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
427 lines
14 KiB
LLVM
427 lines
14 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
|
|
; RUN: llc < %s -mtriple=i686-- -mattr=+avx | FileCheck %s --check-prefixes=AVX
|
|
; RUN: llc < %s -mtriple=i686-- -mcpu=x86-64-v4 | FileCheck %s --check-prefixes=AVX512
|
|
|
|
define zeroext i8 @oeq_f64_i32(double %x) nounwind readnone {
|
|
; SSE-LABEL: oeq_f64_i32:
|
|
; SSE: # %bb.0: # %entry
|
|
; SSE-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
|
|
; SSE-NEXT: cvttpd2dq %xmm0, %xmm1
|
|
; SSE-NEXT: cvtdq2pd %xmm1, %xmm1
|
|
; SSE-NEXT: cmpeqsd %xmm0, %xmm1
|
|
; SSE-NEXT: movd %xmm1, %eax
|
|
; SSE-NEXT: andl $1, %eax
|
|
; SSE-NEXT: # kill: def $al killed $al killed $eax
|
|
; SSE-NEXT: retl
|
|
;
|
|
; AVX-LABEL: oeq_f64_i32:
|
|
; AVX: # %bb.0: # %entry
|
|
; AVX-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
|
|
; AVX-NEXT: vroundsd $11, %xmm0, %xmm0, %xmm1
|
|
; AVX-NEXT: vcmpeqsd %xmm0, %xmm1, %xmm0
|
|
; AVX-NEXT: vmovd %xmm0, %eax
|
|
; AVX-NEXT: andl $1, %eax
|
|
; AVX-NEXT: # kill: def $al killed $al killed $eax
|
|
; AVX-NEXT: retl
|
|
;
|
|
; AVX512-LABEL: oeq_f64_i32:
|
|
; AVX512: # %bb.0: # %entry
|
|
; AVX512-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
|
|
; AVX512-NEXT: vroundsd $11, %xmm0, %xmm0, %xmm1
|
|
; AVX512-NEXT: vcmpeqsd %xmm0, %xmm1, %k0
|
|
; AVX512-NEXT: kmovd %k0, %eax
|
|
; AVX512-NEXT: # kill: def $al killed $al killed $eax
|
|
; AVX512-NEXT: retl
|
|
entry:
|
|
%0 = fptosi double %x to i32
|
|
%1 = sitofp i32 %0 to double
|
|
%2 = fcmp oeq double %1, %x
|
|
%retval12 = zext i1 %2 to i8
|
|
ret i8 %retval12
|
|
}
|
|
|
|
define zeroext i8 @oeq_f64_u32(double %x) nounwind readnone {
|
|
; SSE-LABEL: oeq_f64_u32:
|
|
; SSE: # %bb.0: # %entry
|
|
; SSE-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
|
|
; SSE-NEXT: cvttsd2si %xmm0, %eax
|
|
; SSE-NEXT: movl %eax, %ecx
|
|
; SSE-NEXT: sarl $31, %ecx
|
|
; SSE-NEXT: movapd %xmm0, %xmm1
|
|
; SSE-NEXT: subsd {{\.?LCPI[0-9]+_[0-9]+}}, %xmm1
|
|
; SSE-NEXT: cvttsd2si %xmm1, %edx
|
|
; SSE-NEXT: andl %ecx, %edx
|
|
; SSE-NEXT: orl %eax, %edx
|
|
; SSE-NEXT: movd %edx, %xmm1
|
|
; SSE-NEXT: por {{\.?LCPI[0-9]+_[0-9]+}}, %xmm1
|
|
; SSE-NEXT: subsd {{\.?LCPI[0-9]+_[0-9]+}}, %xmm1
|
|
; SSE-NEXT: cmpeqsd %xmm0, %xmm1
|
|
; SSE-NEXT: movd %xmm1, %eax
|
|
; SSE-NEXT: andl $1, %eax
|
|
; SSE-NEXT: # kill: def $al killed $al killed $eax
|
|
; SSE-NEXT: retl
|
|
;
|
|
; AVX-LABEL: oeq_f64_u32:
|
|
; AVX: # %bb.0: # %entry
|
|
; AVX-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
|
|
; AVX-NEXT: vroundsd $11, %xmm0, %xmm0, %xmm1
|
