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
188 lines
7.3 KiB
LLVM
188 lines
7.3 KiB
LLVM
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
|
|
; RUN: llc -enable-subreg-liveness -mattr=+sve -force-streaming-compatible < %s | FileCheck %s
|
|
; RUN: llc -enable-subreg-liveness -mattr=+sme -force-streaming < %s | FileCheck %s
|
|
; RUN: llc -enable-subreg-liveness -force-streaming-compatible < %s | FileCheck %s --check-prefix=NONEON-NOSVE
|
|
|
|
target triple = "aarch64-unknown-linux-gnu"
|
|
|
|
define <4 x i32> @test(ptr %arg1, ptr %arg2) {
|
|
; CHECK-LABEL: test:
|
|
; CHECK: // %bb.0: // %entry
|
|
; CHECK-NEXT: ldp q0, q1, [x0, #32]
|
|
; CHECK-NEXT: ldp q3, q4, [x0]
|
|
; CHECK-NEXT: add z2.s, z0.s, z0.s
|
|
; CHECK-NEXT: add z5.s, z1.s, z1.s
|
|
; CHECK-NEXT: mov z0.s, z1.s[2]
|
|
; CHECK-NEXT: add z1.s, z3.s, z3.s
|
|
; CHECK-NEXT: add z3.s, z4.s, z4.s
|
|
; CHECK-NEXT: stp q2, q5, [x0, #32]
|
|
; CHECK-NEXT: stp q1, q3, [x0]
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: test:
|
|
; NONEON-NOSVE: // %bb.0: // %entry
|
|
; NONEON-NOSVE-NEXT: sub sp, sp, #144
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 144
|
|
; NONEON-NOSVE-NEXT: ldp q1, q0, [x0, #32]
|
|
; NONEON-NOSVE-NEXT: ldp q2, q3, [x0]
|
|
; NONEON-NOSVE-NEXT: str q1, [sp, #64]
|
|
; NONEON-NOSVE-NEXT: stp q0, q3, [sp, #32]
|
|
; NONEON-NOSVE-NEXT: ldr w9, [sp, #60]
|
|
; NONEON-NOSVE-NEXT: str q2, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: ldr w8, [sp, #40]
|
|
; NONEON-NOSVE-NEXT: add w9, w9, w9
|
|
; NONEON-NOSVE-NEXT: stp w8, w8, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: str w9, [sp, #124]
|
|
; NONEON-NOSVE-NEXT: ldr w9, [sp, #56]
|
|
; NONEON-NOSVE-NEXT: stp w8, w8, [sp]
|
|
; NONEON-NOSVE-NEXT: add w8, w8, w8
|
|
; NONEON-NOSVE-NEXT: add w9, w9, w9
|
|
; NONEON-NOSVE-NEXT: ldr q0, [sp]
|
|
; NONEON-NOSVE-NEXT: str w9, [sp, #120]
|
|
; NONEON-NOSVE-NEXT: ldr w9, [sp, #52]
|
|
; NONEON-NOSVE-NEXT: add w9, w9, w9
|
|
; NONEON-NOSVE-NEXT: str w9, [sp, #116]
|
|
; NONEON-NOSVE-NEXT: ldr w9, [sp, #48]
|
|
; NONEON-NOSVE-NEXT: add w9, w9, w9
|
|
; NONEON-NOSVE-NEXT: str w9, [sp, #112]
|
|
; NONEON-NOSVE-NEXT: ldr w9, [sp, #28]
|
|
; NONEON-NOSVE-NEXT: add w9, w9, w9
|
|
; NONEON-NOSVE-NEXT: str w9, [sp, #92]
|
|
; NONEON-NOSVE-NEXT: ldr w9, [sp, #24]
|
|
; NONEON-NOSVE-NEXT: add w9, w9, w9
|
|
; NONEON-NOSVE-NEXT: str w9, [sp, #88]
|
|
; NONEON-NOSVE-NEXT: ldr w9, [sp, #20]
|
|
; NONEON-NOSVE-NEXT: add w9, w9, w9
|
|
; NONEON-NOSVE-NEXT: str w9, [sp, #84]
|
|
; NONEON-NOSVE-NEXT: ldr w9, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: add w9, w9, w9
|
|
; NONEON-NOSVE-NEXT: str w9, [sp, #80]
