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
2100 lines
77 KiB
LLVM
2100 lines
77 KiB
LLVM
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
|
|
; RUN: llc < %s -mtriple=i686-- -mattr=+sse2 | FileCheck %s --check-prefixes=SSE,X86-SSE,SSE2,X86-SSE2
|
|
; RUN: llc < %s -mtriple=x86_64-- -mattr=+sse2 | FileCheck %s --check-prefixes=SSE,X64-SSE,SSE2,X64-SSE2
|
|
; RUN: llc < %s -mtriple=i686-- -mattr=+sse4.1 | FileCheck %s --check-prefixes=SSE,X86-SSE,SSE4,X86-SSE4
|
|
; RUN: llc < %s -mtriple=x86_64-- -mattr=+sse4.1 | FileCheck %s --check-prefixes=SSE,X64-SSE,SSE4,X64-SSE4
|
|
; RUN: llc < %s -mtriple=i686-- -mattr=+sse4.2 | FileCheck %s --check-prefixes=SSE,X86-SSE,SSE4,X86-SSE4
|
|
; RUN: llc < %s -mtriple=x86_64-- -mattr=+sse4.2 | FileCheck %s --check-prefixes=SSE,X64-SSE,SSE4,X64-SSE4
|
|
; RUN: llc < %s -mtriple=i686-- -mattr=+avx | FileCheck %s --check-prefixes=AVX,AVX1,AVX1-SLOW,X86-AVX1,X86-AVX1-SLOW
|
|
; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx | FileCheck %s --check-prefixes=AVX,AVX1,AVX1-SLOW,X64-AVX1,X64-AVX1-SLOW
|
|
; RUN: llc < %s -mtriple=i686-- -mattr=+avx,+fast-hops | FileCheck %s --check-prefixes=AVX,AVX1,AVX1-FAST,X86-AVX1,X86-AVX1-FAST
|
|
; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx,+fast-hops | FileCheck %s --check-prefixes=AVX,AVX1,AVX1-FAST,X64-AVX1,X64-AVX1-FAST
|
|
; RUN: llc < %s -mtriple=i686-- -mattr=+avx2 | FileCheck %s --check-prefixes=AVX,AVX2,X86-AVX2
|
|
; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx2 | FileCheck %s --check-prefixes=AVX,AVX2,X64-AVX2
|
|
; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx512f,+avx512bw | FileCheck %s --check-prefix=AVX512
|
|
; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx512f,+avx512bw,+avx512vl | FileCheck %s --check-prefix=AVX512
|
|
|
|
;
|
|
; vXi64
|
|
;
|
|
|
|
define i64 @test_v2i64(<2 x i64> %a0) nounwind {
|
|
; X86-SSE2-LABEL: test_v2i64:
|
|
; X86-SSE2: # %bb.0:
|
|
; X86-SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
|
|
; X86-SSE2-NEXT: paddq %xmm0, %xmm1
|
|
; X86-SSE2-NEXT: movd %xmm1, %eax
|
|
; X86-SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm1[1,1,1,1]
|
|
; X86-SSE2-NEXT: movd %xmm0, %edx
|
|
; X86-SSE2-NEXT: retl
|
|
;
|
|
; X64-SSE-LABEL: test_v2i64:
|
|
; X64-SSE: # %bb.0:
|
|
; X64-SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
|
|
; X64-SSE-NEXT: paddq %xmm0, %xmm1
|
|
; X64-SSE-NEXT: movq %xmm1, %rax
|
|
; X64-SSE-NEXT: retq
|
|
;
|
|
; X86-SSE4-LABEL: test_v2i64:
|
|
; X86-SSE4: # %bb.0:
|
|
; X86-SSE4-NEXT: pshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
|
|
; X86-SSE4-NEXT: paddq %xmm0, %xmm1
|
|
; X86-SSE4-NEXT: movd %xmm1, %eax
|
|
; X86-SSE4-NEXT: pextrd $1, %xmm1, %edx
|
|
; X86-SSE4-NEXT: retl
|
|
;
|
|
; X86-AVX1-LABEL: test_v2i64:
|
|
; X86-AVX1: # %bb.0:
|
|
; X86-AVX1-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
|
|
; X86-AVX1-NEXT: vpaddq %xmm1, %xmm0, %xmm0
|
|
; X86-AVX1-NEXT: vmovd %xmm0, %eax
|
|
; X86-AVX1-NEXT: vpextrd $1, %xmm0, %edx
|
|
; X86-AVX1-NEXT: retl
|
|
;
|
|
; X64-AVX1-LABEL: test_v2i64:
|
|
; X64-AVX1: # %bb.0:
|
|
; X64-AVX1-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
|
|
; X64-AVX1-NEXT: vpaddq %xmm1, %xmm0, %xmm0
|
|
; X64-AVX1-NEXT: vmovq %xmm0, %rax
|
|
; X64-AVX1-NEXT: retq
|
|
;
|
|
; X86-AVX2-LABEL: test_v2i64:
|
|
; X86-AVX2: # %bb.0:
|
|
; X86-AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
|
|
; X86-AVX2-NEXT: vpaddq %xmm1, %xmm0, %xmm0
|
|
; X86-AVX2-NEXT: vmovd %xmm0, %eax
|
|
; X86-AVX2-NEXT: vpextrd $1, %xmm0, %edx
|
|
; X86-AVX2-NEXT: retl
|
|
;
|
|
; X64-AVX2-LABEL: test_v2i64:
|
|
; X64-AVX2: # %bb.0:
|
|
; X64-AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
|
|
; X64-AVX2-NEXT: vpaddq %xmm1, %xmm0, %xmm0
|
|
; X64-AVX2-NEXT: vmovq %xmm0, %rax
|
|
; X64-AVX2-NEXT: retq
|
|
;
|
|
; AVX512-LABEL: test_v2i64:
|
|
; AVX512: # %bb.0:
|
|
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
|
|
; AVX512-NEXT: vpaddq %xmm1, %xmm0, %xmm0
|
|
; AVX512-NEXT: vmovq %xmm0, %rax
|
|
; AVX512-NEXT: retq
|
|
%1 = call i64 @llvm.vector.reduce.add.v2i64(<2 x i64> %a0)
|
|
ret i64 %1
|
|
}
|
|
|
|
define i64 @test_v4i64(<4 x i64> %a0) nounwind {
|
|
; X86-SSE2-LABEL: test_v4i64:
|
|
; X86-SSE2: # %bb.0:
|
|
; X86-SSE2-NEXT: paddq %xmm1, %xmm0
|
|
; X86-SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
|
|
; X86-SSE2-NEXT: paddq %xmm0, %xmm1
|
|
; X86-SSE2-NEXT: movd %xmm1, %eax
|
|
; X86-SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm1[1,1,1,1]
|
|
; X86-SSE2-NEXT: movd %xmm0, %edx
|
|
; X86-SSE2-NEXT: retl
|
|
;
|
|
; X64-SSE-LABEL: test_v4i64:
|
|
; X64-SSE: # %bb.0:
|
|
; X64-SSE-NEXT: paddq %xmm1, %xmm0
|
|
; X64-SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
|
|
; X64-SSE-NEXT: paddq %xmm0, %xmm1
|
|
; X64-SSE-NEXT: movq %xmm1, %rax
|
|
; X64-SSE-NEXT: retq
|
|
;
|
|
; X86-SSE4-LABEL: test_v4i64:
|
|
; X86-SSE4: # %bb.0:
|
|
; X86-SSE4-NEXT: paddq %xmm1, %xmm0
|
|
; X86-SSE4-NEXT: pshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
|
|
; X86-SSE4-NEXT: paddq %xmm0, %xmm1
|
|
; X86-SSE4-NEXT: movd %xmm1, %eax
|
|
; X86-SSE4-NEXT: pextrd $1, %xmm1, %edx
|
|
; X86-SSE4-NEXT: retl
|
|
;
|
|
; X86-AVX1-LABEL: test_v4i64:
|
|
; X86-AVX1: # %bb.0:
|
|
; X86-AVX1-NEXT: vextractf128 $1, %ymm0, %xmm1
|
|
; X86-AVX1-NEXT: vpaddq %xmm1, %xmm0, %xmm0
|
|
; X86-AVX1-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
|
|
; X86-AVX1-NEXT: vpaddq %xmm1, %xmm0, %xmm0
|
|
; X86-AVX1-NEXT: vmovd %xmm0, %eax
|
|
; X86-AVX1-NEXT: vpextrd $1, %xmm0, %edx
|
|
; X86-AVX1-NEXT: vzeroupper
|
|
; X86-AVX1-NEXT: retl
|
|
;
|
|
; X64-AVX1-LABEL: test_v4i64:
|
|
; X64-AVX1: # %bb.0:
|
|
; X64-AVX1-NEXT: vextractf128 $1, %ymm0, %xmm1
|
|
; X64-AVX1-NEXT: vpaddq %xmm1, %xmm0, %xmm0
|
|
; X64-AVX1-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
|
|
; X64-AVX1-NEXT: vpaddq %xmm1, %xmm0, %xmm0
|
|
; X64-AVX1-NEXT: vmovq %xmm0, %rax
|
|
; X64-AVX1-NEXT: vzeroupper
|
|
; X64-AVX1-NEXT: retq
|
|
;
|
|
; X86-AVX2-LABEL: test_v4i64:
|
|
; X86-AVX2: # %bb.0:
|
|
; X86-AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1
|
|
; X86-AVX2-NEXT: vpaddq %xmm1, %xmm0, %xmm0
|
|
; X86-AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
|
|
; X86-AVX2-NEXT: vpaddq %xmm1, %xmm0, %xmm0
|
|
; X86-AVX2-NEXT: vmovd %xmm0, %eax
|
|
; X86-AVX2-NEXT: vpextrd $1, %xmm0, %edx
|
|
; X86-AVX2-NEXT: vzeroupper
|
|
; X86-AVX2-NEXT: retl
|
|
;
|
|
; X64-AVX2-LABEL: test_v4i64:
|
|
; X64-AVX2: # %bb.0:
|
|
; X64-AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1
|
|
; X64-AVX2-NEXT: vpaddq %xmm1, %xmm0, %xmm0
|
|
; X64-AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
|
|
; X64-AVX2-NEXT: vpaddq %xmm1, %xmm0, %xmm0
|
|
; X64-AVX2-NEXT: vmovq %xmm0, %rax
|
|
; X64-AVX2-NEXT: vzeroupper
|
|
; X64-AVX2-NEXT: retq
|
|
;
|
|
; AVX512-LABEL: test_v4i64:
|
|
; AVX512: # %bb.0:
|
|
; AVX512-NEXT: vextracti128 $1, %ymm0, %xmm1
|
|
; AVX512-NEXT: vpaddq %xmm1, %xmm0, %xmm0
|
|
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
|
|
; AVX512-NEXT: vpaddq %xmm1, %xmm0, %xmm0
|
|
; AVX512-NEXT: vmovq %xmm0, %rax
|
|
; AVX512-NEXT: vzeroupper
|
|
; AVX512-NEXT: retq
|
|
%1 = call i64 @llvm.vector.reduce.add.v4i64(<4 x i64> %a0)
|
|
ret i64 %1
|
|
}
|
|
|
|
define i64 @test_v8i64(<8 x i64> %a0) nounwind {
|
|
; X86-SSE2-LABEL: test_v8i64:
|
|
; X86-SSE2: # %bb.0:
|
|
; X86-SSE2-NEXT: pushl %ebp
|
|
; X86-SSE2-NEXT: movl %esp, %ebp
|
|
; X86-SSE2-NEXT: andl $-16, %esp
|
|
; X86-SSE2-NEXT: subl $16, %esp
|
|
; X86-SSE2-NEXT: paddq %xmm2, %xmm0
|
|
; X86-SSE2-NEXT: paddq 8(%ebp), %xmm1
|
|
; X86-SSE2-NEXT: paddq %xmm0, %xmm1
|
|
; X86-SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm1[2,3,2,3]
|
|
; X86-SSE2-NEXT: paddq %xmm1, %xmm0
|
|
; X86-SSE2-NEXT: movd %xmm0, %eax
|
|
; X86-SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm0[1,1,1,1]
|
|
; X86-SSE2-NEXT: movd %xmm0, %edx
|
|
; X86-SSE2-NEXT: movl %ebp, %esp
|
|
; X86-SSE2-NEXT: popl %ebp
|
|
; X86-SSE2-NEXT: retl
|
|
;
|
|
; X64-SSE-LABEL: test_v8i64:
|
|
; X64-SSE: # %bb.0:
|
|
; X64-SSE-NEXT: paddq %xmm3, %xmm1
|
|
; X64-SSE-NEXT: paddq %xmm2, %xmm0
|
|
; X64-SSE-NEXT: paddq %xmm1, %xmm0
|
|
; X64-SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
|
|
; X64-SSE-NEXT: paddq %xmm0, %xmm1
|
|
; X64-SSE-NEXT: movq %xmm1, %rax
|
|
; X64-SSE-NEXT: retq
|
|
;
|
|
; X86-SSE4-LABEL: test_v8i64:
|
|
; X86-SSE4: # %bb.0:
|
|
; X86-SSE4-NEXT: pushl %ebp
|
|
; X86-SSE4-NEXT: movl %esp, %ebp
|
|
; X86-SSE4-NEXT: andl $-16, %esp
|
|
; X86-SSE4-NEXT: subl $16, %esp
|
|
; X86-SSE4-NEXT: paddq %xmm2, %xmm0
|
|
; X86-SSE4-NEXT: paddq 8(%ebp), %xmm1
|
|
; X86-SSE4-NEXT: paddq %xmm0, %xmm1
|
|
; X86-SSE4-NEXT: pshufd {{.*#+}} xmm0 = xmm1[2,3,2,3]
|
|
; X86-SSE4-NEXT: paddq %xmm1, %xmm0
|
|
; X86-SSE4-NEXT: movd %xmm0, %eax
|
|
; X86-SSE4-NEXT: pextrd $1, %xmm0, %edx
|
|
; X86-SSE4-NEXT: movl %ebp, %esp
|
|
; X86-SSE4-NEXT: popl %ebp
|
|
; X86-SSE4-NEXT: retl
|
|
;
|
|
; X86-AVX1-LABEL: test_v8i64:
|
|
; X86-AVX1: # %bb.0:
|
|
; X86-AVX1-NEXT: vextractf128 $1, %ymm1, %xmm2
|
|
; X86-AVX1-NEXT: vextractf128 $1, %ymm0, %xmm3
|
|
; X86-AVX1-NEXT: vpaddq %xmm2, %xmm3, %xmm2
|
|
; X86-AVX1-NEXT: vpaddq %xmm1, %xmm0, %xmm0
|
|
; X86-AVX1-NEXT: vpaddq %xmm2, %xmm0, %xmm0
|
|
; X86-AVX1-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
|
|
; X86-AVX1-NEXT: vpaddq %xmm1, %xmm0, %xmm0
|
|
; X86-AVX1-NEXT: vmovd %xmm0, %eax
|
|
; X86-AVX1-NEXT: vpextrd $1, %xmm0, %edx
|
|
; X86-AVX1-NEXT: vzeroupper
|
|
; X86-AVX1-NEXT: retl
|
|
;
|
|
; X64-AVX1-LABEL: test_v8i64:
|
|
; X64-AVX1: # %bb.0:
|
|
; X64-AVX1-NEXT: vextractf128 $1, %ymm1, %xmm2
|
|
; X64-AVX1-NEXT: vextractf128 $1, %ymm0, %xmm3
|
|
; X64-AVX1-NEXT: vpaddq %xmm2, %xmm3, %xmm2
|
|
; X64-AVX1-NEXT: vpaddq %xmm1, %xmm0, %xmm0
|
|
; X64-AVX1-NEXT: vpaddq %xmm2, %xmm0, %xmm0
|
|
; X64-AVX1-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
|
|
; X64-AVX1-NEXT: vpaddq %xmm1, %xmm0, %xmm0
|
|
; X64-AVX1-NEXT: vmovq %xmm0, %rax
|
|
; X64-AVX1-NEXT: vzeroupper
|
|
; X64-AVX1-NEXT: retq
|
|
;
|
|
; X86-AVX2-LABEL: test_v8i64:
|
|
; X86-AVX2: # %bb.0:
|
|
; X86-AVX2-NEXT: vpaddq %ymm1, %ymm0, %ymm0
|
|
; X86-AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1
|
|
; X86-AVX2-NEXT: vpaddq %xmm1, %xmm0, %xmm0
|
|
; X86-AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
|
|
; X86-AVX2-NEXT: vpaddq %xmm1, %xmm0, %xmm0
|
|
; X86-AVX2-NEXT: vmovd %xmm0, %eax
|
|
; X86-AVX2-NEXT: vpextrd $1, %xmm0, %edx