|
; AVX-NEXT: vcmpeqsd %xmm0, %xmm1, %xmm0
|
|
; AVX-NEXT: vmovd %xmm0, %eax
|
|
; AVX-NEXT: andl $1, %eax
|
|
; AVX-NEXT: # kill: def $al killed $al killed $eax
|
|
; AVX-NEXT: retl
|
|
;
|
|
; AVX512-LABEL: oeq_f64_u32:
|
|
; AVX512: # %bb.0: # %entry
|
|
; AVX512-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
|
|
; AVX512-NEXT: vroundsd $11, %xmm0, %xmm0, %xmm1
|
|
; AVX512-NEXT: vcmpeqsd %xmm0, %xmm1, %k0
|
|
; AVX512-NEXT: kmovd %k0, %eax
|
|
; AVX512-NEXT: # kill: def $al killed $al killed $eax
|
|
; AVX512-NEXT: retl
|
|
entry:
|
|
%0 = fptoui double %x to i32
|
|
%1 = uitofp i32 %0 to double
|
|
%2 = fcmp oeq double %1, %x
|
|
%retval12 = zext i1 %2 to i8
|
|
ret i8 %retval12
|
|
}
|
|
|
|
define zeroext i8 @oeq_f64_i64(double %x) nounwind readnone {
|
|
; SSE-LABEL: oeq_f64_i64:
|
|
; SSE: # %bb.0: # %entry
|
|
; SSE-NEXT: pushl %ebp
|
|
; SSE-NEXT: movl %esp, %ebp
|
|
; SSE-NEXT: andl $-8, %esp
|
|
; SSE-NEXT: subl $32, %esp
|
|
; SSE-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
|
|
; SSE-NEXT: movsd %xmm0, {{[0-9]+}}(%esp)
|
|
; SSE-NEXT: fldl {{[0-9]+}}(%esp)
|
|
; SSE-NEXT: fnstcw {{[0-9]+}}(%esp)
|
|
; SSE-NEXT: movzwl {{[0-9]+}}(%esp), %eax
|
|
; SSE-NEXT: orl $3072, %eax # imm = 0xC00
|
|
; SSE-NEXT: movw %ax, {{[0-9]+}}(%esp)
|
|
; SSE-NEXT: fldcw {{[0-9]+}}(%esp)
|
|
; SSE-NEXT: fistpll {{[0-9]+}}(%esp)
|
|
; SSE-NEXT: fldcw {{[0-9]+}}(%esp)
|
|
; SSE-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero
|
|
; SSE-NEXT: movlps %xmm1, {{[0-9]+}}(%esp)
|
|
; SSE-NEXT: fildll {{[0-9]+}}(%esp)
|
|
; SSE-NEXT: fstpl {{[0-9]+}}(%esp)
|
|
; SSE-NEXT: cmpeqsd {{[0-9]+}}(%esp), %xmm0
|
|
; SSE-NEXT: movd %xmm0, %eax
|
|
; SSE-NEXT: andl $1, %eax
|
|
; SSE-NEXT: # kill: def $al killed $al killed $eax
|
|
; SSE-NEXT: movl %ebp, %esp
|
|
; SSE-NEXT: popl %ebp
|
|
; SSE-NEXT: retl
|
|
;
|
|
; AVX-LABEL: oeq_f64_i64:
|
|
; AVX: # %bb.0: # %entry
|
|
; AVX-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
|
|
; AVX-NEXT: vroundsd $11, %xmm0, %xmm0, %xmm1
|
|
; AVX-NEXT: vcmpeqsd %xmm0, %xmm1, %xmm0
|
|
; AVX-NEXT: vmovd %xmm0, %eax
|
|
; AVX-NEXT: andl $1, %eax
|
|
; AVX-NEXT: # kill: def $al killed $al killed $eax
|
|
; AVX-NEXT: retl
|
|
;
|
|
; AVX512-LABEL: oeq_f64_i64:
|
|
; AVX512: # %bb.0: # %entry
|
|
; AVX512-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
|
|
; AVX512-NEXT: vroundsd $11, %xmm0, %xmm0, %xmm1
|
|
; AVX512-NEXT: vcmpeqsd %xmm0, %xmm1, %k0
|
|
; AVX512-NEXT: kmovd %k0, %eax
|
|
; AVX512-NEXT: # kill: def $al killed $al killed $eax
|
|
; AVX512-NEXT: retl
|
|
entry:
|
|
%0 = fptosi double %x to i64
|
|
%1 = sitofp i64 %0 to double
