468 changes: 67 additions & 401 deletions llvm/test/CodeGen/X86/vector-half-conversions.ll
View file

Large diffs are not rendered by default.

73 changes: 51 additions & 22 deletions llvm/test/CodeGen/X86/vector-reduce-fmax-nnan.ll
View file
Original file line number Diff line number Diff line change
Expand Up @@ -3,8 +3,8 @@
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+sse4.1 | FileCheck %s --check-prefixes=ALL,SSE,SSE41
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx | FileCheck %s --check-prefixes=ALL,AVX
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2 | FileCheck %s --check-prefixes=ALL,AVX
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512f,+avx512bw | FileCheck %s --check-prefixes=ALL,AVX512,AVX512BW
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512f,+avx512bw,+avx512vl | FileCheck %s --check-prefixes=ALL,AVX512,AVX512BW
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512f,+avx512bw | FileCheck %s --check-prefixes=ALL,AVX512,AVX512F
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512f,+avx512bw,+avx512vl | FileCheck %s --check-prefixes=ALL,AVX512,AVX512VL
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512fp16,+avx512vl | FileCheck %s --check-prefixes=ALL,AVX512,AVX512FP16

;
Expand Down Expand Up @@ -370,10 +370,11 @@ define half @test_v2f16(<2 x half> %a0) nounwind {
; SSE-NEXT: pushq %rbp
; SSE-NEXT: pushq %rbx
; SSE-NEXT: subq $40, %rsp
; SSE-NEXT: movdqa %xmm0, %xmm1
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pextrw $0, %xmm1, %ebx
; SSE-NEXT: pextrw $0, %xmm0, %ebp
; SSE-NEXT: movdqa %xmm1, %xmm0
; SSE-NEXT: psrld $16, %xmm0
; SSE-NEXT: pextrw $0, %xmm0, %ebx
; SSE-NEXT: pextrw $0, %xmm1, %ebp
; SSE-NEXT: callq __extendhfsf2@PLT
; SSE-NEXT: movd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Folded Spill
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
Expand All @@ -391,10 +392,11 @@ define half @test_v2f16(<2 x half> %a0) nounwind {
; AVX-NEXT: pushq %rbp
; AVX-NEXT: pushq %rbx
; AVX-NEXT: subq $40, %rsp
; AVX-NEXT: vmovdqa %xmm0, %xmm1
; AVX-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT: vpextrw $0, %xmm1, %ebx
; AVX-NEXT: vpextrw $0, %xmm0, %ebp
; AVX-NEXT: vmovdqa %xmm1, %xmm0
; AVX-NEXT: vpsrld $16, %xmm0, %xmm0
; AVX-NEXT: vpextrw $0, %xmm0, %ebx
; AVX-NEXT: vpextrw $0, %xmm1, %ebp
; AVX-NEXT: callq __extendhfsf2@PLT
; AVX-NEXT: vmovd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Folded Spill
; AVX-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
Expand All @@ -407,20 +409,47 @@ define half @test_v2f16(<2 x half> %a0) nounwind {
; AVX-NEXT: popq %rbp
; AVX-NEXT: retq
;
; AVX512BW-LABEL: test_v2f16:
; AVX512BW: # %bb.0:
; AVX512BW-NEXT: vpextrw $0, %xmm0, %eax
; AVX512BW-NEXT: vpextrw $0, %xmm1, %ecx
; AVX512BW-NEXT: movzwl %cx, %ecx
; AVX512BW-NEXT: vmovd %ecx, %xmm0
; AVX512BW-NEXT: vcvtph2ps %xmm0, %xmm0
; AVX512BW-NEXT: movzwl %ax, %eax
; AVX512BW-NEXT: vmovd %eax, %xmm1
; AVX512BW-NEXT: vcvtph2ps %xmm1, %xmm1
; AVX512BW-NEXT: vucomiss %xmm0, %xmm1