|
|
; X86-AVX2-NEXT: vzeroupper
|
|
; X86-AVX2-NEXT: retl
|
|
;
|
|
; X64-AVX2-LABEL: test_v8i64:
|
|
; X64-AVX2: # %bb.0:
|
|
; X64-AVX2-NEXT: vpaddq %ymm1, %ymm0, %ymm0
|
|
; X64-AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1
|
|
; X64-AVX2-NEXT: vpaddq %xmm1, %xmm0, %xmm0
|
|
; X64-AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
|
|
; X64-AVX2-NEXT: vpaddq %xmm1, %xmm0, %xmm0
|
|
; X64-AVX2-NEXT: vmovq %xmm0, %rax
|
|
; X64-AVX2-NEXT: vzeroupper
|
|
; X64-AVX2-NEXT: retq
|
|
;
|
|
; AVX512-LABEL: test_v8i64:
|
|
; AVX512: # %bb.0:
|
|
; AVX512-NEXT: vextracti64x4 $1, %zmm0, %ymm1
|
|
; AVX512-NEXT: vpaddq %zmm1, %zmm0, %zmm0
|
|
; AVX512-NEXT: vextracti128 $1, %ymm0, %xmm1
|
|
; AVX512-NEXT: vpaddq %xmm1, %xmm0, %xmm0
|
|
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
|
|
; AVX512-NEXT: vpaddq %xmm1, %xmm0, %xmm0
|
|
; AVX512-NEXT: vmovq %xmm0, %rax
|
|
; AVX512-NEXT: vzeroupper
|
|
; AVX512-NEXT: retq
|
|
%1 = call i64 @llvm.vector.reduce.add.v8i64(<8 x i64> %a0)
|
|
ret i64 %1
|
|
}
|
|
|
|
define i64 @test_v16i64(<16 x i64> %a0) nounwind {
|
|
; X86-SSE2-LABEL: test_v16i64:
|
|
; X86-SSE2: # %bb.0:
|
|
; X86-SSE2-NEXT: pushl %ebp
|
|
; X86-SSE2-NEXT: movl %esp, %ebp
|
|
; X86-SSE2-NEXT: andl $-16, %esp
|
|
; X86-SSE2-NEXT: subl $16, %esp
|
|
; X86-SSE2-NEXT: movdqa 8(%ebp), %xmm3
|
|
; X86-SSE2-NEXT: paddq 56(%ebp), %xmm2
|
|
; X86-SSE2-NEXT: paddq 24(%ebp), %xmm0
|
|
; X86-SSE2-NEXT: paddq %xmm2, %xmm0
|
|
; X86-SSE2-NEXT: paddq 72(%ebp), %xmm3
|
|
; X86-SSE2-NEXT: paddq 40(%ebp), %xmm1
|
|
; X86-SSE2-NEXT: paddq %xmm3, %xmm1
|
|
; X86-SSE2-NEXT: paddq %xmm0, %xmm1
|
|
; X86-SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm1[2,3,2,3]
|
|
; X86-SSE2-NEXT: paddq %xmm1, %xmm0
|
|
; X86-SSE2-NEXT: movd %xmm0, %eax
|
|
; X86-SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm0[1,1,1,1]
|
|
; X86-SSE2-NEXT: movd %xmm0, %edx
|
|
; X86-SSE2-NEXT: movl %ebp, %esp
|
|
; X86-SSE2-NEXT: popl %ebp
|
|
; X86-SSE2-NEXT: retl
|
|
;
|
|
; X64-SSE-LABEL: test_v16i64:
|
|
; X64-SSE: # %bb.0:
|
|
; X64-SSE-NEXT: paddq %xmm6, %xmm2
|
|
; X64-SSE-NEXT: paddq %xmm4, %xmm0
|
|
; X64-SSE-NEXT: paddq %xmm2, %xmm0
|
|
; X64-SSE-NEXT: paddq %xmm7, %xmm3
|
|
; X64-SSE-NEXT: paddq %xmm5, %xmm1
|
|
; X64-SSE-NEXT: paddq %xmm3, %xmm1
|
|
; X64-SSE-NEXT: paddq %xmm0, %xmm1
|
|
; X64-SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm1[2,3,2,3]
|
|
; X64-SSE-NEXT: paddq %xmm1, %xmm0
|
|
; X64-SSE-NEXT: movq %xmm0, %rax
|
|
; X64-SSE-NEXT: retq
|
|
;
|
|
; X86-SSE4-LABEL: test_v16i64:
|
|
; X86-SSE4: # %bb.0:
|
|
; X86-SSE4-NEXT: pushl %ebp
|
|
; X86-SSE4-NEXT: movl %esp, %ebp
|
|
; X86-SSE4-NEXT: andl $-16, %esp
|
|
; X86-SSE4-NEXT: subl $16, %esp
|
|
; X86-SSE4-NEXT: movdqa 8(%ebp), %xmm3
|
|
; X86-SSE4-NEXT: paddq 56(%ebp), %xmm2
|
|
; X86-SSE4-NEXT: paddq 24(%ebp), %xmm0
|
|
; X86-SSE4-NEXT: paddq %xmm2, %xmm0
|
|
; X86-SSE4-NEXT: paddq 72(%ebp), %xmm3
|
|
; X86-SSE4-NEXT: paddq 40(%ebp), %xmm1
|
|
; X86-SSE4-NEXT: paddq %xmm3, %xmm1
|
|
; X86-SSE4-NEXT: paddq %xmm0, %xmm1
|
|
; X86-SSE4-NEXT: pshufd {{.*#+}} xmm0 = xmm1[2,3,2,3]
|
|
; X86-SSE4-NEXT: paddq %xmm1, %xmm0
|
|
; X86-SSE4-NEXT: movd %xmm0, %eax
|
|
; X86-SSE4-NEXT: pextrd $1, %xmm0, %edx
|
|
; X86-SSE4-NEXT: movl %ebp, %esp
|
|
; X86-SSE4-NEXT: popl %ebp
|
|
; X86-SSE4-NEXT: retl
|
|
;
|
|
; X86-AVX1-LABEL: test_v16i64:
|
|
; X86-AVX1: # %bb.0:
|
|
; X86-AVX1-NEXT: pushl %ebp
|
|
; X86-AVX1-NEXT: movl %esp, %ebp
|
|
; X86-AVX1-NEXT: andl $-32, %esp
|
|
; X86-AVX1-NEXT: subl $32, %esp
|
|
; X86-AVX1-NEXT: vpaddq %xmm2, %xmm0, %xmm3
|
|
; X86-AVX1-NEXT: vextractf128 $1, %ymm2, %xmm2
|
|
; X86-AVX1-NEXT: vextractf128 $1, %ymm0, %xmm0
|
|
; X86-AVX1-NEXT: vpaddq %xmm2, %xmm0, %xmm0
|
|
; X86-AVX1-NEXT: vpaddq 8(%ebp), %xmm1, %xmm2
|
|
; X86-AVX1-NEXT: vpaddq %xmm2, %xmm3, %xmm2
|
|
; X86-AVX1-NEXT: vextractf128 $1, %ymm1, %xmm1
|
|
; X86-AVX1-NEXT: vpaddq 24(%ebp), %xmm1, %xmm1
|
|
; X86-AVX1-NEXT: vpaddq %xmm1, %xmm0, %xmm0
|
|
; X86-AVX1-NEXT: vpaddq %xmm0, %xmm2, %xmm0
|
|
; X86-AVX1-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
|
|
; X86-AVX1-NEXT: vpaddq %xmm1, %xmm0, %xmm0
|
|
; X86-AVX1-NEXT: vmovd %xmm0, %eax
|
|
; X86-AVX1-NEXT: vpextrd $1, %xmm0, %edx
|
|
; X86-AVX1-NEXT: movl %ebp, %esp
|
|
; X86-AVX1-NEXT: popl %ebp
|
|
; X86-AVX1-NEXT: vzeroupper
|
|
; X86-AVX1-NEXT: retl
|
|
;
|
|
; X64-AVX1-LABEL: test_v16i64:
|
|
; X64-AVX1: # %bb.0:
|
|
; X64-AVX1-NEXT: vpaddq %xmm3, %xmm1, %xmm4
|
|
; X64-AVX1-NEXT: vpaddq %xmm2, %xmm0, %xmm5
|
|
; X64-AVX1-NEXT: vpaddq %xmm4, %xmm5, %xmm4
|
|
; X64-AVX1-NEXT: vextractf128 $1, %ymm3, %xmm3
|
|
; X64-AVX1-NEXT: vextractf128 $1, %ymm1, %xmm1
|
|
; X64-AVX1-NEXT: vpaddq %xmm3, %xmm1, %xmm1
|
|
; X64-AVX1-NEXT: vextractf128 $1, %ymm2, %xmm2
|
|
; X64-AVX1-NEXT: vextractf128 $1, %ymm0, %xmm0
|
|
; X64-AVX1-NEXT: vpaddq %xmm2, %xmm0, %xmm0
|
|
; X64-AVX1-NEXT: vpaddq %xmm1, %xmm0, %xmm0
|
|
; X64-AVX1-NEXT: vpaddq %xmm0, %xmm4, %xmm0
|
|
; X64-AVX1-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
|
|
; X64-AVX1-NEXT: vpaddq %xmm1, %xmm0, %xmm0
|
|
; X64-AVX1-NEXT: vmovq %xmm0, %rax
|
|
; X64-AVX1-NEXT: vzeroupper
|
|
; X64-AVX1-NEXT: retq
|
|
;
|
|
; X86-AVX2-LABEL: test_v16i64:
|
|
; X86-AVX2: # %bb.0:
|
|
; X86-AVX2-NEXT: pushl %ebp
|
|
; X86-AVX2-NEXT: movl %esp, %ebp
|
|
; X86-AVX2-NEXT: andl $-32, %esp
|
|
; X86-AVX2-NEXT: subl $32, %esp
|
|
; X86-AVX2-NEXT: vpaddq %ymm2, %ymm0, %ymm0
|
|
; X86-AVX2-NEXT: vpaddq 8(%ebp), %ymm1, %ymm1
|
|
; X86-AVX2-NEXT: vpaddq %ymm1, %ymm0, %ymm0
|
|
; X86-AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1
|
|
; X86-AVX2-NEXT: vpaddq %xmm1, %xmm0, %xmm0
|
|
; X86-AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
|
|
; X86-AVX2-NEXT: vpaddq %xmm1, %xmm0, %xmm0
|
|
; X86-AVX2-NEXT: vmovd %xmm0, %eax
|
|
; X86-AVX2-NEXT: vpextrd $1, %xmm0, %edx
|
|
; X86-AVX2-NEXT: movl %ebp, %esp
|
|
; X86-AVX2-NEXT: popl %ebp
|
|
; X86-AVX2-NEXT: vzeroupper
|
|
; X86-AVX2-NEXT: retl
|
|
;
|
|
; X64-AVX2-LABEL: test_v16i64:
|
|
; X64-AVX2: # %bb.0:
|
|
; X64-AVX2-NEXT: vpaddq %ymm3, %ymm1, %ymm1
|
|
; X64-AVX2-NEXT: vpaddq %ymm2, %ymm0, %ymm0
|
|
; X64-AVX2-NEXT: vpaddq %ymm1, %ymm0, %ymm0
|
|
; X64-AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1
|
|
; X64-AVX2-NEXT: vpaddq %xmm1, %xmm0, %xmm0
|
|
; X64-AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
|
|
; X64-AVX2-NEXT: vpaddq %xmm1, %xmm0, %xmm0
|
|
; X64-AVX2-NEXT: vmovq %xmm0, %rax
|
|
; X64-AVX2-NEXT: vzeroupper
|
|
; X64-AVX2-NEXT: retq
|
|
;
|
|
; AVX512-LABEL: test_v16i64:
|
|
; AVX512: # %bb.0:
|
|
; AVX512-NEXT: vpaddq %zmm1, %zmm0, %zmm0
|
|
; AVX512-NEXT: vextracti64x4 $1, %zmm0, %ymm1
|
|
; AVX512-NEXT: vpaddq %zmm1, %zmm0, %zmm0
|
|
; AVX512-NEXT: vextracti128 $1, %ymm0, %xmm1
|
|
; AVX512-NEXT: vpaddq %xmm1, %xmm0, %xmm0
|
|
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
|
|
; AVX512-NEXT: vpaddq %xmm1, %xmm0, %xmm0
|
|
; AVX512-NEXT: vmovq %xmm0, %rax
|
|
; AVX512-NEXT: vzeroupper
|
|
; AVX512-NEXT: retq
|
|
%1 = call i64 @llvm.vector.reduce.add.v16i64(<16 x i64> %a0)
|
|
ret i64 %1
|
|
}
|
|
|
|
;
|
|
; vXi32
|
|
;
|
|
|
|
define i32 @test_v2i32(<2 x i32> %a0) nounwind {
|
|
; SSE-LABEL: test_v2i32:
|
|
; SSE: # %bb.0:
|
|
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
|
|
; SSE-NEXT: paddd %xmm0, %xmm1
|
|
; SSE-NEXT: movd %xmm1, %eax
|
|
; SSE-NEXT: ret{{[l|q]}}
|
|
;
|
|
; AVX1-SLOW-LABEL: test_v2i32:
|
|
; AVX1-SLOW: # %bb.0:
|
|
; AVX1-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
|
|
; AVX1-SLOW-NEXT: vpaddd %xmm1, %xmm0, %xmm0
|
|
; AVX1-SLOW-NEXT: vmovd %xmm0, %eax
|
|
; AVX1-SLOW-NEXT: ret{{[l|q]}}
|
|
;
|
|
; AVX1-FAST-LABEL: test_v2i32:
|
|
; AVX1-FAST: # %bb.0:
|
|
; AVX1-FAST-NEXT: vphaddd %xmm0, %xmm0, %xmm0
|
|
; AVX1-FAST-NEXT: vmovd %xmm0, %eax
|
|
; AVX1-FAST-NEXT: ret{{[l|q]}}
|
|
;
|
|
; AVX2-LABEL: test_v2i32:
|
|
; AVX2: # %bb.0:
|
|
; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
|
|
; AVX2-NEXT: vpaddd %xmm1, %xmm0, %xmm0
|
|
; AVX2-NEXT: vmovd %xmm0, %eax
|
|
; AVX2-NEXT: ret{{[l|q]}}
|
|
;
|
|
; AVX512-LABEL: test_v2i32:
|
|
; AVX512: # %bb.0:
|
|
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
|
|
; AVX512-NEXT: vpaddd %xmm1, %xmm0, %xmm0
|
|
; AVX512-NEXT: vmovd %xmm0, %eax
|
|
; AVX512-NEXT: retq
|
|
%1 = call i32 @llvm.vector.reduce.add.v2i32(<2 x i32> %a0)
|
|
ret i32 %1
|
|
}
|
|
|
|
define i32 @test_v4i32(<4 x i32> %a0) nounwind {
|
|
; SSE-LABEL: test_v4i32:
|
|
; SSE: # %bb.0:
|
|
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
|
|
; SSE-NEXT: paddd %xmm0, %xmm1
|
|
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm1[1,1,1,1]
|
|
; SSE-NEXT: paddd %xmm1, %xmm0
|
|
; SSE-NEXT: movd %xmm0, %eax
|
|
; SSE-NEXT: ret{{[l|q]}}
|
|
;
|
|
; AVX1-SLOW-LABEL: test_v4i32:
|
|
; AVX1-SLOW: # %bb.0:
|
|
; AVX1-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
|
|
; AVX1-SLOW-NEXT: vpaddd %xmm1, %xmm0, %xmm0
|
|
; AVX1-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
|
|
; AVX1-SLOW-NEXT: vpaddd %xmm1, %xmm0, %xmm0
|
|
; AVX1-SLOW-NEXT: vmovd %xmm0, %eax
|
|
; AVX1-SLOW-NEXT: ret{{[l|q]}}
|
|
;
|
|
; AVX1-FAST-LABEL: test_v4i32:
|
|
; AVX1-FAST: # %bb.0:
|
|
; AVX1-FAST-NEXT: vphaddd %xmm0, %xmm0, %xmm0
|
|
; AVX1-FAST-NEXT: vphaddd %xmm0, %xmm0, %xmm0
|
|
; AVX1-FAST-NEXT: vmovd %xmm0, %eax
|
|
; AVX1-FAST-NEXT: ret{{[l|q]}}
|
|
;
|
|
; AVX2-LABEL: test_v4i32:
|
|
; AVX2: # %bb.0:
|
|
; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
|
|
; AVX2-NEXT: vpaddd %xmm1, %xmm0, %xmm0
|
|
; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
|
|
; AVX2-NEXT: vpaddd %xmm1, %xmm0, %xmm0
|
|
; AVX2-NEXT: vmovd %xmm0, %eax
|
|
; AVX2-NEXT: ret{{[l|q]}}
|
|
;
|
|
; AVX512-LABEL: test_v4i32:
|
|
; AVX512: # %bb.0:
|
|
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
|
|
; AVX512-NEXT: vpaddd %xmm1, %xmm0, %xmm0
|
|
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
|
|
; AVX512-NEXT: vpaddd %xmm1, %xmm0, %xmm0
|
|
; AVX512-NEXT: vmovd %xmm0, %eax
|
|
; AVX512-NEXT: retq
|
|
%1 = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> %a0)
|
|
ret i32 %1
|
|
}
|
|
|
|
define i32 @test_v8i32(<8 x i32> %a0) nounwind {
|
|
; SSE-LABEL: test_v8i32:
|
|
; SSE: # %bb.0:
|
|
; SSE-NEXT: paddd %xmm1, %xmm0
|
|
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
|
|
; SSE-NEXT: paddd %xmm0, %xmm1
|
|
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm1[1,1,1,1]
|
|
; SSE-NEXT: paddd %xmm1, %xmm0
|
|