|
|
%2 = fcmp oeq double %1, %x
|
|
%retval12 = zext i1 %2 to i8
|
|
ret i8 %retval12
|
|
}
|
|
|
|
define zeroext i8 @oeq_f64_u64(double %x) nounwind readnone {
|
|
; SSE-LABEL: oeq_f64_u64:
|
|
; SSE: # %bb.0: # %entry
|
|
; SSE-NEXT: pushl %ebp
|
|
; SSE-NEXT: movl %esp, %ebp
|
|
; SSE-NEXT: andl $-8, %esp
|
|
; SSE-NEXT: subl $16, %esp
|
|
; SSE-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
|
|
; SSE-NEXT: movsd {{.*#+}} xmm1 = [9.2233720368547758E+18,0.0E+0]
|
|
; SSE-NEXT: ucomisd %xmm0, %xmm1
|
|
; SSE-NEXT: jbe .LBB3_2
|
|
; SSE-NEXT: # %bb.1: # %entry
|
|
; SSE-NEXT: xorpd %xmm1, %xmm1
|
|
; SSE-NEXT: .LBB3_2: # %entry
|
|
; SSE-NEXT: movapd %xmm0, %xmm2
|
|
; SSE-NEXT: subsd %xmm1, %xmm2
|
|
; SSE-NEXT: movsd %xmm2, {{[0-9]+}}(%esp)
|
|
; SSE-NEXT: setbe %al
|
|
; SSE-NEXT: fldl {{[0-9]+}}(%esp)
|
|
; SSE-NEXT: fnstcw {{[0-9]+}}(%esp)
|
|
; SSE-NEXT: movzwl {{[0-9]+}}(%esp), %ecx
|
|
; SSE-NEXT: orl $3072, %ecx # imm = 0xC00
|
|
; SSE-NEXT: movw %cx, {{[0-9]+}}(%esp)
|
|
; SSE-NEXT: fldcw {{[0-9]+}}(%esp)
|
|
; SSE-NEXT: fistpll {{[0-9]+}}(%esp)
|
|
; SSE-NEXT: fldcw {{[0-9]+}}(%esp)
|
|
; SSE-NEXT: movzbl %al, %eax
|
|
; SSE-NEXT: shll $31, %eax
|
|
; SSE-NEXT: xorl {{[0-9]+}}(%esp), %eax
|
|
; SSE-NEXT: movd %eax, %xmm1
|
|
; SSE-NEXT: movd {{.*#+}} xmm2 = mem[0],zero,zero,zero
|
|
; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
|
|
; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],mem[0],xmm2[1],mem[1]
|
|
; SSE-NEXT: subpd {{\.?LCPI[0-9]+_[0-9]+}}, %xmm2
|
|
; SSE-NEXT: movapd %xmm2, %xmm1
|
|
; SSE-NEXT: unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm2[1]
|
|
; SSE-NEXT: addsd %xmm2, %xmm1
|
|
; SSE-NEXT: cmpeqsd %xmm0, %xmm1
|
|
; SSE-NEXT: movd %xmm1, %eax
|
|
; SSE-NEXT: andl $1, %eax
|
|
; SSE-NEXT: # kill: def $al killed $al killed $eax
|
|
; SSE-NEXT: movl %ebp, %esp
|
|
; SSE-NEXT: popl %ebp
|
|
; SSE-NEXT: retl
|
|
;
|
|
; AVX-LABEL: oeq_f64_u64:
|
|
; AVX: # %bb.0: # %entry
|
|
; AVX-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
|
|
; AVX-NEXT: vroundsd $11, %xmm0, %xmm0, %xmm1
|
|
; AVX-NEXT: vcmpeqsd %xmm0, %xmm1, %xmm0
|
|
; AVX-NEXT: vmovd %xmm0, %eax
|
|
; AVX-NEXT: andl $1, %eax
|
|
; AVX-NEXT: # kill: def $al killed $al killed $eax
|
|
; AVX-NEXT: retl
|
|
;
|
|
; AVX512-LABEL: oeq_f64_u64:
|
|
; AVX512: # %bb.0: # %entry
|
|
; AVX512-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
|
|
; AVX512-NEXT: vroundsd $11, %xmm0, %xmm0, %xmm1
|
|
; AVX512-NEXT: vcmpeqsd %xmm0, %xmm1, %k0
|
|
; AVX512-NEXT: kmovd %k0, %eax
|
|
; AVX512-NEXT: # kill: def $al killed $al killed $eax
|
|
; AVX512-NEXT: retl
|
|
entry:
|
|
