64ZVE32F-NEXT: vse32.v v11, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB80_8: # %else6
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 16
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 4, e8, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v10, 4
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB80_10
|
|
; RV64ZVE32F-NEXT: # %bb.9: # %cond.store7
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: zext.b a2, a2
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 2
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e32, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v8, 4
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v12, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB80_10: # %else8
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 32
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB80_12
|
|
; RV64ZVE32F-NEXT: # %bb.11: # %cond.store9
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v10, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v11
|
|
; RV64ZVE32F-NEXT: zext.b a2, a2
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 2
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v8, 5
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v12, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB80_12: # %else10
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 64
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v10, 2
|
|
; RV64ZVE32F-NEXT: bnez a2, .LBB80_15
|
|
; RV64ZVE32F-NEXT: # %bb.13: # %else12
|
|
; RV64ZVE32F-NEXT: andi a1, a1, -128
|
|
; RV64ZVE32F-NEXT: bnez a1, .LBB80_16
|
|
; RV64ZVE32F-NEXT: .LBB80_14: # %else14
|
|
; RV64ZVE32F-NEXT: ret
|
|
; RV64ZVE32F-NEXT: .LBB80_15: # %cond.store11
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: zext.b a2, a2
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 2
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v8, 6
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v12, (a2)
|
|
; RV64ZVE32F-NEXT: andi a1, a1, -128
|
|
; RV64ZVE32F-NEXT: beqz a1, .LBB80_14
|
|
; RV64ZVE32F-NEXT: .LBB80_16: # %cond.store13
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v10, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a1, v10
|
|
; RV64ZVE32F-NEXT: zext.b a1, a1
|
|
; RV64ZVE32F-NEXT: slli a1, a1, 2
|
|
; RV64ZVE32F-NEXT: add a0, a0, a1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 7
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v8, (a0)
|
|
; RV64ZVE32F-NEXT: ret
|
|
%eidxs = zext <8 x i8> %idxs to <8 x i32>
|
|
%ptrs = getelementptr inbounds float, ptr %base, <8 x i32> %eidxs
|
|
call void @llvm.masked.scatter.v8f32.v8p0(<8 x float> %val, <8 x ptr> %ptrs, i32 4, <8 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_baseidx_v8i16_v8f32(<8 x float> %val, ptr %base, <8 x i16> %idxs, <8 x i1> %m) {
|
|
; RV32-LABEL: mscatter_baseidx_v8i16_v8f32:
|
|
; RV32: # %bb.0:
|
|
; RV32-NEXT: li a1, 4
|
|
; RV32-NEXT: vsetivli zero, 8, e16, m1, ta, ma
|
|
; RV32-NEXT: vwmulsu.vx v12, v10, a1
|
|
; RV32-NEXT: vsetvli zero, zero, e32, m2, ta, ma
|
|
; RV32-NEXT: vsoxei32.v v8, (a0), v12, v0.t
|
|
; RV32-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_baseidx_v8i16_v8f32:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: vsetivli zero, 8, e64, m4, ta, ma
|
|
; RV64V-NEXT: vsext.vf4 v12, v10
|
|
; RV64V-NEXT: vsll.vi v12, v12, 2
|
|
; RV64V-NEXT: vsetvli zero, zero, e32, m2, ta, ma
|
|
; RV64V-NEXT: vsoxei64.v v8, (a0), v12, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_baseidx_v8i16_v8f32:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a1, v0
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 1
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB81_2
|
|
; RV64ZVE32F-NEXT: # %bb.1: # %cond.store
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e16, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 2
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v8, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB81_2: # %else
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 2
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB81_4
|
|
; RV64ZVE32F-NEXT: # %bb.3: # %cond.store1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v10, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v11
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 2
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v8, 1
|
|
; RV64ZVE32F-NEXT: vse32.v v11, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB81_4: # %else2
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 4
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v10, 2
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB81_6
|
|
; RV64ZVE32F-NEXT: # %bb.5: # %cond.store3
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v11
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 2
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v8, 2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v12, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB81_6: # %else4
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 8
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB81_8
|
|
; RV64ZVE32F-NEXT: # %bb.7: # %cond.store5
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v11, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v11
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 2
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v8, 3
|
|
; RV64ZVE32F-NEXT: vse32.v v11, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB81_8: # %else6
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 16
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 4, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v10, 4
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB81_10
|
|
; RV64ZVE32F-NEXT: # %bb.9: # %cond.store7
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 2
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e32, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v8, 4
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v12, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB81_10: # %else8
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 32
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB81_12
|
|
; RV64ZVE32F-NEXT: # %bb.11: # %cond.store9
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v10, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v11
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 2
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v8, 5
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v12, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB81_12: # %else10
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 64
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v10, 2
|
|
; RV64ZVE32F-NEXT: bnez a2, .LBB81_15
|
|
; RV64ZVE32F-NEXT: # %bb.13: # %else12
|
|
; RV64ZVE32F-NEXT: andi a1, a1, -128
|
|
; RV64ZVE32F-NEXT: bnez a1, .LBB81_16
|
|
; RV64ZVE32F-NEXT: .LBB81_14: # %else14
|
|
; RV64ZVE32F-NEXT: ret
|
|
; RV64ZVE32F-NEXT: .LBB81_15: # %cond.store11
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 2
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v8, 6
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v12, (a2)
|
|
; RV64ZVE32F-NEXT: andi a1, a1, -128
|
|
; RV64ZVE32F-NEXT: beqz a1, .LBB81_14
|
|
; RV64ZVE32F-NEXT: .LBB81_16: # %cond.store13
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v10, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a1, v10
|
|
; RV64ZVE32F-NEXT: slli a1, a1, 2
|
|
; RV64ZVE32F-NEXT: add a0, a0, a1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 7
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v8, (a0)
|
|
; RV64ZVE32F-NEXT: ret
|
|
%ptrs = getelementptr inbounds float, ptr %base, <8 x i16> %idxs
|
|
call void @llvm.masked.scatter.v8f32.v8p0(<8 x float> %val, <8 x ptr> %ptrs, i32 4, <8 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_baseidx_sext_v8i16_v8f32(<8 x float> %val, ptr %base, <8 x i16> %idxs, <8 x i1> %m) {
|
|
; RV32-LABEL: mscatter_baseidx_sext_v8i16_v8f32:
|
|
; RV32: # %bb.0:
|
|
; RV32-NEXT: li a1, 4
|
|
; RV32-NEXT: vsetivli zero, 8, e16, m1, ta, ma
|
|
; RV32-NEXT: vwmulsu.vx v12, v10, a1
|
|
; RV32-NEXT: vsetvli zero, zero, e32, m2, ta, ma
|
|
; RV32-NEXT: vsoxei32.v v8, (a0), v12, v0.t
|
|
; RV32-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_baseidx_sext_v8i16_v8f32:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: vsetivli zero, 8, e64, m4, ta, ma
|
|
; RV64V-NEXT: vsext.vf4 v12, v10
|
|
; RV64V-NEXT: vsll.vi v12, v12, 2
|
|
; RV64V-NEXT: vsetvli zero, zero, e32, m2, ta, ma
|
|
; RV64V-NEXT: vsoxei64.v v8, (a0), v12, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_baseidx_sext_v8i16_v8f32:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a1, v0
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 1
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB82_2
|
|
; RV64ZVE32F-NEXT: # %bb.1: # %cond.store
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e16, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 2
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v8, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB82_2: # %else
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 2
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB82_4
|
|
; RV64ZVE32F-NEXT: # %bb.3: # %cond.store1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v10, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v11
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 2
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v8, 1
|
|
; RV64ZVE32F-NEXT: vse32.v v11, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB82_4: # %else2
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 4
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v10, 2
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB82_6
|
|
; RV64ZVE32F-NEXT: # %bb.5: # %cond.store3
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v11
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 2
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v8, 2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v12, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB82_6: # %else4
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 8
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB82_8
|
|
; RV64ZVE32F-NEXT: # %bb.7: # %cond.store5
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v11, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v11
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 2
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v8, 3
|
|
; RV64ZVE32F-NEXT: vse32.v v11, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB82_8: # %else6
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 16
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 4, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v10, 4
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB82_10
|
|
; RV64ZVE32F-NEXT: # %bb.9: # %cond.store7
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 2
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e32, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v8, 4
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v12, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB82_10: # %else8
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 32
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB82_12
|
|
; RV64ZVE32F-NEXT: # %bb.11: # %cond.store9
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v10, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v11
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 2
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v8, 5
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v12, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB82_12: # %else10
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 64
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v10, 2
|
|
; RV64ZVE32F-NEXT: bnez a2, .LBB82_15
|
|
; RV64ZVE32F-NEXT: # %bb.13: # %else12
|
|
; RV64ZVE32F-NEXT: andi a1, a1, -128
|
|
; RV64ZVE32F-NEXT: bnez a1, .LBB82_16
|
|
; RV64ZVE32F-NEXT: .LBB82_14: # %else14
|
|
; RV64ZVE32F-NEXT: ret
|
|
; RV64ZVE32F-NEXT: .LBB82_15: # %cond.store11
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 2
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v8, 6
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v12, (a2)
|
|
; RV64ZVE32F-NEXT: andi a1, a1, -128
|
|
; RV64ZVE32F-NEXT: beqz a1, .LBB82_14
|
|
; RV64ZVE32F-NEXT: .LBB82_16: # %cond.store13
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v10, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a1, v10
|
|
; RV64ZVE32F-NEXT: slli a1, a1, 2
|
|
; RV64ZVE32F-NEXT: add a0, a0, a1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 7
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v8, (a0)
|
|
; RV64ZVE32F-NEXT: ret
|
|
%eidxs = sext <8 x i16> %idxs to <8 x i32>
|
|
%ptrs = getelementptr inbounds float, ptr %base, <8 x i32> %eidxs
|
|
call void @llvm.masked.scatter.v8f32.v8p0(<8 x float> %val, <8 x ptr> %ptrs, i32 4, <8 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_baseidx_zext_v8i16_v8f32(<8 x float> %val, ptr %base, <8 x i16> %idxs, <8 x i1> %m) {
|
|
; RV32-LABEL: mscatter_baseidx_zext_v8i16_v8f32:
|
|
; RV32: # %bb.0:
|
|
; RV32-NEXT: li a1, 4
|
|
; RV32-NEXT: vsetivli zero, 8, e16, m1, ta, ma
|
|
; RV32-NEXT: vwmulu.vx v12, v10, a1
|
|
; RV32-NEXT: vsetvli zero, zero, e32, m2, ta, ma
|
|
; RV32-NEXT: vsoxei32.v v8, (a0), v12, v0.t
|
|
; RV32-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_baseidx_zext_v8i16_v8f32:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: li a1, 4
|
|
; RV64V-NEXT: vsetivli zero, 8, e16, m1, ta, ma
|
|
; RV64V-NEXT: vwmulu.vx v12, v10, a1
|
|
; RV64V-NEXT: vsetvli zero, zero, e32, m2, ta, ma
|
|
; RV64V-NEXT: vsoxei32.v v8, (a0), v12, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_baseidx_zext_v8i16_v8f32:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a1, v0
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 1
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB83_2
|
|
; RV64ZVE32F-NEXT: # %bb.1: # %cond.store
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e16, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 48
|
|
; RV64ZVE32F-NEXT: srli a2, a2, 46
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v8, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB83_2: # %else
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 2
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB83_4
|
|
; RV64ZVE32F-NEXT: # %bb.3: # %cond.store1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v10, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v11
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 48
|
|
; RV64ZVE32F-NEXT: srli a2, a2, 46
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v8, 1
|
|
; RV64ZVE32F-NEXT: vse32.v v11, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB83_4: # %else2
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 4
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v10, 2
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB83_6
|
|
; RV64ZVE32F-NEXT: # %bb.5: # %cond.store3
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v11
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 48
|
|
; RV64ZVE32F-NEXT: srli a2, a2, 46
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v8, 2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v12, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB83_6: # %else4
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 8
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB83_8
|
|
; RV64ZVE32F-NEXT: # %bb.7: # %cond.store5
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v11, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v11
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 48
|
|
; RV64ZVE32F-NEXT: srli a2, a2, 46
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v8, 3
|
|
; RV64ZVE32F-NEXT: vse32.v v11, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB83_8: # %else6
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 16
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 4, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v10, 4
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB83_10
|
|
; RV64ZVE32F-NEXT: # %bb.9: # %cond.store7
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 48
|
|
; RV64ZVE32F-NEXT: srli a2, a2, 46
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e32, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v8, 4
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v12, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB83_10: # %else8
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 32
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB83_12
|
|
; RV64ZVE32F-NEXT: # %bb.11: # %cond.store9
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v10, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v11
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 48
|
|
; RV64ZVE32F-NEXT: srli a2, a2, 46
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v8, 5
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v12, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB83_12: # %else10
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 64
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v10, 2
|
|
; RV64ZVE32F-NEXT: bnez a2, .LBB83_15
|
|
; RV64ZVE32F-NEXT: # %bb.13: # %else12
|
|
; RV64ZVE32F-NEXT: andi a1, a1, -128
|
|
; RV64ZVE32F-NEXT: bnez a1, .LBB83_16
|
|
; RV64ZVE32F-NEXT: .LBB83_14: # %else14
|
|
; RV64ZVE32F-NEXT: ret
|
|
; RV64ZVE32F-NEXT: .LBB83_15: # %cond.store11
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 48
|
|
; RV64ZVE32F-NEXT: srli a2, a2, 46
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v8, 6
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v12, (a2)
|
|
; RV64ZVE32F-NEXT: andi a1, a1, -128
|
|
; RV64ZVE32F-NEXT: beqz a1, .LBB83_14
|
|
; RV64ZVE32F-NEXT: .LBB83_16: # %cond.store13
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v10, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a1, v10
|
|
; RV64ZVE32F-NEXT: slli a1, a1, 48
|
|
; RV64ZVE32F-NEXT: srli a1, a1, 46
|
|
; RV64ZVE32F-NEXT: add a0, a0, a1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 7
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v8, (a0)
|
|
; RV64ZVE32F-NEXT: ret
|
|
%eidxs = zext <8 x i16> %idxs to <8 x i32>
|
|
%ptrs = getelementptr inbounds float, ptr %base, <8 x i32> %eidxs
|
|
call void @llvm.masked.scatter.v8f32.v8p0(<8 x float> %val, <8 x ptr> %ptrs, i32 4, <8 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_baseidx_v8f32(<8 x float> %val, ptr %base, <8 x i32> %idxs, <8 x i1> %m) {
|
|
; RV32-LABEL: mscatter_baseidx_v8f32:
|
|
; RV32: # %bb.0:
|
|
; RV32-NEXT: vsetivli zero, 8, e32, m2, ta, ma
|
|
; RV32-NEXT: vsll.vi v10, v10, 2
|
|
; RV32-NEXT: vsoxei32.v v8, (a0), v10, v0.t
|
|
; RV32-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_baseidx_v8f32:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: li a1, 4
|
|
; RV64V-NEXT: vsetivli zero, 8, e32, m2, ta, ma
|
|
; RV64V-NEXT: vwmulsu.vx v12, v10, a1
|
|
; RV64V-NEXT: vsoxei64.v v8, (a0), v12, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_baseidx_v8f32:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a1, v0