; SSE-NEXT: movd %xmm0, %eax
|
|
; SSE-NEXT: ret{{[l|q]}}
|
|
;
|
|
; AVX1-SLOW-LABEL: test_v8i32:
|
|
; AVX1-SLOW: # %bb.0:
|
|
; AVX1-SLOW-NEXT: vextractf128 $1, %ymm0, %xmm1
|
|
; AVX1-SLOW-NEXT: vpaddd %xmm1, %xmm0, %xmm0
|
|
; AVX1-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
|
|
; AVX1-SLOW-NEXT: vpaddd %xmm1, %xmm0, %xmm0
|
|
; AVX1-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
|
|
; AVX1-SLOW-NEXT: vpaddd %xmm1, %xmm0, %xmm0
|
|
; AVX1-SLOW-NEXT: vmovd %xmm0, %eax
|
|
; AVX1-SLOW-NEXT: vzeroupper
|
|
; AVX1-SLOW-NEXT: ret{{[l|q]}}
|
|
;
|
|
; AVX1-FAST-LABEL: test_v8i32:
|
|
; AVX1-FAST: # %bb.0:
|
|
; AVX1-FAST-NEXT: vextractf128 $1, %ymm0, %xmm1
|
|
; AVX1-FAST-NEXT: vphaddd %xmm0, %xmm1, %xmm0
|
|
; AVX1-FAST-NEXT: vphaddd %xmm0, %xmm0, %xmm0
|
|
; AVX1-FAST-NEXT: vphaddd %xmm0, %xmm0, %xmm0
|
|
; AVX1-FAST-NEXT: vmovd %xmm0, %eax
|
|
; AVX1-FAST-NEXT: vzeroupper
|
|
; AVX1-FAST-NEXT: ret{{[l|q]}}
|
|
;
|
|
; AVX2-LABEL: test_v8i32:
|
|
; AVX2: # %bb.0:
|
|
; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1
|
|
; AVX2-NEXT: vpaddd %xmm1, %xmm0, %xmm0
|
|
; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
|
|
; AVX2-NEXT: vpaddd %xmm1, %xmm0, %xmm0
|
|
; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
|
|
; AVX2-NEXT: vpaddd %xmm1, %xmm0, %xmm0
|
|
; AVX2-NEXT: vmovd %xmm0, %eax
|
|
; AVX2-NEXT: vzeroupper
|
|
; AVX2-NEXT: ret{{[l|q]}}
|
|
;
|
|
; AVX512-LABEL: test_v8i32:
|
|
; AVX512: # %bb.0:
|
|
; AVX512-NEXT: vextracti128 $1, %ymm0, %xmm1
|
|
; AVX512-NEXT: vpaddd %xmm1, %xmm0, %xmm0
|
|
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
|
|
; AVX512-NEXT: vpaddd %xmm1, %xmm0, %xmm0
|
|
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
|
|
; AVX512-NEXT: vpaddd %xmm1, %xmm0, %xmm0
|
|
; AVX512-NEXT: vmovd %xmm0, %eax
|
|
; AVX512-NEXT: vzeroupper
|
|
; AVX512-NEXT: retq
|
|
%1 = call i32 @llvm.vector.reduce.add.v8i32(<8 x i32> %a0)
|
|
ret i32 %1
|
|
}
|
|
|
|
define i32 @test_v16i32(<16 x i32> %a0) nounwind {
|
|
; X86-SSE-LABEL: test_v16i32:
|
|
; X86-SSE: # %bb.0:
|
|
; X86-SSE-NEXT: pushl %ebp
|
|
; X86-SSE-NEXT: movl %esp, %ebp
|
|
; X86-SSE-NEXT: andl $-16, %esp
|
|
; X86-SSE-NEXT: subl $16, %esp
|
|
; X86-SSE-NEXT: paddd %xmm2, %xmm0
|
|
; X86-SSE-NEXT: paddd 8(%ebp), %xmm1
|
|
; X86-SSE-NEXT: paddd %xmm0, %xmm1
|
|
; X86-SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm1[2,3,2,3]
|
|
; X86-SSE-NEXT: paddd %xmm1, %xmm0
|
|
; X86-SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
|
|
; X86-SSE-NEXT: paddd %xmm0, %xmm1
|
|
; X86-SSE-NEXT: movd %xmm1, %eax
|
|
; X86-SSE-NEXT: movl %ebp, %esp
|
|
; X86-SSE-NEXT: popl %ebp
|
|
; X86-SSE-NEXT: retl
|
|
;
|
|
; X64-SSE-LABEL: test_v16i32:
|
|
; X64-SSE: # %bb.0:
|
|
; X64-SSE-NEXT: paddd %xmm3, %xmm1
|
|
; X64-SSE-NEXT: paddd %xmm2, %xmm0
|
|
; X64-SSE-NEXT: paddd %xmm1, %xmm0
|
|
; X64-SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
|
|
; X64-SSE-NEXT: paddd %xmm0, %xmm1
|
|
; X64-SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm1[1,1,1,1]
|
|
; X64-SSE-NEXT: paddd %xmm1, %xmm0
|
|
; X64-SSE-NEXT: movd %xmm0, %eax
|
|
; X64-SSE-NEXT: retq
|
|
;
|
|
; AVX1-SLOW-LABEL: test_v16i32:
|
|
; AVX1-SLOW: # %bb.0:
|
|
; AVX1-SLOW-NEXT: vextractf128 $1, %ymm1, %xmm2
|
|
; AVX1-SLOW-NEXT: vextractf128 $1, %ymm0, %xmm3
|
|
; AVX1-SLOW-NEXT: vpaddd %xmm2, %xmm3, %xmm2
|
|
; AVX1-SLOW-NEXT: vpaddd %xmm1, %xmm0, %xmm0
|
|
; AVX1-SLOW-NEXT: vpaddd %xmm2, %xmm0, %xmm0
|
|
; AVX1-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
|
|
; AVX1-SLOW-NEXT: vpaddd %xmm1, %xmm0, %xmm0
|
|
; AVX1-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
|
|
; AVX1-SLOW-NEXT: vpaddd %xmm1, %xmm0, %xmm0
|
|
; AVX1-SLOW-NEXT: vmovd %xmm0, %eax
|
|
; AVX1-SLOW-NEXT: vzeroupper
|
|
; AVX1-SLOW-NEXT: ret{{[l|q]}}
|
|
;
|
|
; AVX1-FAST-LABEL: test_v16i32:
|
|
; AVX1-FAST: # %bb.0:
|
|
; AVX1-FAST-NEXT: vextractf128 $1, %ymm1, %xmm2
|
|
; AVX1-FAST-NEXT: vextractf128 $1, %ymm0, %xmm3
|
|
; AVX1-FAST-NEXT: vpaddd %xmm2, %xmm3, %xmm2
|
|
; AVX1-FAST-NEXT: vpaddd %xmm1, %xmm0, %xmm0
|
|
; AVX1-FAST-NEXT: vpaddd %xmm2, %xmm0, %xmm0
|
|
; AVX1-FAST-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
|
|
; AVX1-FAST-NEXT: vpaddd %xmm1, %xmm0, %xmm0
|
|
; AVX1-FAST-NEXT: vphaddd %xmm0, %xmm0, %xmm0
|
|
; AVX1-FAST-NEXT: vmovd %xmm0, %eax
|
|
; AVX1-FAST-NEXT: vzeroupper
|
|
; AVX1-FAST-NEXT: ret{{[l|q]}}
|
|
;
|
|
; AVX2-LABEL: test_v16i32:
|
|
; AVX2: # %bb.0:
|
|
; AVX2-NEXT: vpaddd %ymm1, %ymm0, %ymm0
|
|
; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1
|
|
; AVX2-NEXT: vpaddd %xmm1, %xmm0, %xmm0
|
|
; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
|
|
; AVX2-NEXT: vpaddd %xmm1, %xmm0, %xmm0
|
|
; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
|
|
; AVX2-NEXT: vpaddd %xmm1, %xmm0, %xmm0
|
|
; AVX2-NEXT: vmovd %xmm0, %eax
|
|
; AVX2-NEXT: vzeroupper
|
|
; AVX2-NEXT: ret{{[l|q]}}
|
|
;
|
|
; AVX512-LABEL: test_v16i32:
|
|
; AVX512: # %bb.0:
|
|
; AVX512-NEXT: vextracti64x4 $1, %zmm0, %ymm1
|
|
; AVX512-NEXT: vpaddd %zmm1, %zmm0, %zmm0
|
|
; AVX512-NEXT: vextracti128 $1, %ymm0, %xmm1
|
|
; AVX512-NEXT: vpaddd %xmm1, %xmm0, %xmm0
|
|
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
|
|
; AVX512-NEXT: vpaddd %xmm1, %xmm0, %xmm0
|
|
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
|
|
; AVX512-NEXT: vpaddd %xmm1, %xmm0, %xmm0
|
|
; AVX512-NEXT: vmovd %xmm0, %eax
|
|
; AVX512-NEXT: vzeroupper
|
|
; AVX512-NEXT: retq
|
|
%1 = call i32 @llvm.vector.reduce.add.v16i32(<16 x i32> %a0)
|
|
ret i32 %1
|
|
}
|
|
|
|
define i32 @test_v32i32(<32 x i32> %a0) nounwind {
|
|
; X86-SSE-LABEL: test_v32i32:
|
|
; X86-SSE: # %bb.0:
|
|
; X86-SSE-NEXT: pushl %ebp
|
|
; X86-SSE-NEXT: movl %esp, %ebp
|
|
; X86-SSE-NEXT: andl $-16, %esp
|
|
; X86-SSE-NEXT: subl $16, %esp
|
|
; X86-SSE-NEXT: movdqa 8(%ebp), %xmm3
|
|
; X86-SSE-NEXT: paddd 56(%ebp), %xmm2
|
|
; X86-SSE-NEXT: paddd 24(%ebp), %xmm0
|
|
; X86-SSE-NEXT: paddd %xmm2, %xmm0
|
|
; X86-SSE-NEXT: paddd 72(%ebp), %xmm3
|
|
; X86-SSE-NEXT: paddd 40(%ebp), %xmm1
|
|
; X86-SSE-NEXT: paddd %xmm3, %xmm1
|
|
; X86-SSE-NEXT: paddd %xmm0, %xmm1
|
|
; X86-SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm1[2,3,2,3]
|
|
; X86-SSE-NEXT: paddd %xmm1, %xmm0
|
|
; X86-SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
|
|
; X86-SSE-NEXT: paddd %xmm0, %xmm1
|
|
; X86-SSE-NEXT: movd %xmm1, %eax
|
|
; X86-SSE-NEXT: movl %ebp, %esp
|
|
; X86-SSE-NEXT: popl %ebp
|
|
; X86-SSE-NEXT: retl
|
|
;
|
|
; X64-SSE-LABEL: test_v32i32:
|
|
; X64-SSE: # %bb.0:
|
|
; X64-SSE-NEXT: paddd %xmm6, %xmm2
|
|
; X64-SSE-NEXT: paddd %xmm4, %xmm0
|
|
; X64-SSE-NEXT: paddd %xmm2, %xmm0
|
|
; X64-SSE-NEXT: paddd %xmm7, %xmm3
|
|
; X64-SSE-NEXT: paddd %xmm5, %xmm1
|
|
; X64-SSE-NEXT: paddd %xmm3, %xmm1
|
|
; X64-SSE-NEXT: paddd %xmm0, %xmm1
|
|
; X64-SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm1[2,3,2,3]
|
|
; X64-SSE-NEXT: paddd %xmm1, %xmm0
|
|
; X64-SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
|
|
; X64-SSE-NEXT: paddd %xmm0, %xmm1
|
|
; X64-SSE-NEXT: movd %xmm1, %eax
|
|
; X64-SSE-NEXT: retq
|
|
;
|
|
; X86-AVX1-SLOW-LABEL: test_v32i32:
|
|
; X86-AVX1-SLOW: # %bb.0:
|
|
; X86-AVX1-SLOW-NEXT: pushl %ebp
|
|
; X86-AVX1-SLOW-NEXT: movl %esp, %ebp
|
|
; X86-AVX1-SLOW-NEXT: andl $-32, %esp
|
|
; X86-AVX1-SLOW-NEXT: subl $32, %esp
|
|
; X86-AVX1-SLOW-NEXT: vpaddd %xmm2, %xmm0, %xmm3
|
|
; X86-AVX1-SLOW-NEXT: vextractf128 $1, %ymm2, %xmm2
|
|
; X86-AVX1-SLOW-NEXT: vextractf128 $1, %ymm0, %xmm0
|
|
; X86-AVX1-SLOW-NEXT: vpaddd %xmm2, %xmm0, %xmm0
|
|
; X86-AVX1-SLOW-NEXT: vpaddd 8(%ebp), %xmm1, %xmm2
|
|
; X86-AVX1-SLOW-NEXT: vpaddd %xmm2, %xmm3, %xmm2
|
|
; X86-AVX1-SLOW-NEXT: vextractf128 $1, %ymm1, %xmm1
|
|
; X86-AVX1-SLOW-NEXT: vpaddd 24(%ebp), %xmm1, %xmm1
|
|
; X86-AVX1-SLOW-NEXT: vpaddd %xmm1, %xmm0, %xmm0
|
|
; X86-AVX1-SLOW-NEXT: vpaddd %xmm0, %xmm2, %xmm0
|
|
; X86-AVX1-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
|
|
; X86-AVX1-SLOW-NEXT: vpaddd %xmm1, %xmm0, %xmm0
|
|
; X86-AVX1-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
|
|
; X86-AVX1-SLOW-NEXT: vpaddd %xmm1, %xmm0, %xmm0
|
|
; X86-AVX1-SLOW-NEXT: vmovd %xmm0, %eax
|
|
; X86-AVX1-SLOW-NEXT: movl %ebp, %esp
|
|
; X86-AVX1-SLOW-NEXT: popl %ebp
|
|
; X86-AVX1-SLOW-NEXT: vzeroupper
|
|
; X86-AVX1-SLOW-NEXT: retl
|
|
;
|
|
; X64-AVX1-SLOW-LABEL: test_v32i32:
|
|
; X64-AVX1-SLOW: # %bb.0:
|
|
; X64-AVX1-SLOW-NEXT: vpaddd %xmm3, %xmm1, %xmm4
|
|
; X64-AVX1-SLOW-NEXT: vpaddd %xmm2, %xmm0, %xmm5
|
|
; X64-AVX1-SLOW-NEXT: vpaddd %xmm4, %xmm5, %xmm4
|
|
; X64-AVX1-SLOW-NEXT: vextractf128 $1, %ymm3, %xmm3
|
|
; X64-AVX1-SLOW-NEXT: vextractf128 $1, %ymm1, %xmm1
|
|
; X64-AVX1-SLOW-NEXT: vpaddd %xmm3, %xmm1, %xmm1
|
|
; X64-AVX1-SLOW-NEXT: vextractf128 $1, %ymm2, %xmm2
|
|
; X64-AVX1-SLOW-NEXT: vextractf128 $1, %ymm0, %xmm0
|
|
; X64-AVX1-SLOW-NEXT: vpaddd %xmm2, %xmm0, %xmm0
|
|
; X64-AVX1-SLOW-NEXT: vpaddd %xmm1, %xmm0, %xmm0
|
|
; X64-AVX1-SLOW-NEXT: vpaddd %xmm0, %xmm4, %xmm0
|
|
; X64-AVX1-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
|
|
; X64-AVX1-SLOW-NEXT: vpaddd %xmm1, %xmm0, %xmm0
|
|
; X64-AVX1-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
|
|
; X64-AVX1-SLOW-NEXT: vpaddd %xmm1, %xmm0, %xmm0
|
|
; X64-AVX1-SLOW-NEXT: vmovd %xmm0, %eax
|
|
; X64-AVX1-SLOW-NEXT: vzeroupper
|
|
; X64-AVX1-SLOW-NEXT: retq
|
|
;
|
|
; X86-AVX1-FAST-LABEL: test_v32i32:
|
|
; X86-AVX1-FAST: # %bb.0:
|
|
; X86-AVX1-FAST-NEXT: pushl %ebp
|
|
; X86-AVX1-FAST-NEXT: movl %esp, %ebp
|
|
; X86-AVX1-FAST-NEXT: andl $-32, %esp
|
|
; X86-AVX1-FAST-NEXT: subl $32, %esp
|
|
; X86-AVX1-FAST-NEXT: vpaddd %xmm2, %xmm0, %xmm3
|
|
; X86-AVX1-FAST-NEXT: vextractf128 $1, %ymm2, %xmm2
|
|
; X86-AVX1-FAST-NEXT: vextractf128 $1, %ymm0, %xmm0
|
|
; X86-AVX1-FAST-NEXT: vpaddd %xmm2, %xmm0, %xmm0
|
|
; X86-AVX1-FAST-NEXT: vpaddd 8(%ebp), %xmm1, %xmm2
|
|
; X86-AVX1-FAST-NEXT: vpaddd %xmm2, %xmm3, %xmm2
|
|
; X86-AVX1-FAST-NEXT: vextractf128 $1, %ymm1, %xmm1
|
|
; X86-AVX1-FAST-NEXT: vpaddd 24(%ebp), %xmm1, %xmm1
|
|
; X86-AVX1-FAST-NEXT: vpaddd %xmm1, %xmm0, %xmm0
|
|
; X86-AVX1-FAST-NEXT: vpaddd %xmm0, %xmm2, %xmm0
|
|
; X86-AVX1-FAST-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
|
|
; X86-AVX1-FAST-NEXT: vpaddd %xmm1, %xmm0, %xmm0
|
|
; X86-AVX1-FAST-NEXT: vphaddd %xmm0, %xmm0, %xmm0
|
|
; X86-AVX1-FAST-NEXT: vmovd %xmm0, %eax
|
|
; X86-AVX1-FAST-NEXT: movl %ebp, %esp
|
|
; X86-AVX1-FAST-NEXT: popl %ebp
|
|
; X86-AVX1-FAST-NEXT: vzeroupper