%0 = fptoui double %x to i64
|
|
%1 = uitofp i64 %0 to double
|
|
%2 = fcmp oeq double %1, %x
|
|
%retval12 = zext i1 %2 to i8
|
|
ret i8 %retval12
|
|
}
|
|
|
|
define zeroext i8 @une_f64_i32(double %x) nounwind readnone {
|
|
; SSE-LABEL: une_f64_i32:
|
|
; SSE: # %bb.0: # %entry
|
|
; SSE-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
|
|
; SSE-NEXT: cvttpd2dq %xmm0, %xmm1
|
|
; SSE-NEXT: cvtdq2pd %xmm1, %xmm1
|
|
; SSE-NEXT: cmpneqsd %xmm0, %xmm1
|
|
; SSE-NEXT: movd %xmm1, %eax
|
|
; SSE-NEXT: andl $1, %eax
|
|
; SSE-NEXT: # kill: def $al killed $al killed $eax
|
|
; SSE-NEXT: retl
|
|
;
|
|
; AVX-LABEL: une_f64_i32:
|
|
; AVX: # %bb.0: # %entry
|
|
; AVX-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
|
|
; AVX-NEXT: vroundsd $11, %xmm0, %xmm0, %xmm1
|
|
; AVX-NEXT: vcmpneqsd %xmm0, %xmm1, %xmm0
|
|
; AVX-NEXT: vmovd %xmm0, %eax
|
|
; AVX-NEXT: andl $1, %eax
|
|
; AVX-NEXT: # kill: def $al killed $al killed $eax
|
|
; AVX-NEXT: retl
|
|
;
|
|
; AVX512-LABEL: une_f64_i32:
|
|
; AVX512: # %bb.0: # %entry
|
|
; AVX512-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
|
|
; AVX512-NEXT: vroundsd $11, %xmm0, %xmm0, %xmm1
|
|
; AVX512-NEXT: vcmpneqsd %xmm0, %xmm1, %k0
|
|
; AVX512-NEXT: kmovd %k0, %eax
|
|
; AVX512-NEXT: # kill: def $al killed $al killed $eax
|
|
; AVX512-NEXT: retl
|
|
entry:
|
|
%0 = fptosi double %x to i32
|
|
%1 = sitofp i32 %0 to double
|
|
%2 = fcmp une double %1, %x
|
|
%retval12 = zext i1 %2 to i8
|
|
ret i8 %retval12
|
|
}
|
|
|
|
define zeroext i8 @une_f64_u32(double %x) nounwind readnone {
|
|
; SSE-LABEL: une_f64_u32:
|
|
; SSE: # %bb.0: # %entry
|
|
; SSE-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
|
|
; SSE-NEXT: cvttsd2si %xmm0, %eax
|
|
; SSE-NEXT: movl %eax, %ecx
|
|
; SSE-NEXT: sarl $31, %ecx
|
|
; SSE-NEXT: movapd %xmm0, %xmm1
|
|
; SSE-NEXT: subsd {{\.?LCPI[0-9]+_[0-9]+}}, %xmm1
|
|
; SSE-NEXT: cvttsd2si %xmm1, %edx
|
|
; SSE-NEXT: andl %ecx, %edx
|
|
; SSE-NEXT: orl %eax, %edx
|
|
; SSE-NEXT: movd %edx, %xmm1
|
|
; SSE-NEXT: por {{\.?LCPI[0-9]+_[0-9]+}}, %xmm1
|
|
; SSE-NEXT: subsd {{\.?LCPI[0-9]+_[0-9]+}}, %xmm1
|
|
; SSE-NEXT: cmpneqsd %xmm0, %xmm1
|
|
; SSE-NEXT: movd %xmm1, %eax
|
|
; SSE-NEXT: andl $1, %eax
|
|
; SSE-NEXT: # kill: def $al killed $al killed $eax
|
|
; SSE-NEXT: retl
|
|
;
|
|
; AVX-LABEL: une_f64_u32:
|
|
; AVX: # %bb.0: # %entry
|
|
; AVX-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
|
|
; AVX-NEXT: vroundsd $11, %xmm0, %xmm0, %xmm1
|
|
; AVX-NEXT: vcmpneqsd %xmm0, %xmm1, %xmm0
|
|
; AVX-NEXT: vmovd %xmm0, %eax
|
|
; AVX-NEXT: andl $1, %eax
|
|
; AVX-NEXT: # kill: def $al killed $al killed $eax