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 1
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB84_2
|
|
; RV64ZVE32F-NEXT: # %bb.1: # %cond.store
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 2
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vse32.v v8, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB84_2: # %else
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 2
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB84_4
|
|
; RV64ZVE32F-NEXT: # %bb.3: # %cond.store1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v10, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v12
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 2
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v8, 1
|
|
; RV64ZVE32F-NEXT: vse32.v v12, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB84_4: # %else2
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 4
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v10, 2
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB84_6
|
|
; RV64ZVE32F-NEXT: # %bb.5: # %cond.store3
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v12
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 2
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v13, v8, 2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v13, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB84_6: # %else4
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 8
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 4, e32, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v10, 4
|
|
; RV64ZVE32F-NEXT: bnez a2, .LBB84_13
|
|
; RV64ZVE32F-NEXT: # %bb.7: # %else6
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 16
|
|
; RV64ZVE32F-NEXT: bnez a2, .LBB84_14
|
|
; RV64ZVE32F-NEXT: .LBB84_8: # %else8
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 32
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB84_10
|
|
; RV64ZVE32F-NEXT: .LBB84_9: # %cond.store9
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v10, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v11
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 2
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v8, 5
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v12, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB84_10: # %else10
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 64
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v10, 2
|
|
; RV64ZVE32F-NEXT: bnez a2, .LBB84_15
|
|
; RV64ZVE32F-NEXT: # %bb.11: # %else12
|
|
; RV64ZVE32F-NEXT: andi a1, a1, -128
|
|
; RV64ZVE32F-NEXT: bnez a1, .LBB84_16
|
|
; RV64ZVE32F-NEXT: .LBB84_12: # %else14
|
|
; RV64ZVE32F-NEXT: ret
|
|
; RV64ZVE32F-NEXT: .LBB84_13: # %cond.store5
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v12, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v11
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 2
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v8, 3
|
|
; RV64ZVE32F-NEXT: vse32.v v11, (a2)
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 16
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB84_8
|
|
; RV64ZVE32F-NEXT: .LBB84_14: # %cond.store7
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 2
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v8, 4
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v12, (a2)
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 32
|
|
; RV64ZVE32F-NEXT: bnez a2, .LBB84_9
|
|
; RV64ZVE32F-NEXT: j .LBB84_10
|
|
; RV64ZVE32F-NEXT: .LBB84_15: # %cond.store11
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 2
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v8, 6
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v12, (a2)
|
|
; RV64ZVE32F-NEXT: andi a1, a1, -128
|
|
; RV64ZVE32F-NEXT: beqz a1, .LBB84_12
|
|
; RV64ZVE32F-NEXT: .LBB84_16: # %cond.store13
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v10, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a1, v10
|
|
; RV64ZVE32F-NEXT: slli a1, a1, 2
|
|
; RV64ZVE32F-NEXT: add a0, a0, a1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 7
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse32.v v8, (a0)
|
|
; RV64ZVE32F-NEXT: ret
|
|
%ptrs = getelementptr inbounds float, ptr %base, <8 x i32> %idxs
|
|
call void @llvm.masked.scatter.v8f32.v8p0(<8 x float> %val, <8 x ptr> %ptrs, i32 4, <8 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_v1f64(<1 x double> %val, <1 x ptr> %ptrs, <1 x i1> %m) {
|
|
; RV32V-LABEL: mscatter_v1f64:
|
|
; RV32V: # %bb.0:
|
|
; RV32V-NEXT: vsetivli zero, 1, e64, m1, ta, ma
|
|
; RV32V-NEXT: vsoxei32.v v8, (zero), v9, v0.t
|
|
; RV32V-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_v1f64:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: vsetivli zero, 1, e64, m1, ta, ma
|
|
; RV64V-NEXT: vsoxei64.v v8, (zero), v9, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV32ZVE32F-LABEL: mscatter_v1f64:
|
|
; RV32ZVE32F: # %bb.0:
|
|
; RV32ZVE32F-NEXT: vsetvli a0, zero, e8, mf4, ta, ma
|
|
; RV32ZVE32F-NEXT: vfirst.m a0, v0
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB85_2
|
|
; RV32ZVE32F-NEXT: # %bb.1: # %cond.store
|
|
; RV32ZVE32F-NEXT: vsetvli zero, zero, e32, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v8
|
|
; RV32ZVE32F-NEXT: fsd fa0, 0(a0)
|
|
; RV32ZVE32F-NEXT: .LBB85_2: # %else
|
|
; RV32ZVE32F-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_v1f64:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: vsetvli a1, zero, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vfirst.m a1, v0
|
|
; RV64ZVE32F-NEXT: bnez a1, .LBB85_2
|
|
; RV64ZVE32F-NEXT: # %bb.1: # %cond.store
|
|
; RV64ZVE32F-NEXT: fsd fa0, 0(a0)
|
|
; RV64ZVE32F-NEXT: .LBB85_2: # %else
|
|
; RV64ZVE32F-NEXT: ret
|
|
call void @llvm.masked.scatter.v1f64.v1p0(<1 x double> %val, <1 x ptr> %ptrs, i32 8, <1 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_v2f64(<2 x double> %val, <2 x ptr> %ptrs, <2 x i1> %m) {
|
|
; RV32V-LABEL: mscatter_v2f64:
|
|
; RV32V: # %bb.0:
|
|
; RV32V-NEXT: vsetivli zero, 2, e64, m1, ta, ma
|
|
; RV32V-NEXT: vsoxei32.v v8, (zero), v9, v0.t
|
|
; RV32V-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_v2f64:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: vsetivli zero, 2, e64, m1, ta, ma
|
|
; RV64V-NEXT: vsoxei64.v v8, (zero), v9, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV32ZVE32F-LABEL: mscatter_v2f64:
|
|
; RV32ZVE32F: # %bb.0:
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v0
|
|
; RV32ZVE32F-NEXT: andi a1, a0, 1
|
|
; RV32ZVE32F-NEXT: bnez a1, .LBB86_3
|
|
; RV32ZVE32F-NEXT: # %bb.1: # %else
|
|
; RV32ZVE32F-NEXT: andi a0, a0, 2
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB86_4
|
|
; RV32ZVE32F-NEXT: .LBB86_2: # %else2
|
|
; RV32ZVE32F-NEXT: ret
|
|
; RV32ZVE32F-NEXT: .LBB86_3: # %cond.store
|
|
; RV32ZVE32F-NEXT: vsetvli zero, zero, e32, m4, ta, ma
|
|
; RV32ZVE32F-NEXT: vmv.x.s a1, v8
|
|
; RV32ZVE32F-NEXT: fsd fa0, 0(a1)
|
|
; RV32ZVE32F-NEXT: andi a0, a0, 2
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB86_2
|
|
; RV32ZVE32F-NEXT: .LBB86_4: # %cond.store1
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v8, v8, 1
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v8
|
|
; RV32ZVE32F-NEXT: fsd fa1, 0(a0)
|
|
; RV32ZVE32F-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_v2f64:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v0
|
|
; RV64ZVE32F-NEXT: andi a3, a2, 1
|
|
; RV64ZVE32F-NEXT: bnez a3, .LBB86_3
|
|
; RV64ZVE32F-NEXT: # %bb.1: # %else
|
|
; RV64ZVE32F-NEXT: andi a2, a2, 2
|
|
; RV64ZVE32F-NEXT: bnez a2, .LBB86_4
|
|
; RV64ZVE32F-NEXT: .LBB86_2: # %else2
|
|
; RV64ZVE32F-NEXT: ret
|
|
; RV64ZVE32F-NEXT: .LBB86_3: # %cond.store
|
|
; RV64ZVE32F-NEXT: fsd fa0, 0(a0)
|
|
; RV64ZVE32F-NEXT: andi a2, a2, 2
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB86_2
|
|
; RV64ZVE32F-NEXT: .LBB86_4: # %cond.store1
|
|
; RV64ZVE32F-NEXT: fsd fa1, 0(a1)
|
|
; RV64ZVE32F-NEXT: ret
|
|
call void @llvm.masked.scatter.v2f64.v2p0(<2 x double> %val, <2 x ptr> %ptrs, i32 8, <2 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_v4f64(<4 x double> %val, <4 x ptr> %ptrs, <4 x i1> %m) {
|
|
; RV32V-LABEL: mscatter_v4f64:
|
|
; RV32V: # %bb.0:
|
|
; RV32V-NEXT: vsetivli zero, 4, e64, m2, ta, ma
|
|
; RV32V-NEXT: vsoxei32.v v8, (zero), v10, v0.t
|
|
; RV32V-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_v4f64:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: vsetivli zero, 4, e64, m2, ta, ma
|
|
; RV64V-NEXT: vsoxei64.v v8, (zero), v10, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV32ZVE32F-LABEL: mscatter_v4f64:
|
|
; RV32ZVE32F: # %bb.0:
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v0
|
|
; RV32ZVE32F-NEXT: andi a1, a0, 1
|
|
; RV32ZVE32F-NEXT: bnez a1, .LBB87_5
|
|
; RV32ZVE32F-NEXT: # %bb.1: # %else
|
|
; RV32ZVE32F-NEXT: andi a1, a0, 2
|
|
; RV32ZVE32F-NEXT: bnez a1, .LBB87_6
|
|
; RV32ZVE32F-NEXT: .LBB87_2: # %else2
|
|
; RV32ZVE32F-NEXT: andi a1, a0, 4
|
|
; RV32ZVE32F-NEXT: bnez a1, .LBB87_7
|
|
; RV32ZVE32F-NEXT: .LBB87_3: # %else4
|
|
; RV32ZVE32F-NEXT: andi a0, a0, 8
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB87_8
|
|
; RV32ZVE32F-NEXT: .LBB87_4: # %else6
|
|
; RV32ZVE32F-NEXT: ret
|
|
; RV32ZVE32F-NEXT: .LBB87_5: # %cond.store
|
|
; RV32ZVE32F-NEXT: vsetvli zero, zero, e32, m4, ta, ma
|
|
; RV32ZVE32F-NEXT: vmv.x.s a1, v8
|
|
; RV32ZVE32F-NEXT: fsd fa0, 0(a1)
|
|
; RV32ZVE32F-NEXT: andi a1, a0, 2
|
|
; RV32ZVE32F-NEXT: beqz a1, .LBB87_2
|
|
; RV32ZVE32F-NEXT: .LBB87_6: # %cond.store1
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v9, v8, 1
|
|
; RV32ZVE32F-NEXT: vmv.x.s a1, v9
|
|
; RV32ZVE32F-NEXT: fsd fa1, 0(a1)
|
|
; RV32ZVE32F-NEXT: andi a1, a0, 4
|
|
; RV32ZVE32F-NEXT: beqz a1, .LBB87_3
|
|
; RV32ZVE32F-NEXT: .LBB87_7: # %cond.store3
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v9, v8, 2
|
|
; RV32ZVE32F-NEXT: vmv.x.s a1, v9
|
|
; RV32ZVE32F-NEXT: fsd fa2, 0(a1)
|
|
; RV32ZVE32F-NEXT: andi a0, a0, 8
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB87_4
|
|
; RV32ZVE32F-NEXT: .LBB87_8: # %cond.store5
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v8, v8, 3
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v8
|
|
; RV32ZVE32F-NEXT: fsd fa3, 0(a0)
|
|
; RV32ZVE32F-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_v4f64:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: ld a4, 8(a0)
|
|
; RV64ZVE32F-NEXT: ld a2, 16(a0)
|
|
; RV64ZVE32F-NEXT: ld a1, 24(a0)
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a3, v0
|
|
; RV64ZVE32F-NEXT: andi a5, a3, 1
|
|
; RV64ZVE32F-NEXT: bnez a5, .LBB87_5
|
|
; RV64ZVE32F-NEXT: # %bb.1: # %else
|
|
; RV64ZVE32F-NEXT: andi a0, a3, 2
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB87_6
|
|
; RV64ZVE32F-NEXT: .LBB87_2: # %else2
|
|
; RV64ZVE32F-NEXT: andi a0, a3, 4
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB87_7
|
|
; RV64ZVE32F-NEXT: .LBB87_3: # %else4
|
|
; RV64ZVE32F-NEXT: andi a3, a3, 8
|
|
; RV64ZVE32F-NEXT: bnez a3, .LBB87_8
|
|
; RV64ZVE32F-NEXT: .LBB87_4: # %else6
|
|
; RV64ZVE32F-NEXT: ret
|
|
; RV64ZVE32F-NEXT: .LBB87_5: # %cond.store
|
|
; RV64ZVE32F-NEXT: ld a0, 0(a0)
|
|
; RV64ZVE32F-NEXT: fsd fa0, 0(a0)
|
|
; RV64ZVE32F-NEXT: andi a0, a3, 2
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB87_2
|
|
; RV64ZVE32F-NEXT: .LBB87_6: # %cond.store1
|
|
; RV64ZVE32F-NEXT: fsd fa1, 0(a4)
|
|
; RV64ZVE32F-NEXT: andi a0, a3, 4
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB87_3
|
|
; RV64ZVE32F-NEXT: .LBB87_7: # %cond.store3
|
|
; RV64ZVE32F-NEXT: fsd fa2, 0(a2)
|
|
; RV64ZVE32F-NEXT: andi a3, a3, 8
|
|
; RV64ZVE32F-NEXT: beqz a3, .LBB87_4
|
|
; RV64ZVE32F-NEXT: .LBB87_8: # %cond.store5
|
|
; RV64ZVE32F-NEXT: fsd fa3, 0(a1)
|
|
; RV64ZVE32F-NEXT: ret
|
|
call void @llvm.masked.scatter.v4f64.v4p0(<4 x double> %val, <4 x ptr> %ptrs, i32 8, <4 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_truemask_v4f64(<4 x double> %val, <4 x ptr> %ptrs) {
|
|
; RV32V-LABEL: mscatter_truemask_v4f64:
|
|
; RV32V: # %bb.0:
|
|
; RV32V-NEXT: vsetivli zero, 4, e64, m2, ta, ma
|
|
; RV32V-NEXT: vsoxei32.v v8, (zero), v10
|
|
; RV32V-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_truemask_v4f64:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: vsetivli zero, 4, e64, m2, ta, ma
|
|
; RV64V-NEXT: vsoxei64.v v8, (zero), v10
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV32ZVE32F-LABEL: mscatter_truemask_v4f64:
|
|
; RV32ZVE32F: # %bb.0:
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v8
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v9, v8, 1
|
|
; RV32ZVE32F-NEXT: fsd fa0, 0(a0)
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v9
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v9, v8, 2
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v8, v8, 3
|
|
; RV32ZVE32F-NEXT: fsd fa1, 0(a0)
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v9
|
|
; RV32ZVE32F-NEXT: fsd fa2, 0(a0)
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v8
|
|
; RV32ZVE32F-NEXT: fsd fa3, 0(a0)
|
|
; RV32ZVE32F-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_truemask_v4f64:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: ld a1, 0(a0)
|
|
; RV64ZVE32F-NEXT: ld a2, 8(a0)
|
|
; RV64ZVE32F-NEXT: ld a3, 16(a0)
|
|
; RV64ZVE32F-NEXT: ld a0, 24(a0)
|
|
; RV64ZVE32F-NEXT: fsd fa0, 0(a1)
|
|
; RV64ZVE32F-NEXT: fsd fa1, 0(a2)
|
|
; RV64ZVE32F-NEXT: fsd fa2, 0(a3)
|
|
; RV64ZVE32F-NEXT: fsd fa3, 0(a0)
|
|
; RV64ZVE32F-NEXT: ret
|
|
call void @llvm.masked.scatter.v4f64.v4p0(<4 x double> %val, <4 x ptr> %ptrs, i32 8, <4 x i1> splat (i1 1))
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_falsemask_v4f64(<4 x double> %val, <4 x ptr> %ptrs) {
|
|
; CHECK-LABEL: mscatter_falsemask_v4f64:
|
|
; CHECK: # %bb.0:
|
|
; CHECK-NEXT: ret
|
|
call void @llvm.masked.scatter.v4f64.v4p0(<4 x double> %val, <4 x ptr> %ptrs, i32 8, <4 x i1> zeroinitializer)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_v8f64(<8 x double> %val, <8 x ptr> %ptrs, <8 x i1> %m) {
|
|
; RV32V-LABEL: mscatter_v8f64:
|
|
; RV32V: # %bb.0:
|
|
; RV32V-NEXT: vsetivli zero, 8, e64, m4, ta, ma
|
|
; RV32V-NEXT: vsoxei32.v v8, (zero), v12, v0.t
|
|
; RV32V-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_v8f64:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: vsetivli zero, 8, e64, m4, ta, ma
|
|
; RV64V-NEXT: vsoxei64.v v8, (zero), v12, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV32ZVE32F-LABEL: mscatter_v8f64:
|
|
; RV32ZVE32F: # %bb.0:
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v0
|
|
; RV32ZVE32F-NEXT: andi a1, a0, 1
|
|
; RV32ZVE32F-NEXT: bnez a1, .LBB90_9
|
|
; RV32ZVE32F-NEXT: # %bb.1: # %else
|
|
; RV32ZVE32F-NEXT: andi a1, a0, 2
|
|
; RV32ZVE32F-NEXT: bnez a1, .LBB90_10
|
|
; RV32ZVE32F-NEXT: .LBB90_2: # %else2
|
|
; RV32ZVE32F-NEXT: andi a1, a0, 4
|
|
; RV32ZVE32F-NEXT: bnez a1, .LBB90_11
|
|
; RV32ZVE32F-NEXT: .LBB90_3: # %else4
|
|
; RV32ZVE32F-NEXT: andi a1, a0, 8
|
|
; RV32ZVE32F-NEXT: bnez a1, .LBB90_12
|
|
; RV32ZVE32F-NEXT: .LBB90_4: # %else6
|
|
; RV32ZVE32F-NEXT: andi a1, a0, 16
|
|
; RV32ZVE32F-NEXT: bnez a1, .LBB90_13
|
|
; RV32ZVE32F-NEXT: .LBB90_5: # %else8
|
|
; RV32ZVE32F-NEXT: andi a1, a0, 32
|
|
; RV32ZVE32F-NEXT: bnez a1, .LBB90_14
|
|
; RV32ZVE32F-NEXT: .LBB90_6: # %else10
|
|
; RV32ZVE32F-NEXT: andi a1, a0, 64
|
|
; RV32ZVE32F-NEXT: bnez a1, .LBB90_15
|
|
; RV32ZVE32F-NEXT: .LBB90_7: # %else12
|
|
; RV32ZVE32F-NEXT: andi a0, a0, -128
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB90_16
|
|
; RV32ZVE32F-NEXT: .LBB90_8: # %else14
|
|
; RV32ZVE32F-NEXT: ret
|
|
; RV32ZVE32F-NEXT: .LBB90_9: # %cond.store
|
|
; RV32ZVE32F-NEXT: vsetvli zero, zero, e32, m4, ta, ma
|
|
; RV32ZVE32F-NEXT: vmv.x.s a1, v8
|
|
; RV32ZVE32F-NEXT: fsd fa0, 0(a1)
|
|
; RV32ZVE32F-NEXT: andi a1, a0, 2
|
|
; RV32ZVE32F-NEXT: beqz a1, .LBB90_2
|
|
; RV32ZVE32F-NEXT: .LBB90_10: # %cond.store1
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 1
|
|
; RV32ZVE32F-NEXT: vmv.x.s a1, v10
|
|
; RV32ZVE32F-NEXT: fsd fa1, 0(a1)
|
|
; RV32ZVE32F-NEXT: andi a1, a0, 4
|
|
; RV32ZVE32F-NEXT: beqz a1, .LBB90_3
|
|
; RV32ZVE32F-NEXT: .LBB90_11: # %cond.store3
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 2
|
|
; RV32ZVE32F-NEXT: vmv.x.s a1, v10
|
|
; RV32ZVE32F-NEXT: fsd fa2, 0(a1)
|
|
; RV32ZVE32F-NEXT: andi a1, a0, 8
|
|
; RV32ZVE32F-NEXT: beqz a1, .LBB90_4
|
|
; RV32ZVE32F-NEXT: .LBB90_12: # %cond.store5
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 3
|
|
; RV32ZVE32F-NEXT: vmv.x.s a1, v10
|
|
; RV32ZVE32F-NEXT: fsd fa3, 0(a1)
|
|
; RV32ZVE32F-NEXT: andi a1, a0, 16
|
|
; RV32ZVE32F-NEXT: beqz a1, .LBB90_5
|
|
; RV32ZVE32F-NEXT: .LBB90_13: # %cond.store7
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 4
|
|
; RV32ZVE32F-NEXT: vmv.x.s a1, v10
|
|
; RV32ZVE32F-NEXT: fsd fa4, 0(a1)
|
|
; RV32ZVE32F-NEXT: andi a1, a0, 32
|
|
; RV32ZVE32F-NEXT: beqz a1, .LBB90_6
|
|
; RV32ZVE32F-NEXT: .LBB90_14: # %cond.store9
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 5
|
|
; RV32ZVE32F-NEXT: vmv.x.s a1, v10
|
|
; RV32ZVE32F-NEXT: fsd fa5, 0(a1)
|
|
; RV32ZVE32F-NEXT: andi a1, a0, 64
|
|
; RV32ZVE32F-NEXT: beqz a1, .LBB90_7
|
|
; RV32ZVE32F-NEXT: .LBB90_15: # %cond.store11
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 6
|
|
; RV32ZVE32F-NEXT: vmv.x.s a1, v10
|
|
; RV32ZVE32F-NEXT: fsd fa6, 0(a1)
|
|
; RV32ZVE32F-NEXT: andi a0, a0, -128
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB90_8
|
|
; RV32ZVE32F-NEXT: .LBB90_16: # %cond.store13
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v8, v8, 7
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v8
|
|
; RV32ZVE32F-NEXT: fsd fa7, 0(a0)
|
|
; RV32ZVE32F-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_v8f64:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: ld a3, 40(a0)
|
|
; RV64ZVE32F-NEXT: ld a2, 48(a0)
|
|
; RV64ZVE32F-NEXT: ld a1, 56(a0)
|
|
; RV64ZVE32F-NEXT: ld t0, 8(a0)
|
|
; RV64ZVE32F-NEXT: ld a7, 16(a0)
|
|
; RV64ZVE32F-NEXT: ld a6, 24(a0)
|
|
; RV64ZVE32F-NEXT: ld a5, 32(a0)
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a4, v0
|
|
; RV64ZVE32F-NEXT: andi t1, a4, 1
|
|
; RV64ZVE32F-NEXT: bnez t1, .LBB90_9
|
|
; RV64ZVE32F-NEXT: # %bb.1: # %else
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 2
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB90_10
|
|
; RV64ZVE32F-NEXT: .LBB90_2: # %else2
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 4
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB90_11
|
|
; RV64ZVE32F-NEXT: .LBB90_3: # %else4
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 8
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB90_12
|
|
; RV64ZVE32F-NEXT: .LBB90_4: # %else6
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 16
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB90_13
|
|
; RV64ZVE32F-NEXT: .LBB90_5: # %else8
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 32
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB90_14
|
|
; RV64ZVE32F-NEXT: .LBB90_6: # %else10
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 64
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB90_15
|
|
; RV64ZVE32F-NEXT: .LBB90_7: # %else12
|
|
; RV64ZVE32F-NEXT: andi a0, a4, -128
|
|
; RV64ZVE32F-NEXT: bnez a0, .LBB90_16
|
|
; RV64ZVE32F-NEXT: .LBB90_8: # %else14
|
|
