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
343 lines
11 KiB
LLVM
343 lines
11 KiB
LLVM
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
|
|
; RUN: llc < %s -mtriple=x86_64-unknown -mcpu=x86-64 | FileCheck %s --check-prefixes=SSE
|
|
; RUN: llc < %s -mtriple=x86_64-unknown -mcpu=x86-64-v2 | FileCheck %s --check-prefixes=SSE
|
|
; RUN: llc < %s -mtriple=x86_64-unknown -mcpu=x86-64-v3 | FileCheck %s --check-prefixes=AVX,AVX2
|
|
; RUN: llc < %s -mtriple=x86_64-unknown -mcpu=x86-64-v4 | FileCheck %s --check-prefixes=AVX,AVX512,AVX512-VL
|
|
; RUN: llc < %s -mtriple=x86_64-unknown -mcpu=x86-64-v4 -mattr=-avx512vl | FileCheck %s --check-prefixes=AVX,AVX512,AVX512-NOVL
|
|
; RUN: llc < %s -mtriple=x86_64-unknown -mcpu=knl | FileCheck %s --check-prefixes=AVX,AVX512,AVX512F
|
|
|
|
;
|
|
; fptosi -> sitofp
|
|
;
|
|
|
|
define double @scvtf64_i32(double %a0) {
|
|
; SSE-LABEL: scvtf64_i32:
|
|
; SSE: # %bb.0:
|
|
; SSE-NEXT: cvttpd2dq %xmm0, %xmm0
|
|
; SSE-NEXT: cvtdq2pd %xmm0, %xmm0
|
|
; SSE-NEXT: retq
|
|
;
|
|
; AVX-LABEL: scvtf64_i32:
|
|
; AVX: # %bb.0:
|
|
; AVX-NEXT: vcvttpd2dq %xmm0, %xmm0
|
|
; AVX-NEXT: vcvtdq2pd %xmm0, %xmm0
|
|
; AVX-NEXT: retq
|
|
%ii = fptosi double %a0 to i32
|
|
%ff = sitofp i32 %ii to double
|
|
ret double %ff
|
|
}
|
|
|
|
define double @scvtf64_i64(double %a0) {
|
|
; SSE-LABEL: scvtf64_i64:
|
|
; SSE: # %bb.0:
|
|
; SSE-NEXT: cvttsd2si %xmm0, %rax
|
|
; SSE-NEXT: xorps %xmm0, %xmm0
|
|
; SSE-NEXT: cvtsi2sd %rax, %xmm0
|
|
; SSE-NEXT: retq
|
|
;
|
|
; AVX2-LABEL: scvtf64_i64:
|
|
; AVX2: # %bb.0:
|
|
; AVX2-NEXT: vcvttsd2si %xmm0, %rax
|
|
; AVX2-NEXT: vcvtsi2sd %rax, %xmm15, %xmm0
|
|
; AVX2-NEXT: retq
|
|
;
|
|
; AVX512-VL-LABEL: scvtf64_i64:
|
|
; AVX512-VL: # %bb.0:
|
|
; AVX512-VL-NEXT: vcvttpd2qq %xmm0, %xmm0
|
|
; AVX512-VL-NEXT: vcvtqq2pd %xmm0, %xmm0
|
|
; AVX512-VL-NEXT: retq
|
|
;
|
|
; AVX512-NOVL-LABEL: scvtf64_i64:
|
|
; AVX512-NOVL: # %bb.0:
|
|
; AVX512-NOVL-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0
|
|
; AVX512-NOVL-NEXT: vcvttpd2qq %zmm0, %zmm0
|
|
; AVX512-NOVL-NEXT: vcvtqq2pd %zmm0, %zmm0
|
|
; AVX512-NOVL-NEXT: # kill: def $xmm0 killed $xmm0 killed $zmm0
|
|
; AVX512-NOVL-NEXT: vzeroupper
|
|
; AVX512-NOVL-NEXT: retq
|
|
;
|
|
; AVX512F-LABEL: scvtf64_i64:
|
|
; AVX512F: # %bb.0:
|
|
; AVX512F-NEXT: vcvttsd2si %xmm0, %rax
|
|
; AVX512F-NEXT: vcvtsi2sd %rax, %xmm15, %xmm0
|
|
; AVX512F-NEXT: retq
|
|
%ii = fptosi double %a0 to i64
|
|
%ff = sitofp i64 %ii to double
|
|
ret double %ff
|
|
}
|
|
|
|
define float @scvtf32_i32(float %a0) {
|
|
; SSE-LABEL: scvtf32_i32:
|
|
; SSE: # %bb.0:
|
|
; SSE-NEXT: cvttps2dq %xmm0, %xmm0
|
|
; SSE-NEXT: cvtdq2ps %xmm0, %xmm0
|
|
; SSE-NEXT: retq
|
|
;
|
|
; AVX-LABEL: scvtf32_i32:
|
|
; AVX: # %bb.0:
|
|
; AVX-NEXT: vcvttps2dq %xmm0, %xmm0
|
|
; AVX-NEXT: vcvtdq2ps %xmm0, %xmm0
|
|
; AVX-NEXT: retq
|
|
%ii = fptosi float %a0 to i32
|
|
%ff = sitofp i32 %ii to float
|
|
ret float %ff
|
|
}
|
|
|
|
define float @scvtf32_i64(float %a0) {
|
|
; SSE-LABEL: scvtf32_i64:
|
|
; SSE: # %bb.0:
|
|
; SSE-NEXT: cvttss2si %xmm0, %rax
|
|
; SSE-NEXT: xorps %xmm0, %xmm0
|
|
; SSE-NEXT: cvtsi2ss %rax, %xmm0
|
|
; SSE-NEXT: retq
|
|
;
|
|
; AVX2-LABEL: scvtf32_i64:
|
|
; AVX2: # %bb.0:
|
|
; AVX2-NEXT: vcvttss2si %xmm0, %rax
|
|
; AVX2-NEXT: vcvtsi2ss %rax, %xmm15, %xmm0
|
|
; AVX2-NEXT: retq
|
|
;
|
|
; AVX512-VL-LABEL: scvtf32_i64:
|
|
; AVX512-VL: # %bb.0:
|
|
; AVX512-VL-NEXT: vcvttps2qq %xmm0, %xmm0
|
|
; AVX512-VL-NEXT: vcvtqq2ps %xmm0, %xmm0
|
|
; AVX512-VL-NEXT: retq
|
|
;
|
|
; AVX512-NOVL-LABEL: scvtf32_i64:
|
|
; AVX512-NOVL: # %bb.0:
|
|
; AVX512-NOVL-NEXT: # kill: def $xmm0 killed $xmm0 def $ymm0
|
|
; AVX512-NOVL-NEXT: vcvttps2qq %ymm0, %zmm0
|
|
; AVX512-NOVL-NEXT: vcvtqq2ps %zmm0, %ymm0
|
|
; AVX512-NOVL-NEXT: # kill: def $xmm0 killed $xmm0 killed $ymm0
|
|
; AVX512-NOVL-NEXT: vzeroupper
|
|
; AVX512-NOVL-NEXT: retq
|
|
;
|
|
; AVX512F-LABEL: scvtf32_i64:
|
|
; AVX512F: # %bb.0:
|
|
; AVX512F-NEXT: vcvttss2si %xmm0, %rax
|
|
; AVX512F-NEXT: vcvtsi2ss %rax, %xmm15, %xmm0
|
|
; AVX512F-NEXT: retq
|
|
%ii = fptosi float %a0 to i64
|
|
%ff = sitofp i64 %ii to float
|
|
ret float %ff
|
|
}
|
|
|
|
;
|
|
; fptoui -> uitofp
|
|
;
|
|
|
|
define double @ucvtf64_i32(double %a0) {
|
|
; SSE-LABEL: ucvtf64_i32:
|
|
; SSE: # %bb.0:
|
|
; SSE-NEXT: cvttsd2si %xmm0, %rax
|
|
; SSE-NEXT: movl %eax, %eax
|
|
; SSE-NEXT: xorps %xmm0, %xmm0
|
|
; SSE-NEXT: cvtsi2sd %rax, %xmm0
|
|
; SSE-NEXT: retq
|
|
;
|
|
; AVX2-LABEL: ucvtf64_i32:
|
|
; AVX2: # %bb.0:
|
|
; AVX2-NEXT: vcvttsd2si %xmm0, %rax
|
|