|
|
; AVX-NEXT: retl
|
|
;
|
|
; AVX512-LABEL: une_f64_u32:
|
|
; AVX512: # %bb.0: # %entry
|
|
; AVX512-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
|
|
; AVX512-NEXT: vroundsd $11, %xmm0, %xmm0, %xmm1
|
|
; AVX512-NEXT: vcmpneqsd %xmm0, %xmm1, %k0
|
|
; AVX512-NEXT: kmovd %k0, %eax
|
|
; AVX512-NEXT: # kill: def $al killed $al killed $eax
|
|
; AVX512-NEXT: retl
|
|
entry:
|
|
%0 = fptoui double %x to i32
|
|
%1 = uitofp i32 %0 to double
|
|
%2 = fcmp une double %1, %x
|
|
%retval12 = zext i1 %2 to i8
|
|
ret i8 %retval12
|
|
}
|
|
|
|
define zeroext i8 @une_f64_i64(double %x) nounwind readnone {
|
|
; SSE-LABEL: une_f64_i64:
|
|
; SSE: # %bb.0: # %entry
|
|
; SSE-NEXT: pushl %ebp
|
|
; SSE-NEXT: movl %esp, %ebp
|
|
; SSE-NEXT: andl $-8, %esp
|
|
; SSE-NEXT: subl $32, %esp
|
|
; SSE-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
|
|
; SSE-NEXT: movsd %xmm0, {{[0-9]+}}(%esp)
|
|
; SSE-NEXT: fldl {{[0-9]+}}(%esp)
|
|
; SSE-NEXT: fnstcw {{[0-9]+}}(%esp)
|
|
; SSE-NEXT: movzwl {{[0-9]+}}(%esp), %eax
|
|
; SSE-NEXT: orl $3072, %eax # imm = 0xC00
|
|
; SSE-NEXT: movw %ax, {{[0-9]+}}(%esp)
|
|
; SSE-NEXT: fldcw {{[0-9]+}}(%esp)
|
|
; SSE-NEXT: fistpll {{[0-9]+}}(%esp)
|
|
; SSE-NEXT: fldcw {{[0-9]+}}(%esp)
|
|
; SSE-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero
|
|
; SSE-NEXT: movlps %xmm1, {{[0-9]+}}(%esp)
|
|
; SSE-NEXT: fildll {{[0-9]+}}(%esp)
|
|
; SSE-NEXT: fstpl {{[0-9]+}}(%esp)
|
|
; SSE-NEXT: cmpneqsd {{[0-9]+}}(%esp), %xmm0
|
|
; SSE-NEXT: movd %xmm0, %eax
|
|
; SSE-NEXT: andl $1, %eax
|
|
; SSE-NEXT: # kill: def $al killed $al killed $eax
|
|
; SSE-NEXT: movl %ebp, %esp
|
|
; SSE-NEXT: popl %ebp
|
|
; SSE-NEXT: retl
|
|
;
|
|
; AVX-LABEL: une_f64_i64:
|
|
; AVX: # %bb.0: # %entry
|
|
; AVX-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
|
|
; AVX-NEXT: vroundsd $11, %xmm0, %xmm0, %xmm1
|
|
; AVX-NEXT: vcmpneqsd %xmm0, %xmm1, %xmm0
|
|
; AVX-NEXT: vmovd %xmm0, %eax
|
|
; AVX-NEXT: andl $1, %eax
|
|
; AVX-NEXT: # kill: def $al killed $al killed $eax
|
|
; AVX-NEXT: retl
|
|
;
|
|
; AVX512-LABEL: une_f64_i64:
|
|
; AVX512: # %bb.0: # %entry
|
|
; AVX512-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
|
|
; AVX512-NEXT: vroundsd $11, %xmm0, %xmm0, %xmm1
|
|
; AVX512-NEXT: vcmpneqsd %xmm0, %xmm1, %k0
|
|
; AVX512-NEXT: kmovd %k0, %eax
|
|
; AVX512-NEXT: # kill: def $al killed $al killed $eax
|
|
; AVX512-NEXT: retl
|
|
entry:
|
|
%0 = fptosi double %x to i64
|
|
%1 = sitofp i64 %0 to double
|
|
%2 = fcmp une double %1, %x
|
|
%retval12 = zext i1 %2 to i8
|
|
ret i8 %retval12
|
|
}
|
|
|
|
define zeroext i8 @une_f64_u64(double %x) nounwind readnone {