|
|
; X86-AVX1-FAST-NEXT: retl
|
|
;
|
|
; X64-AVX1-FAST-LABEL: test_v32i32:
|
|
; X64-AVX1-FAST: # %bb.0:
|
|
; X64-AVX1-FAST-NEXT: vpaddd %xmm3, %xmm1, %xmm4
|
|
; X64-AVX1-FAST-NEXT: vpaddd %xmm2, %xmm0, %xmm5
|
|
; X64-AVX1-FAST-NEXT: vpaddd %xmm4, %xmm5, %xmm4
|
|
; X64-AVX1-FAST-NEXT: vextractf128 $1, %ymm3, %xmm3
|
|
; X64-AVX1-FAST-NEXT: vextractf128 $1, %ymm1, %xmm1
|
|
; X64-AVX1-FAST-NEXT: vpaddd %xmm3, %xmm1, %xmm1
|
|
; X64-AVX1-FAST-NEXT: vextractf128 $1, %ymm2, %xmm2
|
|
; X64-AVX1-FAST-NEXT: vextractf128 $1, %ymm0, %xmm0
|
|
; X64-AVX1-FAST-NEXT: vpaddd %xmm2, %xmm0, %xmm0
|
|
; X64-AVX1-FAST-NEXT: vpaddd %xmm1, %xmm0, %xmm0
|
|
; X64-AVX1-FAST-NEXT: vpaddd %xmm0, %xmm4, %xmm0
|
|
; X64-AVX1-FAST-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
|
|
; X64-AVX1-FAST-NEXT: vpaddd %xmm1, %xmm0, %xmm0
|
|
; X64-AVX1-FAST-NEXT: vphaddd %xmm0, %xmm0, %xmm0
|
|
; X64-AVX1-FAST-NEXT: vmovd %xmm0, %eax
|
|
; X64-AVX1-FAST-NEXT: vzeroupper
|
|
; X64-AVX1-FAST-NEXT: retq
|
|
;
|
|
; X86-AVX2-LABEL: test_v32i32:
|
|
; X86-AVX2: # %bb.0:
|
|
; X86-AVX2-NEXT: pushl %ebp
|
|
; X86-AVX2-NEXT: movl %esp, %ebp
|
|
; X86-AVX2-NEXT: andl $-32, %esp
|
|
; X86-AVX2-NEXT: subl $32, %esp
|
|
; X86-AVX2-NEXT: vpaddd %ymm2, %ymm0, %ymm0
|
|
; X86-AVX2-NEXT: vpaddd 8(%ebp), %ymm1, %ymm1
|
|
; X86-AVX2-NEXT: vpaddd %ymm1, %ymm0, %ymm0
|
|
; X86-AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1
|
|
; X86-AVX2-NEXT: vpaddd %xmm1, %xmm0, %xmm0
|
|
; X86-AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
|
|
; X86-AVX2-NEXT: vpaddd %xmm1, %xmm0, %xmm0
|
|
; X86-AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
|
|
; X86-AVX2-NEXT: vpaddd %xmm1, %xmm0, %xmm0
|
|
; X86-AVX2-NEXT: vmovd %xmm0, %eax
|
|
; X86-AVX2-NEXT: movl %ebp, %esp
|
|
; X86-AVX2-NEXT: popl %ebp
|
|
; X86-AVX2-NEXT: vzeroupper
|
|
; X86-AVX2-NEXT: retl
|
|
;
|
|
; X64-AVX2-LABEL: test_v32i32:
|
|
; X64-AVX2: # %bb.0:
|
|
; X64-AVX2-NEXT: vpaddd %ymm3, %ymm1, %ymm1
|
|
; X64-AVX2-NEXT: vpaddd %ymm2, %ymm0, %ymm0
|
|
; X64-AVX2-NEXT: vpaddd %ymm1, %ymm0, %ymm0
|
|
; X64-AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1
|
|
; X64-AVX2-NEXT: vpaddd %xmm1, %xmm0, %xmm0
|
|
; X64-AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
|
|
; X64-AVX2-NEXT: vpaddd %xmm1, %xmm0, %xmm0
|
|
; X64-AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
|
|
; X64-AVX2-NEXT: vpaddd %xmm1, %xmm0, %xmm0
|
|
; X64-AVX2-NEXT: vmovd %xmm0, %eax
|
|
; X64-AVX2-NEXT: vzeroupper
|
|
; X64-AVX2-NEXT: retq
|
|
;
|
|
; AVX512-LABEL: test_v32i32:
|
|
; AVX512: # %bb.0:
|
|
; AVX512-NEXT: vpaddd %zmm1, %zmm0, %zmm0
|
|
; AVX512-NEXT: vextracti64x4 $1, %zmm0, %ymm1
|
|
; AVX512-NEXT: vpaddd %zmm1, %zmm0, %zmm0
|
|
; AVX512-NEXT: vextracti128 $1, %ymm0, %xmm1
|
|
; AVX512-NEXT: vpaddd %xmm1, %xmm0, %xmm0
|
|
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
|
|
; AVX512-NEXT: vpaddd %xmm1, %xmm0, %xmm0
|
|
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
|
|
; AVX512-NEXT: vpaddd %xmm1, %xmm0, %xmm0
|
|
; AVX512-NEXT: vmovd %xmm0, %eax
|
|
; AVX512-NEXT: vzeroupper
|
|
; AVX512-NEXT: retq
|
|
%1 = call i32 @llvm.vector.reduce.add.v32i32(<32 x i32> %a0)
|
|
ret i32 %1
|
|
}
|
|
|
|
;
|
|
; vXi16
|
|
;
|
|
|
|
define i16 @test_v2i16(<2 x i16> %a0) nounwind {
|
|
; SSE-LABEL: test_v2i16:
|
|
; SSE: # %bb.0:
|
|
; SSE-NEXT: movdqa %xmm0, %xmm1
|
|
; SSE-NEXT: psrld $16, %xmm1
|
|
; SSE-NEXT: paddw %xmm0, %xmm1
|
|
; SSE-NEXT: movd %xmm1, %eax
|
|
; SSE-NEXT: # kill: def $ax killed $ax killed $eax
|
|
; SSE-NEXT: ret{{[l|q]}}
|
|
;
|
|
; AVX1-SLOW-LABEL: test_v2i16:
|
|
; AVX1-SLOW: # %bb.0:
|
|
; AVX1-SLOW-NEXT: vpsrld $16, %xmm0, %xmm1
|
|
; AVX1-SLOW-NEXT: vpaddw %xmm1, %xmm0, %xmm0
|
|
; AVX1-SLOW-NEXT: vmovd %xmm0, %eax
|
|
; AVX1-SLOW-NEXT: # kill: def $ax killed $ax killed $eax
|
|
; AVX1-SLOW-NEXT: ret{{[l|q]}}
|
|
;
|
|
; AVX1-FAST-LABEL: test_v2i16:
|
|
; AVX1-FAST: # %bb.0:
|
|
; AVX1-FAST-NEXT: vphaddw %xmm0, %xmm0, %xmm0
|
|
; AVX1-FAST-NEXT: vmovd %xmm0, %eax
|
|
; AVX1-FAST-NEXT: # kill: def $ax killed $ax killed $eax
|
|
; AVX1-FAST-NEXT: ret{{[l|q]}}
|
|
;
|
|
; AVX2-LABEL: test_v2i16:
|
|
; AVX2: # %bb.0:
|
|
; AVX2-NEXT: vpsrld $16, %xmm0, %xmm1
|
|
; AVX2-NEXT: vpaddw %xmm1, %xmm0, %xmm0
|
|
; AVX2-NEXT: vmovd %xmm0, %eax
|
|
; AVX2-NEXT: # kill: def $ax killed $ax killed $eax
|
|
; AVX2-NEXT: ret{{[l|q]}}
|
|
;
|
|
; AVX512-LABEL: test_v2i16:
|
|
; AVX512: # %bb.0:
|
|
; AVX512-NEXT: vpsrld $16, %xmm0, %xmm1
|
|
; AVX512-NEXT: vpaddw %xmm1, %xmm0, %xmm0
|
|
; AVX512-NEXT: vmovd %xmm0, %eax
|
|
; AVX512-NEXT: # kill: def $ax killed $ax killed $eax
|
|
; AVX512-NEXT: retq
|
|
%1 = call i16 @llvm.vector.reduce.add.v2i16(<2 x i16> %a0)
|
|
ret i16 %1
|
|
}
|
|
|
|
define i16 @test_v4i16(<4 x i16> %a0) nounwind {
|
|
; SSE-LABEL: test_v4i16:
|
|
; SSE: # %bb.0:
|
|
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
|
|
; SSE-NEXT: paddw %xmm0, %xmm1
|
|
; SSE-NEXT: movdqa %xmm1, %xmm0
|
|
; SSE-NEXT: psrld $16, %xmm0
|
|
; SSE-NEXT: paddw %xmm1, %xmm0
|
|
; SSE-NEXT: movd %xmm0, %eax
|
|
; SSE-NEXT: # kill: def $ax killed $ax killed $eax
|
|
; SSE-NEXT: ret{{[l|q]}}
|
|
;
|
|
; AVX1-SLOW-LABEL: test_v4i16:
|
|
; AVX1-SLOW: # %bb.0:
|
|
; AVX1-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
|
|
; AVX1-SLOW-NEXT: vpaddw %xmm1, %xmm0, %xmm0
|
|
; AVX1-SLOW-NEXT: vpsrld $16, %xmm0, %xmm1
|
|
; AVX1-SLOW-NEXT: vpaddw %xmm1, %xmm0, %xmm0
|
|
; AVX1-SLOW-NEXT: vmovd %xmm0, %eax
|
|
; AVX1-SLOW-NEXT: # kill: def $ax killed $ax killed $eax
|
|
; AVX1-SLOW-NEXT: ret{{[l|q]}}
|
|
;
|
|
; AVX1-FAST-LABEL: test_v4i16:
|
|
; AVX1-FAST: # %bb.0:
|
|
; AVX1-FAST-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
|
|
; AVX1-FAST-NEXT: vpaddw %xmm1, %xmm0, %xmm0
|
|
; AVX1-FAST-NEXT: vphaddw %xmm0, %xmm0, %xmm0
|
|
; AVX1-FAST-NEXT: vmovd %xmm0, %eax
|
|
; AVX1-FAST-NEXT: # kill: def $ax killed $ax killed $eax
|
|
; AVX1-FAST-NEXT: ret{{[l|q]}}
|
|
;
|
|
; AVX2-LABEL: test_v4i16:
|
|
; AVX2: # %bb.0:
|
|
; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
|
|
; AVX2-NEXT: vpaddw %xmm1, %xmm0, %xmm0
|
|
; AVX2-NEXT: vpsrld $16, %xmm0, %xmm1
|
|
; AVX2-NEXT: vpaddw %xmm1, %xmm0, %xmm0
|
|
; AVX2-NEXT: vmovd %xmm0, %eax
|
|
; AVX2-NEXT: # kill: def $ax killed $ax killed $eax
|
|
; AVX2-NEXT: ret{{[l|q]}}
|
|
;
|
|
; AVX512-LABEL: test_v4i16:
|
|
; AVX512: # %bb.0:
|
|
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
|
|
; AVX512-NEXT: vpaddw %xmm1, %xmm0, %xmm0
|
|
; AVX512-NEXT: vpsrld $16, %xmm0, %xmm1
|
|
; AVX512-NEXT: vpaddw %xmm1, %xmm0, %xmm0
|
|
; AVX512-NEXT: vmovd %xmm0, %eax
|
|
; AVX512-NEXT: # kill: def $ax killed $ax killed $eax
|
|
; AVX512-NEXT: retq
|
|
%1 = call i16 @llvm.vector.reduce.add.v4i16(<4 x i16> %a0)
|
|
ret i16 %1
|
|
}
|
|
|
|
define i16 @test_v8i16(<8 x i16> %a0) nounwind {
|
|
; SSE-LABEL: test_v8i16:
|
|
; SSE: # %bb.0:
|
|
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
|
|
; SSE-NEXT: paddw %xmm0, %xmm1
|
|
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm1[1,1,1,1]
|
|
; SSE-NEXT: paddw %xmm1, %xmm0
|
|
; SSE-NEXT: movdqa %xmm0, %xmm1
|
|
; SSE-NEXT: psrld $16, %xmm1
|
|
; SSE-NEXT: paddw %xmm0, %xmm1
|
|
; SSE-NEXT: movd %xmm1, %eax
|
|
; SSE-NEXT: # kill: def $ax killed $ax killed $eax
|
|
; SSE-NEXT: ret{{[l|q]}}
|
|
;
|
|
; AVX1-SLOW-LABEL: test_v8i16:
|
|
; AVX1-SLOW: # %bb.0:
|
|
; AVX1-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
|
|
; AVX1-SLOW-NEXT: vpaddw %xmm1, %xmm0, %xmm0
|
|
; AVX1-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
|
|
; AVX1-SLOW-NEXT: vpaddw %xmm1, %xmm0, %xmm0
|
|
; AVX1-SLOW-NEXT: vpsrld $16, %xmm0, %xmm1
|
|
; AVX1-SLOW-NEXT: vpaddw %xmm1, %xmm0, %xmm0
|
|
; AVX1-SLOW-NEXT: vmovd %xmm0, %eax
|
|
; AVX1-SLOW-NEXT: # kill: def $ax killed $ax killed $eax
|
|
; AVX1-SLOW-NEXT: ret{{[l|q]}}
|
|
;
|
|
; AVX1-FAST-LABEL: test_v8i16:
|
|
; AVX1-FAST: # %bb.0:
|
|
; AVX1-FAST-NEXT: vphaddw %xmm0, %xmm0, %xmm0
|
|
; AVX1-FAST-NEXT: vphaddw %xmm0, %xmm0, %xmm0
|
|
; AVX1-FAST-NEXT: vphaddw %xmm0, %xmm0, %xmm0
|
|
; AVX1-FAST-NEXT: vmovd %xmm0, %eax
|
|
; AVX1-FAST-NEXT: # kill: def $ax killed $ax killed $eax
|
|
; AVX1-FAST-NEXT: ret{{[l|q]}}
|
|
;
|
|
; AVX2-LABEL: test_v8i16:
|
|
; AVX2: # %bb.0:
|
|
; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
|
|
; AVX2-NEXT: vpaddw %xmm1, %xmm0, %xmm0
|
|
; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
|
|
; AVX2-NEXT: vpaddw %xmm1, %xmm0, %xmm0
|
|
; AVX2-NEXT: vpsrld $16, %xmm0, %xmm1
|
|
; AVX2-NEXT: vpaddw %xmm1, %xmm0, %xmm0
|
|
; AVX2-NEXT: vmovd %xmm0, %eax
|
|
; AVX2-NEXT: # kill: def $ax killed $ax killed $eax
|
|
; AVX2-NEXT: ret{{[l|q]}}
|
|
;
|
|
; AVX512-LABEL: test_v8i16:
|
|
; AVX512: # %bb.0:
|
|
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
|
|
; AVX512-NEXT: vpaddw %xmm1, %xmm0, %xmm0
|
|
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
|
|
; AVX512-NEXT: vpaddw %xmm1, %xmm0, %xmm0
|
|
; AVX512-NEXT: vpsrld $16, %xmm0, %xmm1
|
|
; AVX512-NEXT: vpaddw %xmm1, %xmm0, %xmm0
|
|
; AVX512-NEXT: vmovd %xmm0, %eax
|
|
; AVX512-NEXT: # kill: def $ax killed $ax killed $eax
|
|
; AVX512-NEXT: retq
|
|
%1 = call i16 @llvm.vector.reduce.add.v8i16(<8 x i16> %a0)
|
|
ret i16 %1
|
|
}
|
|
|
|
define i16 @test_v16i16(<16 x i16> %a0) nounwind {
|
|
; SSE-LABEL: test_v16i16:
|
|
; SSE: # %bb.0:
|
|
; SSE-NEXT: paddw %xmm1, %xmm0
|
|
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
|
|
; SSE-NEXT: paddw %xmm0, %xmm1
|
|
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm1[1,1,1,1]
|
|
; SSE-NEXT: paddw %xmm1, %xmm0
|
|
; SSE-NEXT: movdqa %xmm0, %xmm1
|
|
; SSE-NEXT: psrld $16, %xmm1
|
|
; SSE-NEXT: paddw %xmm0, %xmm1
|
|
; SSE-NEXT: movd %xmm1, %eax
|
|
; SSE-NEXT: # kill: def $ax killed $ax killed $eax
|
|
; SSE-NEXT: ret{{[l|q]}}
|
|
;
|
|
; AVX1-SLOW-LABEL: test_v16i16:
|
|
; AVX1-SLOW: # %bb.0:
|
|
; AVX1-SLOW-NEXT: vextractf128 $1, %ymm0, %xmm1
|
|
; AVX1-SLOW-NEXT: vpaddw %xmm1, %xmm0, %xmm0
|
|
; AVX1-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
|
|
; AVX1-SLOW-NEXT: vpaddw %xmm1, %xmm0, %xmm0
|
|
; AVX1-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
|
|
; AVX1-SLOW-NEXT: vpaddw %xmm1, %xmm0, %xmm0
|
|
; AVX1-SLOW-NEXT: vpsrld $16, %xmm0, %xmm1
|
|
; AVX1-SLOW-NEXT: vpaddw %xmm1, %xmm0, %xmm0
|
|
; AVX1-SLOW-NEXT: vmovd %xmm0, %eax
|
|
; AVX1-SLOW-NEXT: # kill: def $ax killed $ax killed $eax
|
|
; AVX1-SLOW-NEXT: vzeroupper
|
|
; AVX1-SLOW-NEXT: ret{{[l|q]}}
|
|
;
