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
315 lines
11 KiB
LLVM
315 lines
11 KiB
LLVM
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
|
|
; RUN: llc -aarch64-sve-vector-bits-min=1024 < %s | FileCheck %s -check-prefixes=CHECK,VBITS_GE_1024
|
|
; RUN: llc -aarch64-sve-vector-bits-min=2048 < %s | FileCheck %s -check-prefixes=CHECK,VBITS_GE_2048
|
|
|
|
target triple = "aarch64-unknown-linux-gnu"
|
|
|
|
define <4 x i32> @load_zext_v4i16i32(ptr %ap) vscale_range(2,0) #0 {
|
|
; CHECK-LABEL: load_zext_v4i16i32:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ldp s0, s1, [x0]
|
|
; CHECK-NEXT: ushll v0.4s, v0.4h, #0
|
|
; CHECK-NEXT: ushll v1.4s, v1.4h, #0
|
|
; CHECK-NEXT: mov v0.d[1], v1.d[0]
|
|
; CHECK-NEXT: ret
|
|
%a = load <4 x i16>, ptr %ap
|
|
%val = zext <4 x i16> %a to <4 x i32>
|
|
ret <4 x i32> %val
|
|
}
|
|
|
|
; Don't try to use SVE for irregular types.
|
|
define <2 x i256> @load_zext_v2i64i256(ptr %ap) #0 {
|
|
; CHECK-LABEL: load_zext_v2i64i256:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ldp x0, x4, [x0]
|
|
; CHECK-NEXT: mov x1, xzr
|
|
; CHECK-NEXT: mov x2, xzr
|
|
; CHECK-NEXT: mov x3, xzr
|
|
; CHECK-NEXT: mov x5, xzr
|
|
; CHECK-NEXT: mov x6, xzr
|
|
; CHECK-NEXT: mov x7, xzr
|
|
; CHECK-NEXT: ret
|
|
%a = load <2 x i64>, ptr %ap
|
|
%val = zext <2 x i64> %a to <2 x i256>
|
|
ret <2 x i256> %val
|
|
}
|
|
|
|
define void @load_zext_v8i16i32(ptr %ap, ptr %b) vscale_range(2,0) #0 {
|
|
; CHECK-LABEL: load_zext_v8i16i32:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.s, vl8
|
|
; CHECK-NEXT: ld1h { z0.s }, p0/z, [x0]
|
|
; CHECK-NEXT: st1w { z0.s }, p0, [x1]
|
|
; CHECK-NEXT: ret
|
|
%a = load <8 x i16>, ptr %ap
|
|
%val = zext <8 x i16> %a to <8 x i32>
|
|
store <8 x i32> %val, ptr %b
|
|
ret void
|
|
}
|
|
|
|
define void @load_zext_v16i16i32(ptr %ap, ptr %b) vscale_range(4,0) #0 {
|
|
; CHECK-LABEL: load_zext_v16i16i32:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.s, vl16
|
|
; CHECK-NEXT: ld1h { z0.s }, p0/z, [x0]
|
|
; CHECK-NEXT: st1w { z0.s }, p0, [x1]
|
|
; CHECK-NEXT: ret
|
|
%a = load <16 x i16>, ptr %ap
|
|
%val = zext <16 x i16> %a to <16 x i32>
|
|
store <16 x i32> %val, ptr %b
|
|
ret void
|
|
}
|
|
|
|
define void @load_zext_v32i16i32(ptr %ap, ptr %b) vscale_range(8,0) #0 {
|
|
