1; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py 2; RUN: llc -fast-isel-sink-local-values < %s -O0 -mtriple=x86_64-apple-darwin -mcpu=skx | FileCheck %s 3 4; ModuleID = 'mask_set.c' 5source_filename = "mask_set.c" 6target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128" 7target triple = "x86_64-unknown-linux-gnu" 8 9declare void @llvm.dbg.declare(metadata, metadata, metadata) 10 11; Function Attrs: nounwind uwtable 12declare i64 @calc_expected_mask_val(i8* %valp, i32 %el_size, i32 %length) 13; Function Attrs: nounwind uwtable 14declare i32 @check_mask16(i16 zeroext %res_mask, i16 zeroext %exp_mask, i8* %fname, i8* %input) 15 16; Function Attrs: nounwind uwtable 17define void @test_xmm(i32 %shift, i32 %mulp, <2 x i64> %a,i8* %arraydecay,i8* %fname){ 18; CHECK-LABEL: test_xmm: 19; CHECK: ## %bb.0: 20; CHECK-NEXT: subq $56, %rsp 21; CHECK-NEXT: .cfi_def_cfa_offset 64 22; CHECK-NEXT: vpmovw2m %xmm0, %k0 23; CHECK-NEXT: movl $2, %esi 24; CHECK-NEXT: movl $8, %eax 25; CHECK-NEXT: movq %rdx, %rdi 26; CHECK-NEXT: movq %rdx, {{[0-9]+}}(%rsp) ## 8-byte Spill 27; CHECK-NEXT: movl %eax, %edx 28; CHECK-NEXT: kmovw %k0, {{[0-9]+}}(%rsp) ## 2-byte Spill 29; CHECK-NEXT: movq %rcx, {{[0-9]+}}(%rsp) ## 8-byte Spill 30; CHECK-NEXT: vmovaps %xmm0, {{[0-9]+}}(%rsp) ## 16-byte Spill 31; CHECK-NEXT: callq _calc_expected_mask_val 32; CHECK-NEXT: movl %eax, %edx 33; CHECK-NEXT: movw %dx, %r8w 34; CHECK-NEXT: movzwl %r8w, %esi 35; CHECK-NEXT: kmovw {{[0-9]+}}(%rsp), %k0 ## 2-byte Reload 36; CHECK-NEXT: kmovb %k0, %edi 37; CHECK-NEXT: movq {{[0-9]+}}(%rsp), %rdx ## 8-byte Reload 38; CHECK-NEXT: movq {{[0-9]+}}(%rsp), %rcx ## 8-byte Reload 39; CHECK-NEXT: callq _check_mask16 40; CHECK-NEXT: vmovaps {{[0-9]+}}(%rsp), %xmm0 ## 16-byte Reload 41; CHECK-NEXT: vpmovd2m %xmm0, %k0 42; CHECK-NEXT: kmovq %k0, %k1 43; CHECK-NEXT: kmovd %k0, %esi 44; CHECK-NEXT: movb %sil, %r9b 45; CHECK-NEXT: movzbl %r9b, %esi 46; CHECK-NEXT: movw %si, %r8w 47; CHECK-NEXT: movq {{[0-9]+}}(%rsp), %rdi ## 8-byte Reload 48; CHECK-NEXT: movl $4, %esi 49; CHECK-NEXT: movl %esi, {{[0-9]+}}(%rsp) ## 4-byte Spill 50; CHECK-NEXT: movl {{[0-9]+}}(%rsp), %edx ## 4-byte Reload 51; CHECK-NEXT: movl %eax, {{[0-9]+}}(%rsp) ## 4-byte Spill 52; CHECK-NEXT: kmovw %k1, {{[0-9]+}}(%rsp) ## 2-byte Spill 53; CHECK-NEXT: movw %r8w, {{[0-9]+}}(%rsp) ## 2-byte Spill 54; CHECK-NEXT: callq _calc_expected_mask_val 55; CHECK-NEXT: movw %ax, %r8w 56; CHECK-NEXT: movw {{[0-9]+}}(%rsp), %r10w ## 2-byte Reload 57; CHECK-NEXT: movzwl %r10w, %edi 58; CHECK-NEXT: movzwl %r8w, %esi 59; CHECK-NEXT: movq {{[0-9]+}}(%rsp), %rdx ## 8-byte Reload 60; CHECK-NEXT: movq {{[0-9]+}}(%rsp), %rcx ## 8-byte Reload 61; CHECK-NEXT: callq _check_mask16 62; CHECK-NEXT: movl %eax, (%rsp) ## 4-byte Spill 63; CHECK-NEXT: addq $56, %rsp 64; CHECK-NEXT: retq 65 %d2 = bitcast <2 x i64> %a to <8 x i16> 66 %m2 = call i8 @llvm.x86.avx512.cvtw2mask.128(<8 x i16> %d2) 67 %conv7 = zext i8 %m2 to i16 68 %call9 = call i64 @calc_expected_mask_val(i8* %arraydecay, i32 2, i32 8) 69 %conv10 = trunc i64 %call9 to i16 70 %call12 = call i32 @check_mask16(i16 zeroext %conv7, i16 zeroext %conv10, i8* %fname, i8* %arraydecay) 71 %d3 = bitcast <2 x i64> %a to <4 x i32> 72 %m3 = call i8 @llvm.x86.avx512.cvtd2mask.128(<4 x i32> %d3) 73 %conv14 = zext i8 %m3 to i16 74 %call16 = call i64 @calc_expected_mask_val(i8* %arraydecay, i32 4, i32 4) 75 %conv17 = trunc i64 %call16 to i16 76 %call19 = call i32 @check_mask16(i16 zeroext %conv14, i16 zeroext %conv17, i8* %fname, i8* %arraydecay) 77 ret void 78} 79 80; Function Attrs: nounwind readnone 81declare i8 @llvm.x86.avx512.cvtw2mask.128(<8 x i16>) 82 83; Function Attrs: nounwind readnone 84declare i8 @llvm.x86.avx512.cvtd2mask.128(<4 x i32>) 85 86