|
|
; SSE-LABEL: une_f64_u64:
|
|
; SSE: # %bb.0: # %entry
|
|
; SSE-NEXT: pushl %ebp
|
|
; SSE-NEXT: movl %esp, %ebp
|
|
; SSE-NEXT: andl $-8, %esp
|
|
; SSE-NEXT: subl $16, %esp
|
|
; SSE-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
|
|
; SSE-NEXT: movsd {{.*#+}} xmm1 = [9.2233720368547758E+18,0.0E+0]
|
|
; SSE-NEXT: ucomisd %xmm0, %xmm1
|
|
; SSE-NEXT: jbe .LBB7_2
|
|
; SSE-NEXT: # %bb.1: # %entry
|
|
; SSE-NEXT: xorpd %xmm1, %xmm1
|
|
; SSE-NEXT: .LBB7_2: # %entry
|
|
; SSE-NEXT: movapd %xmm0, %xmm2
|
|
; SSE-NEXT: subsd %xmm1, %xmm2
|
|
; SSE-NEXT: movsd %xmm2, {{[0-9]+}}(%esp)
|
|
; SSE-NEXT: setbe %al
|
|
; SSE-NEXT: fldl {{[0-9]+}}(%esp)
|
|
; SSE-NEXT: fnstcw {{[0-9]+}}(%esp)
|
|
; SSE-NEXT: movzwl {{[0-9]+}}(%esp), %ecx
|
|
; SSE-NEXT: orl $3072, %ecx # imm = 0xC00
|
|
; SSE-NEXT: movw %cx, {{[0-9]+}}(%esp)
|
|
; SSE-NEXT: fldcw {{[0-9]+}}(%esp)
|
|
; SSE-NEXT: fistpll {{[0-9]+}}(%esp)
|
|
; SSE-NEXT: fldcw {{[0-9]+}}(%esp)
|
|
; SSE-NEXT: movzbl %al, %eax
|
|
; SSE-NEXT: shll $31, %eax
|
|
; SSE-NEXT: xorl {{[0-9]+}}(%esp), %eax
|
|
; SSE-NEXT: movd %eax, %xmm1
|
|
; SSE-NEXT: movd {{.*#+}} xmm2 = mem[0],zero,zero,zero
|
|
; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
|
|
; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],mem[0],xmm2[1],mem[1]
|
|
; SSE-NEXT: subpd {{\.?LCPI[0-9]+_[0-9]+}}, %xmm2
|
|
; SSE-NEXT: movapd %xmm2, %xmm1
|
|
; SSE-NEXT: unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm2[1]
|
|
; SSE-NEXT: addsd %xmm2, %xmm1
|
|
; SSE-NEXT: cmpneqsd %xmm0, %xmm1
|
|
; SSE-NEXT: movd %xmm1, %eax
|
|
; SSE-NEXT: andl $1, %eax
|
|
; SSE-NEXT: # kill: def $al killed $al killed $eax
|
|
; SSE-NEXT: movl %ebp, %esp
|
|
; SSE-NEXT: popl %ebp
|
|
; SSE-NEXT: retl
|
|
;
|
|
; AVX-LABEL: une_f64_u64:
|
|
; AVX: # %bb.0: # %entry
|
|
; AVX-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
|
|
; AVX-NEXT: vroundsd $11, %xmm0, %xmm0, %xmm1
|
|
; AVX-NEXT: vcmpneqsd %xmm0, %xmm1, %xmm0
|
|
; AVX-NEXT: vmovd %xmm0, %eax
|
|
; AVX-NEXT: andl $1, %eax
|
|
; AVX-NEXT: # kill: def $al killed $al killed $eax
|
|
; AVX-NEXT: retl
|
|
;
|
|
; AVX512-LABEL: une_f64_u64:
|
|
; AVX512: # %bb.0: # %entry
|
|
; AVX512-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
|
|
; AVX512-NEXT: vroundsd $11, %xmm0, %xmm0, %xmm1
|
|
; AVX512-NEXT: vcmpneqsd %xmm0, %xmm1, %k0
|
|
; AVX512-NEXT: kmovd %k0, %eax
|
|
; AVX512-NEXT: # kill: def $al killed $al killed $eax
|
|
; AVX512-NEXT: retl
|
|
entry:
|
|
%0 = fptoui double %x to i64
|
|
%1 = uitofp i64 %0 to double
|
|
%2 = fcmp une double %1, %x
|
|
%retval12 = zext i1 %2 to i8
|
|
ret i8 %retval12
|
|
}
|