; CHECK-LABEL: load_zext_v32i16i32:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.s, vl32
|
|
; CHECK-NEXT: ld1h { z0.s }, p0/z, [x0]
|
|
; CHECK-NEXT: st1w { z0.s }, p0, [x1]
|
|
; CHECK-NEXT: ret
|
|
%a = load <32 x i16>, ptr %ap
|
|
%val = zext <32 x i16> %a to <32 x i32>
|
|
store <32 x i32> %val, ptr %b
|
|
ret void
|
|
}
|
|
|
|
define void @load_zext_v64i16i32(ptr %ap, ptr %b) #0 {
|
|
; VBITS_GE_1024-LABEL: load_zext_v64i16i32:
|
|
; VBITS_GE_1024: // %bb.0:
|
|
; VBITS_GE_1024-NEXT: ptrue p0.s, vl32
|
|
; VBITS_GE_1024-NEXT: mov x8, #32 // =0x20
|
|
; VBITS_GE_1024-NEXT: ld1h { z0.s }, p0/z, [x0, x8, lsl #1]
|
|
; VBITS_GE_1024-NEXT: ld1h { z1.s }, p0/z, [x0]
|
|
; VBITS_GE_1024-NEXT: st1w { z0.s }, p0, [x1, x8, lsl #2]
|
|
; VBITS_GE_1024-NEXT: st1w { z1.s }, p0, [x1]
|
|
; VBITS_GE_1024-NEXT: ret
|
|
;
|
|
; VBITS_GE_2048-LABEL: load_zext_v64i16i32:
|
|
; VBITS_GE_2048: // %bb.0:
|
|
; VBITS_GE_2048-NEXT: ptrue p0.s, vl64
|
|
; VBITS_GE_2048-NEXT: ld1h { z0.s }, p0/z, [x0]
|
|
; VBITS_GE_2048-NEXT: st1w { z0.s }, p0, [x1]
|
|
; VBITS_GE_2048-NEXT: ret
|
|
%a = load <64 x i16>, ptr %ap
|
|
%val = zext <64 x i16> %a to <64 x i32>
|
|
store <64 x i32> %val, ptr %b
|
|
ret void
|
|
}
|
|
|
|
define <4 x i32> @load_sext_v4i16i32(ptr %ap) vscale_range(2,0) #0 {
|
|
; CHECK-LABEL: load_sext_v4i16i32:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ldp s0, s1, [x0]
|
|
; CHECK-NEXT: sshll v0.4s, v0.4h, #0
|
|
; CHECK-NEXT: sshll v1.4s, v1.4h, #0
|
|
; CHECK-NEXT: mov v0.d[1], v1.d[0]
|
|
; CHECK-NEXT: ret
|
|
%a = load <4 x i16>, ptr %ap
|
|
%val = sext <4 x i16> %a to <4 x i32>
|
|
ret <4 x i32> %val
|
|
}
|
|
|
|
define void @load_sext_v8i16i32(ptr %ap, ptr %b) vscale_range(2,0) #0 {
|
|
; CHECK-LABEL: load_sext_v8i16i32:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.s, vl8
|
|
; CHECK-NEXT: ld1sh { z0.s }, p0/z, [x0]
|
|
; CHECK-NEXT: st1w { z0.s }, p0, [x1]
|
|
; CHECK-NEXT: ret
|
|
%a = load <8 x i16>, ptr %ap
|
|
%val = sext <8 x i16> %a to <8 x i32>
|
|
store <8 x i32> %val, ptr %b
|
|
ret void
|
|
}
|
|
|
|
define void @load_sext_v16i16i32(ptr %ap, ptr %b) vscale_range(4,0) #0 {
|
|
; CHECK-LABEL: load_sext_v16i16i32:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.s, vl16
|
|
; CHECK-NEXT: ld1sh { z0.s }, p0/z, [x0]
|
|