; AVX2-NEXT: movl %eax, %eax
|
|
; AVX2-NEXT: vcvtsi2sd %rax, %xmm15, %xmm0
|
|
; AVX2-NEXT: retq
|
|
;
|
|
; AVX512-VL-LABEL: ucvtf64_i32:
|
|
; AVX512-VL: # %bb.0:
|
|
; AVX512-VL-NEXT: vcvttpd2udq %xmm0, %xmm0
|
|
; AVX512-VL-NEXT: vcvtudq2pd %xmm0, %xmm0
|
|
; AVX512-VL-NEXT: retq
|
|
;
|
|
; AVX512-NOVL-LABEL: ucvtf64_i32:
|
|
; AVX512-NOVL: # %bb.0:
|
|
; AVX512-NOVL-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0
|
|
; AVX512-NOVL-NEXT: vcvttpd2udq %zmm0, %ymm0
|
|
; AVX512-NOVL-NEXT: vcvtudq2pd %ymm0, %zmm0
|
|
; AVX512-NOVL-NEXT: # kill: def $xmm0 killed $xmm0 killed $zmm0
|
|
; AVX512-NOVL-NEXT: vzeroupper
|
|
; AVX512-NOVL-NEXT: retq
|
|
;
|
|
; AVX512F-LABEL: ucvtf64_i32:
|
|
; AVX512F: # %bb.0:
|
|
; AVX512F-NEXT: vcvttsd2usi %xmm0, %eax
|
|
; AVX512F-NEXT: vcvtusi2sd %eax, %xmm15, %xmm0
|
|
; AVX512F-NEXT: retq
|
|
%ii = fptoui double %a0 to i32
|
|
%ff = uitofp i32 %ii to double
|
|
ret double %ff
|
|
}
|
|
|
|
define double @ucvtf64_i64(double %a0) {
|
|
; SSE-LABEL: ucvtf64_i64:
|
|
; SSE: # %bb.0:
|
|
; SSE-NEXT: cvttsd2si %xmm0, %rax
|
|
; SSE-NEXT: movq %rax, %rcx
|
|
; SSE-NEXT: subsd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
|
|
; SSE-NEXT: cvttsd2si %xmm0, %rdx
|
|
; SSE-NEXT: sarq $63, %rcx
|
|
; SSE-NEXT: andq %rcx, %rdx
|
|
; SSE-NEXT: orq %rax, %rdx
|
|
; SSE-NEXT: movq %rdx, %xmm1
|
|
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],mem[0],xmm1[1],mem[1]
|
|
; SSE-NEXT: subpd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1
|
|
; SSE-NEXT: movapd %xmm1, %xmm0
|
|
; SSE-NEXT: unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
|
|
; SSE-NEXT: addsd %xmm1, %xmm0
|
|
; SSE-NEXT: retq
|
|
;
|
|
; AVX2-LABEL: ucvtf64_i64:
|
|
; AVX2: # %bb.0:
|
|
; AVX2-NEXT: vcvttsd2si %xmm0, %rax
|
|
; AVX2-NEXT: movq %rax, %rcx
|
|
; AVX2-NEXT: vsubsd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
|
|
; AVX2-NEXT: sarq $63, %rcx
|
|
; AVX2-NEXT: vcvttsd2si %xmm0, %rdx
|
|
; AVX2-NEXT: andq %rcx, %rdx
|
|
; AVX2-NEXT: orq %rax, %rdx
|
|
; AVX2-NEXT: vmovq %rdx, %xmm0
|
|
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
|
|
; AVX2-NEXT: vsubpd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
|
|
; AVX2-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
|
|
; AVX2-NEXT: vaddsd %xmm0, %xmm1, %xmm0
|
|