|
|
; AVX1-FAST-LABEL: test_v16i16:
|
|
; AVX1-FAST: # %bb.0:
|
|
; AVX1-FAST-NEXT: vextractf128 $1, %ymm0, %xmm1
|
|
; AVX1-FAST-NEXT: vphaddw %xmm0, %xmm1, %xmm0
|
|
; AVX1-FAST-NEXT: vphaddw %xmm0, %xmm0, %xmm0
|
|
; AVX1-FAST-NEXT: vphaddw %xmm0, %xmm0, %xmm0
|
|
; AVX1-FAST-NEXT: vphaddw %xmm0, %xmm0, %xmm0
|
|
; AVX1-FAST-NEXT: vmovd %xmm0, %eax
|
|
; AVX1-FAST-NEXT: # kill: def $ax killed $ax killed $eax
|
|
; AVX1-FAST-NEXT: vzeroupper
|
|
; AVX1-FAST-NEXT: ret{{[l|q]}}
|
|
;
|
|
; AVX2-LABEL: test_v16i16:
|
|
; AVX2: # %bb.0:
|
|
; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1
|
|
; AVX2-NEXT: vpaddw %xmm1, %xmm0, %xmm0
|
|
; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
|
|
; AVX2-NEXT: vpaddw %xmm1, %xmm0, %xmm0
|
|
; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
|
|
; AVX2-NEXT: vpaddw %xmm1, %xmm0, %xmm0
|
|
; AVX2-NEXT: vpsrld $16, %xmm0, %xmm1
|
|
; AVX2-NEXT: vpaddw %xmm1, %xmm0, %xmm0
|
|
; AVX2-NEXT: vmovd %xmm0, %eax
|
|
; AVX2-NEXT: # kill: def $ax killed $ax killed $eax
|
|
; AVX2-NEXT: vzeroupper
|
|
; AVX2-NEXT: ret{{[l|q]}}
|
|
;
|
|
; AVX512-LABEL: test_v16i16:
|
|
; AVX512: # %bb.0:
|
|
; AVX512-NEXT: vextracti128 $1, %ymm0, %xmm1
|
|
; AVX512-NEXT: vpaddw %xmm1, %xmm0, %xmm0
|
|
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
|
|
; AVX512-NEXT: vpaddw %xmm1, %xmm0, %xmm0
|
|
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
|
|
; AVX512-NEXT: vpaddw %xmm1, %xmm0, %xmm0
|
|
; AVX512-NEXT: vpsrld $16, %xmm0, %xmm1
|
|
; AVX512-NEXT: vpaddw %xmm1, %xmm0, %xmm0
|
|
; AVX512-NEXT: vmovd %xmm0, %eax
|
|
; AVX512-NEXT: # kill: def $ax killed $ax killed $eax
|
|
; AVX512-NEXT: vzeroupper
|
|
; AVX512-NEXT: retq
|
|
%1 = call i16 @llvm.vector.reduce.add.v16i16(<16 x i16> %a0)
|
|
ret i16 %1
|
|
}
|
|
|
|
define i16 @test_v32i16(<32 x i16> %a0) nounwind {
|
|
; X86-SSE-LABEL: test_v32i16:
|
|
; X86-SSE: # %bb.0:
|
|
; X86-SSE-NEXT: pushl %ebp
|
|
; X86-SSE-NEXT: movl %esp, %ebp
|
|
; X86-SSE-NEXT: andl $-16, %esp
|
|
; X86-SSE-NEXT: subl $16, %esp
|
|
; X86-SSE-NEXT: paddw %xmm2, %xmm0
|
|
; X86-SSE-NEXT: paddw 8(%ebp), %xmm1
|
|
; X86-SSE-NEXT: paddw %xmm0, %xmm1
|
|
; X86-SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm1[2,3,2,3]
|
|
; X86-SSE-NEXT: paddw %xmm1, %xmm0
|
|
; X86-SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
|
|
; X86-SSE-NEXT: paddw %xmm0, %xmm1
|
|
; X86-SSE-NEXT: movdqa %xmm1, %xmm0
|
|
; X86-SSE-NEXT: psrld $16, %xmm0
|
|
; X86-SSE-NEXT: paddw %xmm1, %xmm0
|
|
; X86-SSE-NEXT: movd %xmm0, %eax
|
|
; X86-SSE-NEXT: # kill: def $ax killed $ax killed $eax
|
|
; X86-SSE-NEXT: movl %ebp, %esp
|
|
; X86-SSE-NEXT: popl %ebp
|
|
; X86-SSE-NEXT: retl
|
|
;
|
|
; X64-SSE-LABEL: test_v32i16:
|
|
; X64-SSE: # %bb.0:
|
|
; X64-SSE-NEXT: paddw %xmm3, %xmm1
|
|
; X64-SSE-NEXT: paddw %xmm2, %xmm0
|
|
; X64-SSE-NEXT: paddw %xmm1, %xmm0
|
|
; X64-SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
|
|
; X64-SSE-NEXT: paddw %xmm0, %xmm1
|
|
; X64-SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm1[1,1,1,1]
|
|
; X64-SSE-NEXT: paddw %xmm1, %xmm0
|
|
; X64-SSE-NEXT: movdqa %xmm0, %xmm1
|
|
; X64-SSE-NEXT: psrld $16, %xmm1
|
|
; X64-SSE-NEXT: paddw %xmm0, %xmm1
|
|
; X64-SSE-NEXT: movd %xmm1, %eax
|
|
; X64-SSE-NEXT: # kill: def $ax killed $ax killed $eax
|
|
; X64-SSE-NEXT: retq
|
|
;
|
|
; AVX1-SLOW-LABEL: test_v32i16:
|
|
; AVX1-SLOW: # %bb.0:
|
|
; AVX1-SLOW-NEXT: vextractf128 $1, %ymm1, %xmm2
|
|
; AVX1-SLOW-NEXT: vextractf128 $1, %ymm0, %xmm3
|
|
; AVX1-SLOW-NEXT: vpaddw %xmm2, %xmm3, %xmm2
|
|
; AVX1-SLOW-NEXT: vpaddw %xmm1, %xmm0, %xmm0
|
|
; AVX1-SLOW-NEXT: vpaddw %xmm2, %xmm0, %xmm0
|
|
; AVX1-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
|
|
; AVX1-SLOW-NEXT: vpaddw %xmm1, %xmm0, %xmm0
|
|
; AVX1-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
|
|
; AVX1-SLOW-NEXT: vpaddw %xmm1, %xmm0, %xmm0
|
|
; AVX1-SLOW-NEXT: vpsrld $16, %xmm0, %xmm1
|
|
; AVX1-SLOW-NEXT: vpaddw %xmm1, %xmm0, %xmm0
|
|
; AVX1-SLOW-NEXT: vmovd %xmm0, %eax
|
|
; AVX1-SLOW-NEXT: # kill: def $ax killed $ax killed $eax
|
|
; AVX1-SLOW-NEXT: vzeroupper
|
|
; AVX1-SLOW-NEXT: ret{{[l|q]}}
|
|
;
|
|
; AVX1-FAST-LABEL: test_v32i16:
|
|
; AVX1-FAST: # %bb.0:
|
|
; AVX1-FAST-NEXT: vextractf128 $1, %ymm1, %xmm2
|
|
; AVX1-FAST-NEXT: vextractf128 $1, %ymm0, %xmm3
|
|
; AVX1-FAST-NEXT: vpaddw %xmm2, %xmm3, %xmm2
|
|
; AVX1-FAST-NEXT: vpaddw %xmm1, %xmm0, %xmm0
|
|
; AVX1-FAST-NEXT: vpaddw %xmm2, %xmm0, %xmm0
|
|
; AVX1-FAST-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
|
|
; AVX1-FAST-NEXT: vpaddw %xmm1, %xmm0, %xmm0
|
|
; AVX1-FAST-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
|
|
; AVX1-FAST-NEXT: vpaddw %xmm1, %xmm0, %xmm0
|
|
; AVX1-FAST-NEXT: vphaddw %xmm0, %xmm0, %xmm0
|
|
; AVX1-FAST-NEXT: vmovd %xmm0, %eax
|
|
; AVX1-FAST-NEXT: # kill: def $ax killed $ax killed $eax
|
|
; AVX1-FAST-NEXT: vzeroupper
|
|
; AVX1-FAST-NEXT: ret{{[l|q]}}
|
|
;
|
|
; AVX2-LABEL: test_v32i16:
|
|
; AVX2: # %bb.0:
|
|
; AVX2-NEXT: vpaddw %ymm1, %ymm0, %ymm0
|
|
; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1
|
|
; AVX2-NEXT: vpaddw %xmm1, %xmm0, %xmm0
|
|
; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
|
|
; AVX2-NEXT: vpaddw %xmm1, %xmm0, %xmm0
|
|
; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
|
|
; AVX2-NEXT: vpaddw %xmm1, %xmm0, %xmm0
|
|
; AVX2-NEXT: vpsrld $16, %xmm0, %xmm1
|
|
; AVX2-NEXT: vpaddw %xmm1, %xmm0, %xmm0
|
|
; AVX2-NEXT: vmovd %xmm0, %eax
|
|
; AVX2-NEXT: # kill: def $ax killed $ax killed $eax
|
|
; AVX2-NEXT: vzeroupper
|
|
; AVX2-NEXT: ret{{[l|q]}}
|
|
;
|
|
; AVX512-LABEL: test_v32i16:
|
|
; AVX512: # %bb.0:
|
|
; AVX512-NEXT: vextracti64x4 $1, %zmm0, %ymm1
|
|
; AVX512-NEXT: vpaddw %zmm1, %zmm0, %zmm0
|
|
; AVX512-NEXT: vextracti128 $1, %ymm0, %xmm1
|
|
; AVX512-NEXT: vpaddw %xmm1, %xmm0, %xmm0
|
|
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
|
|
; AVX512-NEXT: vpaddw %xmm1, %xmm0, %xmm0
|
|
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
|
|
; AVX512-NEXT: vpaddw %xmm1, %xmm0, %xmm0
|
|
; AVX512-NEXT: vpsrld $16, %xmm0, %xmm1
|
|
; AVX512-NEXT: vpaddw %xmm1, %xmm0, %xmm0
|
|
; AVX512-NEXT: vmovd %xmm0, %eax
|
|
; AVX512-NEXT: # kill: def $ax killed $ax killed $eax
|
|
; AVX512-NEXT: vzeroupper
|
|
; AVX512-NEXT: retq
|
|
%1 = call i16 @llvm.vector.reduce.add.v32i16(<32 x i16> %a0)
|
|
ret i16 %1
|
|
}
|
|
|
|
define i16 @test_v64i16(<64 x i16> %a0) nounwind {
|
|
; X86-SSE-LABEL: test_v64i16:
|
|
; X86-SSE: # %bb.0:
|
|
; X86-SSE-NEXT: pushl %ebp
|
|
; X86-SSE-NEXT: movl %esp, %ebp
|
|
; X86-SSE-NEXT: andl $-16, %esp
|
|
; X86-SSE-NEXT: subl $16, %esp
|
|
; X86-SSE-NEXT: movdqa 8(%ebp), %xmm3
|
|
; X86-SSE-NEXT: paddw 56(%ebp), %xmm2
|
|
; X86-SSE-NEXT: paddw 24(%ebp), %xmm0
|
|
; X86-SSE-NEXT: paddw %xmm2, %xmm0
|
|
; X86-SSE-NEXT: paddw 72(%ebp), %xmm3
|
|
; X86-SSE-NEXT: paddw 40(%ebp), %xmm1
|
|
; X86-SSE-NEXT: paddw %xmm3, %xmm1
|
|
; X86-SSE-NEXT: paddw %xmm0, %xmm1
|
|
; X86-SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm1[2,3,2,3]
|
|
; X86-SSE-NEXT: paddw %xmm1, %xmm0
|
|
; X86-SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
|
|
; X86-SSE-NEXT: paddw %xmm0, %xmm1
|
|
; X86-SSE-NEXT: movdqa %xmm1, %xmm0
|
|
; X86-SSE-NEXT: psrld $16, %xmm0
|
|
; X86-SSE-NEXT: paddw %xmm1, %xmm0
|
|
; X86-SSE-NEXT: movd %xmm0, %eax
|
|
; X86-SSE-NEXT: # kill: def $ax killed $ax killed $eax
|
|
; X86-SSE-NEXT: movl %ebp, %esp
|
|
; X86-SSE-NEXT: popl %ebp
|
|
; X86-SSE-NEXT: retl
|
|
;
|
|
; X64-SSE-LABEL: test_v64i16:
|
|
; X64-SSE: # %bb.0:
|
|
; X64-SSE-NEXT: paddw %xmm6, %xmm2
|
|
; X64-SSE-NEXT: paddw %xmm4, %xmm0
|
|
; X64-SSE-NEXT: paddw %xmm2, %xmm0
|
|
; X64-SSE-NEXT: paddw %xmm7, %xmm3
|
|
; X64-SSE-NEXT: paddw %xmm5, %xmm1
|
|
; X64-SSE-NEXT: paddw %xmm3, %xmm1
|
|
; X64-SSE-NEXT: paddw %xmm0, %xmm1
|
|
; X64-SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm1[2,3,2,3]
|
|
; X64-SSE-NEXT: paddw %xmm1, %xmm0
|
|
; X64-SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
|
|
; X64-SSE-NEXT: paddw %xmm0, %xmm1
|
|
; X64-SSE-NEXT: movdqa %xmm1, %xmm0
|
|
; X64-SSE-NEXT: psrld $16, %xmm0
|
|
; X64-SSE-NEXT: paddw %xmm1, %xmm0
|
|
; X64-SSE-NEXT: movd %xmm0, %eax
|
|
; X64-SSE-NEXT: # kill: def $ax killed $ax killed $eax
|
|
; X64-SSE-NEXT: retq
|
|
;
|
|
; X86-AVX1-SLOW-LABEL: test_v64i16:
|
|
; X86-AVX1-SLOW: # %bb.0:
|
|
; X86-AVX1-SLOW-NEXT: pushl %ebp
|
|
; X86-AVX1-SLOW-NEXT: movl %esp, %ebp
|
|
; X86-AVX1-SLOW-NEXT: andl $-32, %esp
|
|
; X86-AVX1-SLOW-NEXT: subl $32, %esp
|
|
; X86-AVX1-SLOW-NEXT: vpaddw %xmm2, %xmm0, %xmm3
|
|
; X86-AVX1-SLOW-NEXT: vextractf128 $1, %ymm2, %xmm2
|
|
; X86-AVX1-SLOW-NEXT: vextractf128 $1, %ymm0, %xmm0
|
|
; X86-AVX1-SLOW-NEXT: vpaddw %xmm2, %xmm0, %xmm0
|
|
; X86-AVX1-SLOW-NEXT: vpaddw 8(%ebp), %xmm1, %xmm2
|
|
; X86-AVX1-SLOW-NEXT: vpaddw %xmm2, %xmm3, %xmm2
|
|
; X86-AVX1-SLOW-NEXT: vextractf128 $1, %ymm1, %xmm1
|
|
; X86-AVX1-SLOW-NEXT: vpaddw 24(%ebp), %xmm1, %xmm1
|
|
; X86-AVX1-SLOW-NEXT: vpaddw %xmm1, %xmm0, %xmm0
|
|
; X86-AVX1-SLOW-NEXT: vpaddw %xmm0, %xmm2, %xmm0
|
|
; X86-AVX1-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
|
|
; X86-AVX1-SLOW-NEXT: vpaddw %xmm1, %xmm0, %xmm0
|
|
; X86-AVX1-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
|
|
; X86-AVX1-SLOW-NEXT: vpaddw %xmm1, %xmm0, %xmm0
|
|
; X86-AVX1-SLOW-NEXT: vpsrld $16, %xmm0, %xmm1
|
|
; X86-AVX1-SLOW-NEXT: vpaddw %xmm1, %xmm0, %xmm0
|
|
; X86-AVX1-SLOW-NEXT: vmovd %xmm0, %eax
|
|
; X86-AVX1-SLOW-NEXT: # kill: def $ax killed $ax killed $eax
|
|
; X86-AVX1-SLOW-NEXT: movl %ebp, %esp
|
|
; X86-AVX1-SLOW-NEXT: popl %ebp
|
|
; X86-AVX1-SLOW-NEXT: vzeroupper
|
|
; X86-AVX1-SLOW-NEXT: retl
|
|
;
|
|
; X64-AVX1-SLOW-LABEL: test_v64i16:
|
|
; X64-AVX1-SLOW: # %bb.0:
|
|
; X64-AVX1-SLOW-NEXT: vpaddw %xmm3, %xmm1, %xmm4
|
|
; X64-AVX1-SLOW-NEXT: vpaddw %xmm2, %xmm0, %xmm5
|
|
; X64-AVX1-SLOW-NEXT: vpaddw %xmm4, %xmm5, %xmm4
|
|
; X64-AVX1-SLOW-NEXT: vextractf128 $1, %ymm3, %xmm3
|
|
; X64-AVX1-SLOW-NEXT: vextractf128 $1, %ymm1, %xmm1
|
|
; X64-AVX1-SLOW-NEXT: vpaddw %xmm3, %xmm1, %xmm1
|
|
; X64-AVX1-SLOW-NEXT: vextractf128 $1, %ymm2, %xmm2
|
|
; X64-AVX1-SLOW-NEXT: vextractf128 $1, %ymm0, %xmm0
|
|
; X64-AVX1-SLOW-NEXT: vpaddw %xmm2, %xmm0, %xmm0
|
|
; X64-AVX1-SLOW-NEXT: vpaddw %xmm1, %xmm0, %xmm0
|
|
; X64-AVX1-SLOW-NEXT: vpaddw %xmm0, %xmm4, %xmm0
|
|
; X64-AVX1-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
|
|
; X64-AVX1-SLOW-NEXT: vpaddw %xmm1, %xmm0, %xmm0
|
|