; CHECK-NEXT: st1w { z0.s }, p0, [x1]
|
|
; CHECK-NEXT: ret
|
|
%a = load <16 x i16>, ptr %ap
|
|
%val = sext <16 x i16> %a to <16 x i32>
|
|
store <16 x i32> %val, ptr %b
|
|
ret void
|
|
}
|
|
|
|
define void @load_sext_v32i16i32(ptr %ap, ptr %b) vscale_range(8,0) #0 {
|
|
; CHECK-LABEL: load_sext_v32i16i32:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.s, vl32
|
|
; CHECK-NEXT: ld1sh { z0.s }, p0/z, [x0]
|
|
; CHECK-NEXT: st1w { z0.s }, p0, [x1]
|
|
; CHECK-NEXT: ret
|
|
%a = load <32 x i16>, ptr %ap
|
|
%val = sext <32 x i16> %a to <32 x i32>
|
|
store <32 x i32> %val, ptr %b
|
|
ret void
|
|
}
|
|
|
|
define void @load_sext_v64i16i32(ptr %ap, ptr %b) #0 {
|
|
; VBITS_GE_1024-LABEL: load_sext_v64i16i32:
|
|
; VBITS_GE_1024: // %bb.0:
|
|
; VBITS_GE_1024-NEXT: ptrue p0.s, vl32
|
|
; VBITS_GE_1024-NEXT: mov x8, #32 // =0x20
|
|
; VBITS_GE_1024-NEXT: ld1sh { z0.s }, p0/z, [x0, x8, lsl #1]
|
|
; VBITS_GE_1024-NEXT: ld1sh { z1.s }, p0/z, [x0]
|
|
; VBITS_GE_1024-NEXT: st1w { z0.s }, p0, [x1, x8, lsl #2]
|
|
; VBITS_GE_1024-NEXT: st1w { z1.s }, p0, [x1]
|
|
; VBITS_GE_1024-NEXT: ret
|
|
;
|
|
; VBITS_GE_2048-LABEL: load_sext_v64i16i32:
|
|
; VBITS_GE_2048: // %bb.0:
|
|
; VBITS_GE_2048-NEXT: ptrue p0.s, vl64
|
|
; VBITS_GE_2048-NEXT: ld1sh { z0.s }, p0/z, [x0]
|
|
; VBITS_GE_2048-NEXT: st1w { z0.s }, p0, [x1]
|
|
; VBITS_GE_2048-NEXT: ret
|
|
%a = load <64 x i16>, ptr %ap
|
|
%val = sext <64 x i16> %a to <64 x i32>
|
|
store <64 x i32> %val, ptr %b
|
|
ret void
|
|
}
|
|
|
|
define void @load_zext_v32i8i64(ptr %ap, ptr %b) #0 {
|
|
; VBITS_GE_1024-LABEL: load_zext_v32i8i64:
|
|
; VBITS_GE_1024: // %bb.0:
|
|
; VBITS_GE_1024-NEXT: ptrue p0.d, vl16
|
|
; VBITS_GE_1024-NEXT: mov w8, #16 // =0x10
|
|
; VBITS_GE_1024-NEXT: ld1b { z0.d }, p0/z, [x0, x8]
|
|
; VBITS_GE_1024-NEXT: ld1b { z1.d }, p0/z, [x0]
|
|
; VBITS_GE_1024-NEXT: mov x8, #16 // =0x10
|
|
; VBITS_GE_1024-NEXT: st1d { z0.d }, p0, [x1, x8, lsl #3]
|
|
; VBITS_GE_1024-NEXT: st1d { z1.d }, p0, [x1]
|
|
; VBITS_GE_1024-NEXT: ret
|
|
;
|
|
; VBITS_GE_2048-LABEL: load_zext_v32i8i64:
|
|
; VBITS_GE_2048: // %bb.0:
|
|
; VBITS_GE_2048-NEXT: ptrue p0.d, vl32
|
|
; VBITS_GE_2048-NEXT: ld1b { z0.d }, p0/z, [x0]
|
|
; VBITS_GE_2048-NEXT: st1d { z0.d }, p0, [x1]
|
|
; VBITS_GE_2048-NEXT: ret
|
|
%a = load <32 x i8>, ptr %ap