; RV64ZVE32F-NEXT: ret
|
|
; RV64ZVE32F-NEXT: .LBB90_9: # %cond.store
|
|
; RV64ZVE32F-NEXT: ld a0, 0(a0)
|
|
; RV64ZVE32F-NEXT: fsd fa0, 0(a0)
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 2
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB90_2
|
|
; RV64ZVE32F-NEXT: .LBB90_10: # %cond.store1
|
|
; RV64ZVE32F-NEXT: fsd fa1, 0(t0)
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 4
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB90_3
|
|
; RV64ZVE32F-NEXT: .LBB90_11: # %cond.store3
|
|
; RV64ZVE32F-NEXT: fsd fa2, 0(a7)
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 8
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB90_4
|
|
; RV64ZVE32F-NEXT: .LBB90_12: # %cond.store5
|
|
; RV64ZVE32F-NEXT: fsd fa3, 0(a6)
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 16
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB90_5
|
|
; RV64ZVE32F-NEXT: .LBB90_13: # %cond.store7
|
|
; RV64ZVE32F-NEXT: fsd fa4, 0(a5)
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 32
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB90_6
|
|
; RV64ZVE32F-NEXT: .LBB90_14: # %cond.store9
|
|
; RV64ZVE32F-NEXT: fsd fa5, 0(a3)
|
|
; RV64ZVE32F-NEXT: andi a0, a4, 64
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB90_7
|
|
; RV64ZVE32F-NEXT: .LBB90_15: # %cond.store11
|
|
; RV64ZVE32F-NEXT: fsd fa6, 0(a2)
|
|
; RV64ZVE32F-NEXT: andi a0, a4, -128
|
|
; RV64ZVE32F-NEXT: beqz a0, .LBB90_8
|
|
; RV64ZVE32F-NEXT: .LBB90_16: # %cond.store13
|
|
; RV64ZVE32F-NEXT: fsd fa7, 0(a1)
|
|
; RV64ZVE32F-NEXT: ret
|
|
call void @llvm.masked.scatter.v8f64.v8p0(<8 x double> %val, <8 x ptr> %ptrs, i32 8, <8 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_baseidx_v8i8_v8f64(<8 x double> %val, ptr %base, <8 x i8> %idxs, <8 x i1> %m) {
|
|
; RV32V-LABEL: mscatter_baseidx_v8i8_v8f64:
|
|
; RV32V: # %bb.0:
|
|
; RV32V-NEXT: vsetivli zero, 8, e32, m2, ta, ma
|
|
; RV32V-NEXT: vsext.vf4 v14, v12
|
|
; RV32V-NEXT: vsll.vi v12, v14, 3
|
|
; RV32V-NEXT: vsetvli zero, zero, e64, m4, ta, ma
|
|
; RV32V-NEXT: vsoxei32.v v8, (a0), v12, v0.t
|
|
; RV32V-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_baseidx_v8i8_v8f64:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: vsetivli zero, 8, e64, m4, ta, ma
|
|
; RV64V-NEXT: vsext.vf8 v16, v12
|
|
; RV64V-NEXT: vsll.vi v12, v16, 3
|
|
; RV64V-NEXT: vsoxei64.v v8, (a0), v12, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV32ZVE32F-LABEL: mscatter_baseidx_v8i8_v8f64:
|
|
; RV32ZVE32F: # %bb.0:
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 8, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vsext.vf4 v10, v8
|
|
; RV32ZVE32F-NEXT: vsetvli zero, zero, e8, mf2, ta, ma
|
|
; RV32ZVE32F-NEXT: vmv.x.s a1, v0
|
|
; RV32ZVE32F-NEXT: vsetvli zero, zero, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vsll.vi v8, v10, 3
|
|
; RV32ZVE32F-NEXT: andi a2, a1, 1
|
|
; RV32ZVE32F-NEXT: vadd.vx v8, v8, a0
|
|
; RV32ZVE32F-NEXT: bnez a2, .LBB91_9
|
|
; RV32ZVE32F-NEXT: # %bb.1: # %else
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 2
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB91_10
|
|
; RV32ZVE32F-NEXT: .LBB91_2: # %else2
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 4
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB91_11
|
|
; RV32ZVE32F-NEXT: .LBB91_3: # %else4
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 8
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB91_12
|
|
; RV32ZVE32F-NEXT: .LBB91_4: # %else6
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 16
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB91_13
|
|
; RV32ZVE32F-NEXT: .LBB91_5: # %else8
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 32
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB91_14
|
|
; RV32ZVE32F-NEXT: .LBB91_6: # %else10
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 64
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB91_15
|
|
; RV32ZVE32F-NEXT: .LBB91_7: # %else12
|
|
; RV32ZVE32F-NEXT: andi a0, a1, -128
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB91_16
|
|
; RV32ZVE32F-NEXT: .LBB91_8: # %else14
|
|
; RV32ZVE32F-NEXT: ret
|
|
; RV32ZVE32F-NEXT: .LBB91_9: # %cond.store
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v8
|
|
; RV32ZVE32F-NEXT: fsd fa0, 0(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 2
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB91_2
|
|
; RV32ZVE32F-NEXT: .LBB91_10: # %cond.store1
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 1
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: fsd fa1, 0(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 4
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB91_3
|
|
; RV32ZVE32F-NEXT: .LBB91_11: # %cond.store3
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 2
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: fsd fa2, 0(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 8
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB91_4
|
|
; RV32ZVE32F-NEXT: .LBB91_12: # %cond.store5
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 3
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: fsd fa3, 0(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 16
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB91_5
|
|
; RV32ZVE32F-NEXT: .LBB91_13: # %cond.store7
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 4
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: fsd fa4, 0(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 32
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB91_6
|
|
; RV32ZVE32F-NEXT: .LBB91_14: # %cond.store9
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 5
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: fsd fa5, 0(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 64
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB91_7
|
|
; RV32ZVE32F-NEXT: .LBB91_15: # %cond.store11
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 6
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: fsd fa6, 0(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, a1, -128
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB91_8
|
|
; RV32ZVE32F-NEXT: .LBB91_16: # %cond.store13
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v8, v8, 7
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v8
|
|
; RV32ZVE32F-NEXT: fsd fa7, 0(a0)
|
|
; RV32ZVE32F-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_baseidx_v8i8_v8f64:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a1, v0
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 1
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB91_2
|
|
; RV64ZVE32F-NEXT: # %bb.1: # %cond.store
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v8
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 3
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fsd fa0, 0(a2)
|
|
; RV64ZVE32F-NEXT: .LBB91_2: # %else
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 2
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB91_4
|
|
; RV64ZVE32F-NEXT: # %bb.3: # %cond.store1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v9
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 3
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fsd fa1, 0(a2)
|
|
; RV64ZVE32F-NEXT: .LBB91_4: # %else2
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 4
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 2
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB91_6
|
|
; RV64ZVE32F-NEXT: # %bb.5: # %cond.store3
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v9
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 3
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fsd fa2, 0(a2)
|
|
; RV64ZVE32F-NEXT: .LBB91_6: # %else4
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 8
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB91_8
|
|
; RV64ZVE32F-NEXT: # %bb.7: # %cond.store5
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v9, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v9
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 3
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fsd fa3, 0(a2)
|
|
; RV64ZVE32F-NEXT: .LBB91_8: # %else6
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 16
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 4, e8, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 4
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB91_10
|
|
; RV64ZVE32F-NEXT: # %bb.9: # %cond.store7
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v8
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 3
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fsd fa4, 0(a2)
|
|
; RV64ZVE32F-NEXT: .LBB91_10: # %else8
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 32
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB91_12
|
|
; RV64ZVE32F-NEXT: # %bb.11: # %cond.store9
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v9
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 3
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fsd fa5, 0(a2)
|
|
; RV64ZVE32F-NEXT: .LBB91_12: # %else10
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 64
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 2
|
|
; RV64ZVE32F-NEXT: bnez a2, .LBB91_15
|
|
; RV64ZVE32F-NEXT: # %bb.13: # %else12
|
|
; RV64ZVE32F-NEXT: andi a1, a1, -128
|
|
; RV64ZVE32F-NEXT: bnez a1, .LBB91_16
|
|
; RV64ZVE32F-NEXT: .LBB91_14: # %else14
|
|
; RV64ZVE32F-NEXT: ret
|
|
; RV64ZVE32F-NEXT: .LBB91_15: # %cond.store11
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v8
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 3
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fsd fa6, 0(a2)
|
|
; RV64ZVE32F-NEXT: andi a1, a1, -128
|
|
; RV64ZVE32F-NEXT: beqz a1, .LBB91_14
|
|
; RV64ZVE32F-NEXT: .LBB91_16: # %cond.store13
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a1, v8
|
|
; RV64ZVE32F-NEXT: slli a1, a1, 3
|
|
; RV64ZVE32F-NEXT: add a0, a0, a1
|
|
; RV64ZVE32F-NEXT: fsd fa7, 0(a0)
|
|
; RV64ZVE32F-NEXT: ret
|
|
%ptrs = getelementptr inbounds double, ptr %base, <8 x i8> %idxs
|
|
call void @llvm.masked.scatter.v8f64.v8p0(<8 x double> %val, <8 x ptr> %ptrs, i32 8, <8 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_baseidx_sext_v8i8_v8f64(<8 x double> %val, ptr %base, <8 x i8> %idxs, <8 x i1> %m) {
|
|
; RV32V-LABEL: mscatter_baseidx_sext_v8i8_v8f64:
|
|
; RV32V: # %bb.0:
|
|
; RV32V-NEXT: vsetivli zero, 8, e32, m2, ta, ma
|
|
; RV32V-NEXT: vsext.vf4 v14, v12
|
|
; RV32V-NEXT: vsll.vi v12, v14, 3
|
|
; RV32V-NEXT: vsetvli zero, zero, e64, m4, ta, ma
|
|
; RV32V-NEXT: vsoxei32.v v8, (a0), v12, v0.t
|
|
; RV32V-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_baseidx_sext_v8i8_v8f64:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: vsetivli zero, 8, e64, m4, ta, ma
|
|
; RV64V-NEXT: vsext.vf8 v16, v12
|
|
; RV64V-NEXT: vsll.vi v12, v16, 3
|
|
; RV64V-NEXT: vsoxei64.v v8, (a0), v12, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV32ZVE32F-LABEL: mscatter_baseidx_sext_v8i8_v8f64:
|
|
; RV32ZVE32F: # %bb.0:
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 8, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vsext.vf4 v10, v8
|
|
; RV32ZVE32F-NEXT: vsetvli zero, zero, e8, mf2, ta, ma
|
|
; RV32ZVE32F-NEXT: vmv.x.s a1, v0
|
|
; RV32ZVE32F-NEXT: vsetvli zero, zero, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vsll.vi v8, v10, 3
|
|
; RV32ZVE32F-NEXT: andi a2, a1, 1
|
|
; RV32ZVE32F-NEXT: vadd.vx v8, v8, a0
|
|
; RV32ZVE32F-NEXT: bnez a2, .LBB92_9
|
|
; RV32ZVE32F-NEXT: # %bb.1: # %else
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 2
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB92_10
|
|
; RV32ZVE32F-NEXT: .LBB92_2: # %else2
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 4
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB92_11
|
|
; RV32ZVE32F-NEXT: .LBB92_3: # %else4
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 8
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB92_12
|
|
; RV32ZVE32F-NEXT: .LBB92_4: # %else6
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 16
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB92_13
|
|
; RV32ZVE32F-NEXT: .LBB92_5: # %else8
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 32
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB92_14
|
|
; RV32ZVE32F-NEXT: .LBB92_6: # %else10
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 64
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB92_15
|
|
; RV32ZVE32F-NEXT: .LBB92_7: # %else12
|
|
; RV32ZVE32F-NEXT: andi a0, a1, -128
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB92_16
|
|
; RV32ZVE32F-NEXT: .LBB92_8: # %else14
|
|
; RV32ZVE32F-NEXT: ret
|
|
; RV32ZVE32F-NEXT: .LBB92_9: # %cond.store
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v8
|
|
; RV32ZVE32F-NEXT: fsd fa0, 0(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 2
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB92_2
|
|
; RV32ZVE32F-NEXT: .LBB92_10: # %cond.store1
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 1
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: fsd fa1, 0(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 4
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB92_3
|
|
; RV32ZVE32F-NEXT: .LBB92_11: # %cond.store3
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 2
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: fsd fa2, 0(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 8
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB92_4
|
|
; RV32ZVE32F-NEXT: .LBB92_12: # %cond.store5
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 3
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: fsd fa3, 0(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 16
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB92_5
|
|
; RV32ZVE32F-NEXT: .LBB92_13: # %cond.store7
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 4
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: fsd fa4, 0(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 32
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB92_6
|
|
; RV32ZVE32F-NEXT: .LBB92_14: # %cond.store9
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 5
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: fsd fa5, 0(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 64
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB92_7
|
|
; RV32ZVE32F-NEXT: .LBB92_15: # %cond.store11
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 6
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: fsd fa6, 0(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, a1, -128
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB92_8
|
|
; RV32ZVE32F-NEXT: .LBB92_16: # %cond.store13
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v8, v8, 7
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v8
|
|
; RV32ZVE32F-NEXT: fsd fa7, 0(a0)
|
|
; RV32ZVE32F-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_baseidx_sext_v8i8_v8f64:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a1, v0
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 1
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB92_2
|
|
; RV64ZVE32F-NEXT: # %bb.1: # %cond.store
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v8
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 3
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fsd fa0, 0(a2)
|
|
; RV64ZVE32F-NEXT: .LBB92_2: # %else
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 2
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB92_4
|
|
; RV64ZVE32F-NEXT: # %bb.3: # %cond.store1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v9
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 3
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fsd fa1, 0(a2)
|
|
; RV64ZVE32F-NEXT: .LBB92_4: # %else2
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 4
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 2
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB92_6
|
|
; RV64ZVE32F-NEXT: # %bb.5: # %cond.store3
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v9
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 3
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fsd fa2, 0(a2)
|
|
; RV64ZVE32F-NEXT: .LBB92_6: # %else4
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 8
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB92_8
|
|
; RV64ZVE32F-NEXT: # %bb.7: # %cond.store5
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v9, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v9
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 3
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fsd fa3, 0(a2)
|
|
; RV64ZVE32F-NEXT: .LBB92_8: # %else6
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 16
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 4, e8, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 4
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB92_10
|
|
; RV64ZVE32F-NEXT: # %bb.9: # %cond.store7
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v8
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 3
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fsd fa4, 0(a2)
|
|
; RV64ZVE32F-NEXT: .LBB92_10: # %else8
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 32
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB92_12
|
|
; RV64ZVE32F-NEXT: # %bb.11: # %cond.store9
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v9
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 3
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fsd fa5, 0(a2)
|
|
; RV64ZVE32F-NEXT: .LBB92_12: # %else10
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 64
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 2
|
|
; RV64ZVE32F-NEXT: bnez a2, .LBB92_15
|
|
; RV64ZVE32F-NEXT: # %bb.13: # %else12
|
|
; RV64ZVE32F-NEXT: andi a1, a1, -128
|
|
; RV64ZVE32F-NEXT: bnez a1, .LBB92_16
|
|
; RV64ZVE32F-NEXT: .LBB92_14: # %else14
|
|
; RV64ZVE32F-NEXT: ret
|
|
; RV64ZVE32F-NEXT: .LBB92_15: # %cond.store11
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v8
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 3
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fsd fa6, 0(a2)
|
|
; RV64ZVE32F-NEXT: andi a1, a1, -128
|
|
; RV64ZVE32F-NEXT: beqz a1, .LBB92_14
|
|
; RV64ZVE32F-NEXT: .LBB92_16: # %cond.store13
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a1, v8
|
|
; RV64ZVE32F-NEXT: slli a1, a1, 3
|
|
; RV64ZVE32F-NEXT: add a0, a0, a1
|
|
; RV64ZVE32F-NEXT: fsd fa7, 0(a0)
|
|
; RV64ZVE32F-NEXT: ret
|
|
%eidxs = sext <8 x i8> %idxs to <8 x i64>
|
|
%ptrs = getelementptr inbounds double, ptr %base, <8 x i64> %eidxs
|
|
call void @llvm.masked.scatter.v8f64.v8p0(<8 x double> %val, <8 x ptr> %ptrs, i32 8, <8 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_baseidx_zext_v8i8_v8f64(<8 x double> %val, ptr %base, <8 x i8> %idxs, <8 x i1> %m) {
|
|
; RV32V-LABEL: mscatter_baseidx_zext_v8i8_v8f64:
|
|
; RV32V: # %bb.0:
|
|
; RV32V-NEXT: li a1, 8
|
|
; RV32V-NEXT: vsetivli zero, 8, e8, mf2, ta, ma
|
|
; RV32V-NEXT: vwmulu.vx v13, v12, a1
|
|
; RV32V-NEXT: vsetvli zero, zero, e64, m4, ta, ma
|
|
; RV32V-NEXT: vsoxei16.v v8, (a0), v13, v0.t
|
|
; RV32V-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_baseidx_zext_v8i8_v8f64:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: li a1, 8
|
|
; RV64V-NEXT: vsetivli zero, 8, e8, mf2, ta, ma
|
|