; X64-AVX1-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
|
|
; X64-AVX1-SLOW-NEXT: vpaddw %xmm1, %xmm0, %xmm0
|
|
; X64-AVX1-SLOW-NEXT: vpsrld $16, %xmm0, %xmm1
|
|
; X64-AVX1-SLOW-NEXT: vpaddw %xmm1, %xmm0, %xmm0
|
|
; X64-AVX1-SLOW-NEXT: vmovd %xmm0, %eax
|
|
; X64-AVX1-SLOW-NEXT: # kill: def $ax killed $ax killed $eax
|
|
; X64-AVX1-SLOW-NEXT: vzeroupper
|
|
; X64-AVX1-SLOW-NEXT: retq
|
|
;
|
|
; X86-AVX1-FAST-LABEL: test_v64i16:
|
|
; X86-AVX1-FAST: # %bb.0:
|
|
; X86-AVX1-FAST-NEXT: pushl %ebp
|
|
; X86-AVX1-FAST-NEXT: movl %esp, %ebp
|
|
; X86-AVX1-FAST-NEXT: andl $-32, %esp
|
|
; X86-AVX1-FAST-NEXT: subl $32, %esp
|
|
; X86-AVX1-FAST-NEXT: vpaddw %xmm2, %xmm0, %xmm3
|
|
; X86-AVX1-FAST-NEXT: vextractf128 $1, %ymm2, %xmm2
|
|
; X86-AVX1-FAST-NEXT: vextractf128 $1, %ymm0, %xmm0
|
|
; X86-AVX1-FAST-NEXT: vpaddw %xmm2, %xmm0, %xmm0
|
|
; X86-AVX1-FAST-NEXT: vpaddw 8(%ebp), %xmm1, %xmm2
|
|
; X86-AVX1-FAST-NEXT: vpaddw %xmm2, %xmm3, %xmm2
|
|
; X86-AVX1-FAST-NEXT: vextractf128 $1, %ymm1, %xmm1
|
|
; X86-AVX1-FAST-NEXT: vpaddw 24(%ebp), %xmm1, %xmm1
|
|
; X86-AVX1-FAST-NEXT: vpaddw %xmm1, %xmm0, %xmm0
|
|
; X86-AVX1-FAST-NEXT: vpaddw %xmm0, %xmm2, %xmm0
|
|
; X86-AVX1-FAST-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
|
|
; X86-AVX1-FAST-NEXT: vpaddw %xmm1, %xmm0, %xmm0
|
|
; X86-AVX1-FAST-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
|
|
; X86-AVX1-FAST-NEXT: vpaddw %xmm1, %xmm0, %xmm0
|
|
; X86-AVX1-FAST-NEXT: vphaddw %xmm0, %xmm0, %xmm0
|
|
; X86-AVX1-FAST-NEXT: vmovd %xmm0, %eax
|
|
; X86-AVX1-FAST-NEXT: # kill: def $ax killed $ax killed $eax
|
|
; X86-AVX1-FAST-NEXT: movl %ebp, %esp
|
|
; X86-AVX1-FAST-NEXT: popl %ebp
|
|
; X86-AVX1-FAST-NEXT: vzeroupper
|
|
; X86-AVX1-FAST-NEXT: retl
|
|
;
|
|
; X64-AVX1-FAST-LABEL: test_v64i16:
|
|
; X64-AVX1-FAST: # %bb.0:
|
|
; X64-AVX1-FAST-NEXT: vpaddw %xmm3, %xmm1, %xmm4
|
|
; X64-AVX1-FAST-NEXT: vpaddw %xmm2, %xmm0, %xmm5
|
|
; X64-AVX1-FAST-NEXT: vpaddw %xmm4, %xmm5, %xmm4
|
|
; X64-AVX1-FAST-NEXT: vextractf128 $1, %ymm3, %xmm3
|
|
; X64-AVX1-FAST-NEXT: vextractf128 $1, %ymm1, %xmm1
|
|
; X64-AVX1-FAST-NEXT: vpaddw %xmm3, %xmm1, %xmm1
|
|
; X64-AVX1-FAST-NEXT: vextractf128 $1, %ymm2, %xmm2
|
|
; X64-AVX1-FAST-NEXT: vextractf128 $1, %ymm0, %xmm0
|
|
; X64-AVX1-FAST-NEXT: vpaddw %xmm2, %xmm0, %xmm0
|
|
; X64-AVX1-FAST-NEXT: vpaddw %xmm1, %xmm0, %xmm0
|
|
; X64-AVX1-FAST-NEXT: vpaddw %xmm0, %xmm4, %xmm0
|
|
; X64-AVX1-FAST-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
|
|
; X64-AVX1-FAST-NEXT: vpaddw %xmm1, %xmm0, %xmm0
|
|
; X64-AVX1-FAST-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
|
|
; X64-AVX1-FAST-NEXT: vpaddw %xmm1, %xmm0, %xmm0
|
|
; X64-AVX1-FAST-NEXT: vphaddw %xmm0, %xmm0, %xmm0
|
|
; X64-AVX1-FAST-NEXT: vmovd %xmm0, %eax
|
|
; X64-AVX1-FAST-NEXT: # kill: def $ax killed $ax killed $eax
|
|
; X64-AVX1-FAST-NEXT: vzeroupper
|
|
; X64-AVX1-FAST-NEXT: retq
|
|
;
|
|
; X86-AVX2-LABEL: test_v64i16:
|
|
; X86-AVX2: # %bb.0:
|
|
; X86-AVX2-NEXT: pushl %ebp
|
|
; X86-AVX2-NEXT: movl %esp, %ebp
|
|
; X86-AVX2-NEXT: andl $-32, %esp
|
|
; X86-AVX2-NEXT: subl $32, %esp
|
|
; X86-AVX2-NEXT: vpaddw %ymm2, %ymm0, %ymm0
|
|
; X86-AVX2-NEXT: vpaddw 8(%ebp), %ymm1, %ymm1
|
|
; X86-AVX2-NEXT: vpaddw %ymm1, %ymm0, %ymm0
|
|
; X86-AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1
|
|
; X86-AVX2-NEXT: vpaddw %xmm1, %xmm0, %xmm0
|
|
; X86-AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
|
|
; X86-AVX2-NEXT: vpaddw %xmm1, %xmm0, %xmm0
|
|
; X86-AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
|
|
; X86-AVX2-NEXT: vpaddw %xmm1, %xmm0, %xmm0
|
|
; X86-AVX2-NEXT: vpsrld $16, %xmm0, %xmm1
|
|
; X86-AVX2-NEXT: vpaddw %xmm1, %xmm0, %xmm0
|
|
; X86-AVX2-NEXT: vmovd %xmm0, %eax
|
|
; X86-AVX2-NEXT: # kill: def $ax killed $ax killed $eax
|
|
; X86-AVX2-NEXT: movl %ebp, %esp
|
|
; X86-AVX2-NEXT: popl %ebp
|
|
; X86-AVX2-NEXT: vzeroupper
|
|
; X86-AVX2-NEXT: retl
|
|
;
|
|
; X64-AVX2-LABEL: test_v64i16:
|
|
; X64-AVX2: # %bb.0:
|
|
; X64-AVX2-NEXT: vpaddw %ymm3, %ymm1, %ymm1
|
|
; X64-AVX2-NEXT: vpaddw %ymm2, %ymm0, %ymm0
|
|
; X64-AVX2-NEXT: vpaddw %ymm1, %ymm0, %ymm0
|
|
; X64-AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1
|
|
; X64-AVX2-NEXT: vpaddw %xmm1, %xmm0, %xmm0
|
|
; X64-AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
|
|
; X64-AVX2-NEXT: vpaddw %xmm1, %xmm0, %xmm0
|
|
; X64-AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
|
|
; X64-AVX2-NEXT: vpaddw %xmm1, %xmm0, %xmm0
|
|
; X64-AVX2-NEXT: vpsrld $16, %xmm0, %xmm1
|
|
; X64-AVX2-NEXT: vpaddw %xmm1, %xmm0, %xmm0
|
|
; X64-AVX2-NEXT: vmovd %xmm0, %eax
|
|
; X64-AVX2-NEXT: # kill: def $ax killed $ax killed $eax
|
|
; X64-AVX2-NEXT: vzeroupper
|
|
; X64-AVX2-NEXT: retq
|
|
;
|
|
; AVX512-LABEL: test_v64i16:
|
|
; AVX512: # %bb.0:
|
|
; AVX512-NEXT: vpaddw %zmm1, %zmm0, %zmm0
|
|
; AVX512-NEXT: vextracti64x4 $1, %zmm0, %ymm1
|
|
; AVX512-NEXT: vpaddw %zmm1, %zmm0, %zmm0
|
|
; AVX512-NEXT: vextracti128 $1, %ymm0, %xmm1
|
|
; AVX512-NEXT: vpaddw %xmm1, %xmm0, %xmm0
|
|
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
|
|
; AVX512-NEXT: vpaddw %xmm1, %xmm0, %xmm0
|
|
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
|
|
; AVX512-NEXT: vpaddw %xmm1, %xmm0, %xmm0
|
|
; AVX512-NEXT: vpsrld $16, %xmm0, %xmm1
|
|
; AVX512-NEXT: vpaddw %xmm1, %xmm0, %xmm0
|
|
; AVX512-NEXT: vmovd %xmm0, %eax
|
|
; AVX512-NEXT: # kill: def $ax killed $ax killed $eax
|
|
; AVX512-NEXT: vzeroupper
|
|
; AVX512-NEXT: retq
|
|
%1 = call i16 @llvm.vector.reduce.add.v64i16(<64 x i16> %a0)
|
|
ret i16 %1
|
|
}
|
|
|
|
;
|
|
; vXi8
|
|
;
|
|
|
|
define i8 @test_v2i8(<2 x i8> %a0) nounwind {
|
|
; SSE-LABEL: test_v2i8:
|
|
; SSE: # %bb.0:
|
|
; SSE-NEXT: movdqa %xmm0, %xmm1
|
|
; SSE-NEXT: psrlw $8, %xmm1
|
|
; SSE-NEXT: paddb %xmm0, %xmm1
|
|
; SSE-NEXT: movd %xmm1, %eax
|
|
; SSE-NEXT: # kill: def $al killed $al killed $eax
|
|
; SSE-NEXT: ret{{[l|q]}}
|
|
;
|
|
; AVX-LABEL: test_v2i8:
|
|
; AVX: # %bb.0:
|
|
; AVX-NEXT: vpsrlw $8, %xmm0, %xmm1
|
|
; AVX-NEXT: vpaddb %xmm1, %xmm0, %xmm0
|
|
; AVX-NEXT: vmovd %xmm0, %eax
|
|
; AVX-NEXT: # kill: def $al killed $al killed $eax
|
|
; AVX-NEXT: ret{{[l|q]}}
|
|
;
|
|
; AVX512-LABEL: test_v2i8:
|
|
; AVX512: # %bb.0:
|
|
; AVX512-NEXT: vpsrlw $8, %xmm0, %xmm1
|
|
; AVX512-NEXT: vpaddb %xmm1, %xmm0, %xmm0
|
|
; AVX512-NEXT: vmovd %xmm0, %eax
|
|
; AVX512-NEXT: # kill: def $al killed $al killed $eax
|
|
; AVX512-NEXT: retq
|
|
%1 = call i8 @llvm.vector.reduce.add.v2i8(<2 x i8> %a0)
|
|
ret i8 %1
|
|
}
|
|
|
|
define i8 @test_v2i8_load(ptr %p) nounwind {
|
|
; X86-SSE-LABEL: test_v2i8_load:
|
|
; X86-SSE: # %bb.0:
|
|
; X86-SSE-NEXT: movl {{[0-9]+}}(%esp), %eax
|
|
; X86-SSE-NEXT: movzwl (%eax), %eax
|
|
; X86-SSE-NEXT: movd %eax, %xmm0
|
|
; X86-SSE-NEXT: movdqa %xmm0, %xmm1
|
|
; X86-SSE-NEXT: psrlw $8, %xmm1
|
|
; X86-SSE-NEXT: paddb %xmm0, %xmm1
|
|
; X86-SSE-NEXT: movd %xmm1, %eax
|
|
; X86-SSE-NEXT: # kill: def $al killed $al killed $eax
|
|
; X86-SSE-NEXT: retl
|
|
;
|
|
; X64-SSE-LABEL: test_v2i8_load:
|
|
; X64-SSE: # %bb.0:
|
|
; X64-SSE-NEXT: movzwl (%rdi), %eax
|
|
; X64-SSE-NEXT: movd %eax, %xmm0
|
|
; X64-SSE-NEXT: movdqa %xmm0, %xmm1
|
|
; X64-SSE-NEXT: psrlw $8, %xmm1
|
|
; X64-SSE-NEXT: paddb %xmm0, %xmm1
|
|
; X64-SSE-NEXT: movd %xmm1, %eax
|
|
; X64-SSE-NEXT: # kill: def $al killed $al killed $eax
|
|
; X64-SSE-NEXT: retq
|
|
;
|
|
; X86-AVX1-LABEL: test_v2i8_load:
|
|
; X86-AVX1: # %bb.0:
|
|
; X86-AVX1-NEXT: movl {{[0-9]+}}(%esp), %eax
|
|
; X86-AVX1-NEXT: movzwl (%eax), %eax
|
|
; X86-AVX1-NEXT: vmovd %eax, %xmm0
|
|
; X86-AVX1-NEXT: vpsrlw $8, %xmm0, %xmm1
|
|
; X86-AVX1-NEXT: vpaddb %xmm1, %xmm0, %xmm0
|
|
; X86-AVX1-NEXT: vmovd %xmm0, %eax
|
|
; X86-AVX1-NEXT: # kill: def $al killed $al killed $eax
|
|
; X86-AVX1-NEXT: retl
|
|
;
|
|
; X64-AVX1-LABEL: test_v2i8_load:
|
|
; X64-AVX1: # %bb.0:
|
|
; X64-AVX1-NEXT: movzwl (%rdi), %eax
|
|
; X64-AVX1-NEXT: vmovd %eax, %xmm0
|
|
; X64-AVX1-NEXT: vpsrlw $8, %xmm0, %xmm1
|
|
; X64-AVX1-NEXT: vpaddb %xmm1, %xmm0, %xmm0
|
|
; X64-AVX1-NEXT: vmovd %xmm0, %eax
|
|
; X64-AVX1-NEXT: # kill: def $al killed $al killed $eax
|
|
; X64-AVX1-NEXT: retq
|
|
;
|
|
; X86-AVX2-LABEL: test_v2i8_load:
|
|
; X86-AVX2: # %bb.0:
|
|
; X86-AVX2-NEXT: movl {{[0-9]+}}(%esp), %eax
|
|
; X86-AVX2-NEXT: movzwl (%eax), %eax
|
|
; X86-AVX2-NEXT: vmovd %eax, %xmm0
|
|
; X86-AVX2-NEXT: vpsrlw $8, %xmm0, %xmm1
|
|
; X86-AVX2-NEXT: vpaddb %xmm1, %xmm0, %xmm0
|
|
; X86-AVX2-NEXT: vmovd %xmm0, %eax
|
|
; X86-AVX2-NEXT: # kill: def $al killed $al killed $eax
|
|
; X86-AVX2-NEXT: retl
|
|
;
|
|
; X64-AVX2-LABEL: test_v2i8_load:
|
|
; X64-AVX2: # %bb.0:
|
|
; X64-AVX2-NEXT: movzwl (%rdi), %eax
|
|
; X64-AVX2-NEXT: vmovd %eax, %xmm0
|
|
; X64-AVX2-NEXT: vpsrlw $8, %xmm0, %xmm1
|
|
; X64-AVX2-NEXT: vpaddb %xmm1, %xmm0, %xmm0
|
|
; X64-AVX2-NEXT: vmovd %xmm0, %eax
|
|
; X64-AVX2-NEXT: # kill: def $al killed $al killed $eax
|
|
; X64-AVX2-NEXT: retq
|
|
;
|
|
; AVX512-LABEL: test_v2i8_load:
|
|
; AVX512: # %bb.0:
|
|
; AVX512-NEXT: movzwl (%rdi), %eax
|
|
; AVX512-NEXT: vmovd %eax, %xmm0
|
|
; AVX512-NEXT: vpsrlw $8, %xmm0, %xmm1
|
|
; AVX512-NEXT: vpaddb %xmm1, %xmm0, %xmm0
|
|
; AVX512-NEXT: vmovd %xmm0, %eax
|
|
; AVX512-NEXT: # kill: def $al killed $al killed $eax
|
|
; AVX512-NEXT: retq
|
|
%a0 = load <2 x i8>, ptr %p
|
|
%1 = call i8 @llvm.vector.reduce.add.v2i8(<2 x i8> %a0)
|
|
ret i8 %1
|
|
}
|
|
|
|
define i8 @test_v4i8(<4 x i8> %a0) nounwind {
|
|
; SSE2-LABEL: test_v4i8:
|
|
; SSE2: # %bb.0:
|
|
; SSE2-NEXT: pxor %xmm1, %xmm1
|
|
; SSE2-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
|
|
; SSE2-NEXT: psadbw %xmm1, %xmm0
|
|
; SSE2-NEXT: movd %xmm0, %eax
|
|
; SSE2-NEXT: # kill: def $al killed $al killed $eax
|
|
; SSE2-NEXT: ret{{[l|q]}}
|
|
;
|
|
; SSE4-LABEL: test_v4i8:
|
|
; SSE4: # %bb.0:
|
|
; SSE4-NEXT: pxor %xmm1, %xmm1
|
|
; SSE4-NEXT: pblendw {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3,4,5,6,7]
|
|
; SSE4-NEXT: psadbw %xmm1, %xmm0
|
|
; SSE4-NEXT: movd %xmm0, %eax
|
|
; SSE4-NEXT: # kill: def $al killed $al killed $eax
|
|
; SSE4-NEXT: ret{{[l|q]}}
|
|
;
|
|
; AVX1-LABEL: test_v4i8:
|
|
; AVX1: # %bb.0:
|
|
; AVX1-NEXT: vpxor %xmm1, %xmm1, %xmm1
|
|
; AVX1-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3,4,5,6,7]
|
|
; AVX1-NEXT: vpsadbw %xmm1, %xmm0, %xmm0
|
|
; AVX1-NEXT: vmovd %xmm0, %eax
|
|
; AVX1-NEXT: # kill: def $al killed $al killed $eax
|
|
; AVX1-NEXT: ret{{[l|q]}}
|