|
|
; NONEON-NOSVE-NEXT: ldr w9, [sp, #44]
|
|
; NONEON-NOSVE-NEXT: add w9, w9, w9
|
|
; NONEON-NOSVE-NEXT: stp w8, w9, [sp, #104]
|
|
; NONEON-NOSVE-NEXT: ldr w8, [sp, #36]
|
|
; NONEON-NOSVE-NEXT: add w8, w8, w8
|
|
; NONEON-NOSVE-NEXT: str w8, [sp, #100]
|
|
; NONEON-NOSVE-NEXT: ldr w8, [sp, #32]
|
|
; NONEON-NOSVE-NEXT: add w8, w8, w8
|
|
; NONEON-NOSVE-NEXT: str w8, [sp, #96]
|
|
; NONEON-NOSVE-NEXT: ldr w8, [sp, #76]
|
|
; NONEON-NOSVE-NEXT: ldp q4, q2, [sp, #80]
|
|
; NONEON-NOSVE-NEXT: add w8, w8, w8
|
|
; NONEON-NOSVE-NEXT: str w8, [sp, #140]
|
|
; NONEON-NOSVE-NEXT: ldr w8, [sp, #72]
|
|
; NONEON-NOSVE-NEXT: add w8, w8, w8
|
|
; NONEON-NOSVE-NEXT: str w8, [sp, #136]
|
|
; NONEON-NOSVE-NEXT: ldr w8, [sp, #68]
|
|
; NONEON-NOSVE-NEXT: add w8, w8, w8
|
|
; NONEON-NOSVE-NEXT: str w8, [sp, #132]
|
|
; NONEON-NOSVE-NEXT: ldr w8, [sp, #64]
|
|
; NONEON-NOSVE-NEXT: add w8, w8, w8
|
|
; NONEON-NOSVE-NEXT: str w8, [sp, #128]
|
|
; NONEON-NOSVE-NEXT: ldp q3, q1, [sp, #112]
|
|
; NONEON-NOSVE-NEXT: stp q4, q3, [x0]
|
|
; NONEON-NOSVE-NEXT: stp q1, q2, [x0, #32]
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #144
|
|
; NONEON-NOSVE-NEXT: ret
|
|
entry:
|
|
%0 = load <16 x i32>, ptr %arg1, align 256
|
|
%1 = load <16 x i32>, ptr %arg2, align 256
|
|
%shvec = shufflevector <16 x i32> %0, <16 x i32> %1, <4 x i32> <i32 14, i32 14, i32 14, i32 14>
|
|
%2 = add <16 x i32> %0, %0
|
|
store <16 x i32> %2, ptr %arg1, align 256
|
|
ret <4 x i32> %shvec
|
|
}
|
|
|
|
define <2 x i32> @test2(ptr %arg1, ptr %arg2) {
|
|
; CHECK-LABEL: test2:
|
|
; CHECK: // %bb.0: // %entry
|
|
; CHECK-NEXT: ldp q1, q0, [x0, #32]
|
|
; CHECK-NEXT: ldp q3, q4, [x0]
|
|
; CHECK-NEXT: movprfx z2, z0
|
|
; CHECK-NEXT: ext z2.b, z2.b, z0.b, #8
|
|
; CHECK-NEXT: add z1.s, z1.s, z1.s
|
|
; CHECK-NEXT: add z5.s, z0.s, z0.s
|
|
; CHECK-NEXT: mov z0.s, s2
|
|
; CHECK-NEXT: add z2.s, z3.s, z3.s
|
|
; CHECK-NEXT: add z3.s, z4.s, z4.s
|
|
; CHECK-NEXT: stp q1, q5, [x0, #32]
|
|
; CHECK-NEXT: stp q2, q3, [x0]
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: test2:
|
|
; NONEON-NOSVE: // %bb.0: // %entry
|
|
; NONEON-NOSVE-NEXT: sub sp, sp, #144
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 144
|
|
; NONEON-NOSVE-NEXT: ldp q1, q0, [x0, #32]
|
|
; NONEON-NOSVE-NEXT: ldp q2, q3, [x0]
|
|
; NONEON-NOSVE-NEXT: str q1, [sp, #64]
|
|
; NONEON-NOSVE-NEXT: stp q0, q3, [sp, #32]
|
|
; NONEON-NOSVE-NEXT: ldr d0, [sp, #40]
|
|
; NONEON-NOSVE-NEXT: str q2, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: str d0, [sp]
|
|
; NONEON-NOSVE-NEXT: ldr w8, [sp]
|
|
; NONEON-NOSVE-NEXT: stp w8, w8, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: ldr w8, [sp, #60]