; RV64V-NEXT: vwmulu.vx v13, v12, a1
|
|
; RV64V-NEXT: vsetvli zero, zero, e64, m4, ta, ma
|
|
; RV64V-NEXT: vsoxei16.v v8, (a0), v13, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV32ZVE32F-LABEL: mscatter_baseidx_zext_v8i8_v8f64:
|
|
; RV32ZVE32F: # %bb.0:
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 8, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vzext.vf4 v10, v8
|
|
; RV32ZVE32F-NEXT: vsetvli zero, zero, e8, mf2, ta, ma
|
|
; RV32ZVE32F-NEXT: vmv.x.s a1, v0
|
|
; RV32ZVE32F-NEXT: vsetvli zero, zero, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vsll.vi v8, v10, 3
|
|
; RV32ZVE32F-NEXT: andi a2, a1, 1
|
|
; RV32ZVE32F-NEXT: vadd.vx v8, v8, a0
|
|
; RV32ZVE32F-NEXT: bnez a2, .LBB93_9
|
|
; RV32ZVE32F-NEXT: # %bb.1: # %else
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 2
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB93_10
|
|
; RV32ZVE32F-NEXT: .LBB93_2: # %else2
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 4
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB93_11
|
|
; RV32ZVE32F-NEXT: .LBB93_3: # %else4
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 8
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB93_12
|
|
; RV32ZVE32F-NEXT: .LBB93_4: # %else6
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 16
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB93_13
|
|
; RV32ZVE32F-NEXT: .LBB93_5: # %else8
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 32
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB93_14
|
|
; RV32ZVE32F-NEXT: .LBB93_6: # %else10
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 64
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB93_15
|
|
; RV32ZVE32F-NEXT: .LBB93_7: # %else12
|
|
; RV32ZVE32F-NEXT: andi a0, a1, -128
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB93_16
|
|
; RV32ZVE32F-NEXT: .LBB93_8: # %else14
|
|
; RV32ZVE32F-NEXT: ret
|
|
; RV32ZVE32F-NEXT: .LBB93_9: # %cond.store
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v8
|
|
; RV32ZVE32F-NEXT: fsd fa0, 0(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 2
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB93_2
|
|
; RV32ZVE32F-NEXT: .LBB93_10: # %cond.store1
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 1
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: fsd fa1, 0(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 4
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB93_3
|
|
; RV32ZVE32F-NEXT: .LBB93_11: # %cond.store3
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 2
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: fsd fa2, 0(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 8
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB93_4
|
|
; RV32ZVE32F-NEXT: .LBB93_12: # %cond.store5
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 3
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: fsd fa3, 0(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 16
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB93_5
|
|
; RV32ZVE32F-NEXT: .LBB93_13: # %cond.store7
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 4
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: fsd fa4, 0(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 32
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB93_6
|
|
; RV32ZVE32F-NEXT: .LBB93_14: # %cond.store9
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 5
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: fsd fa5, 0(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 64
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB93_7
|
|
; RV32ZVE32F-NEXT: .LBB93_15: # %cond.store11
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 6
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: fsd fa6, 0(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, a1, -128
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB93_8
|
|
; RV32ZVE32F-NEXT: .LBB93_16: # %cond.store13
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v8, v8, 7
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v8
|
|
; RV32ZVE32F-NEXT: fsd fa7, 0(a0)
|
|
; RV32ZVE32F-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_baseidx_zext_v8i8_v8f64:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a1, v0
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 1
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB93_2
|
|
; RV64ZVE32F-NEXT: # %bb.1: # %cond.store
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v8
|
|
; RV64ZVE32F-NEXT: zext.b a2, a2
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 3
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fsd fa0, 0(a2)
|
|
; RV64ZVE32F-NEXT: .LBB93_2: # %else
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 2
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB93_4
|
|
; RV64ZVE32F-NEXT: # %bb.3: # %cond.store1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v9
|
|
; RV64ZVE32F-NEXT: zext.b a2, a2
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 3
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fsd fa1, 0(a2)
|
|
; RV64ZVE32F-NEXT: .LBB93_4: # %else2
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 4
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 2
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB93_6
|
|
; RV64ZVE32F-NEXT: # %bb.5: # %cond.store3
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v9
|
|
; RV64ZVE32F-NEXT: zext.b a2, a2
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 3
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fsd fa2, 0(a2)
|
|
; RV64ZVE32F-NEXT: .LBB93_6: # %else4
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 8
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB93_8
|
|
; RV64ZVE32F-NEXT: # %bb.7: # %cond.store5
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v9, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v9
|
|
; RV64ZVE32F-NEXT: zext.b a2, a2
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 3
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fsd fa3, 0(a2)
|
|
; RV64ZVE32F-NEXT: .LBB93_8: # %else6
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 16
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 4, e8, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 4
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB93_10
|
|
; RV64ZVE32F-NEXT: # %bb.9: # %cond.store7
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v8
|
|
; RV64ZVE32F-NEXT: zext.b a2, a2
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 3
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fsd fa4, 0(a2)
|
|
; RV64ZVE32F-NEXT: .LBB93_10: # %else8
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 32
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB93_12
|
|
; RV64ZVE32F-NEXT: # %bb.11: # %cond.store9
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v9
|
|
; RV64ZVE32F-NEXT: zext.b a2, a2
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 3
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fsd fa5, 0(a2)
|
|
; RV64ZVE32F-NEXT: .LBB93_12: # %else10
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 64
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 2
|
|
; RV64ZVE32F-NEXT: bnez a2, .LBB93_15
|
|
; RV64ZVE32F-NEXT: # %bb.13: # %else12
|
|
; RV64ZVE32F-NEXT: andi a1, a1, -128
|
|
; RV64ZVE32F-NEXT: bnez a1, .LBB93_16
|
|
; RV64ZVE32F-NEXT: .LBB93_14: # %else14
|
|
; RV64ZVE32F-NEXT: ret
|
|
; RV64ZVE32F-NEXT: .LBB93_15: # %cond.store11
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v8
|
|
; RV64ZVE32F-NEXT: zext.b a2, a2
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 3
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fsd fa6, 0(a2)
|
|
; RV64ZVE32F-NEXT: andi a1, a1, -128
|
|
; RV64ZVE32F-NEXT: beqz a1, .LBB93_14
|
|
; RV64ZVE32F-NEXT: .LBB93_16: # %cond.store13
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a1, v8
|
|
; RV64ZVE32F-NEXT: zext.b a1, a1
|
|
; RV64ZVE32F-NEXT: slli a1, a1, 3
|
|
; RV64ZVE32F-NEXT: add a0, a0, a1
|
|
; RV64ZVE32F-NEXT: fsd fa7, 0(a0)
|
|
; RV64ZVE32F-NEXT: ret
|
|
%eidxs = zext <8 x i8> %idxs to <8 x i64>
|
|
%ptrs = getelementptr inbounds double, ptr %base, <8 x i64> %eidxs
|
|
call void @llvm.masked.scatter.v8f64.v8p0(<8 x double> %val, <8 x ptr> %ptrs, i32 8, <8 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_baseidx_v8i16_v8f64(<8 x double> %val, ptr %base, <8 x i16> %idxs, <8 x i1> %m) {
|
|
; RV32V-LABEL: mscatter_baseidx_v8i16_v8f64:
|
|
; RV32V: # %bb.0:
|
|
; RV32V-NEXT: li a1, 8
|
|
; RV32V-NEXT: vsetivli zero, 8, e16, m1, ta, ma
|
|
; RV32V-NEXT: vwmulsu.vx v14, v12, a1
|
|
; RV32V-NEXT: vsetvli zero, zero, e64, m4, ta, ma
|
|
; RV32V-NEXT: vsoxei32.v v8, (a0), v14, v0.t
|
|
; RV32V-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_baseidx_v8i16_v8f64:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: vsetivli zero, 8, e64, m4, ta, ma
|
|
; RV64V-NEXT: vsext.vf4 v16, v12
|
|
; RV64V-NEXT: vsll.vi v12, v16, 3
|
|
; RV64V-NEXT: vsoxei64.v v8, (a0), v12, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV32ZVE32F-LABEL: mscatter_baseidx_v8i16_v8f64:
|
|
; RV32ZVE32F: # %bb.0:
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 8, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vmv.v.x v10, a0
|
|
; RV32ZVE32F-NEXT: li a1, 8
|
|
; RV32ZVE32F-NEXT: vsetvli zero, zero, e8, mf2, ta, ma
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v0
|
|
; RV32ZVE32F-NEXT: andi a2, a0, 1
|
|
; RV32ZVE32F-NEXT: vsetvli zero, zero, e16, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vwmaccus.vx v10, a1, v8
|
|
; RV32ZVE32F-NEXT: bnez a2, .LBB94_9
|
|
; RV32ZVE32F-NEXT: # %bb.1: # %else
|
|
; RV32ZVE32F-NEXT: andi a1, a0, 2
|
|
; RV32ZVE32F-NEXT: bnez a1, .LBB94_10
|
|
; RV32ZVE32F-NEXT: .LBB94_2: # %else2
|
|
; RV32ZVE32F-NEXT: andi a1, a0, 4
|
|
; RV32ZVE32F-NEXT: bnez a1, .LBB94_11
|
|
; RV32ZVE32F-NEXT: .LBB94_3: # %else4
|
|
; RV32ZVE32F-NEXT: andi a1, a0, 8
|
|
; RV32ZVE32F-NEXT: bnez a1, .LBB94_12
|
|
; RV32ZVE32F-NEXT: .LBB94_4: # %else6
|
|
; RV32ZVE32F-NEXT: andi a1, a0, 16
|
|
; RV32ZVE32F-NEXT: bnez a1, .LBB94_13
|
|
; RV32ZVE32F-NEXT: .LBB94_5: # %else8
|
|
; RV32ZVE32F-NEXT: andi a1, a0, 32
|
|
; RV32ZVE32F-NEXT: bnez a1, .LBB94_14
|
|
; RV32ZVE32F-NEXT: .LBB94_6: # %else10
|
|
; RV32ZVE32F-NEXT: andi a1, a0, 64
|
|
; RV32ZVE32F-NEXT: bnez a1, .LBB94_15
|
|
; RV32ZVE32F-NEXT: .LBB94_7: # %else12
|
|
; RV32ZVE32F-NEXT: andi a0, a0, -128
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB94_16
|
|
; RV32ZVE32F-NEXT: .LBB94_8: # %else14
|
|
; RV32ZVE32F-NEXT: ret
|
|
; RV32ZVE32F-NEXT: .LBB94_9: # %cond.store
|
|
; RV32ZVE32F-NEXT: vsetvli zero, zero, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vmv.x.s a1, v10
|
|
; RV32ZVE32F-NEXT: fsd fa0, 0(a1)
|
|
; RV32ZVE32F-NEXT: andi a1, a0, 2
|
|
; RV32ZVE32F-NEXT: beqz a1, .LBB94_2
|
|
; RV32ZVE32F-NEXT: .LBB94_10: # %cond.store1
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v8, v10, 1
|
|
; RV32ZVE32F-NEXT: vmv.x.s a1, v8
|
|
; RV32ZVE32F-NEXT: fsd fa1, 0(a1)
|
|
; RV32ZVE32F-NEXT: andi a1, a0, 4
|
|
; RV32ZVE32F-NEXT: beqz a1, .LBB94_3
|
|
; RV32ZVE32F-NEXT: .LBB94_11: # %cond.store3
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v8, v10, 2
|
|
; RV32ZVE32F-NEXT: vmv.x.s a1, v8
|
|
; RV32ZVE32F-NEXT: fsd fa2, 0(a1)
|
|
; RV32ZVE32F-NEXT: andi a1, a0, 8
|
|
; RV32ZVE32F-NEXT: beqz a1, .LBB94_4
|
|
; RV32ZVE32F-NEXT: .LBB94_12: # %cond.store5
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v8, v10, 3
|
|
; RV32ZVE32F-NEXT: vmv.x.s a1, v8
|
|
; RV32ZVE32F-NEXT: fsd fa3, 0(a1)
|
|
; RV32ZVE32F-NEXT: andi a1, a0, 16
|
|
; RV32ZVE32F-NEXT: beqz a1, .LBB94_5
|
|
; RV32ZVE32F-NEXT: .LBB94_13: # %cond.store7
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v8, v10, 4
|
|
; RV32ZVE32F-NEXT: vmv.x.s a1, v8
|
|
; RV32ZVE32F-NEXT: fsd fa4, 0(a1)
|
|
; RV32ZVE32F-NEXT: andi a1, a0, 32
|
|
; RV32ZVE32F-NEXT: beqz a1, .LBB94_6
|
|
; RV32ZVE32F-NEXT: .LBB94_14: # %cond.store9
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v8, v10, 5
|
|
; RV32ZVE32F-NEXT: vmv.x.s a1, v8
|
|
; RV32ZVE32F-NEXT: fsd fa5, 0(a1)
|
|
; RV32ZVE32F-NEXT: andi a1, a0, 64
|
|
; RV32ZVE32F-NEXT: beqz a1, .LBB94_7
|
|
; RV32ZVE32F-NEXT: .LBB94_15: # %cond.store11
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v8, v10, 6
|
|
; RV32ZVE32F-NEXT: vmv.x.s a1, v8
|
|
; RV32ZVE32F-NEXT: fsd fa6, 0(a1)
|
|
; RV32ZVE32F-NEXT: andi a0, a0, -128
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB94_8
|
|
; RV32ZVE32F-NEXT: .LBB94_16: # %cond.store13
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v8, v10, 7
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v8
|
|
; RV32ZVE32F-NEXT: fsd fa7, 0(a0)
|
|
; RV32ZVE32F-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_baseidx_v8i16_v8f64:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a1, v0
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 1
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB94_2
|
|
; RV64ZVE32F-NEXT: # %bb.1: # %cond.store
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e16, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v8
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 3
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fsd fa0, 0(a2)
|
|
; RV64ZVE32F-NEXT: .LBB94_2: # %else
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 2
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB94_4
|
|
; RV64ZVE32F-NEXT: # %bb.3: # %cond.store1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v9
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 3
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fsd fa1, 0(a2)
|
|
; RV64ZVE32F-NEXT: .LBB94_4: # %else2
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 4
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 2
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB94_6
|
|
; RV64ZVE32F-NEXT: # %bb.5: # %cond.store3
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v9
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 3
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fsd fa2, 0(a2)
|
|
; RV64ZVE32F-NEXT: .LBB94_6: # %else4
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 8
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB94_8
|
|
; RV64ZVE32F-NEXT: # %bb.7: # %cond.store5
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v9, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v9
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 3
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fsd fa3, 0(a2)
|
|
; RV64ZVE32F-NEXT: .LBB94_8: # %else6
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 16
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 4, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 4
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB94_10
|
|
; RV64ZVE32F-NEXT: # %bb.9: # %cond.store7
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v8
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 3
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fsd fa4, 0(a2)
|
|
; RV64ZVE32F-NEXT: .LBB94_10: # %else8
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 32
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB94_12
|
|
; RV64ZVE32F-NEXT: # %bb.11: # %cond.store9
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v9
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 3
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fsd fa5, 0(a2)
|
|
; RV64ZVE32F-NEXT: .LBB94_12: # %else10
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 64
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 2
|
|
; RV64ZVE32F-NEXT: bnez a2, .LBB94_15
|
|
; RV64ZVE32F-NEXT: # %bb.13: # %else12
|
|
; RV64ZVE32F-NEXT: andi a1, a1, -128
|
|
; RV64ZVE32F-NEXT: bnez a1, .LBB94_16
|
|
; RV64ZVE32F-NEXT: .LBB94_14: # %else14
|
|
; RV64ZVE32F-NEXT: ret
|
|
; RV64ZVE32F-NEXT: .LBB94_15: # %cond.store11
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v8
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 3
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fsd fa6, 0(a2)
|
|
; RV64ZVE32F-NEXT: andi a1, a1, -128
|
|
; RV64ZVE32F-NEXT: beqz a1, .LBB94_14
|
|
; RV64ZVE32F-NEXT: .LBB94_16: # %cond.store13
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a1, v8
|
|
; RV64ZVE32F-NEXT: slli a1, a1, 3
|
|
; RV64ZVE32F-NEXT: add a0, a0, a1
|
|
; RV64ZVE32F-NEXT: fsd fa7, 0(a0)
|
|
; RV64ZVE32F-NEXT: ret
|
|
%ptrs = getelementptr inbounds double, ptr %base, <8 x i16> %idxs
|
|
call void @llvm.masked.scatter.v8f64.v8p0(<8 x double> %val, <8 x ptr> %ptrs, i32 8, <8 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_baseidx_sext_v8i16_v8f64(<8 x double> %val, ptr %base, <8 x i16> %idxs, <8 x i1> %m) {
|
|
; RV32V-LABEL: mscatter_baseidx_sext_v8i16_v8f64:
|
|
; RV32V: # %bb.0:
|
|
; RV32V-NEXT: li a1, 8
|
|
; RV32V-NEXT: vsetivli zero, 8, e16, m1, ta, ma
|
|
; RV32V-NEXT: vwmulsu.vx v14, v12, a1
|
|
; RV32V-NEXT: vsetvli zero, zero, e64, m4, ta, ma
|
|
; RV32V-NEXT: vsoxei32.v v8, (a0), v14, v0.t
|
|
; RV32V-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_baseidx_sext_v8i16_v8f64:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: vsetivli zero, 8, e64, m4, ta, ma
|
|
; RV64V-NEXT: vsext.vf4 v16, v12
|
|
; RV64V-NEXT: vsll.vi v12, v16, 3
|
|
; RV64V-NEXT: vsoxei64.v v8, (a0), v12, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV32ZVE32F-LABEL: mscatter_baseidx_sext_v8i16_v8f64:
|
|
; RV32ZVE32F: # %bb.0:
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 8, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vmv.v.x v10, a0
|
|
; RV32ZVE32F-NEXT: li a1, 8
|
|
; RV32ZVE32F-NEXT: vsetvli zero, zero, e8, mf2, ta, ma
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v0
|
|
; RV32ZVE32F-NEXT: andi a2, a0, 1
|
|
; RV32ZVE32F-NEXT: vsetvli zero, zero, e16, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vwmaccus.vx v10, a1, v8
|
|
; RV32ZVE32F-NEXT: bnez a2, .LBB95_9
|
|
; RV32ZVE32F-NEXT: # %bb.1: # %else
|
|
; RV32ZVE32F-NEXT: andi a1, a0, 2
|
|
; RV32ZVE32F-NEXT: bnez a1, .LBB95_10
|
|
; RV32ZVE32F-NEXT: .LBB95_2: # %else2
|
|
; RV32ZVE32F-NEXT: andi a1, a0, 4
|
|
; RV32ZVE32F-NEXT: bnez a1, .LBB95_11
|
|
; RV32ZVE32F-NEXT: .LBB95_3: # %else4
|
|
; RV32ZVE32F-NEXT: andi a1, a0, 8
|
|
; RV32ZVE32F-NEXT: bnez a1, .LBB95_12
|
|
; RV32ZVE32F-NEXT: .LBB95_4: # %else6
|
|
; RV32ZVE32F-NEXT: andi a1, a0, 16
|
|
; RV32ZVE32F-NEXT: bnez a1, .LBB95_13
|
|
; RV32ZVE32F-NEXT: .LBB95_5: # %else8
|
|
; RV32ZVE32F-NEXT: andi a1, a0, 32
|
|
; RV32ZVE32F-NEXT: bnez a1, .LBB95_14
|
|
; RV32ZVE32F-NEXT: .LBB95_6: # %else10
|
|
; RV32ZVE32F-NEXT: andi a1, a0, 64
|
|
; RV32ZVE32F-NEXT: bnez a1, .LBB95_15
|
|
; RV32ZVE32F-NEXT: .LBB95_7: # %else12
|
|
; RV32ZVE32F-NEXT: andi a0, a0, -128
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB95_16
|
|
; RV32ZVE32F-NEXT: .LBB95_8: # %else14
|
|
; RV32ZVE32F-NEXT: ret
|
|
; RV32ZVE32F-NEXT: .LBB95_9: # %cond.store
|
|
; RV32ZVE32F-NEXT: vsetvli zero, zero, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vmv.x.s a1, v10
|
|
; RV32ZVE32F-NEXT: fsd fa0, 0(a1)
|
|
; RV32ZVE32F-NEXT: andi a1, a0, 2
|
|
; RV32ZVE32F-NEXT: beqz a1, .LBB95_2
|
|
; RV32ZVE32F-NEXT: .LBB95_10: # %cond.store1
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v8, v10, 1
|
|
; RV32ZVE32F-NEXT: vmv.x.s a1, v8
|
|
; RV32ZVE32F-NEXT: fsd fa1, 0(a1)
|
|
; RV32ZVE32F-NEXT: andi a1, a0, 4
|
|