; AVX512BW-NEXT: cmoval %eax, %ecx
; AVX512BW-NEXT: vpinsrw $0, %ecx, %xmm0, %xmm0
; AVX512BW-NEXT: retq
; AVX512F-LABEL: test_v2f16:
; AVX512F: # %bb.0:
; AVX512F-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0
; AVX512F-NEXT: vpsrld $16, %xmm0, %xmm1
; AVX512F-NEXT: vpextrw $0, %xmm0, %eax
; AVX512F-NEXT: movzwl %ax, %eax
; AVX512F-NEXT: vmovd %eax, %xmm2
; AVX512F-NEXT: vcvtph2ps %xmm2, %xmm2
; AVX512F-NEXT: vpextrw $0, %xmm1, %eax
; AVX512F-NEXT: movzwl %ax, %eax
; AVX512F-NEXT: vmovd %eax, %xmm3
; AVX512F-NEXT: vcvtph2ps %xmm3, %xmm3
; AVX512F-NEXT: xorl %eax, %eax
; AVX512F-NEXT: vucomiss %xmm3, %xmm2
; AVX512F-NEXT: movl $255, %ecx
; AVX512F-NEXT: cmovbel %eax, %ecx
; AVX512F-NEXT: kmovd %ecx, %k1
; AVX512F-NEXT: vmovdqu16 %zmm0, %zmm1 {%k1}
; AVX512F-NEXT: vmovdqa %xmm1, %xmm0
; AVX512F-NEXT: vzeroupper
; AVX512F-NEXT: retq
;
; AVX512VL-LABEL: test_v2f16:
; AVX512VL: # %bb.0:
; AVX512VL-NEXT: vpsrld $16, %xmm0, %xmm1
; AVX512VL-NEXT: vpextrw $0, %xmm0, %eax
; AVX512VL-NEXT: movzwl %ax, %eax
; AVX512VL-NEXT: vmovd %eax, %xmm2
; AVX512VL-NEXT: vcvtph2ps %xmm2, %xmm2
; AVX512VL-NEXT: vpextrw $0, %xmm1, %eax
; AVX512VL-NEXT: movzwl %ax, %eax
; AVX512VL-NEXT: vmovd %eax, %xmm3
; AVX512VL-NEXT: vcvtph2ps %xmm3, %xmm3
; AVX512VL-NEXT: xorl %eax, %eax
; AVX512VL-NEXT: vucomiss %xmm3, %xmm2
; AVX512VL-NEXT: movl $255, %ecx
; AVX512VL-NEXT: cmovbel %eax, %ecx
; AVX512VL-NEXT: kmovd %ecx, %k1
; AVX512VL-NEXT: vmovdqu16 %xmm0, %xmm1 {%k1}
; AVX512VL-NEXT: vmovdqa %xmm1, %xmm0
; AVX512VL-NEXT: retq
;
; AVX512FP16-LABEL: test_v2f16:
; AVX512FP16: # %bb.0:
Expand Down
73 changes: 51 additions & 22 deletions llvm/test/CodeGen/X86/vector-reduce-fmin-nnan.ll
View file
Original file line number Diff line number Diff line change
Expand Up @@ -3,8 +3,8 @@
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+sse4.1 | FileCheck %s --check-prefixes=ALL,SSE,SSE41
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx | FileCheck %s --check-prefixes=ALL,AVX
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2 | FileCheck %s --check-prefixes=ALL,AVX
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512f,+avx512bw | FileCheck %s --check-prefixes=ALL,AVX512,AVX512BW
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512f,+avx512bw,+avx512vl | FileCheck %s --check-prefixes=ALL,AVX512,AVX512BW
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512f,+avx512bw | FileCheck %s --check-prefixes=ALL,AVX512,AVX512F
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512f,+avx512bw,+avx512vl | FileCheck %s --check-prefixes=ALL,AVX512,AVX512VL
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512fp16,+avx512vl | FileCheck %s --check-prefixes=ALL,AVX512,AVX512FP16

;
Expand Down Expand Up @@ -369,10 +369,11 @@ define half @test_v2f16(<2 x half> %a0) nounwind {
; SSE-NEXT: pushq %rbp
; SSE-NEXT: pushq %rbx
; SSE-NEXT: subq $40, %rsp
; SSE-NEXT: movdqa %xmm0, %xmm1
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pextrw $0, %xmm1, %ebx
; SSE-NEXT: pextrw $0, %xmm0, %ebp
; SSE-NEXT: movdqa %xmm1, %xmm0