|
;
|
|
; AVX2-LABEL: test_v4i8:
|
|
; AVX2: # %bb.0:
|
|
; AVX2-NEXT: vpxor %xmm1, %xmm1, %xmm1
|
|
; AVX2-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0],xmm1[1,2,3]
|
|
; AVX2-NEXT: vpsadbw %xmm1, %xmm0, %xmm0
|
|
; AVX2-NEXT: vmovd %xmm0, %eax
|
|
; AVX2-NEXT: # kill: def $al killed $al killed $eax
|
|
; AVX2-NEXT: ret{{[l|q]}}
|
|
;
|
|
; AVX512-LABEL: test_v4i8:
|
|
; AVX512: # %bb.0:
|
|
; AVX512-NEXT: vpxor %xmm1, %xmm1, %xmm1
|
|
; AVX512-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0],xmm1[1,2,3]
|
|
; AVX512-NEXT: vpxor %xmm1, %xmm1, %xmm1
|
|
; AVX512-NEXT: vpsadbw %xmm1, %xmm0, %xmm0
|
|
; AVX512-NEXT: vmovd %xmm0, %eax
|
|
; AVX512-NEXT: # kill: def $al killed $al killed $eax
|
|
; AVX512-NEXT: retq
|
|
%1 = call i8 @llvm.vector.reduce.add.v4i8(<4 x i8> %a0)
|
|
ret i8 %1
|
|
}
|
|
|
|
define i8 @test_v4i8_load(ptr %p) nounwind {
|
|
; X86-SSE-LABEL: test_v4i8_load:
|
|
; X86-SSE: # %bb.0:
|
|
; X86-SSE-NEXT: movl {{[0-9]+}}(%esp), %eax
|
|
; X86-SSE-NEXT: movd {{.*#+}} xmm0 = mem[0],zero,zero,zero
|
|
; X86-SSE-NEXT: pxor %xmm1, %xmm1
|
|
; X86-SSE-NEXT: psadbw %xmm0, %xmm1
|
|
; X86-SSE-NEXT: movd %xmm1, %eax
|
|
; X86-SSE-NEXT: # kill: def $al killed $al killed $eax
|
|
; X86-SSE-NEXT: retl
|
|
;
|
|
; X64-SSE-LABEL: test_v4i8_load:
|
|
; X64-SSE: # %bb.0:
|
|
; X64-SSE-NEXT: movd {{.*#+}} xmm0 = mem[0],zero,zero,zero
|
|
; X64-SSE-NEXT: pxor %xmm1, %xmm1
|
|
; X64-SSE-NEXT: psadbw %xmm0, %xmm1
|
|
; X64-SSE-NEXT: movd %xmm1, %eax
|
|
; X64-SSE-NEXT: # kill: def $al killed $al killed $eax
|
|
; X64-SSE-NEXT: retq
|
|
;
|
|
; X86-AVX1-LABEL: test_v4i8_load:
|
|
; X86-AVX1: # %bb.0:
|
|
; X86-AVX1-NEXT: movl {{[0-9]+}}(%esp), %eax
|
|
; X86-AVX1-NEXT: vmovd {{.*#+}} xmm0 = mem[0],zero,zero,zero
|
|
; X86-AVX1-NEXT: vpxor %xmm1, %xmm1, %xmm1
|
|
; X86-AVX1-NEXT: vpsadbw %xmm1, %xmm0, %xmm0
|
|
; X86-AVX1-NEXT: vmovd %xmm0, %eax
|
|
; X86-AVX1-NEXT: # kill: def $al killed $al killed $eax
|
|
; X86-AVX1-NEXT: retl
|
|
;
|
|
; X64-AVX1-LABEL: test_v4i8_load:
|
|
; X64-AVX1: # %bb.0:
|
|
; X64-AVX1-NEXT: vmovd {{.*#+}} xmm0 = mem[0],zero,zero,zero
|
|
; X64-AVX1-NEXT: vpxor %xmm1, %xmm1, %xmm1
|
|
; X64-AVX1-NEXT: vpsadbw %xmm1, %xmm0, %xmm0
|
|
; X64-AVX1-NEXT: vmovd %xmm0, %eax
|
|
; X64-AVX1-NEXT: # kill: def $al killed $al killed $eax
|
|
; X64-AVX1-NEXT: retq
|
|
;
|
|
; X86-AVX2-LABEL: test_v4i8_load:
|
|
; X86-AVX2: # %bb.0:
|
|
; X86-AVX2-NEXT: movl {{[0-9]+}}(%esp), %eax
|
|
; X86-AVX2-NEXT: vmovd {{.*#+}} xmm0 = mem[0],zero,zero,zero
|
|
; X86-AVX2-NEXT: vpxor %xmm1, %xmm1, %xmm1
|
|
; X86-AVX2-NEXT: vpsadbw %xmm1, %xmm0, %xmm0
|
|
; X86-AVX2-NEXT: vmovd %xmm0, %eax
|
|
; X86-AVX2-NEXT: # kill: def $al killed $al killed $eax
|
|
; X86-AVX2-NEXT: retl
|
|
;
|
|
; X64-AVX2-LABEL: test_v4i8_load:
|
|
; X64-AVX2: # %bb.0:
|
|
; X64-AVX2-NEXT: vmovd {{.*#+}} xmm0 = mem[0],zero,zero,zero
|
|
; X64-AVX2-NEXT: vpxor %xmm1, %xmm1, %xmm1
|
|
; X64-AVX2-NEXT: vpsadbw %xmm1, %xmm0, %xmm0
|
|
; X64-AVX2-NEXT: vmovd %xmm0, %eax
|
|
; X64-AVX2-NEXT: # kill: def $al killed $al killed $eax
|
|
; X64-AVX2-NEXT: retq
|
|
;
|
|
; AVX512-LABEL: test_v4i8_load:
|
|
; AVX512: # %bb.0:
|
|
; AVX512-NEXT: vmovd {{.*#+}} xmm0 = mem[0],zero,zero,zero
|
|
; AVX512-NEXT: vpxor %xmm1, %xmm1, %xmm1
|
|
; AVX512-NEXT: vpsadbw %xmm1, %xmm0, %xmm0
|
|
; AVX512-NEXT: vmovd %xmm0, %eax
|
|
; AVX512-NEXT: # kill: def $al killed $al killed $eax
|
|
; AVX512-NEXT: retq
|
|
%a0 = load <4 x i8>, ptr %p
|
|
%1 = call i8 @llvm.vector.reduce.add.v4i8(<4 x i8> %a0)
|
|
ret i8 %1
|
|
}
|
|
|
|
define i8 @test_v8i8(<8 x i8> %a0) nounwind {
|
|
; SSE-LABEL: test_v8i8:
|
|
; SSE: # %bb.0:
|
|
; SSE-NEXT: pxor %xmm1, %xmm1
|
|
; SSE-NEXT: psadbw %xmm0, %xmm1
|
|
; SSE-NEXT: movd %xmm1, %eax
|
|
; SSE-NEXT: # kill: def $al killed $al killed $eax
|
|
; SSE-NEXT: ret{{[l|q]}}
|
|
;
|
|
; AVX-LABEL: test_v8i8:
|
|
; AVX: # %bb.0:
|
|
; AVX-NEXT: vpxor %xmm1, %xmm1, %xmm1
|
|
; AVX-NEXT: vpsadbw %xmm1, %xmm0, %xmm0
|
|
; AVX-NEXT: vmovd %xmm0, %eax
|
|
; AVX-NEXT: # kill: def $al killed $al killed $eax
|
|
; AVX-NEXT: ret{{[l|q]}}
|
|
;
|
|
; AVX512-LABEL: test_v8i8:
|
|
; AVX512: # %bb.0:
|
|
; AVX512-NEXT: vpxor %xmm1, %xmm1, %xmm1
|
|
; AVX512-NEXT: vpsadbw %xmm1, %xmm0, %xmm0
|
|
; AVX512-NEXT: vmovd %xmm0, %eax
|
|
; AVX512-NEXT: # kill: def $al killed $al killed $eax
|
|
; AVX512-NEXT: retq
|
|
%1 = call i8 @llvm.vector.reduce.add.v8i8(<8 x i8> %a0)
|
|
ret i8 %1
|
|
}
|
|
|
|
define i8 @test_v8i8_load(ptr %p) nounwind {
|
|
; X86-SSE-LABEL: test_v8i8_load:
|
|
; X86-SSE: # %bb.0:
|
|
; X86-SSE-NEXT: movl {{[0-9]+}}(%esp), %eax
|
|
; X86-SSE-NEXT: movq {{.*#+}} xmm0 = mem[0],zero
|
|
; X86-SSE-NEXT: pxor %xmm1, %xmm1
|
|
; X86-SSE-NEXT: psadbw %xmm0, %xmm1
|
|
; X86-SSE-NEXT: movd %xmm1, %eax
|
|
; X86-SSE-NEXT: # kill: def $al killed $al killed $eax
|
|
; X86-SSE-NEXT: retl
|
|
;
|
|
; X64-SSE-LABEL: test_v8i8_load:
|
|
; X64-SSE: # %bb.0:
|
|
; X64-SSE-NEXT: movq {{.*#+}} xmm0 = mem[0],zero
|
|
; X64-SSE-NEXT: pxor %xmm1, %xmm1
|
|
; X64-SSE-NEXT: psadbw %xmm0, %xmm1
|
|
; X64-SSE-NEXT: movd %xmm1, %eax
|
|
; X64-SSE-NEXT: # kill: def $al killed $al killed $eax
|
|
; X64-SSE-NEXT: retq
|
|
;
|
|
; X86-AVX1-LABEL: test_v8i8_load:
|
|
; X86-AVX1: # %bb.0:
|
|
; X86-AVX1-NEXT: movl {{[0-9]+}}(%esp), %eax
|
|
; X86-AVX1-NEXT: vmovq {{.*#+}} xmm0 = mem[0],zero
|
|
; X86-AVX1-NEXT: vpxor %xmm1, %xmm1, %xmm1
|
|
; X86-AVX1-NEXT: vpsadbw %xmm1, %xmm0, %xmm0
|
|
; X86-AVX1-NEXT: vmovd %xmm0, %eax
|
|
; X86-AVX1-NEXT: # kill: def $al killed $al killed $eax
|
|
; X86-AVX1-NEXT: retl
|
|
;
|
|
; X64-AVX1-LABEL: test_v8i8_load:
|
|
; X64-AVX1: # %bb.0:
|
|
; X64-AVX1-NEXT: vmovq {{.*#+}} xmm0 = mem[0],zero
|
|
; X64-AVX1-NEXT: vpxor %xmm1, %xmm1, %xmm1
|
|
; X64-AVX1-NEXT: vpsadbw %xmm1, %xmm0, %xmm0
|
|
; X64-AVX1-NEXT: vmovd %xmm0, %eax
|
|
; X64-AVX1-NEXT: # kill: def $al killed $al killed $eax
|
|
; X64-AVX1-NEXT: retq
|
|
;
|
|
; X86-AVX2-LABEL: test_v8i8_load:
|
|
; X86-AVX2: # %bb.0:
|
|
; X86-AVX2-NEXT: movl {{[0-9]+}}(%esp), %eax
|
|
; X86-AVX2-NEXT: vmovq {{.*#+}} xmm0 = mem[0],zero
|
|
; X86-AVX2-NEXT: vpxor %xmm1, %xmm1, %xmm1
|
|
; X86-AVX2-NEXT: vpsadbw %xmm1, %xmm0, %xmm0
|
|
; X86-AVX2-NEXT: vmovd %xmm0, %eax
|
|
; X86-AVX2-NEXT: # kill: def $al killed $al killed $eax
|
|
; X86-AVX2-NEXT: retl
|
|
;
|
|
; X64-AVX2-LABEL: test_v8i8_load:
|
|
; X64-AVX2: # %bb.0:
|
|
; X64-AVX2-NEXT: vmovq {{.*#+}} xmm0 = mem[0],zero
|
|
; X64-AVX2-NEXT: vpxor %xmm1, %xmm1, %xmm1
|
|
; X64-AVX2-NEXT: vpsadbw %xmm1, %xmm0, %xmm0
|
|
; X64-AVX2-NEXT: vmovd %xmm0, %eax
|
|
; X64-AVX2-NEXT: # kill: def $al killed $al killed $eax
|
|
; X64-AVX2-NEXT: retq
|
|
;
|
|
; AVX512-LABEL: test_v8i8_load:
|
|
; AVX512: # %bb.0:
|
|
; AVX512-NEXT: vmovq {{.*#+}} xmm0 = mem[0],zero
|
|
; AVX512-NEXT: vpxor %xmm1, %xmm1, %xmm1
|
|
; AVX512-NEXT: vpsadbw %xmm1, %xmm0, %xmm0
|
|
; AVX512-NEXT: vmovd %xmm0, %eax
|
|
; AVX512-NEXT: # kill: def $al killed $al killed $eax
|
|
; AVX512-NEXT: retq
|
|
%a0 = load <8 x i8>, ptr %p
|
|
%1 = call i8 @llvm.vector.reduce.add.v8i8(<8 x i8> %a0)
|
|
ret i8 %1
|
|
}
|
|
|
|
define i8 @test_v16i8(<16 x i8> %a0) nounwind {
|
|
; SSE-LABEL: test_v16i8:
|
|
; SSE: # %bb.0:
|
|
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
|
|
; SSE-NEXT: paddb %xmm0, %xmm1
|
|
; SSE-NEXT: pxor %xmm0, %xmm0
|
|
; SSE-NEXT: psadbw %xmm1, %xmm0
|
|
; SSE-NEXT: movd %xmm0, %eax
|
|
; SSE-NEXT: # kill: def $al killed $al killed $eax
|
|
; SSE-NEXT: ret{{[l|q]}}
|
|
;
|
|
; AVX-LABEL: test_v16i8:
|
|
; AVX: # %bb.0:
|
|
; AVX-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
|
|
; AVX-NEXT: vpaddb %xmm1, %xmm0, %xmm0
|
|
; AVX-NEXT: vpxor %xmm1, %xmm1, %xmm1
|
|
; AVX-NEXT: vpsadbw %xmm1, %xmm0, %xmm0
|
|
; AVX-NEXT: vmovd %xmm0, %eax
|
|
; AVX-NEXT: # kill: def $al killed $al killed $eax
|
|
; AVX-NEXT: ret{{[l|q]}}
|
|
;
|
|
; AVX512-LABEL: test_v16i8:
|
|
; AVX512: # %bb.0:
|
|
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
|
|
; AVX512-NEXT: vpaddb %xmm1, %xmm0, %xmm0
|
|
; AVX512-NEXT: vpxor %xmm1, %xmm1, %xmm1
|
|
; AVX512-NEXT: vpsadbw %xmm1, %xmm0, %xmm0
|
|
; AVX512-NEXT: vmovd %xmm0, %eax
|
|
; AVX512-NEXT: # kill: def $al killed $al killed $eax
|
|
; AVX512-NEXT: retq
|
|
%1 = call i8 @llvm.vector.reduce.add.v16i8(<16 x i8> %a0)
|
|
ret i8 %1
|
|
}
|
|
|
|
define i8 @test_v32i8(<32 x i8> %a0) nounwind {
|
|
; SSE-LABEL: test_v32i8:
|
|
; SSE: # %bb.0:
|
|
; SSE-NEXT: paddb %xmm1, %xmm0
|
|
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
|
|
; SSE-NEXT: paddb %xmm0, %xmm1
|
|
; SSE-NEXT: pxor %xmm0, %xmm0
|
|
; SSE-NEXT: psadbw %xmm1, %xmm0
|
|
; SSE-NEXT: movd %xmm0, %eax
|
|
; SSE-NEXT: # kill: def $al killed $al killed $eax
|
|
; SSE-NEXT: ret{{[l|q]}}
|
|
;
|
|
; AVX1-LABEL: test_v32i8:
|
|
; AVX1: # %bb.0:
|
|
; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm1
|
|
; AVX1-NEXT: vpaddb %xmm1, %xmm0, %xmm0
|
|
; AVX1-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
|
|
; AVX1-NEXT: vpaddb %xmm1, %xmm0, %xmm0
|
|
; AVX1-NEXT: vpxor %xmm1, %xmm1, %xmm1
|
|
; AVX1-NEXT: vpsadbw %xmm1, %xmm0, %xmm0
|
|
; AVX1-NEXT: vmovd %xmm0, %eax
|
|
; AVX1-NEXT: # kill: def $al killed $al killed $eax
|
|
; AVX1-NEXT: vzeroupper
|
|
; AVX1-NEXT: ret{{[l|q]}}
|
|
;
|
|
; AVX2-LABEL: test_v32i8:
|
|
; AVX2: # %bb.0:
|
|
; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1
|
|
; AVX2-NEXT: vpaddb %xmm1, %xmm0, %xmm0
|
|
; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
|
|
; AVX2-NEXT: vpaddb %xmm1, %xmm0, %xmm0
|
|
; AVX2-NEXT: vpxor %xmm1, %xmm1, %xmm1
|
|
; AVX2-NEXT: vpsadbw %xmm1, %xmm0, %xmm0
|
|
; AVX2-NEXT: vmovd %xmm0, %eax
|
|
; AVX2-NEXT: # kill: def $al killed $al killed $eax
|
|
; AVX2-NEXT: vzeroupper
|
|
; AVX2-NEXT: ret{{[l|q]}}
|
|
;
|
|
; AVX512-LABEL: test_v32i8:
|
|
; AVX512: # %bb.0:
|
|
; AVX512-NEXT: vextracti128 $1, %ymm0, %xmm1
|
|
; AVX512-NEXT: vpaddb %xmm1, %xmm0, %xmm0
|
|
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
|
|
; AVX512-NEXT: vpaddb %xmm1, %xmm0, %xmm0
|
|
; AVX512-NEXT: vpxor %xmm1, %xmm1, %xmm1
|
|
; AVX512-NEXT: vpsadbw %xmm1, %xmm0, %xmm0
|
|
; AVX512-NEXT: vmovd %xmm0, %eax
|
|
; AVX512-NEXT: # kill: def $al killed $al killed $eax
|
|
; AVX512-NEXT: vzeroupper
|
|
; AVX512-NEXT: retq
|
|
%1 = call i8 @llvm.vector.reduce.add.v32i8(<32 x i8> %a0)
|
|
ret i8 %1
|
|
}
|
|
|
|
define i8 @test_v64i8(<64 x i8> %a0) nounwind {
|
|
; X86-SSE-LABEL: test_v64i8:
|
|