; RV32ZVE32F-NEXT: beqz a1, .LBB95_3
|
|
; RV32ZVE32F-NEXT: .LBB95_11: # %cond.store3
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v8, v10, 2
|
|
; RV32ZVE32F-NEXT: vmv.x.s a1, v8
|
|
; RV32ZVE32F-NEXT: fsd fa2, 0(a1)
|
|
; RV32ZVE32F-NEXT: andi a1, a0, 8
|
|
; RV32ZVE32F-NEXT: beqz a1, .LBB95_4
|
|
; RV32ZVE32F-NEXT: .LBB95_12: # %cond.store5
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v8, v10, 3
|
|
; RV32ZVE32F-NEXT: vmv.x.s a1, v8
|
|
; RV32ZVE32F-NEXT: fsd fa3, 0(a1)
|
|
; RV32ZVE32F-NEXT: andi a1, a0, 16
|
|
; RV32ZVE32F-NEXT: beqz a1, .LBB95_5
|
|
; RV32ZVE32F-NEXT: .LBB95_13: # %cond.store7
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v8, v10, 4
|
|
; RV32ZVE32F-NEXT: vmv.x.s a1, v8
|
|
; RV32ZVE32F-NEXT: fsd fa4, 0(a1)
|
|
; RV32ZVE32F-NEXT: andi a1, a0, 32
|
|
; RV32ZVE32F-NEXT: beqz a1, .LBB95_6
|
|
; RV32ZVE32F-NEXT: .LBB95_14: # %cond.store9
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v8, v10, 5
|
|
; RV32ZVE32F-NEXT: vmv.x.s a1, v8
|
|
; RV32ZVE32F-NEXT: fsd fa5, 0(a1)
|
|
; RV32ZVE32F-NEXT: andi a1, a0, 64
|
|
; RV32ZVE32F-NEXT: beqz a1, .LBB95_7
|
|
; RV32ZVE32F-NEXT: .LBB95_15: # %cond.store11
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v8, v10, 6
|
|
; RV32ZVE32F-NEXT: vmv.x.s a1, v8
|
|
; RV32ZVE32F-NEXT: fsd fa6, 0(a1)
|
|
; RV32ZVE32F-NEXT: andi a0, a0, -128
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB95_8
|
|
; RV32ZVE32F-NEXT: .LBB95_16: # %cond.store13
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v8, v10, 7
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v8
|
|
; RV32ZVE32F-NEXT: fsd fa7, 0(a0)
|
|
; RV32ZVE32F-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_baseidx_sext_v8i16_v8f64:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a1, v0
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 1
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB95_2
|
|
; RV64ZVE32F-NEXT: # %bb.1: # %cond.store
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e16, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v8
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 3
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fsd fa0, 0(a2)
|
|
; RV64ZVE32F-NEXT: .LBB95_2: # %else
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 2
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB95_4
|
|
; RV64ZVE32F-NEXT: # %bb.3: # %cond.store1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v9
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 3
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fsd fa1, 0(a2)
|
|
; RV64ZVE32F-NEXT: .LBB95_4: # %else2
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 4
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 2
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB95_6
|
|
; RV64ZVE32F-NEXT: # %bb.5: # %cond.store3
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v9
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 3
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fsd fa2, 0(a2)
|
|
; RV64ZVE32F-NEXT: .LBB95_6: # %else4
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 8
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB95_8
|
|
; RV64ZVE32F-NEXT: # %bb.7: # %cond.store5
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v9, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v9
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 3
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fsd fa3, 0(a2)
|
|
; RV64ZVE32F-NEXT: .LBB95_8: # %else6
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 16
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 4, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 4
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB95_10
|
|
; RV64ZVE32F-NEXT: # %bb.9: # %cond.store7
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v8
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 3
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fsd fa4, 0(a2)
|
|
; RV64ZVE32F-NEXT: .LBB95_10: # %else8
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 32
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB95_12
|
|
; RV64ZVE32F-NEXT: # %bb.11: # %cond.store9
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v9
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 3
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fsd fa5, 0(a2)
|
|
; RV64ZVE32F-NEXT: .LBB95_12: # %else10
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 64
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 2
|
|
; RV64ZVE32F-NEXT: bnez a2, .LBB95_15
|
|
; RV64ZVE32F-NEXT: # %bb.13: # %else12
|
|
; RV64ZVE32F-NEXT: andi a1, a1, -128
|
|
; RV64ZVE32F-NEXT: bnez a1, .LBB95_16
|
|
; RV64ZVE32F-NEXT: .LBB95_14: # %else14
|
|
; RV64ZVE32F-NEXT: ret
|
|
; RV64ZVE32F-NEXT: .LBB95_15: # %cond.store11
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v8
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 3
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fsd fa6, 0(a2)
|
|
; RV64ZVE32F-NEXT: andi a1, a1, -128
|
|
; RV64ZVE32F-NEXT: beqz a1, .LBB95_14
|
|
; RV64ZVE32F-NEXT: .LBB95_16: # %cond.store13
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a1, v8
|
|
; RV64ZVE32F-NEXT: slli a1, a1, 3
|
|
; RV64ZVE32F-NEXT: add a0, a0, a1
|
|
; RV64ZVE32F-NEXT: fsd fa7, 0(a0)
|
|
; RV64ZVE32F-NEXT: ret
|
|
%eidxs = sext <8 x i16> %idxs to <8 x i64>
|
|
%ptrs = getelementptr inbounds double, ptr %base, <8 x i64> %eidxs
|
|
call void @llvm.masked.scatter.v8f64.v8p0(<8 x double> %val, <8 x ptr> %ptrs, i32 8, <8 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_baseidx_zext_v8i16_v8f64(<8 x double> %val, ptr %base, <8 x i16> %idxs, <8 x i1> %m) {
|
|
; RV32V-LABEL: mscatter_baseidx_zext_v8i16_v8f64:
|
|
; RV32V: # %bb.0:
|
|
; RV32V-NEXT: li a1, 8
|
|
; RV32V-NEXT: vsetivli zero, 8, e16, m1, ta, ma
|
|
; RV32V-NEXT: vwmulu.vx v14, v12, a1
|
|
; RV32V-NEXT: vsetvli zero, zero, e64, m4, ta, ma
|
|
; RV32V-NEXT: vsoxei32.v v8, (a0), v14, v0.t
|
|
; RV32V-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_baseidx_zext_v8i16_v8f64:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: li a1, 8
|
|
; RV64V-NEXT: vsetivli zero, 8, e16, m1, ta, ma
|
|
; RV64V-NEXT: vwmulu.vx v14, v12, a1
|
|
; RV64V-NEXT: vsetvli zero, zero, e64, m4, ta, ma
|
|
; RV64V-NEXT: vsoxei32.v v8, (a0), v14, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV32ZVE32F-LABEL: mscatter_baseidx_zext_v8i16_v8f64:
|
|
; RV32ZVE32F: # %bb.0:
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 8, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vmv.v.x v10, a0
|
|
; RV32ZVE32F-NEXT: li a1, 8
|
|
; RV32ZVE32F-NEXT: vsetvli zero, zero, e8, mf2, ta, ma
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v0
|
|
; RV32ZVE32F-NEXT: andi a2, a0, 1
|
|
; RV32ZVE32F-NEXT: vsetvli zero, zero, e16, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vwmaccu.vx v10, a1, v8
|
|
; RV32ZVE32F-NEXT: bnez a2, .LBB96_9
|
|
; RV32ZVE32F-NEXT: # %bb.1: # %else
|
|
; RV32ZVE32F-NEXT: andi a1, a0, 2
|
|
; RV32ZVE32F-NEXT: bnez a1, .LBB96_10
|
|
; RV32ZVE32F-NEXT: .LBB96_2: # %else2
|
|
; RV32ZVE32F-NEXT: andi a1, a0, 4
|
|
; RV32ZVE32F-NEXT: bnez a1, .LBB96_11
|
|
; RV32ZVE32F-NEXT: .LBB96_3: # %else4
|
|
; RV32ZVE32F-NEXT: andi a1, a0, 8
|
|
; RV32ZVE32F-NEXT: bnez a1, .LBB96_12
|
|
; RV32ZVE32F-NEXT: .LBB96_4: # %else6
|
|
; RV32ZVE32F-NEXT: andi a1, a0, 16
|
|
; RV32ZVE32F-NEXT: bnez a1, .LBB96_13
|
|
; RV32ZVE32F-NEXT: .LBB96_5: # %else8
|
|
; RV32ZVE32F-NEXT: andi a1, a0, 32
|
|
; RV32ZVE32F-NEXT: bnez a1, .LBB96_14
|
|
; RV32ZVE32F-NEXT: .LBB96_6: # %else10
|
|
; RV32ZVE32F-NEXT: andi a1, a0, 64
|
|
; RV32ZVE32F-NEXT: bnez a1, .LBB96_15
|
|
; RV32ZVE32F-NEXT: .LBB96_7: # %else12
|
|
; RV32ZVE32F-NEXT: andi a0, a0, -128
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB96_16
|
|
; RV32ZVE32F-NEXT: .LBB96_8: # %else14
|
|
; RV32ZVE32F-NEXT: ret
|
|
; RV32ZVE32F-NEXT: .LBB96_9: # %cond.store
|
|
; RV32ZVE32F-NEXT: vsetvli zero, zero, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vmv.x.s a1, v10
|
|
; RV32ZVE32F-NEXT: fsd fa0, 0(a1)
|
|
; RV32ZVE32F-NEXT: andi a1, a0, 2
|
|
; RV32ZVE32F-NEXT: beqz a1, .LBB96_2
|
|
; RV32ZVE32F-NEXT: .LBB96_10: # %cond.store1
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v8, v10, 1
|
|
; RV32ZVE32F-NEXT: vmv.x.s a1, v8
|
|
; RV32ZVE32F-NEXT: fsd fa1, 0(a1)
|
|
; RV32ZVE32F-NEXT: andi a1, a0, 4
|
|
; RV32ZVE32F-NEXT: beqz a1, .LBB96_3
|
|
; RV32ZVE32F-NEXT: .LBB96_11: # %cond.store3
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v8, v10, 2
|
|
; RV32ZVE32F-NEXT: vmv.x.s a1, v8
|
|
; RV32ZVE32F-NEXT: fsd fa2, 0(a1)
|
|
; RV32ZVE32F-NEXT: andi a1, a0, 8
|
|
; RV32ZVE32F-NEXT: beqz a1, .LBB96_4
|
|
; RV32ZVE32F-NEXT: .LBB96_12: # %cond.store5
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v8, v10, 3
|
|
; RV32ZVE32F-NEXT: vmv.x.s a1, v8
|
|
; RV32ZVE32F-NEXT: fsd fa3, 0(a1)
|
|
; RV32ZVE32F-NEXT: andi a1, a0, 16
|
|
; RV32ZVE32F-NEXT: beqz a1, .LBB96_5
|
|
; RV32ZVE32F-NEXT: .LBB96_13: # %cond.store7
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v8, v10, 4
|
|
; RV32ZVE32F-NEXT: vmv.x.s a1, v8
|
|
; RV32ZVE32F-NEXT: fsd fa4, 0(a1)
|
|
; RV32ZVE32F-NEXT: andi a1, a0, 32
|
|
; RV32ZVE32F-NEXT: beqz a1, .LBB96_6
|
|
; RV32ZVE32F-NEXT: .LBB96_14: # %cond.store9
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v8, v10, 5
|
|
; RV32ZVE32F-NEXT: vmv.x.s a1, v8
|
|
; RV32ZVE32F-NEXT: fsd fa5, 0(a1)
|
|
; RV32ZVE32F-NEXT: andi a1, a0, 64
|
|
; RV32ZVE32F-NEXT: beqz a1, .LBB96_7
|
|
; RV32ZVE32F-NEXT: .LBB96_15: # %cond.store11
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v8, v10, 6
|
|
; RV32ZVE32F-NEXT: vmv.x.s a1, v8
|
|
; RV32ZVE32F-NEXT: fsd fa6, 0(a1)
|
|
; RV32ZVE32F-NEXT: andi a0, a0, -128
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB96_8
|
|
; RV32ZVE32F-NEXT: .LBB96_16: # %cond.store13
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v8, v10, 7
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v8
|
|
; RV32ZVE32F-NEXT: fsd fa7, 0(a0)
|
|
; RV32ZVE32F-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_baseidx_zext_v8i16_v8f64:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a1, v0
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 1
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB96_2
|
|
; RV64ZVE32F-NEXT: # %bb.1: # %cond.store
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e16, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v8
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 48
|
|
; RV64ZVE32F-NEXT: srli a2, a2, 45
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fsd fa0, 0(a2)
|
|
; RV64ZVE32F-NEXT: .LBB96_2: # %else
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 2
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB96_4
|
|
; RV64ZVE32F-NEXT: # %bb.3: # %cond.store1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v9
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 48
|
|
; RV64ZVE32F-NEXT: srli a2, a2, 45
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fsd fa1, 0(a2)
|
|
; RV64ZVE32F-NEXT: .LBB96_4: # %else2
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 4
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 2
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB96_6
|
|
; RV64ZVE32F-NEXT: # %bb.5: # %cond.store3
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v9
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 48
|
|
; RV64ZVE32F-NEXT: srli a2, a2, 45
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fsd fa2, 0(a2)
|
|
; RV64ZVE32F-NEXT: .LBB96_6: # %else4
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 8
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB96_8
|
|
; RV64ZVE32F-NEXT: # %bb.7: # %cond.store5
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v9, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v9
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 48
|
|
; RV64ZVE32F-NEXT: srli a2, a2, 45
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fsd fa3, 0(a2)
|
|
; RV64ZVE32F-NEXT: .LBB96_8: # %else6
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 16
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 4, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 4
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB96_10
|
|
; RV64ZVE32F-NEXT: # %bb.9: # %cond.store7
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v8
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 48
|
|
; RV64ZVE32F-NEXT: srli a2, a2, 45
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fsd fa4, 0(a2)
|
|
; RV64ZVE32F-NEXT: .LBB96_10: # %else8
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 32
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB96_12
|
|
; RV64ZVE32F-NEXT: # %bb.11: # %cond.store9
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v9
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 48
|
|
; RV64ZVE32F-NEXT: srli a2, a2, 45
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fsd fa5, 0(a2)
|
|
; RV64ZVE32F-NEXT: .LBB96_12: # %else10
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 64
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e16, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 2
|
|
; RV64ZVE32F-NEXT: bnez a2, .LBB96_15
|
|
; RV64ZVE32F-NEXT: # %bb.13: # %else12
|
|
; RV64ZVE32F-NEXT: andi a1, a1, -128
|
|
; RV64ZVE32F-NEXT: bnez a1, .LBB96_16
|
|
; RV64ZVE32F-NEXT: .LBB96_14: # %else14
|
|
; RV64ZVE32F-NEXT: ret
|
|
; RV64ZVE32F-NEXT: .LBB96_15: # %cond.store11
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v8
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 48
|
|
; RV64ZVE32F-NEXT: srli a2, a2, 45
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fsd fa6, 0(a2)
|
|
; RV64ZVE32F-NEXT: andi a1, a1, -128
|
|
; RV64ZVE32F-NEXT: beqz a1, .LBB96_14
|
|
; RV64ZVE32F-NEXT: .LBB96_16: # %cond.store13
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a1, v8
|
|
; RV64ZVE32F-NEXT: slli a1, a1, 48
|
|
; RV64ZVE32F-NEXT: srli a1, a1, 45
|
|
; RV64ZVE32F-NEXT: add a0, a0, a1
|
|
; RV64ZVE32F-NEXT: fsd fa7, 0(a0)
|
|
; RV64ZVE32F-NEXT: ret
|
|
%eidxs = zext <8 x i16> %idxs to <8 x i64>
|
|
%ptrs = getelementptr inbounds double, ptr %base, <8 x i64> %eidxs
|
|
call void @llvm.masked.scatter.v8f64.v8p0(<8 x double> %val, <8 x ptr> %ptrs, i32 8, <8 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_baseidx_v8i32_v8f64(<8 x double> %val, ptr %base, <8 x i32> %idxs, <8 x i1> %m) {
|
|
; RV32V-LABEL: mscatter_baseidx_v8i32_v8f64:
|
|
; RV32V: # %bb.0:
|
|
; RV32V-NEXT: vsetivli zero, 8, e32, m2, ta, ma
|
|
; RV32V-NEXT: vsll.vi v12, v12, 3
|
|
; RV32V-NEXT: vsetvli zero, zero, e64, m4, ta, ma
|
|
; RV32V-NEXT: vsoxei32.v v8, (a0), v12, v0.t
|
|
; RV32V-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_baseidx_v8i32_v8f64:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: li a1, 8
|
|
; RV64V-NEXT: vsetivli zero, 8, e32, m2, ta, ma
|
|
; RV64V-NEXT: vwmulsu.vx v16, v12, a1
|
|
; RV64V-NEXT: vsetvli zero, zero, e64, m4, ta, ma
|
|
; RV64V-NEXT: vsoxei64.v v8, (a0), v16, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV32ZVE32F-LABEL: mscatter_baseidx_v8i32_v8f64:
|
|
; RV32ZVE32F: # %bb.0:
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 8, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vsll.vi v8, v8, 3
|
|
; RV32ZVE32F-NEXT: vsetvli zero, zero, e8, mf2, ta, ma
|
|
; RV32ZVE32F-NEXT: vmv.x.s a1, v0
|
|
; RV32ZVE32F-NEXT: andi a2, a1, 1
|
|
; RV32ZVE32F-NEXT: vsetvli zero, zero, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vadd.vx v8, v8, a0
|
|
; RV32ZVE32F-NEXT: bnez a2, .LBB97_9
|
|
; RV32ZVE32F-NEXT: # %bb.1: # %else
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 2
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB97_10
|
|
; RV32ZVE32F-NEXT: .LBB97_2: # %else2
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 4
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB97_11
|
|
; RV32ZVE32F-NEXT: .LBB97_3: # %else4
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 8
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB97_12
|
|
; RV32ZVE32F-NEXT: .LBB97_4: # %else6
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 16
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB97_13
|
|
; RV32ZVE32F-NEXT: .LBB97_5: # %else8
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 32
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB97_14
|
|
; RV32ZVE32F-NEXT: .LBB97_6: # %else10
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 64
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB97_15
|
|
; RV32ZVE32F-NEXT: .LBB97_7: # %else12
|
|
; RV32ZVE32F-NEXT: andi a0, a1, -128
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB97_16
|
|
; RV32ZVE32F-NEXT: .LBB97_8: # %else14
|
|
; RV32ZVE32F-NEXT: ret
|
|
; RV32ZVE32F-NEXT: .LBB97_9: # %cond.store
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v8
|
|
; RV32ZVE32F-NEXT: fsd fa0, 0(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 2
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB97_2
|
|
; RV32ZVE32F-NEXT: .LBB97_10: # %cond.store1
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 1
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: fsd fa1, 0(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 4
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB97_3
|
|
; RV32ZVE32F-NEXT: .LBB97_11: # %cond.store3
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 2
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: fsd fa2, 0(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 8
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB97_4
|
|
; RV32ZVE32F-NEXT: .LBB97_12: # %cond.store5
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 3
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: fsd fa3, 0(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 16
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB97_5
|
|
; RV32ZVE32F-NEXT: .LBB97_13: # %cond.store7
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 4
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: fsd fa4, 0(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 32
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB97_6
|
|
; RV32ZVE32F-NEXT: .LBB97_14: # %cond.store9
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 5
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: fsd fa5, 0(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 64
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB97_7
|
|
; RV32ZVE32F-NEXT: .LBB97_15: # %cond.store11
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 6
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: fsd fa6, 0(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, a1, -128
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB97_8
|
|
; RV32ZVE32F-NEXT: .LBB97_16: # %cond.store13
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v8, v8, 7
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v8
|
|
; RV32ZVE32F-NEXT: fsd fa7, 0(a0)
|
|
; RV32ZVE32F-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_baseidx_v8i32_v8f64:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a1, v0
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 1
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB97_2
|
|
; RV64ZVE32F-NEXT: # %bb.1: # %cond.store
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e32, m4, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v8