|
|
%val = zext <32 x i8> %a to <32 x i64>
|
|
store <32 x i64> %val, ptr %b
|
|
ret void
|
|
}
|
|
|
|
define void @load_sext_v32i8i64(ptr %ap, ptr %b) #0 {
|
|
; VBITS_GE_1024-LABEL: load_sext_v32i8i64:
|
|
; VBITS_GE_1024: // %bb.0:
|
|
; VBITS_GE_1024-NEXT: ptrue p0.d, vl16
|
|
; VBITS_GE_1024-NEXT: mov w8, #16 // =0x10
|
|
; VBITS_GE_1024-NEXT: ld1sb { z0.d }, p0/z, [x0, x8]
|
|
; VBITS_GE_1024-NEXT: ld1sb { z1.d }, p0/z, [x0]
|
|
; VBITS_GE_1024-NEXT: mov x8, #16 // =0x10
|
|
; VBITS_GE_1024-NEXT: st1d { z0.d }, p0, [x1, x8, lsl #3]
|
|
; VBITS_GE_1024-NEXT: st1d { z1.d }, p0, [x1]
|
|
; VBITS_GE_1024-NEXT: ret
|
|
;
|
|
; VBITS_GE_2048-LABEL: load_sext_v32i8i64:
|
|
; VBITS_GE_2048: // %bb.0:
|
|
; VBITS_GE_2048-NEXT: ptrue p0.d, vl32
|
|
; VBITS_GE_2048-NEXT: ld1sb { z0.d }, p0/z, [x0]
|
|
; VBITS_GE_2048-NEXT: st1d { z0.d }, p0, [x1]
|
|
; VBITS_GE_2048-NEXT: ret
|
|
%a = load <32 x i8>, ptr %ap
|
|
%val = sext <32 x i8> %a to <32 x i64>
|
|
store <32 x i64> %val, ptr %b
|
|
ret void
|
|
}
|
|
|
|
define void @load_zext_v32i16i64(ptr %ap, ptr %b) #0 {
|
|
; VBITS_GE_1024-LABEL: load_zext_v32i16i64:
|
|
; VBITS_GE_1024: // %bb.0:
|
|
; VBITS_GE_1024-NEXT: ptrue p0.d, vl16
|
|
; VBITS_GE_1024-NEXT: mov x8, #16 // =0x10
|
|
; VBITS_GE_1024-NEXT: ld1h { z0.d }, p0/z, [x0, x8, lsl #1]
|
|
; VBITS_GE_1024-NEXT: ld1h { z1.d }, p0/z, [x0]
|
|
; VBITS_GE_1024-NEXT: st1d { z0.d }, p0, [x1, x8, lsl #3]
|
|
; VBITS_GE_1024-NEXT: st1d { z1.d }, p0, [x1]
|
|
; VBITS_GE_1024-NEXT: ret
|
|
;
|
|
; VBITS_GE_2048-LABEL: load_zext_v32i16i64:
|
|
; VBITS_GE_2048: // %bb.0:
|
|
; VBITS_GE_2048-NEXT: ptrue p0.d, vl32
|
|
; VBITS_GE_2048-NEXT: ld1h { z0.d }, p0/z, [x0]
|
|
; VBITS_GE_2048-NEXT: st1d { z0.d }, p0, [x1]
|
|
; VBITS_GE_2048-NEXT: ret
|
|
%a = load <32 x i16>, ptr %ap
|
|
%val = zext <32 x i16> %a to <32 x i64>
|
|
store <32 x i64> %val, ptr %b
|
|
ret void
|
|
}
|
|
|
|
define void @load_sext_v32i16i64(ptr %ap, ptr %b) #0 {
|
|
; VBITS_GE_1024-LABEL: load_sext_v32i16i64:
|
|
; VBITS_GE_1024: // %bb.0:
|
|
; VBITS_GE_1024-NEXT: ptrue p0.d, vl16
|
|
; VBITS_GE_1024-NEXT: mov x8, #16 // =0x10
|
|
; VBITS_GE_1024-NEXT: ld1sh { z0.d }, p0/z, [x0, x8, lsl #1]
|
|
; VBITS_GE_1024-NEXT: ld1sh { z1.d }, p0/z, [x0]
|
|
; VBITS_GE_1024-NEXT: st1d { z0.d }, p0, [x1, x8, lsl #3]