; X86-SSE: # %bb.0:
|
|
; X86-SSE-NEXT: pushl %ebp
|
|
; X86-SSE-NEXT: movl %esp, %ebp
|
|
; X86-SSE-NEXT: andl $-16, %esp
|
|
; X86-SSE-NEXT: subl $16, %esp
|
|
; X86-SSE-NEXT: paddb %xmm2, %xmm0
|
|
; X86-SSE-NEXT: paddb 8(%ebp), %xmm1
|
|
; X86-SSE-NEXT: paddb %xmm0, %xmm1
|
|
; X86-SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm1[2,3,2,3]
|
|
; X86-SSE-NEXT: paddb %xmm1, %xmm0
|
|
; X86-SSE-NEXT: pxor %xmm1, %xmm1
|
|
; X86-SSE-NEXT: psadbw %xmm0, %xmm1
|
|
; X86-SSE-NEXT: movd %xmm1, %eax
|
|
; X86-SSE-NEXT: # kill: def $al killed $al killed $eax
|
|
; X86-SSE-NEXT: movl %ebp, %esp
|
|
; X86-SSE-NEXT: popl %ebp
|
|
; X86-SSE-NEXT: retl
|
|
;
|
|
; X64-SSE-LABEL: test_v64i8:
|
|
; X64-SSE: # %bb.0:
|
|
; X64-SSE-NEXT: paddb %xmm3, %xmm1
|
|
; X64-SSE-NEXT: paddb %xmm2, %xmm0
|
|
; X64-SSE-NEXT: paddb %xmm1, %xmm0
|
|
; X64-SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
|
|
; X64-SSE-NEXT: paddb %xmm0, %xmm1
|
|
; X64-SSE-NEXT: pxor %xmm0, %xmm0
|
|
; X64-SSE-NEXT: psadbw %xmm1, %xmm0
|
|
; X64-SSE-NEXT: movd %xmm0, %eax
|
|
; X64-SSE-NEXT: # kill: def $al killed $al killed $eax
|
|
; X64-SSE-NEXT: retq
|
|
;
|
|
; AVX1-LABEL: test_v64i8:
|
|
; AVX1: # %bb.0:
|
|
; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm2
|
|
; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm3
|
|
; AVX1-NEXT: vpaddb %xmm2, %xmm3, %xmm2
|
|
; AVX1-NEXT: vpaddb %xmm1, %xmm0, %xmm0
|
|
; AVX1-NEXT: vpaddb %xmm2, %xmm0, %xmm0
|
|
; AVX1-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
|
|
; AVX1-NEXT: vpaddb %xmm1, %xmm0, %xmm0
|
|
; AVX1-NEXT: vpxor %xmm1, %xmm1, %xmm1
|
|
; AVX1-NEXT: vpsadbw %xmm1, %xmm0, %xmm0
|
|
; AVX1-NEXT: vmovd %xmm0, %eax
|
|
; AVX1-NEXT: # kill: def $al killed $al killed $eax
|
|
; AVX1-NEXT: vzeroupper
|
|
; AVX1-NEXT: ret{{[l|q]}}
|
|
;
|
|
; AVX2-LABEL: test_v64i8:
|
|
; AVX2: # %bb.0:
|
|
; AVX2-NEXT: vpaddb %ymm1, %ymm0, %ymm0
|
|
; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1
|
|
; AVX2-NEXT: vpaddb %xmm1, %xmm0, %xmm0
|
|
; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
|
|
; AVX2-NEXT: vpaddb %xmm1, %xmm0, %xmm0
|
|
; AVX2-NEXT: vpxor %xmm1, %xmm1, %xmm1
|
|
; AVX2-NEXT: vpsadbw %xmm1, %xmm0, %xmm0
|
|
; AVX2-NEXT: vmovd %xmm0, %eax
|
|
; AVX2-NEXT: # kill: def $al killed $al killed $eax
|
|
; AVX2-NEXT: vzeroupper
|
|
; AVX2-NEXT: ret{{[l|q]}}
|
|
;
|
|
; AVX512-LABEL: test_v64i8:
|
|
; AVX512: # %bb.0:
|
|
; AVX512-NEXT: vextracti64x4 $1, %zmm0, %ymm1
|
|
; AVX512-NEXT: vpaddb %ymm1, %ymm0, %ymm0
|
|
; AVX512-NEXT: vextracti128 $1, %ymm0, %xmm1
|
|
; AVX512-NEXT: vpaddb %xmm1, %xmm0, %xmm0
|
|
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
|
|
; AVX512-NEXT: vpaddb %xmm1, %xmm0, %xmm0
|
|
; AVX512-NEXT: vpxor %xmm1, %xmm1, %xmm1
|
|
; AVX512-NEXT: vpsadbw %xmm1, %xmm0, %xmm0
|
|
; AVX512-NEXT: vmovd %xmm0, %eax
|
|
; AVX512-NEXT: # kill: def $al killed $al killed $eax
|
|
; AVX512-NEXT: vzeroupper
|
|
; AVX512-NEXT: retq
|
|
%1 = call i8 @llvm.vector.reduce.add.v64i8(<64 x i8> %a0)
|
|
ret i8 %1
|
|
}
|
|
|
|
define i8 @test_v128i8(<128 x i8> %a0) nounwind {
|
|
; X86-SSE-LABEL: test_v128i8:
|
|
; X86-SSE: # %bb.0:
|
|
; X86-SSE-NEXT: pushl %ebp
|
|
; X86-SSE-NEXT: movl %esp, %ebp
|
|
; X86-SSE-NEXT: andl $-16, %esp
|
|
; X86-SSE-NEXT: subl $16, %esp
|
|
; X86-SSE-NEXT: movdqa 8(%ebp), %xmm3
|
|
; X86-SSE-NEXT: paddb 72(%ebp), %xmm3
|
|
; X86-SSE-NEXT: paddb 40(%ebp), %xmm1
|
|
; X86-SSE-NEXT: paddb %xmm3, %xmm1
|
|
; X86-SSE-NEXT: paddb 56(%ebp), %xmm2
|
|
; X86-SSE-NEXT: paddb 24(%ebp), %xmm0
|
|
; X86-SSE-NEXT: paddb %xmm2, %xmm0
|
|
; X86-SSE-NEXT: paddb %xmm1, %xmm0
|
|
; X86-SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
|
|
; X86-SSE-NEXT: paddb %xmm0, %xmm1
|
|
; X86-SSE-NEXT: pxor %xmm0, %xmm0
|
|
; X86-SSE-NEXT: psadbw %xmm1, %xmm0
|
|
; X86-SSE-NEXT: movd %xmm0, %eax
|
|
; X86-SSE-NEXT: # kill: def $al killed $al killed $eax
|
|
; X86-SSE-NEXT: movl %ebp, %esp
|
|
; X86-SSE-NEXT: popl %ebp
|
|
; X86-SSE-NEXT: retl
|
|
;
|
|
; X64-SSE-LABEL: test_v128i8:
|
|
; X64-SSE: # %bb.0:
|
|
; X64-SSE-NEXT: paddb %xmm7, %xmm3
|
|
; X64-SSE-NEXT: paddb %xmm5, %xmm1
|
|
; X64-SSE-NEXT: paddb %xmm3, %xmm1
|
|
; X64-SSE-NEXT: paddb %xmm6, %xmm2
|
|
; X64-SSE-NEXT: paddb %xmm4, %xmm0
|
|
; X64-SSE-NEXT: paddb %xmm2, %xmm0
|
|
; X64-SSE-NEXT: paddb %xmm1, %xmm0
|
|
; X64-SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
|
|
; X64-SSE-NEXT: paddb %xmm0, %xmm1
|
|
; X64-SSE-NEXT: pxor %xmm0, %xmm0
|
|
; X64-SSE-NEXT: psadbw %xmm1, %xmm0
|
|
; X64-SSE-NEXT: movd %xmm0, %eax
|
|
; X64-SSE-NEXT: # kill: def $al killed $al killed $eax
|
|
; X64-SSE-NEXT: retq
|
|
;
|
|
; X86-AVX1-LABEL: test_v128i8:
|
|
; X86-AVX1: # %bb.0:
|
|
; X86-AVX1-NEXT: pushl %ebp
|
|
; X86-AVX1-NEXT: movl %esp, %ebp
|
|
; X86-AVX1-NEXT: andl $-32, %esp
|
|
; X86-AVX1-NEXT: subl $32, %esp
|
|
; X86-AVX1-NEXT: vextractf128 $1, %ymm2, %xmm3
|
|
; X86-AVX1-NEXT: vextractf128 $1, %ymm0, %xmm4
|
|
; X86-AVX1-NEXT: vpaddb %xmm3, %xmm4, %xmm3
|
|
; X86-AVX1-NEXT: vextractf128 $1, %ymm1, %xmm4
|
|
; X86-AVX1-NEXT: vpaddb 24(%ebp), %xmm4, %xmm4
|
|
; X86-AVX1-NEXT: vpaddb %xmm4, %xmm3, %xmm3
|
|
; X86-AVX1-NEXT: vpaddb %xmm2, %xmm0, %xmm0
|
|
; X86-AVX1-NEXT: vpaddb 8(%ebp), %xmm1, %xmm1
|
|
; X86-AVX1-NEXT: vpaddb %xmm1, %xmm0, %xmm0
|
|
; X86-AVX1-NEXT: vpaddb %xmm3, %xmm0, %xmm0
|
|
; X86-AVX1-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
|
|
; X86-AVX1-NEXT: vpaddb %xmm1, %xmm0, %xmm0
|
|
; X86-AVX1-NEXT: vpxor %xmm1, %xmm1, %xmm1
|
|
; X86-AVX1-NEXT: vpsadbw %xmm1, %xmm0, %xmm0
|
|
; X86-AVX1-NEXT: vmovd %xmm0, %eax
|
|
; X86-AVX1-NEXT: # kill: def $al killed $al killed $eax
|
|
; X86-AVX1-NEXT: movl %ebp, %esp
|
|
; X86-AVX1-NEXT: popl %ebp
|
|
; X86-AVX1-NEXT: vzeroupper
|
|
; X86-AVX1-NEXT: retl
|
|
;
|
|
; X64-AVX1-LABEL: test_v128i8:
|
|
; X64-AVX1: # %bb.0:
|
|
; X64-AVX1-NEXT: vextractf128 $1, %ymm3, %xmm4
|
|
; X64-AVX1-NEXT: vextractf128 $1, %ymm1, %xmm5
|
|
; X64-AVX1-NEXT: vpaddb %xmm4, %xmm5, %xmm4
|
|
; X64-AVX1-NEXT: vextractf128 $1, %ymm2, %xmm5
|
|
; X64-AVX1-NEXT: vextractf128 $1, %ymm0, %xmm6
|
|
; X64-AVX1-NEXT: vpaddb %xmm5, %xmm6, %xmm5
|
|
; X64-AVX1-NEXT: vpaddb %xmm4, %xmm5, %xmm4
|
|
; X64-AVX1-NEXT: vpaddb %xmm3, %xmm1, %xmm1
|
|
; X64-AVX1-NEXT: vpaddb %xmm2, %xmm0, %xmm0
|
|
; X64-AVX1-NEXT: vpaddb %xmm1, %xmm0, %xmm0
|
|
; X64-AVX1-NEXT: vpaddb %xmm4, %xmm0, %xmm0
|
|
; X64-AVX1-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
|
|
; X64-AVX1-NEXT: vpaddb %xmm1, %xmm0, %xmm0
|
|
; X64-AVX1-NEXT: vpxor %xmm1, %xmm1, %xmm1
|
|
; X64-AVX1-NEXT: vpsadbw %xmm1, %xmm0, %xmm0
|
|
; X64-AVX1-NEXT: vmovd %xmm0, %eax
|
|
; X64-AVX1-NEXT: # kill: def $al killed $al killed $eax
|
|
; X64-AVX1-NEXT: vzeroupper
|
|
; X64-AVX1-NEXT: retq
|
|
;
|
|
; X86-AVX2-LABEL: test_v128i8:
|
|
; X86-AVX2: # %bb.0:
|
|
; X86-AVX2-NEXT: pushl %ebp
|
|
; X86-AVX2-NEXT: movl %esp, %ebp
|
|
; X86-AVX2-NEXT: andl $-32, %esp
|
|
; X86-AVX2-NEXT: subl $32, %esp
|
|
; X86-AVX2-NEXT: vpaddb %ymm2, %ymm0, %ymm0
|
|
; X86-AVX2-NEXT: vpaddb 8(%ebp), %ymm1, %ymm1
|
|
; X86-AVX2-NEXT: vpaddb %ymm1, %ymm0, %ymm0
|
|
; X86-AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1
|
|
; X86-AVX2-NEXT: vpaddb %xmm1, %xmm0, %xmm0
|
|
; X86-AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
|
|
; X86-AVX2-NEXT: vpaddb %xmm1, %xmm0, %xmm0
|
|
; X86-AVX2-NEXT: vpxor %xmm1, %xmm1, %xmm1
|
|
; X86-AVX2-NEXT: vpsadbw %xmm1, %xmm0, %xmm0
|
|
; X86-AVX2-NEXT: vmovd %xmm0, %eax
|
|
; X86-AVX2-NEXT: # kill: def $al killed $al killed $eax
|
|
; X86-AVX2-NEXT: movl %ebp, %esp
|
|
; X86-AVX2-NEXT: popl %ebp
|
|
; X86-AVX2-NEXT: vzeroupper
|
|
; X86-AVX2-NEXT: retl
|
|
;
|
|
; X64-AVX2-LABEL: test_v128i8:
|
|
; X64-AVX2: # %bb.0:
|
|
; X64-AVX2-NEXT: vpaddb %ymm3, %ymm1, %ymm1
|
|
; X64-AVX2-NEXT: vpaddb %ymm2, %ymm0, %ymm0
|
|
; X64-AVX2-NEXT: vpaddb %ymm1, %ymm0, %ymm0
|
|
; X64-AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1
|
|
; X64-AVX2-NEXT: vpaddb %xmm1, %xmm0, %xmm0
|
|
; X64-AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
|
|
; X64-AVX2-NEXT: vpaddb %xmm1, %xmm0, %xmm0
|
|
; X64-AVX2-NEXT: vpxor %xmm1, %xmm1, %xmm1
|
|
; X64-AVX2-NEXT: vpsadbw %xmm1, %xmm0, %xmm0
|
|
; X64-AVX2-NEXT: vmovd %xmm0, %eax
|
|
; X64-AVX2-NEXT: # kill: def $al killed $al killed $eax
|
|
; X64-AVX2-NEXT: vzeroupper
|
|
; X64-AVX2-NEXT: retq
|
|
;
|
|
; AVX512-LABEL: test_v128i8:
|
|
; AVX512: # %bb.0:
|
|
; AVX512-NEXT: vpaddb %zmm1, %zmm0, %zmm0
|
|
; AVX512-NEXT: vextracti64x4 $1, %zmm0, %ymm1
|
|
; AVX512-NEXT: vpaddb %ymm1, %ymm0, %ymm0
|
|
; AVX512-NEXT: vextracti128 $1, %ymm0, %xmm1
|
|
; AVX512-NEXT: vpaddb %xmm1, %xmm0, %xmm0
|
|
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
|
|
; AVX512-NEXT: vpaddb %xmm1, %xmm0, %xmm0
|
|
; AVX512-NEXT: vpxor %xmm1, %xmm1, %xmm1
|
|
; AVX512-NEXT: vpsadbw %xmm1, %xmm0, %xmm0
|
|
; AVX512-NEXT: vmovd %xmm0, %eax
|
|
; AVX512-NEXT: # kill: def $al killed $al killed $eax
|
|
; AVX512-NEXT: vzeroupper
|
|
; AVX512-NEXT: retq
|
|
%1 = call i8 @llvm.vector.reduce.add.v128i8(<128 x i8> %a0)
|
|
ret i8 %1
|
|
}
|
|
|
|
declare i64 @llvm.vector.reduce.add.v2i64(<2 x i64>)
|
|
declare i64 @llvm.vector.reduce.add.v4i64(<4 x i64>)
|
|
declare i64 @llvm.vector.reduce.add.v8i64(<8 x i64>)
|
|
declare i64 @llvm.vector.reduce.add.v16i64(<16 x i64>)
|
|
|
|
declare i32 @llvm.vector.reduce.add.v2i32(<2 x i32>)
|
|
declare i32 @llvm.vector.reduce.add.v4i32(<4 x i32>)
|
|
declare i32 @llvm.vector.reduce.add.v8i32(<8 x i32>)
|
|
declare i32 @llvm.vector.reduce.add.v16i32(<16 x i32>)
|
|
declare i32 @llvm.vector.reduce.add.v32i32(<32 x i32>)
|
|
|
|
declare i16 @llvm.vector.reduce.add.v2i16(<2 x i16>)
|
|
declare i16 @llvm.vector.reduce.add.v4i16(<4 x i16>)
|
|
declare i16 @llvm.vector.reduce.add.v8i16(<8 x i16>)
|
|
declare i16 @llvm.vector.reduce.add.v16i16(<16 x i16>)
|
|
declare i16 @llvm.vector.reduce.add.v32i16(<32 x i16>)
|
|
declare i16 @llvm.vector.reduce.add.v64i16(<64 x i16>)
|
|
|
|
declare i8 @llvm.vector.reduce.add.v2i8(<2 x i8>)
|
|
declare i8 @llvm.vector.reduce.add.v4i8(<4 x i8>)
|
|
declare i8 @llvm.vector.reduce.add.v8i8(<8 x i8>)
|
|
declare i8 @llvm.vector.reduce.add.v16i8(<16 x i8>)
|
|
declare i8 @llvm.vector.reduce.add.v32i8(<32 x i8>)
|
|
declare i8 @llvm.vector.reduce.add.v64i8(<64 x i8>)
|
|
declare i8 @llvm.vector.reduce.add.v128i8(<128 x i8>)
|
|
;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
|
|
; X64-SSE2: {{.*}}
|
|
; X64-SSE4: {{.*}}
|