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 3
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fsd fa0, 0(a2)
|
|
; RV64ZVE32F-NEXT: .LBB97_2: # %else
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 2
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB97_4
|
|
; RV64ZVE32F-NEXT: # %bb.3: # %cond.store1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v8, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 3
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fsd fa1, 0(a2)
|
|
; RV64ZVE32F-NEXT: .LBB97_4: # %else2
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 4
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v8, 2
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB97_6
|
|
; RV64ZVE32F-NEXT: # %bb.5: # %cond.store3
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 3
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fsd fa2, 0(a2)
|
|
; RV64ZVE32F-NEXT: .LBB97_6: # %else4
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 8
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 4, e32, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 4
|
|
; RV64ZVE32F-NEXT: bnez a2, .LBB97_13
|
|
; RV64ZVE32F-NEXT: # %bb.7: # %else6
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 16
|
|
; RV64ZVE32F-NEXT: bnez a2, .LBB97_14
|
|
; RV64ZVE32F-NEXT: .LBB97_8: # %else8
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 32
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB97_10
|
|
; RV64ZVE32F-NEXT: .LBB97_9: # %cond.store9
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v9
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 3
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fsd fa5, 0(a2)
|
|
; RV64ZVE32F-NEXT: .LBB97_10: # %else10
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 64
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 2
|
|
; RV64ZVE32F-NEXT: bnez a2, .LBB97_15
|
|
; RV64ZVE32F-NEXT: # %bb.11: # %else12
|
|
; RV64ZVE32F-NEXT: andi a1, a1, -128
|
|
; RV64ZVE32F-NEXT: bnez a1, .LBB97_16
|
|
; RV64ZVE32F-NEXT: .LBB97_12: # %else14
|
|
; RV64ZVE32F-NEXT: ret
|
|
; RV64ZVE32F-NEXT: .LBB97_13: # %cond.store5
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v10, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v9
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 3
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fsd fa3, 0(a2)
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 16
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB97_8
|
|
; RV64ZVE32F-NEXT: .LBB97_14: # %cond.store7
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v8
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 3
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fsd fa4, 0(a2)
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 32
|
|
; RV64ZVE32F-NEXT: bnez a2, .LBB97_9
|
|
; RV64ZVE32F-NEXT: j .LBB97_10
|
|
; RV64ZVE32F-NEXT: .LBB97_15: # %cond.store11
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v8
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 3
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fsd fa6, 0(a2)
|
|
; RV64ZVE32F-NEXT: andi a1, a1, -128
|
|
; RV64ZVE32F-NEXT: beqz a1, .LBB97_12
|
|
; RV64ZVE32F-NEXT: .LBB97_16: # %cond.store13
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a1, v8
|
|
; RV64ZVE32F-NEXT: slli a1, a1, 3
|
|
; RV64ZVE32F-NEXT: add a0, a0, a1
|
|
; RV64ZVE32F-NEXT: fsd fa7, 0(a0)
|
|
; RV64ZVE32F-NEXT: ret
|
|
%ptrs = getelementptr inbounds double, ptr %base, <8 x i32> %idxs
|
|
call void @llvm.masked.scatter.v8f64.v8p0(<8 x double> %val, <8 x ptr> %ptrs, i32 8, <8 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_baseidx_sext_v8i32_v8f64(<8 x double> %val, ptr %base, <8 x i32> %idxs, <8 x i1> %m) {
|
|
; RV32V-LABEL: mscatter_baseidx_sext_v8i32_v8f64:
|
|
; RV32V: # %bb.0:
|
|
; RV32V-NEXT: vsetivli zero, 8, e32, m2, ta, ma
|
|
; RV32V-NEXT: vsll.vi v12, v12, 3
|
|
; RV32V-NEXT: vsetvli zero, zero, e64, m4, ta, ma
|
|
; RV32V-NEXT: vsoxei32.v v8, (a0), v12, v0.t
|
|
; RV32V-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_baseidx_sext_v8i32_v8f64:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: li a1, 8
|
|
; RV64V-NEXT: vsetivli zero, 8, e32, m2, ta, ma
|
|
; RV64V-NEXT: vwmulsu.vx v16, v12, a1
|
|
; RV64V-NEXT: vsetvli zero, zero, e64, m4, ta, ma
|
|
; RV64V-NEXT: vsoxei64.v v8, (a0), v16, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV32ZVE32F-LABEL: mscatter_baseidx_sext_v8i32_v8f64:
|
|
; RV32ZVE32F: # %bb.0:
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 8, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vsll.vi v8, v8, 3
|
|
; RV32ZVE32F-NEXT: vsetvli zero, zero, e8, mf2, ta, ma
|
|
; RV32ZVE32F-NEXT: vmv.x.s a1, v0
|
|
; RV32ZVE32F-NEXT: andi a2, a1, 1
|
|
; RV32ZVE32F-NEXT: vsetvli zero, zero, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vadd.vx v8, v8, a0
|
|
; RV32ZVE32F-NEXT: bnez a2, .LBB98_9
|
|
; RV32ZVE32F-NEXT: # %bb.1: # %else
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 2
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB98_10
|
|
; RV32ZVE32F-NEXT: .LBB98_2: # %else2
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 4
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB98_11
|
|
; RV32ZVE32F-NEXT: .LBB98_3: # %else4
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 8
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB98_12
|
|
; RV32ZVE32F-NEXT: .LBB98_4: # %else6
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 16
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB98_13
|
|
; RV32ZVE32F-NEXT: .LBB98_5: # %else8
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 32
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB98_14
|
|
; RV32ZVE32F-NEXT: .LBB98_6: # %else10
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 64
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB98_15
|
|
; RV32ZVE32F-NEXT: .LBB98_7: # %else12
|
|
; RV32ZVE32F-NEXT: andi a0, a1, -128
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB98_16
|
|
; RV32ZVE32F-NEXT: .LBB98_8: # %else14
|
|
; RV32ZVE32F-NEXT: ret
|
|
; RV32ZVE32F-NEXT: .LBB98_9: # %cond.store
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v8
|
|
; RV32ZVE32F-NEXT: fsd fa0, 0(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 2
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB98_2
|
|
; RV32ZVE32F-NEXT: .LBB98_10: # %cond.store1
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 1
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: fsd fa1, 0(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 4
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB98_3
|
|
; RV32ZVE32F-NEXT: .LBB98_11: # %cond.store3
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 2
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: fsd fa2, 0(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 8
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB98_4
|
|
; RV32ZVE32F-NEXT: .LBB98_12: # %cond.store5
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 3
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: fsd fa3, 0(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 16
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB98_5
|
|
; RV32ZVE32F-NEXT: .LBB98_13: # %cond.store7
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 4
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: fsd fa4, 0(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 32
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB98_6
|
|
; RV32ZVE32F-NEXT: .LBB98_14: # %cond.store9
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 5
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: fsd fa5, 0(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 64
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB98_7
|
|
; RV32ZVE32F-NEXT: .LBB98_15: # %cond.store11
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 6
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: fsd fa6, 0(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, a1, -128
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB98_8
|
|
; RV32ZVE32F-NEXT: .LBB98_16: # %cond.store13
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v8, v8, 7
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v8
|
|
; RV32ZVE32F-NEXT: fsd fa7, 0(a0)
|
|
; RV32ZVE32F-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_baseidx_sext_v8i32_v8f64:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a1, v0
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 1
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB98_2
|
|
; RV64ZVE32F-NEXT: # %bb.1: # %cond.store
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e32, m4, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v8
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 3
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fsd fa0, 0(a2)
|
|
; RV64ZVE32F-NEXT: .LBB98_2: # %else
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 2
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB98_4
|
|
; RV64ZVE32F-NEXT: # %bb.3: # %cond.store1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v8, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 3
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fsd fa1, 0(a2)
|
|
; RV64ZVE32F-NEXT: .LBB98_4: # %else2
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 4
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v8, 2
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB98_6
|
|
; RV64ZVE32F-NEXT: # %bb.5: # %cond.store3
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 3
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fsd fa2, 0(a2)
|
|
; RV64ZVE32F-NEXT: .LBB98_6: # %else4
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 8
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 4, e32, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 4
|
|
; RV64ZVE32F-NEXT: bnez a2, .LBB98_13
|
|
; RV64ZVE32F-NEXT: # %bb.7: # %else6
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 16
|
|
; RV64ZVE32F-NEXT: bnez a2, .LBB98_14
|
|
; RV64ZVE32F-NEXT: .LBB98_8: # %else8
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 32
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB98_10
|
|
; RV64ZVE32F-NEXT: .LBB98_9: # %cond.store9
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v9
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 3
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fsd fa5, 0(a2)
|
|
; RV64ZVE32F-NEXT: .LBB98_10: # %else10
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 64
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 2
|
|
; RV64ZVE32F-NEXT: bnez a2, .LBB98_15
|
|
; RV64ZVE32F-NEXT: # %bb.11: # %else12
|
|
; RV64ZVE32F-NEXT: andi a1, a1, -128
|
|
; RV64ZVE32F-NEXT: bnez a1, .LBB98_16
|
|
; RV64ZVE32F-NEXT: .LBB98_12: # %else14
|
|
; RV64ZVE32F-NEXT: ret
|
|
; RV64ZVE32F-NEXT: .LBB98_13: # %cond.store5
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v10, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v9
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 3
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fsd fa3, 0(a2)
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 16
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB98_8
|
|
; RV64ZVE32F-NEXT: .LBB98_14: # %cond.store7
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v8
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 3
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fsd fa4, 0(a2)
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 32
|
|
; RV64ZVE32F-NEXT: bnez a2, .LBB98_9
|
|
; RV64ZVE32F-NEXT: j .LBB98_10
|
|
; RV64ZVE32F-NEXT: .LBB98_15: # %cond.store11
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v8
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 3
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fsd fa6, 0(a2)
|
|
; RV64ZVE32F-NEXT: andi a1, a1, -128
|
|
; RV64ZVE32F-NEXT: beqz a1, .LBB98_12
|
|
; RV64ZVE32F-NEXT: .LBB98_16: # %cond.store13
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a1, v8
|
|
; RV64ZVE32F-NEXT: slli a1, a1, 3
|
|
; RV64ZVE32F-NEXT: add a0, a0, a1
|
|
; RV64ZVE32F-NEXT: fsd fa7, 0(a0)
|
|
; RV64ZVE32F-NEXT: ret
|
|
%eidxs = sext <8 x i32> %idxs to <8 x i64>
|
|
%ptrs = getelementptr inbounds double, ptr %base, <8 x i64> %eidxs
|
|
call void @llvm.masked.scatter.v8f64.v8p0(<8 x double> %val, <8 x ptr> %ptrs, i32 8, <8 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_baseidx_zext_v8i32_v8f64(<8 x double> %val, ptr %base, <8 x i32> %idxs, <8 x i1> %m) {
|
|
; RV32V-LABEL: mscatter_baseidx_zext_v8i32_v8f64:
|
|
; RV32V: # %bb.0:
|
|
; RV32V-NEXT: vsetivli zero, 8, e32, m2, ta, ma
|
|
; RV32V-NEXT: vsll.vi v12, v12, 3
|
|
; RV32V-NEXT: vsetvli zero, zero, e64, m4, ta, ma
|
|
; RV32V-NEXT: vsoxei32.v v8, (a0), v12, v0.t
|
|
; RV32V-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_baseidx_zext_v8i32_v8f64:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: li a1, 8
|
|
; RV64V-NEXT: vsetivli zero, 8, e32, m2, ta, ma
|
|
; RV64V-NEXT: vwmulu.vx v16, v12, a1
|
|
; RV64V-NEXT: vsetvli zero, zero, e64, m4, ta, ma
|
|
; RV64V-NEXT: vsoxei64.v v8, (a0), v16, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV32ZVE32F-LABEL: mscatter_baseidx_zext_v8i32_v8f64:
|
|
; RV32ZVE32F: # %bb.0:
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 8, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vsll.vi v8, v8, 3
|
|
; RV32ZVE32F-NEXT: vsetvli zero, zero, e8, mf2, ta, ma
|
|
; RV32ZVE32F-NEXT: vmv.x.s a1, v0
|
|
; RV32ZVE32F-NEXT: andi a2, a1, 1
|
|
; RV32ZVE32F-NEXT: vsetvli zero, zero, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vadd.vx v8, v8, a0
|
|
; RV32ZVE32F-NEXT: bnez a2, .LBB99_9
|
|
; RV32ZVE32F-NEXT: # %bb.1: # %else
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 2
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB99_10
|
|
; RV32ZVE32F-NEXT: .LBB99_2: # %else2
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 4
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB99_11
|
|
; RV32ZVE32F-NEXT: .LBB99_3: # %else4
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 8
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB99_12
|
|
; RV32ZVE32F-NEXT: .LBB99_4: # %else6
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 16
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB99_13
|
|
; RV32ZVE32F-NEXT: .LBB99_5: # %else8
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 32
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB99_14
|
|
; RV32ZVE32F-NEXT: .LBB99_6: # %else10
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 64
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB99_15
|
|
; RV32ZVE32F-NEXT: .LBB99_7: # %else12
|
|
; RV32ZVE32F-NEXT: andi a0, a1, -128
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB99_16
|
|
; RV32ZVE32F-NEXT: .LBB99_8: # %else14
|
|
; RV32ZVE32F-NEXT: ret
|
|
; RV32ZVE32F-NEXT: .LBB99_9: # %cond.store
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v8
|
|
; RV32ZVE32F-NEXT: fsd fa0, 0(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 2
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB99_2
|
|
; RV32ZVE32F-NEXT: .LBB99_10: # %cond.store1
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 1
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: fsd fa1, 0(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 4
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB99_3
|
|
; RV32ZVE32F-NEXT: .LBB99_11: # %cond.store3
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 2
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: fsd fa2, 0(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 8
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB99_4
|
|
; RV32ZVE32F-NEXT: .LBB99_12: # %cond.store5
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 3
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: fsd fa3, 0(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 16
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB99_5
|
|
; RV32ZVE32F-NEXT: .LBB99_13: # %cond.store7
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 4
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: fsd fa4, 0(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 32
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB99_6
|
|
; RV32ZVE32F-NEXT: .LBB99_14: # %cond.store9
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 5
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: fsd fa5, 0(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 64
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB99_7
|
|
; RV32ZVE32F-NEXT: .LBB99_15: # %cond.store11
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 6
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: fsd fa6, 0(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, a1, -128
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB99_8
|
|
; RV32ZVE32F-NEXT: .LBB99_16: # %cond.store13
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v8, v8, 7
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v8
|
|
; RV32ZVE32F-NEXT: fsd fa7, 0(a0)
|
|
; RV32ZVE32F-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_baseidx_zext_v8i32_v8f64:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a1, v0
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 1
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB99_2
|
|
; RV64ZVE32F-NEXT: # %bb.1: # %cond.store
|
|
; RV64ZVE32F-NEXT: vsetvli zero, zero, e32, m4, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v8
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 32
|
|
; RV64ZVE32F-NEXT: srli a2, a2, 29
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fsd fa0, 0(a2)
|
|
; RV64ZVE32F-NEXT: .LBB99_2: # %else
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 2
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB99_4
|
|
; RV64ZVE32F-NEXT: # %bb.3: # %cond.store1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v8, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 32
|
|
; RV64ZVE32F-NEXT: srli a2, a2, 29
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fsd fa1, 0(a2)
|
|
; RV64ZVE32F-NEXT: .LBB99_4: # %else2
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 4
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v8, 2
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB99_6
|
|
; RV64ZVE32F-NEXT: # %bb.5: # %cond.store3
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 32
|
|
; RV64ZVE32F-NEXT: srli a2, a2, 29
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fsd fa2, 0(a2)
|
|
; RV64ZVE32F-NEXT: .LBB99_6: # %else4
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 8
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 4, e32, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 4
|
|
; RV64ZVE32F-NEXT: bnez a2, .LBB99_13
|
|
; RV64ZVE32F-NEXT: # %bb.7: # %else6
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 16
|
|
; RV64ZVE32F-NEXT: bnez a2, .LBB99_14
|
|
; RV64ZVE32F-NEXT: .LBB99_8: # %else8
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 32
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB99_10
|
|
; RV64ZVE32F-NEXT: .LBB99_9: # %cond.store9
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v9
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 32
|
|
; RV64ZVE32F-NEXT: srli a2, a2, 29
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fsd fa5, 0(a2)