; SSE-NEXT: psrld $16, %xmm0
; SSE-NEXT: pextrw $0, %xmm0, %ebx
; SSE-NEXT: pextrw $0, %xmm1, %ebp
; SSE-NEXT: callq __extendhfsf2@PLT
; SSE-NEXT: movd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Folded Spill
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
Expand All @@ -390,10 +391,11 @@ define half @test_v2f16(<2 x half> %a0) nounwind {
; AVX-NEXT: pushq %rbp
; AVX-NEXT: pushq %rbx
; AVX-NEXT: subq $40, %rsp
; AVX-NEXT: vmovdqa %xmm0, %xmm1
; AVX-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT: vpextrw $0, %xmm1, %ebx
; AVX-NEXT: vpextrw $0, %xmm0, %ebp
; AVX-NEXT: vmovdqa %xmm1, %xmm0
; AVX-NEXT: vpsrld $16, %xmm0, %xmm0
; AVX-NEXT: vpextrw $0, %xmm0, %ebx
; AVX-NEXT: vpextrw $0, %xmm1, %ebp
; AVX-NEXT: callq __extendhfsf2@PLT
; AVX-NEXT: vmovd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Folded Spill
; AVX-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
Expand All @@ -406,20 +408,47 @@ define half @test_v2f16(<2 x half> %a0) nounwind {
; AVX-NEXT: popq %rbp
; AVX-NEXT: retq
;
; AVX512BW-LABEL: test_v2f16:
; AVX512BW: # %bb.0:
; AVX512BW-NEXT: vpextrw $0, %xmm0, %eax
; AVX512BW-NEXT: vpextrw $0, %xmm1, %ecx
; AVX512BW-NEXT: movzwl %cx, %ecx
; AVX512BW-NEXT: vmovd %ecx, %xmm0
; AVX512BW-NEXT: vcvtph2ps %xmm0, %xmm0
; AVX512BW-NEXT: movzwl %ax, %eax
; AVX512BW-NEXT: vmovd %eax, %xmm1
; AVX512BW-NEXT: vcvtph2ps %xmm1, %xmm1
; AVX512BW-NEXT: vucomiss %xmm0, %xmm1
; AVX512BW-NEXT: cmovbl %eax, %ecx
; AVX512BW-NEXT: vpinsrw $0, %ecx, %xmm0, %xmm0
; AVX512BW-NEXT: retq
; AVX512F-LABEL: test_v2f16:
; AVX512F: # %bb.0:
; AVX512F-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0
; AVX512F-NEXT: vpsrld $16, %xmm0, %xmm1
; AVX512F-NEXT: vpextrw $0, %xmm0, %eax
; AVX512F-NEXT: movzwl %ax, %eax
; AVX512F-NEXT: vmovd %eax, %xmm2
; AVX512F-NEXT: vcvtph2ps %xmm2, %xmm2
; AVX512F-NEXT: vpextrw $0, %xmm1, %eax
; AVX512F-NEXT: movzwl %ax, %eax
; AVX512F-NEXT: vmovd %eax, %xmm3
; AVX512F-NEXT: vcvtph2ps %xmm3, %xmm3
; AVX512F-NEXT: xorl %eax, %eax
; AVX512F-NEXT: vucomiss %xmm3, %xmm2
; AVX512F-NEXT: movl $255, %ecx
; AVX512F-NEXT: cmovael %eax, %ecx
; AVX512F-NEXT: kmovd %ecx, %k1
; AVX512F-NEXT: vmovdqu16 %zmm0, %zmm1 {%k1}
; AVX512F-NEXT: vmovdqa %xmm1, %xmm0
; AVX512F-NEXT: vzeroupper
; AVX512F-NEXT: retq
;
; AVX512VL-LABEL: test_v2f16:
; AVX512VL: # %bb.0:
; AVX512VL-NEXT: vpsrld $16, %xmm0, %xmm1
; AVX512VL-NEXT: vpextrw $0, %xmm0, %eax
; AVX512VL-NEXT: movzwl %ax, %eax
; AVX512VL-NEXT: vmovd %eax, %xmm2
; AVX512VL-NEXT: vcvtph2ps %xmm2, %xmm2
; AVX512VL-NEXT: vpextrw $0, %xmm1, %eax
; AVX512VL-NEXT: movzwl %ax, %eax
; AVX512VL-NEXT: vmovd %eax, %xmm3
; AVX512VL-NEXT: vcvtph2ps %xmm3, %xmm3
; AVX512VL-NEXT: xorl %eax, %eax
; AVX512VL-NEXT: vucomiss %xmm3, %xmm2
; AVX512VL-NEXT: movl $255, %ecx
; AVX512VL-NEXT: cmovael %eax, %ecx
; AVX512VL-NEXT: kmovd %ecx, %k1
; AVX512VL-NEXT: vmovdqu16 %xmm0, %xmm1 {%k1}
; AVX512VL-NEXT: vmovdqa %xmm1, %xmm0
; AVX512VL-NEXT: retq
;
; AVX512FP16-LABEL: test_v2f16:
; AVX512FP16: # %bb.0:
Expand Down