; AVX2-NEXT: retq
|
|
;
|
|
; AVX512-VL-LABEL: ucvtf64_i64:
|
|
; AVX512-VL: # %bb.0:
|
|
; AVX512-VL-NEXT: vcvttpd2uqq %xmm0, %xmm0
|
|
; AVX512-VL-NEXT: vcvtuqq2pd %xmm0, %xmm0
|
|
; AVX512-VL-NEXT: retq
|
|
;
|
|
; AVX512-NOVL-LABEL: ucvtf64_i64:
|
|
; AVX512-NOVL: # %bb.0:
|
|
; AVX512-NOVL-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0
|
|
; AVX512-NOVL-NEXT: vcvttpd2uqq %zmm0, %zmm0
|
|
; AVX512-NOVL-NEXT: vcvtuqq2pd %zmm0, %zmm0
|
|
; AVX512-NOVL-NEXT: # kill: def $xmm0 killed $xmm0 killed $zmm0
|
|
; AVX512-NOVL-NEXT: vzeroupper
|
|
; AVX512-NOVL-NEXT: retq
|
|
;
|
|
; AVX512F-LABEL: ucvtf64_i64:
|
|
; AVX512F: # %bb.0:
|
|
; AVX512F-NEXT: vcvttsd2usi %xmm0, %rax
|
|
; AVX512F-NEXT: vcvtusi2sd %rax, %xmm15, %xmm0
|
|
; AVX512F-NEXT: retq
|
|
%ii = fptoui double %a0 to i64
|
|
%ff = uitofp i64 %ii to double
|
|
ret double %ff
|
|
}
|
|
|
|
define float @ucvtf32_i32(float %a0) {
|
|
; SSE-LABEL: ucvtf32_i32:
|
|
; SSE: # %bb.0:
|
|
; SSE-NEXT: cvttss2si %xmm0, %rax
|
|
; SSE-NEXT: movl %eax, %eax
|
|
; SSE-NEXT: xorps %xmm0, %xmm0
|
|
; SSE-NEXT: cvtsi2ss %rax, %xmm0
|
|
; SSE-NEXT: retq
|
|
;
|
|
; AVX2-LABEL: ucvtf32_i32:
|
|
; AVX2: # %bb.0:
|
|
; AVX2-NEXT: vcvttss2si %xmm0, %rax
|
|
; AVX2-NEXT: movl %eax, %eax
|
|
; AVX2-NEXT: vcvtsi2ss %rax, %xmm15, %xmm0
|
|
; AVX2-NEXT: retq
|
|
;
|
|
; AVX512-VL-LABEL: ucvtf32_i32:
|
|
; AVX512-VL: # %bb.0:
|
|
; AVX512-VL-NEXT: vcvttps2udq %xmm0, %xmm0
|
|
; AVX512-VL-NEXT: vcvtudq2ps %xmm0, %xmm0
|
|
; AVX512-VL-NEXT: retq
|
|
;
|
|
; AVX512-NOVL-LABEL: ucvtf32_i32:
|
|
; AVX512-NOVL: # %bb.0:
|
|
; AVX512-NOVL-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0
|
|
; AVX512-NOVL-NEXT: vcvttps2udq %zmm0, %zmm0
|
|
; AVX512-NOVL-NEXT: vcvtudq2ps %zmm0, %zmm0
|
|
; AVX512-NOVL-NEXT: # kill: def $xmm0 killed $xmm0 killed $zmm0
|
|
; AVX512-NOVL-NEXT: vzeroupper
|
|
; AVX512-NOVL-NEXT: retq
|
|
;
|
|
; AVX512F-LABEL: ucvtf32_i32:
|
|
; AVX512F: # %bb.0:
|
|
; AVX512F-NEXT: vcvttss2usi %xmm0, %eax
|
|
; AVX512F-NEXT: vcvtusi2ss %eax, %xmm15, %xmm0
|
|
; AVX512F-NEXT: retq
|
|
%ii = fptoui float %a0 to i32
|
|
%ff = uitofp i32 %ii to float
|
|
ret float %ff
|
|
}
|
|
|
|
define float @ucvtf32_i64(float %a0) {
|
|
; SSE-LABEL: ucvtf32_i64:
|
|
; SSE: # %bb.0:
|
|
; SSE-NEXT: cvttss2si %xmm0, %rcx
|
|