|
|
; RV64ZVE32F-NEXT: .LBB99_10: # %else10
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 64
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 2
|
|
; RV64ZVE32F-NEXT: bnez a2, .LBB99_15
|
|
; RV64ZVE32F-NEXT: # %bb.11: # %else12
|
|
; RV64ZVE32F-NEXT: andi a1, a1, -128
|
|
; RV64ZVE32F-NEXT: bnez a1, .LBB99_16
|
|
; RV64ZVE32F-NEXT: .LBB99_12: # %else14
|
|
; RV64ZVE32F-NEXT: ret
|
|
; RV64ZVE32F-NEXT: .LBB99_13: # %cond.store5
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v10, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v9
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 32
|
|
; RV64ZVE32F-NEXT: srli a2, a2, 29
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fsd fa3, 0(a2)
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 16
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB99_8
|
|
; RV64ZVE32F-NEXT: .LBB99_14: # %cond.store7
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v8
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 32
|
|
; RV64ZVE32F-NEXT: srli a2, a2, 29
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fsd fa4, 0(a2)
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 32
|
|
; RV64ZVE32F-NEXT: bnez a2, .LBB99_9
|
|
; RV64ZVE32F-NEXT: j .LBB99_10
|
|
; RV64ZVE32F-NEXT: .LBB99_15: # %cond.store11
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v8
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 32
|
|
; RV64ZVE32F-NEXT: srli a2, a2, 29
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: fsd fa6, 0(a2)
|
|
; RV64ZVE32F-NEXT: andi a1, a1, -128
|
|
; RV64ZVE32F-NEXT: beqz a1, .LBB99_12
|
|
; RV64ZVE32F-NEXT: .LBB99_16: # %cond.store13
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a1, v8
|
|
; RV64ZVE32F-NEXT: slli a1, a1, 32
|
|
; RV64ZVE32F-NEXT: srli a1, a1, 29
|
|
; RV64ZVE32F-NEXT: add a0, a0, a1
|
|
; RV64ZVE32F-NEXT: fsd fa7, 0(a0)
|
|
; RV64ZVE32F-NEXT: ret
|
|
%eidxs = zext <8 x i32> %idxs to <8 x i64>
|
|
%ptrs = getelementptr inbounds double, ptr %base, <8 x i64> %eidxs
|
|
call void @llvm.masked.scatter.v8f64.v8p0(<8 x double> %val, <8 x ptr> %ptrs, i32 8, <8 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_baseidx_v8f64(<8 x double> %val, ptr %base, <8 x i64> %idxs, <8 x i1> %m) {
|
|
; RV32V-LABEL: mscatter_baseidx_v8f64:
|
|
; RV32V: # %bb.0:
|
|
; RV32V-NEXT: vsetivli zero, 8, e32, m2, ta, ma
|
|
; RV32V-NEXT: vnsrl.wi v16, v12, 0
|
|
; RV32V-NEXT: vsll.vi v12, v16, 3
|
|
; RV32V-NEXT: vsetvli zero, zero, e64, m4, ta, ma
|
|
; RV32V-NEXT: vsoxei32.v v8, (a0), v12, v0.t
|
|
; RV32V-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_baseidx_v8f64:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: vsetivli zero, 8, e64, m4, ta, ma
|
|
; RV64V-NEXT: vsll.vi v12, v12, 3
|
|
; RV64V-NEXT: vsoxei64.v v8, (a0), v12, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV32ZVE32F-LABEL: mscatter_baseidx_v8f64:
|
|
; RV32ZVE32F: # %bb.0:
|
|
; RV32ZVE32F-NEXT: lw a2, 32(a1)
|
|
; RV32ZVE32F-NEXT: lw a3, 40(a1)
|
|
; RV32ZVE32F-NEXT: lw a4, 48(a1)
|
|
; RV32ZVE32F-NEXT: lw a5, 56(a1)
|
|
; RV32ZVE32F-NEXT: lw a6, 0(a1)
|
|
; RV32ZVE32F-NEXT: lw a7, 8(a1)
|
|
; RV32ZVE32F-NEXT: lw t0, 16(a1)
|
|
; RV32ZVE32F-NEXT: lw t1, 24(a1)
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 8, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vmv.v.x v8, a6
|
|
; RV32ZVE32F-NEXT: vsetvli zero, zero, e8, mf2, ta, ma
|
|
; RV32ZVE32F-NEXT: vmv.x.s a1, v0
|
|
; RV32ZVE32F-NEXT: vsetvli zero, zero, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslide1down.vx v8, v8, a7
|
|
; RV32ZVE32F-NEXT: vslide1down.vx v8, v8, t0
|
|
; RV32ZVE32F-NEXT: vslide1down.vx v8, v8, t1
|
|
; RV32ZVE32F-NEXT: vslide1down.vx v8, v8, a2
|
|
; RV32ZVE32F-NEXT: vslide1down.vx v8, v8, a3
|
|
; RV32ZVE32F-NEXT: vslide1down.vx v8, v8, a4
|
|
; RV32ZVE32F-NEXT: vslide1down.vx v8, v8, a5
|
|
; RV32ZVE32F-NEXT: vsll.vi v8, v8, 3
|
|
; RV32ZVE32F-NEXT: andi a2, a1, 1
|
|
; RV32ZVE32F-NEXT: vadd.vx v8, v8, a0
|
|
; RV32ZVE32F-NEXT: bnez a2, .LBB100_9
|
|
; RV32ZVE32F-NEXT: # %bb.1: # %else
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 2
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB100_10
|
|
; RV32ZVE32F-NEXT: .LBB100_2: # %else2
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 4
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB100_11
|
|
; RV32ZVE32F-NEXT: .LBB100_3: # %else4
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 8
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB100_12
|
|
; RV32ZVE32F-NEXT: .LBB100_4: # %else6
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 16
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB100_13
|
|
; RV32ZVE32F-NEXT: .LBB100_5: # %else8
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 32
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB100_14
|
|
; RV32ZVE32F-NEXT: .LBB100_6: # %else10
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 64
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB100_15
|
|
; RV32ZVE32F-NEXT: .LBB100_7: # %else12
|
|
; RV32ZVE32F-NEXT: andi a0, a1, -128
|
|
; RV32ZVE32F-NEXT: bnez a0, .LBB100_16
|
|
; RV32ZVE32F-NEXT: .LBB100_8: # %else14
|
|
; RV32ZVE32F-NEXT: ret
|
|
; RV32ZVE32F-NEXT: .LBB100_9: # %cond.store
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v8
|
|
; RV32ZVE32F-NEXT: fsd fa0, 0(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 2
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB100_2
|
|
; RV32ZVE32F-NEXT: .LBB100_10: # %cond.store1
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 1
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: fsd fa1, 0(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 4
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB100_3
|
|
; RV32ZVE32F-NEXT: .LBB100_11: # %cond.store3
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 2
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: fsd fa2, 0(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 8
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB100_4
|
|
; RV32ZVE32F-NEXT: .LBB100_12: # %cond.store5
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 3
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: fsd fa3, 0(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 16
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB100_5
|
|
; RV32ZVE32F-NEXT: .LBB100_13: # %cond.store7
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 4
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: fsd fa4, 0(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 32
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB100_6
|
|
; RV32ZVE32F-NEXT: .LBB100_14: # %cond.store9
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 5
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: fsd fa5, 0(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, a1, 64
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB100_7
|
|
; RV32ZVE32F-NEXT: .LBB100_15: # %cond.store11
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 6
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v10
|
|
; RV32ZVE32F-NEXT: fsd fa6, 0(a0)
|
|
; RV32ZVE32F-NEXT: andi a0, a1, -128
|
|
; RV32ZVE32F-NEXT: beqz a0, .LBB100_8
|
|
; RV32ZVE32F-NEXT: .LBB100_16: # %cond.store13
|
|
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
|
|
; RV32ZVE32F-NEXT: vslidedown.vi v8, v8, 7
|
|
; RV32ZVE32F-NEXT: vmv.x.s a0, v8
|
|
; RV32ZVE32F-NEXT: fsd fa7, 0(a0)
|
|
; RV32ZVE32F-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_baseidx_v8f64:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: ld t1, 8(a1)
|
|
; RV64ZVE32F-NEXT: ld t0, 16(a1)
|
|
; RV64ZVE32F-NEXT: ld a7, 24(a1)
|
|
; RV64ZVE32F-NEXT: ld a6, 32(a1)
|
|
; RV64ZVE32F-NEXT: ld a5, 40(a1)
|
|
; RV64ZVE32F-NEXT: ld a4, 48(a1)
|
|
; RV64ZVE32F-NEXT: ld a2, 56(a1)
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a3, v0
|
|
; RV64ZVE32F-NEXT: andi t2, a3, 1
|
|
; RV64ZVE32F-NEXT: bnez t2, .LBB100_9
|
|
; RV64ZVE32F-NEXT: # %bb.1: # %else
|
|
; RV64ZVE32F-NEXT: andi a1, a3, 2
|
|
; RV64ZVE32F-NEXT: bnez a1, .LBB100_10
|
|
; RV64ZVE32F-NEXT: .LBB100_2: # %else2
|
|
; RV64ZVE32F-NEXT: andi a1, a3, 4
|
|
; RV64ZVE32F-NEXT: bnez a1, .LBB100_11
|
|
; RV64ZVE32F-NEXT: .LBB100_3: # %else4
|
|
; RV64ZVE32F-NEXT: andi a1, a3, 8
|
|
; RV64ZVE32F-NEXT: bnez a1, .LBB100_12
|
|
; RV64ZVE32F-NEXT: .LBB100_4: # %else6
|
|
; RV64ZVE32F-NEXT: andi a1, a3, 16
|
|
; RV64ZVE32F-NEXT: bnez a1, .LBB100_13
|
|
; RV64ZVE32F-NEXT: .LBB100_5: # %else8
|
|
; RV64ZVE32F-NEXT: andi a1, a3, 32
|
|
; RV64ZVE32F-NEXT: bnez a1, .LBB100_14
|
|
; RV64ZVE32F-NEXT: .LBB100_6: # %else10
|
|
; RV64ZVE32F-NEXT: andi a1, a3, 64
|
|
; RV64ZVE32F-NEXT: bnez a1, .LBB100_15
|
|
; RV64ZVE32F-NEXT: .LBB100_7: # %else12
|
|
; RV64ZVE32F-NEXT: andi a1, a3, -128
|
|
; RV64ZVE32F-NEXT: bnez a1, .LBB100_16
|
|
; RV64ZVE32F-NEXT: .LBB100_8: # %else14
|
|
; RV64ZVE32F-NEXT: ret
|
|
; RV64ZVE32F-NEXT: .LBB100_9: # %cond.store
|
|
; RV64ZVE32F-NEXT: ld a1, 0(a1)
|
|
; RV64ZVE32F-NEXT: slli a1, a1, 3
|
|
; RV64ZVE32F-NEXT: add a1, a0, a1
|
|
; RV64ZVE32F-NEXT: fsd fa0, 0(a1)
|
|
; RV64ZVE32F-NEXT: andi a1, a3, 2
|
|
; RV64ZVE32F-NEXT: beqz a1, .LBB100_2
|
|
; RV64ZVE32F-NEXT: .LBB100_10: # %cond.store1
|
|
; RV64ZVE32F-NEXT: slli t1, t1, 3
|
|
; RV64ZVE32F-NEXT: add t1, a0, t1
|
|
; RV64ZVE32F-NEXT: fsd fa1, 0(t1)
|
|
; RV64ZVE32F-NEXT: andi a1, a3, 4
|
|
; RV64ZVE32F-NEXT: beqz a1, .LBB100_3
|
|
; RV64ZVE32F-NEXT: .LBB100_11: # %cond.store3
|
|
; RV64ZVE32F-NEXT: slli t0, t0, 3
|
|
; RV64ZVE32F-NEXT: add t0, a0, t0
|
|
; RV64ZVE32F-NEXT: fsd fa2, 0(t0)
|
|
; RV64ZVE32F-NEXT: andi a1, a3, 8
|
|
; RV64ZVE32F-NEXT: beqz a1, .LBB100_4
|
|
; RV64ZVE32F-NEXT: .LBB100_12: # %cond.store5
|
|
; RV64ZVE32F-NEXT: slli a7, a7, 3
|
|
; RV64ZVE32F-NEXT: add a7, a0, a7
|
|
; RV64ZVE32F-NEXT: fsd fa3, 0(a7)
|
|
; RV64ZVE32F-NEXT: andi a1, a3, 16
|
|
; RV64ZVE32F-NEXT: beqz a1, .LBB100_5
|
|
; RV64ZVE32F-NEXT: .LBB100_13: # %cond.store7
|
|
; RV64ZVE32F-NEXT: slli a6, a6, 3
|
|
; RV64ZVE32F-NEXT: add a6, a0, a6
|
|
; RV64ZVE32F-NEXT: fsd fa4, 0(a6)
|
|
; RV64ZVE32F-NEXT: andi a1, a3, 32
|
|
; RV64ZVE32F-NEXT: beqz a1, .LBB100_6
|
|
; RV64ZVE32F-NEXT: .LBB100_14: # %cond.store9
|
|
; RV64ZVE32F-NEXT: slli a5, a5, 3
|
|
; RV64ZVE32F-NEXT: add a5, a0, a5
|
|
; RV64ZVE32F-NEXT: fsd fa5, 0(a5)
|
|
; RV64ZVE32F-NEXT: andi a1, a3, 64
|
|
; RV64ZVE32F-NEXT: beqz a1, .LBB100_7
|
|
; RV64ZVE32F-NEXT: .LBB100_15: # %cond.store11
|
|
; RV64ZVE32F-NEXT: slli a4, a4, 3
|
|
; RV64ZVE32F-NEXT: add a4, a0, a4
|
|
; RV64ZVE32F-NEXT: fsd fa6, 0(a4)
|
|
; RV64ZVE32F-NEXT: andi a1, a3, -128
|
|
; RV64ZVE32F-NEXT: beqz a1, .LBB100_8
|
|
; RV64ZVE32F-NEXT: .LBB100_16: # %cond.store13
|
|
; RV64ZVE32F-NEXT: slli a2, a2, 3
|
|
; RV64ZVE32F-NEXT: add a0, a0, a2
|
|
; RV64ZVE32F-NEXT: fsd fa7, 0(a0)
|
|
; RV64ZVE32F-NEXT: ret
|
|
%ptrs = getelementptr inbounds double, ptr %base, <8 x i64> %idxs
|
|
call void @llvm.masked.scatter.v8f64.v8p0(<8 x double> %val, <8 x ptr> %ptrs, i32 8, <8 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_baseidx_v16i8(<16 x i8> %val, ptr %base, <16 x i8> %idxs, <16 x i1> %m) {
|
|
; RV32-LABEL: mscatter_baseidx_v16i8:
|
|
; RV32: # %bb.0:
|
|
; RV32-NEXT: vsetivli zero, 16, e32, m4, ta, ma
|
|
; RV32-NEXT: vsext.vf4 v12, v9
|
|
; RV32-NEXT: vsetvli zero, zero, e8, m1, ta, ma
|
|
; RV32-NEXT: vsoxei32.v v8, (a0), v12, v0.t
|
|
; RV32-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_baseidx_v16i8:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: vsetivli zero, 16, e64, m8, ta, ma
|
|
; RV64V-NEXT: vsext.vf8 v16, v9
|
|
; RV64V-NEXT: vsetvli zero, zero, e8, m1, ta, ma
|
|
; RV64V-NEXT: vsoxei64.v v8, (a0), v16, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_baseidx_v16i8:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a1, v0
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 1
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB101_2
|
|
; RV64ZVE32F-NEXT: # %bb.1: # %cond.store
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v9
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vse8.v v8, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB101_2: # %else
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 2
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB101_4
|
|
; RV64ZVE32F-NEXT: # %bb.3: # %cond.store1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v9, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v8, 1
|
|
; RV64ZVE32F-NEXT: vse8.v v10, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB101_4: # %else2
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 4
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v9, 2
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB101_6
|
|
; RV64ZVE32F-NEXT: # %bb.5: # %cond.store3
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v8, 2
|
|
; RV64ZVE32F-NEXT: vse8.v v11, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB101_6: # %else4
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 8
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB101_8
|
|
; RV64ZVE32F-NEXT: # %bb.7: # %cond.store5
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v10, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v8, 3
|
|
; RV64ZVE32F-NEXT: vse8.v v10, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB101_8: # %else6
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 16
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 4, e8, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v9, 4
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB101_10
|
|
; RV64ZVE32F-NEXT: # %bb.9: # %cond.store7
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v8, 4
|
|
; RV64ZVE32F-NEXT: vse8.v v11, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB101_10: # %else8
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 32
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB101_12
|
|
; RV64ZVE32F-NEXT: # %bb.11: # %cond.store9
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v10, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v11
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v8, 5
|
|
; RV64ZVE32F-NEXT: vse8.v v11, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB101_12: # %else10
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 64
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v10, 2
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB101_14
|
|
; RV64ZVE32F-NEXT: # %bb.13: # %cond.store11
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v8, 6
|
|
; RV64ZVE32F-NEXT: vse8.v v11, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB101_14: # %else12
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 128
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB101_16
|
|
; RV64ZVE32F-NEXT: # %bb.15: # %cond.store13
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v10, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v8, 7
|
|
; RV64ZVE32F-NEXT: vse8.v v10, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB101_16: # %else14
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 256
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 8, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v9, 8
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB101_18
|
|
; RV64ZVE32F-NEXT: # %bb.17: # %cond.store15
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v9
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v8, 8
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse8.v v10, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB101_18: # %else16
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 512
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB101_20
|
|
; RV64ZVE32F-NEXT: # %bb.19: # %cond.store17
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v9, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v8, 9
|
|
; RV64ZVE32F-NEXT: vse8.v v10, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB101_20: # %else18
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 1024
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v9, 2
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB101_22
|
|
; RV64ZVE32F-NEXT: # %bb.21: # %cond.store19
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v8, 10
|
|
; RV64ZVE32F-NEXT: vse8.v v11, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB101_22: # %else20
|
|
; RV64ZVE32F-NEXT: slli a2, a1, 52
|
|
; RV64ZVE32F-NEXT: bgez a2, .LBB101_24
|
|
; RV64ZVE32F-NEXT: # %bb.23: # %cond.store21
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v10, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v8, 11
|
|
; RV64ZVE32F-NEXT: vse8.v v10, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB101_24: # %else22
|
|
; RV64ZVE32F-NEXT: slli a2, a1, 51
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 4, e8, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v9, 4
|
|
; RV64ZVE32F-NEXT: bgez a2, .LBB101_26
|
|
; RV64ZVE32F-NEXT: # %bb.25: # %cond.store23
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v9
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v8, 12
|
|
; RV64ZVE32F-NEXT: vse8.v v10, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB101_26: # %else24
|
|
; RV64ZVE32F-NEXT: slli a2, a1, 50
|
|
; RV64ZVE32F-NEXT: bgez a2, .LBB101_28
|
|
; RV64ZVE32F-NEXT: # %bb.27: # %cond.store25
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v9, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v8, 13
|
|
; RV64ZVE32F-NEXT: vse8.v v10, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB101_28: # %else26
|
|
; RV64ZVE32F-NEXT: slli a2, a1, 49
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v9, 2
|
|
; RV64ZVE32F-NEXT: bgez a2, .LBB101_30
|
|
; RV64ZVE32F-NEXT: # %bb.29: # %cond.store27
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v9
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v8, 14
|
|
; RV64ZVE32F-NEXT: vse8.v v10, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB101_30: # %else28
|
|
; RV64ZVE32F-NEXT: lui a2, 1048568
|
|
; RV64ZVE32F-NEXT: and a1, a1, a2
|
|
; RV64ZVE32F-NEXT: beqz a1, .LBB101_32
|
|
; RV64ZVE32F-NEXT: # %bb.31: # %cond.store29
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v9, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a1, v9
|
|
; RV64ZVE32F-NEXT: add a0, a0, a1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 15
|
|
; RV64ZVE32F-NEXT: vse8.v v8, (a0)
|
|
; RV64ZVE32F-NEXT: .LBB101_32: # %else30
|
|
; RV64ZVE32F-NEXT: ret
|
|
%ptrs = getelementptr inbounds i8, ptr %base, <16 x i8> %idxs
|
|
call void @llvm.masked.scatter.v16i8.v16p0(<16 x i8> %val, <16 x ptr> %ptrs, i32 1, <16 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_baseidx_v32i8(<32 x i8> %val, ptr %base, <32 x i8> %idxs, <32 x i1> %m) {
|
|
; RV32-LABEL: mscatter_baseidx_v32i8:
|
|
; RV32: # %bb.0:
|
|
; RV32-NEXT: li a1, 32
|
|
; RV32-NEXT: vsetvli zero, a1, e32, m8, ta, ma
|
|
; RV32-NEXT: vsext.vf4 v16, v10
|
|
; RV32-NEXT: vsetvli zero, zero, e8, m2, ta, ma
|
|
; RV32-NEXT: vsoxei32.v v8, (a0), v16, v0.t
|
|
; RV32-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_baseidx_v32i8:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: vsetivli zero, 16, e64, m8, ta, ma
|
|
; RV64V-NEXT: vsext.vf8 v16, v10
|
|
; RV64V-NEXT: vsetvli zero, zero, e8, m1, ta, ma
|
|
; RV64V-NEXT: vsoxei64.v v8, (a0), v16, v0.t
|
|
