81 lines · plain
1; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py2; RUN: llc < %s -O0 -mtriple=x86_64-apple-darwin -mcpu=skx | FileCheck %s3 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 uwtable12declare i64 @calc_expected_mask_val(ptr %valp, i32 %el_size, i32 %length)13; Function Attrs: nounwind uwtable14declare i32 @check_mask16(i16 zeroext %res_mask, i16 zeroext %exp_mask, ptr %fname, ptr %input)15 16; Function Attrs: nounwind uwtable17define void @test_xmm(i32 %shift, i32 %mulp, <2 x i64> %a,ptr %arraydecay,ptr %fname){18; CHECK-LABEL: test_xmm:19; CHECK: ## %bb.0:20; CHECK-NEXT: subq $56, %rsp21; CHECK-NEXT: .cfi_def_cfa_offset 6422; CHECK-NEXT: movq %rcx, {{[-0-9]+}}(%r{{[sb]}}p) ## 8-byte Spill23; CHECK-NEXT: movq %rdx, %rdi24; CHECK-NEXT: movq %rdi, {{[-0-9]+}}(%r{{[sb]}}p) ## 8-byte Spill25; CHECK-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) ## 16-byte Spill26; CHECK-NEXT: vpmovw2m %xmm0, %k027; CHECK-NEXT: kmovw %k0, {{[-0-9]+}}(%r{{[sb]}}p) ## 2-byte Spill28; CHECK-NEXT: movl $2, %esi29; CHECK-NEXT: movl $8, %edx30; CHECK-NEXT: callq _calc_expected_mask_val31; CHECK-NEXT: kmovw {{[-0-9]+}}(%r{{[sb]}}p), %k0 ## 2-byte Reload32; CHECK-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx ## 8-byte Reload33; CHECK-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rcx ## 8-byte Reload34; CHECK-NEXT: ## kill: def $eax killed $eax killed $rax35; CHECK-NEXT: ## kill: def $ax killed $ax killed $eax36; CHECK-NEXT: movzwl %ax, %esi37; CHECK-NEXT: kmovb %k0, %edi38; CHECK-NEXT: callq _check_mask1639; CHECK-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 ## 16-byte Reload40; CHECK-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdi ## 8-byte Reload41; CHECK-NEXT: vpmovd2m %xmm0, %k042; CHECK-NEXT: ## kill: def $k1 killed $k043; CHECK-NEXT: kmovd %k0, %eax44; CHECK-NEXT: ## kill: def $al killed $al killed $eax45; CHECK-NEXT: movzbl %al, %eax46; CHECK-NEXT: ## kill: def $ax killed $ax killed $eax47; CHECK-NEXT: movw %ax, {{[-0-9]+}}(%r{{[sb]}}p) ## 2-byte Spill48; CHECK-NEXT: movl $4, %edx49; CHECK-NEXT: movl %edx, %esi50; CHECK-NEXT: callq _calc_expected_mask_val51; CHECK-NEXT: movw {{[-0-9]+}}(%r{{[sb]}}p), %si ## 2-byte Reload52; CHECK-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx ## 8-byte Reload53; CHECK-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rcx ## 8-byte Reload54; CHECK-NEXT: ## kill: def $ax killed $ax killed $rax55; CHECK-NEXT: movzwl %si, %edi56; CHECK-NEXT: movzwl %ax, %esi57; CHECK-NEXT: callq _check_mask1658; CHECK-NEXT: addq $56, %rsp59; CHECK-NEXT: retq60 %d2 = bitcast <2 x i64> %a to <8 x i16>61 %m2 = call i8 @llvm.x86.avx512.cvtw2mask.128(<8 x i16> %d2)62 %conv7 = zext i8 %m2 to i1663 %call9 = call i64 @calc_expected_mask_val(ptr %arraydecay, i32 2, i32 8)64 %conv10 = trunc i64 %call9 to i1665 %call12 = call i32 @check_mask16(i16 zeroext %conv7, i16 zeroext %conv10, ptr %fname, ptr %arraydecay)66 %d3 = bitcast <2 x i64> %a to <4 x i32>67 %m3 = call i8 @llvm.x86.avx512.cvtd2mask.128(<4 x i32> %d3)68 %conv14 = zext i8 %m3 to i1669 %call16 = call i64 @calc_expected_mask_val(ptr %arraydecay, i32 4, i32 4)70 %conv17 = trunc i64 %call16 to i1671 %call19 = call i32 @check_mask16(i16 zeroext %conv14, i16 zeroext %conv17, ptr %fname, ptr %arraydecay)72 ret void73}74 75; Function Attrs: nounwind readnone76declare i8 @llvm.x86.avx512.cvtw2mask.128(<8 x i16>)77 78; Function Attrs: nounwind readnone79declare i8 @llvm.x86.avx512.cvtd2mask.128(<4 x i32>)80 81