|
|
; NONEON-NOSVE-NEXT: ldr d0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: add w8, w8, w8
|
|
; NONEON-NOSVE-NEXT: str w8, [sp, #124]
|
|
; NONEON-NOSVE-NEXT: ldr w8, [sp, #56]
|
|
; NONEON-NOSVE-NEXT: add w8, w8, w8
|
|
; NONEON-NOSVE-NEXT: str w8, [sp, #120]
|
|
; NONEON-NOSVE-NEXT: ldr w8, [sp, #52]
|
|
; NONEON-NOSVE-NEXT: add w8, w8, w8
|
|
; NONEON-NOSVE-NEXT: str w8, [sp, #116]
|
|
; NONEON-NOSVE-NEXT: ldr w8, [sp, #48]
|
|
; NONEON-NOSVE-NEXT: add w8, w8, w8
|
|
; NONEON-NOSVE-NEXT: str w8, [sp, #112]
|
|
; NONEON-NOSVE-NEXT: ldr w8, [sp, #28]
|
|
; NONEON-NOSVE-NEXT: add w8, w8, w8
|
|
; NONEON-NOSVE-NEXT: str w8, [sp, #92]
|
|
; NONEON-NOSVE-NEXT: ldr w8, [sp, #24]
|
|
; NONEON-NOSVE-NEXT: add w8, w8, w8
|
|
; NONEON-NOSVE-NEXT: str w8, [sp, #88]
|
|
; NONEON-NOSVE-NEXT: ldr w8, [sp, #20]
|
|
; NONEON-NOSVE-NEXT: add w8, w8, w8
|
|
; NONEON-NOSVE-NEXT: str w8, [sp, #84]
|
|
; NONEON-NOSVE-NEXT: ldr w8, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: add w8, w8, w8
|
|
; NONEON-NOSVE-NEXT: str w8, [sp, #80]
|
|
; NONEON-NOSVE-NEXT: ldr w8, [sp, #44]
|
|
; NONEON-NOSVE-NEXT: add w8, w8, w8
|
|
; NONEON-NOSVE-NEXT: str w8, [sp, #108]
|
|
; NONEON-NOSVE-NEXT: ldr w8, [sp, #40]
|
|
; NONEON-NOSVE-NEXT: add w8, w8, w8
|
|
; NONEON-NOSVE-NEXT: str w8, [sp, #104]
|
|
; NONEON-NOSVE-NEXT: ldr w8, [sp, #36]
|
|
; NONEON-NOSVE-NEXT: add w8, w8, w8
|
|
; NONEON-NOSVE-NEXT: str w8, [sp, #100]
|
|
; NONEON-NOSVE-NEXT: ldr w8, [sp, #32]
|
|
; NONEON-NOSVE-NEXT: add w8, w8, w8
|
|
; NONEON-NOSVE-NEXT: str w8, [sp, #96]
|
|
; NONEON-NOSVE-NEXT: ldr w8, [sp, #76]
|
|
; NONEON-NOSVE-NEXT: ldp q4, q2, [sp, #80]
|
|
; NONEON-NOSVE-NEXT: add w8, w8, w8
|
|
; NONEON-NOSVE-NEXT: str w8, [sp, #140]
|
|
; NONEON-NOSVE-NEXT: ldr w8, [sp, #72]
|
|
; NONEON-NOSVE-NEXT: add w8, w8, w8
|
|
; NONEON-NOSVE-NEXT: str w8, [sp, #136]
|
|
; NONEON-NOSVE-NEXT: ldr w8, [sp, #68]
|
|
; NONEON-NOSVE-NEXT: add w8, w8, w8
|
|
; NONEON-NOSVE-NEXT: str w8, [sp, #132]
|
|
; NONEON-NOSVE-NEXT: ldr w8, [sp, #64]
|
|
; NONEON-NOSVE-NEXT: add w8, w8, w8
|
|
; NONEON-NOSVE-NEXT: str w8, [sp, #128]
|
|
; NONEON-NOSVE-NEXT: ldp q3, q1, [sp, #112]
|
|
; NONEON-NOSVE-NEXT: stp q4, q3, [x0]
|
|
; NONEON-NOSVE-NEXT: stp q1, q2, [x0, #32]
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #144
|
|
; NONEON-NOSVE-NEXT: ret
|
|
entry:
|
|
%0 = load <16 x i32>, ptr %arg1, align 256
|
|
%1 = load <16 x i32>, ptr %arg2, align 256
|
|
%shvec = shufflevector <16 x i32> %0, <16 x i32> %1, <2 x i32> <i32 14, i32 14>
|
|
%2 = add <16 x i32> %0, %0
|
|
store <16 x i32> %2, ptr %arg1, align 256
|
|
ret <2 x i32> %shvec
|
|
}
|