; RV64V-NEXT: vsetivli zero, 16, e8, m2, ta, ma
|
|
; RV64V-NEXT: vslidedown.vi v16, v8, 16
|
|
; RV64V-NEXT: vslidedown.vi v18, v10, 16
|
|
; RV64V-NEXT: vsetivli zero, 2, e8, mf4, ta, ma
|
|
; RV64V-NEXT: vslidedown.vi v0, v0, 2
|
|
; RV64V-NEXT: vsetivli zero, 16, e64, m8, ta, ma
|
|
; RV64V-NEXT: vsext.vf8 v8, v18
|
|
; RV64V-NEXT: vsetvli zero, zero, e8, m1, ta, ma
|
|
; RV64V-NEXT: vsoxei64.v v16, (a0), v8, v0.t
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_baseidx_v32i8:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a1, v0
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 1
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB102_2
|
|
; RV64ZVE32F-NEXT: # %bb.1: # %cond.store
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vse8.v v8, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB102_2: # %else
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 2
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB102_4
|
|
; RV64ZVE32F-NEXT: # %bb.3: # %cond.store1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v10, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v12
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v8, 1
|
|
; RV64ZVE32F-NEXT: vse8.v v12, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB102_4: # %else2
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 4
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v10, 2
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB102_6
|
|
; RV64ZVE32F-NEXT: # %bb.5: # %cond.store3
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v12
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v13, v8, 2
|
|
; RV64ZVE32F-NEXT: vse8.v v13, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB102_6: # %else4
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 8
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB102_8
|
|
; RV64ZVE32F-NEXT: # %bb.7: # %cond.store5
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v12, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v12
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v8, 3
|
|
; RV64ZVE32F-NEXT: vse8.v v12, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB102_8: # %else6
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 16
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 4, e8, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v10, 4
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB102_10
|
|
; RV64ZVE32F-NEXT: # %bb.9: # %cond.store7
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v12
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v13, v8, 4
|
|
; RV64ZVE32F-NEXT: vse8.v v13, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB102_10: # %else8
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 32
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB102_12
|
|
; RV64ZVE32F-NEXT: # %bb.11: # %cond.store9
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v13, v12, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v13
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v13, v8, 5
|
|
; RV64ZVE32F-NEXT: vse8.v v13, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB102_12: # %else10
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 64
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v12, 2
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB102_14
|
|
; RV64ZVE32F-NEXT: # %bb.13: # %cond.store11
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v12
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v13, v8, 6
|
|
; RV64ZVE32F-NEXT: vse8.v v13, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB102_14: # %else12
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 128
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB102_16
|
|
; RV64ZVE32F-NEXT: # %bb.15: # %cond.store13
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v12, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v12
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v8, 7
|
|
; RV64ZVE32F-NEXT: vse8.v v12, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB102_16: # %else14
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 256
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 8, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v10, 8
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB102_18
|
|
; RV64ZVE32F-NEXT: # %bb.17: # %cond.store15
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v12
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v13, v8, 8
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse8.v v13, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB102_18: # %else16
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 512
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB102_20
|
|
; RV64ZVE32F-NEXT: # %bb.19: # %cond.store17
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v13, v12, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v13
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v13, v8, 9
|
|
; RV64ZVE32F-NEXT: vse8.v v13, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB102_20: # %else18
|
|
; RV64ZVE32F-NEXT: andi a2, a1, 1024
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v13, v12, 2
|
|
; RV64ZVE32F-NEXT: beqz a2, .LBB102_22
|
|
; RV64ZVE32F-NEXT: # %bb.21: # %cond.store19
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v13
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v14, v8, 10
|
|
; RV64ZVE32F-NEXT: vse8.v v14, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB102_22: # %else20
|
|
; RV64ZVE32F-NEXT: slli a2, a1, 52
|
|
; RV64ZVE32F-NEXT: bgez a2, .LBB102_24
|
|
; RV64ZVE32F-NEXT: # %bb.23: # %cond.store21
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v13, v13, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v13
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v13, v8, 11
|
|
; RV64ZVE32F-NEXT: vse8.v v13, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB102_24: # %else22
|
|
; RV64ZVE32F-NEXT: slli a2, a1, 51
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 4, e8, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v12, 4
|
|
; RV64ZVE32F-NEXT: bgez a2, .LBB102_26
|
|
; RV64ZVE32F-NEXT: # %bb.25: # %cond.store23
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v12
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v13, v8, 12
|
|
; RV64ZVE32F-NEXT: vse8.v v13, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB102_26: # %else24
|
|
; RV64ZVE32F-NEXT: slli a2, a1, 50
|
|
; RV64ZVE32F-NEXT: bgez a2, .LBB102_28
|
|
; RV64ZVE32F-NEXT: # %bb.27: # %cond.store25
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v13, v12, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v13
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v13, v8, 13
|
|
; RV64ZVE32F-NEXT: vse8.v v13, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB102_28: # %else26
|
|
; RV64ZVE32F-NEXT: slli a2, a1, 49
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v12, 2
|
|
; RV64ZVE32F-NEXT: bgez a2, .LBB102_30
|
|
; RV64ZVE32F-NEXT: # %bb.29: # %cond.store27
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v12
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v13, v8, 14
|
|
; RV64ZVE32F-NEXT: vse8.v v13, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB102_30: # %else28
|
|
; RV64ZVE32F-NEXT: slli a2, a1, 48
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 16, e8, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v10, 16
|
|
; RV64ZVE32F-NEXT: bltz a2, .LBB102_63
|
|
; RV64ZVE32F-NEXT: # %bb.31: # %else30
|
|
; RV64ZVE32F-NEXT: slli a2, a1, 47
|
|
; RV64ZVE32F-NEXT: bltz a2, .LBB102_64
|
|
; RV64ZVE32F-NEXT: .LBB102_32: # %else32
|
|
; RV64ZVE32F-NEXT: slli a2, a1, 46
|
|
; RV64ZVE32F-NEXT: bgez a2, .LBB102_34
|
|
; RV64ZVE32F-NEXT: .LBB102_33: # %cond.store33
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v10, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v11
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v8, 17
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse8.v v12, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB102_34: # %else34
|
|
; RV64ZVE32F-NEXT: slli a2, a1, 45
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v10, 2
|
|
; RV64ZVE32F-NEXT: bgez a2, .LBB102_36
|
|
; RV64ZVE32F-NEXT: # %bb.35: # %cond.store35
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v11
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v8, 18
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse8.v v12, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB102_36: # %else36
|
|
; RV64ZVE32F-NEXT: slli a2, a1, 44
|
|
; RV64ZVE32F-NEXT: bgez a2, .LBB102_38
|
|
; RV64ZVE32F-NEXT: # %bb.37: # %cond.store37
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v11, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v11
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v8, 19
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse8.v v12, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB102_38: # %else38
|
|
; RV64ZVE32F-NEXT: slli a2, a1, 43
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 4, e8, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v10, 4
|
|
; RV64ZVE32F-NEXT: bgez a2, .LBB102_40
|
|
; RV64ZVE32F-NEXT: # %bb.39: # %cond.store39
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v11
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v8, 20
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse8.v v12, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB102_40: # %else40
|
|
; RV64ZVE32F-NEXT: slli a2, a1, 42
|
|
; RV64ZVE32F-NEXT: bgez a2, .LBB102_42
|
|
; RV64ZVE32F-NEXT: # %bb.41: # %cond.store41
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v11, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v12
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v8, 21
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse8.v v12, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB102_42: # %else42
|
|
; RV64ZVE32F-NEXT: slli a2, a1, 41
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v11, 2
|
|
; RV64ZVE32F-NEXT: bgez a2, .LBB102_44
|
|
; RV64ZVE32F-NEXT: # %bb.43: # %cond.store43
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v11
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v8, 22
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse8.v v12, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB102_44: # %else44
|
|
; RV64ZVE32F-NEXT: slli a2, a1, 40
|
|
; RV64ZVE32F-NEXT: bgez a2, .LBB102_46
|
|
; RV64ZVE32F-NEXT: # %bb.45: # %cond.store45
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v11, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v11
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v8, 23
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse8.v v12, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB102_46: # %else46
|
|
; RV64ZVE32F-NEXT: slli a2, a1, 39
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 8, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v10, 8
|
|
; RV64ZVE32F-NEXT: bgez a2, .LBB102_48
|
|
; RV64ZVE32F-NEXT: # %bb.47: # %cond.store47
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v8, 24
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse8.v v12, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB102_48: # %else48
|
|
; RV64ZVE32F-NEXT: slli a2, a1, 38
|
|
; RV64ZVE32F-NEXT: bgez a2, .LBB102_50
|
|
; RV64ZVE32F-NEXT: # %bb.49: # %cond.store49
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v10, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v11
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v8, 25
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse8.v v12, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB102_50: # %else50
|
|
; RV64ZVE32F-NEXT: slli a2, a1, 37
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v10, 2
|
|
; RV64ZVE32F-NEXT: bgez a2, .LBB102_52
|
|
; RV64ZVE32F-NEXT: # %bb.51: # %cond.store51
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v11
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v8, 26
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse8.v v12, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB102_52: # %else52
|
|
; RV64ZVE32F-NEXT: slli a2, a1, 36
|
|
; RV64ZVE32F-NEXT: bgez a2, .LBB102_54
|
|
; RV64ZVE32F-NEXT: # %bb.53: # %cond.store53
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v11, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v11
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v8, 27
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse8.v v12, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB102_54: # %else54
|
|
; RV64ZVE32F-NEXT: slli a2, a1, 35
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 4, e8, mf2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v10, 4
|
|
; RV64ZVE32F-NEXT: bgez a2, .LBB102_56
|
|
; RV64ZVE32F-NEXT: # %bb.55: # %cond.store55
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v8, 28
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse8.v v12, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB102_56: # %else56
|
|
; RV64ZVE32F-NEXT: slli a2, a1, 34
|
|
; RV64ZVE32F-NEXT: bgez a2, .LBB102_58
|
|
; RV64ZVE32F-NEXT: # %bb.57: # %cond.store57
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v10, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v11
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v8, 29
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse8.v v12, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB102_58: # %else58
|
|
; RV64ZVE32F-NEXT: slli a2, a1, 33
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 2, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v10, 2
|
|
; RV64ZVE32F-NEXT: bgez a2, .LBB102_60
|
|
; RV64ZVE32F-NEXT: # %bb.59: # %cond.store59
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v8, 30
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse8.v v12, (a2)
|
|
; RV64ZVE32F-NEXT: .LBB102_60: # %else60
|
|
; RV64ZVE32F-NEXT: lui a2, 524288
|
|
; RV64ZVE32F-NEXT: and a1, a1, a2
|
|
; RV64ZVE32F-NEXT: beqz a1, .LBB102_62
|
|
; RV64ZVE32F-NEXT: # %bb.61: # %cond.store61
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v10, v10, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a1, v10
|
|
; RV64ZVE32F-NEXT: add a0, a0, a1
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 31
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse8.v v8, (a0)
|
|
; RV64ZVE32F-NEXT: .LBB102_62: # %else62
|
|
; RV64ZVE32F-NEXT: ret
|
|
; RV64ZVE32F-NEXT: .LBB102_63: # %cond.store29
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v12, 1
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v11
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v11, v8, 15
|
|
; RV64ZVE32F-NEXT: vse8.v v11, (a2)
|
|
; RV64ZVE32F-NEXT: slli a2, a1, 47
|
|
; RV64ZVE32F-NEXT: bgez a2, .LBB102_32
|
|
; RV64ZVE32F-NEXT: .LBB102_64: # %cond.store31
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m2, ta, ma
|
|
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
|
|
; RV64ZVE32F-NEXT: add a2, a0, a2
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v12, v8, 16
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vse8.v v12, (a2)
|
|
; RV64ZVE32F-NEXT: slli a2, a1, 46
|
|
; RV64ZVE32F-NEXT: bltz a2, .LBB102_33
|
|
; RV64ZVE32F-NEXT: j .LBB102_34
|
|
%ptrs = getelementptr inbounds i8, ptr %base, <32 x i8> %idxs
|
|
call void @llvm.masked.scatter.v32i8.v32p0(<32 x i8> %val, <32 x ptr> %ptrs, i32 1, <32 x i1> %m)
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_unit_stride(<8 x i16> %val, ptr %base) {
|
|
; CHECK-LABEL: mscatter_unit_stride:
|
|
; CHECK: # %bb.0:
|
|
; CHECK-NEXT: vsetivli zero, 8, e16, m1, ta, ma
|
|
; CHECK-NEXT: vse16.v v8, (a0)
|
|
; CHECK-NEXT: ret
|
|
%ptrs = getelementptr inbounds i16, ptr %base, <8 x i64> <i64 0, i64 1, i64 2, i64 3, i64 4, i64 5, i64 6, i64 7>
|
|
call void @llvm.masked.scatter.v8i16.v8p0(<8 x i16> %val, <8 x ptr> %ptrs, i32 2, <8 x i1> splat (i1 true))
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_unit_stride_with_offset(<8 x i16> %val, ptr %base) {
|
|
; CHECK-LABEL: mscatter_unit_stride_with_offset:
|
|
; CHECK: # %bb.0:
|
|
; CHECK-NEXT: addi a0, a0, 10
|
|
; CHECK-NEXT: vsetivli zero, 8, e16, m1, ta, ma
|
|
; CHECK-NEXT: vse16.v v8, (a0)
|
|
; CHECK-NEXT: ret
|
|
%ptrs = getelementptr inbounds i16, ptr %base, <8 x i64> <i64 5, i64 6, i64 7, i64 8, i64 9, i64 10, i64 11, i64 12>
|
|
call void @llvm.masked.scatter.v8i16.v8p0(<8 x i16> %val, <8 x ptr> %ptrs, i32 2, <8 x i1> splat (i1 true))
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_shuffle_reverse(<8 x i16> %val, ptr %base) {
|
|
; CHECK-LABEL: mscatter_shuffle_reverse:
|
|
; CHECK: # %bb.0:
|
|
; CHECK-NEXT: addi a0, a0, 14
|
|
; CHECK-NEXT: li a1, -2
|
|
; CHECK-NEXT: vsetivli zero, 8, e16, m1, ta, ma
|
|
; CHECK-NEXT: vsse16.v v8, (a0), a1
|
|
; CHECK-NEXT: ret
|
|
%ptrs = getelementptr inbounds i16, ptr %base, <8 x i64> <i64 7, i64 6, i64 5, i64 4, i64 3, i64 2, i64 1, i64 0>
|
|
call void @llvm.masked.scatter.v8i16.v8p0(<8 x i16> %val, <8 x ptr> %ptrs, i32 2, <8 x i1> splat (i1 true))
|
|
ret void
|
|
}
|
|
|
|
define void @mscatter_shuffle_rotate(<8 x i16> %val, ptr %base) {
|
|
; RV32-LABEL: mscatter_shuffle_rotate:
|
|
; RV32: # %bb.0:
|
|
; RV32-NEXT: vsetivli zero, 8, e16, m1, ta, ma
|
|
; RV32-NEXT: vslidedown.vi v9, v8, 4
|
|
; RV32-NEXT: vslideup.vi v9, v8, 4
|
|
; RV32-NEXT: vse16.v v9, (a0)
|
|
; RV32-NEXT: ret
|
|
;
|
|
; RV64V-LABEL: mscatter_shuffle_rotate:
|
|
; RV64V: # %bb.0:
|
|
; RV64V-NEXT: vsetivli zero, 8, e16, m1, ta, ma
|
|
; RV64V-NEXT: vslidedown.vi v9, v8, 4
|
|
; RV64V-NEXT: vslideup.vi v9, v8, 4
|
|
; RV64V-NEXT: vse16.v v9, (a0)
|
|
; RV64V-NEXT: ret
|
|
;
|
|
; RV64ZVE32F-LABEL: mscatter_shuffle_rotate:
|
|
; RV64ZVE32F: # %bb.0:
|
|
; RV64ZVE32F-NEXT: addi a1, a0, 6
|
|
; RV64ZVE32F-NEXT: addi a2, a0, 4
|
|
; RV64ZVE32F-NEXT: addi a3, a0, 2
|
|
; RV64ZVE32F-NEXT: addi a4, a0, 14
|
|
; RV64ZVE32F-NEXT: addi a5, a0, 12
|
|
; RV64ZVE32F-NEXT: addi a6, a0, 10
|
|
; RV64ZVE32F-NEXT: addi a7, a0, 8
|
|
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, m1, ta, ma
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 1
|
|
; RV64ZVE32F-NEXT: vse16.v v8, (a7)
|
|
; RV64ZVE32F-NEXT: vse16.v v9, (a6)
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 2
|
|
; RV64ZVE32F-NEXT: vse16.v v9, (a5)
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 3
|
|
; RV64ZVE32F-NEXT: vse16.v v9, (a4)
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 4
|
|
; RV64ZVE32F-NEXT: vse16.v v9, (a0)
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 5
|
|
; RV64ZVE32F-NEXT: vse16.v v9, (a3)
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 6
|
|
; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 7
|
|
; RV64ZVE32F-NEXT: vse16.v v9, (a2)
|
|
; RV64ZVE32F-NEXT: vse16.v v8, (a1)
|
|
; RV64ZVE32F-NEXT: ret
|
|
%ptrs = getelementptr inbounds i16, ptr %base, <8 x i64> <i64 4, i64 5, i64 6, i64 7, i64 0, i64 1, i64 2, i64 3>
|
|
call void @llvm.masked.scatter.v8i16.v8p0(<8 x i16> %val, <8 x ptr> %ptrs, i32 2, <8 x i1> splat (i1 true))
|
|
ret void
|
|
}
|
|
;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
|
|
; RV32V-ZVFH: {{.*}}
|
|
; RV32V-ZVFHMIN: {{.*}}
|
|
; RV32ZVE32F-ZVFH: {{.*}}
|
|
; RV32ZVE32F-ZVFHMIN: {{.*}}
|
|
; RV64: {{.*}}
|
|
; RV64V-ZVFH: {{.*}}
|
|
; RV64V-ZVFHMIN: {{.*}}
|