; SSE-NEXT: movq %rcx, %rdx
|
|
; SSE-NEXT: sarq $63, %rdx
|
|
; SSE-NEXT: subss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
|
|
; SSE-NEXT: cvttss2si %xmm0, %rax
|
|
; SSE-NEXT: andq %rdx, %rax
|
|
; SSE-NEXT: orq %rcx, %rax
|
|
; SSE-NEXT: js .LBB7_1
|
|
; SSE-NEXT: # %bb.2:
|
|
; SSE-NEXT: xorps %xmm0, %xmm0
|
|
; SSE-NEXT: cvtsi2ss %rax, %xmm0
|
|
; SSE-NEXT: retq
|
|
; SSE-NEXT: .LBB7_1:
|
|
; SSE-NEXT: movq %rax, %rcx
|
|
; SSE-NEXT: shrq %rcx
|
|
; SSE-NEXT: andl $1, %eax
|
|
; SSE-NEXT: orq %rcx, %rax
|
|
; SSE-NEXT: xorps %xmm0, %xmm0
|
|
; SSE-NEXT: cvtsi2ss %rax, %xmm0
|
|
; SSE-NEXT: addss %xmm0, %xmm0
|
|
; SSE-NEXT: retq
|
|
;
|
|
; AVX2-LABEL: ucvtf32_i64:
|
|
; AVX2: # %bb.0:
|
|
; AVX2-NEXT: vcvttss2si %xmm0, %rcx
|
|
; AVX2-NEXT: movq %rcx, %rdx
|
|
; AVX2-NEXT: sarq $63, %rdx
|
|
; AVX2-NEXT: vsubss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
|
|
; AVX2-NEXT: vcvttss2si %xmm0, %rax
|
|
; AVX2-NEXT: andq %rdx, %rax
|
|
; AVX2-NEXT: orq %rcx, %rax
|
|
; AVX2-NEXT: js .LBB7_1
|
|
; AVX2-NEXT: # %bb.2:
|
|
; AVX2-NEXT: vcvtsi2ss %rax, %xmm15, %xmm0
|
|
; AVX2-NEXT: retq
|
|
; AVX2-NEXT: .LBB7_1:
|
|
; AVX2-NEXT: movq %rax, %rcx
|
|
; AVX2-NEXT: shrq %rcx
|
|
; AVX2-NEXT: andl $1, %eax
|
|
; AVX2-NEXT: orq %rcx, %rax
|
|
; AVX2-NEXT: vcvtsi2ss %rax, %xmm15, %xmm0
|
|
; AVX2-NEXT: vaddss %xmm0, %xmm0, %xmm0
|
|
; AVX2-NEXT: retq
|
|
;
|
|
; AVX512-VL-LABEL: ucvtf32_i64:
|
|
; AVX512-VL: # %bb.0:
|
|
; AVX512-VL-NEXT: vcvttps2uqq %xmm0, %xmm0
|
|
; AVX512-VL-NEXT: vcvtuqq2ps %xmm0, %xmm0
|
|
; AVX512-VL-NEXT: retq
|
|
;
|
|
; AVX512-NOVL-LABEL: ucvtf32_i64:
|
|
; AVX512-NOVL: # %bb.0:
|
|
; AVX512-NOVL-NEXT: # kill: def $xmm0 killed $xmm0 def $ymm0
|
|
; AVX512-NOVL-NEXT: vcvttps2uqq %ymm0, %zmm0
|
|
; AVX512-NOVL-NEXT: vcvtuqq2ps %zmm0, %ymm0
|
|
; AVX512-NOVL-NEXT: # kill: def $xmm0 killed $xmm0 killed $ymm0
|
|
; AVX512-NOVL-NEXT: vzeroupper
|
|
; AVX512-NOVL-NEXT: retq
|
|
;
|
|
; AVX512F-LABEL: ucvtf32_i64:
|
|
; AVX512F: # %bb.0:
|
|
; AVX512F-NEXT: vcvttss2usi %xmm0, %rax
|
|
; AVX512F-NEXT: vcvtusi2ss %rax, %xmm15, %xmm0
|
|
; AVX512F-NEXT: retq
|
|
%ii = fptoui float %a0 to i64
|
|
%ff = uitofp i64 %ii to float
|
|
ret float %ff
|
|
}
|
|
;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
|
|
; AVX512: {{.*}}
|