|
|
; VBITS_GE_1024-NEXT: st1d { z1.d }, p0, [x1]
|
|
; VBITS_GE_1024-NEXT: ret
|
|
;
|
|
; VBITS_GE_2048-LABEL: load_sext_v32i16i64:
|
|
; VBITS_GE_2048: // %bb.0:
|
|
; VBITS_GE_2048-NEXT: ptrue p0.d, vl32
|
|
; VBITS_GE_2048-NEXT: ld1sh { z0.d }, p0/z, [x0]
|
|
; VBITS_GE_2048-NEXT: st1d { z0.d }, p0, [x1]
|
|
; VBITS_GE_2048-NEXT: ret
|
|
%a = load <32 x i16>, ptr %ap
|
|
%val = sext <32 x i16> %a to <32 x i64>
|
|
store <32 x i64> %val, ptr %b
|
|
ret void
|
|
}
|
|
|
|
define void @load_zext_v32i32i64(ptr %ap, ptr %b) #0 {
|
|
; VBITS_GE_1024-LABEL: load_zext_v32i32i64:
|
|
; VBITS_GE_1024: // %bb.0:
|
|
; VBITS_GE_1024-NEXT: ptrue p0.d, vl16
|
|
; VBITS_GE_1024-NEXT: mov x8, #16 // =0x10
|
|
; VBITS_GE_1024-NEXT: ld1w { z0.d }, p0/z, [x0, x8, lsl #2]
|
|
; VBITS_GE_1024-NEXT: ld1w { z1.d }, p0/z, [x0]
|
|
; VBITS_GE_1024-NEXT: st1d { z0.d }, p0, [x1, x8, lsl #3]
|
|
; VBITS_GE_1024-NEXT: st1d { z1.d }, p0, [x1]
|
|
; VBITS_GE_1024-NEXT: ret
|
|
;
|
|
; VBITS_GE_2048-LABEL: load_zext_v32i32i64:
|
|
; VBITS_GE_2048: // %bb.0:
|
|
; VBITS_GE_2048-NEXT: ptrue p0.d, vl32
|
|
; VBITS_GE_2048-NEXT: ld1w { z0.d }, p0/z, [x0]
|
|
; VBITS_GE_2048-NEXT: st1d { z0.d }, p0, [x1]
|
|
; VBITS_GE_2048-NEXT: ret
|
|
%a = load <32 x i32>, ptr %ap
|
|
%val = zext <32 x i32> %a to <32 x i64>
|
|
store <32 x i64> %val, ptr %b
|
|
ret void
|
|
}
|
|
|
|
define void @load_sext_v32i32i64(ptr %ap, ptr %b) #0 {
|
|
; VBITS_GE_1024-LABEL: load_sext_v32i32i64:
|
|
; VBITS_GE_1024: // %bb.0:
|
|
; VBITS_GE_1024-NEXT: ptrue p0.d, vl16
|
|
; VBITS_GE_1024-NEXT: mov x8, #16 // =0x10
|
|
; VBITS_GE_1024-NEXT: ld1sw { z0.d }, p0/z, [x0, x8, lsl #2]
|
|
; VBITS_GE_1024-NEXT: ld1sw { z1.d }, p0/z, [x0]
|
|
; VBITS_GE_1024-NEXT: st1d { z0.d }, p0, [x1, x8, lsl #3]
|
|
; VBITS_GE_1024-NEXT: st1d { z1.d }, p0, [x1]
|
|
; VBITS_GE_1024-NEXT: ret
|
|
;
|
|
; VBITS_GE_2048-LABEL: load_sext_v32i32i64:
|
|
; VBITS_GE_2048: // %bb.0:
|
|
; VBITS_GE_2048-NEXT: ptrue p0.d, vl32
|
|
; VBITS_GE_2048-NEXT: ld1sw { z0.d }, p0/z, [x0]
|
|
; VBITS_GE_2048-NEXT: st1d { z0.d }, p0, [x1]
|
|
; VBITS_GE_2048-NEXT: ret
|
|
%a = load <32 x i32>, ptr %ap
|
|
%val = sext <32 x i32> %a to <32 x i64>
|
|
store <32 x i64> %val, ptr %b
|
|
ret void
|
|
}
|
|
|
|
attributes #0 = { "target-features"="+sve" }
|