xref: /llvm-project/llvm/test/CodeGen/X86/avx512-mask-zext-bugfix.ll (revision a74c5707be279bdddb51fe49aa9383c1ddc99fbe)
1; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
2; RUN: llc < %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(ptr %valp, i32 %el_size, i32 %length)
13; Function Attrs: nounwind uwtable
14declare i32 @check_mask16(i16 zeroext %res_mask, i16 zeroext %exp_mask, ptr %fname, ptr %input)
15
16; Function Attrs: nounwind uwtable
17define 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, %rsp
21; CHECK-NEXT:    .cfi_def_cfa_offset 64
22; CHECK-NEXT:    movq %rcx, {{[-0-9]+}}(%r{{[sb]}}p) ## 8-byte Spill
23; CHECK-NEXT:    movq %rdx, %rdi
24; CHECK-NEXT:    movq %rdi, {{[-0-9]+}}(%r{{[sb]}}p) ## 8-byte Spill
25; CHECK-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) ## 16-byte Spill
26; CHECK-NEXT:    vpmovw2m %xmm0, %k0
27; CHECK-NEXT:    kmovw %k0, {{[-0-9]+}}(%r{{[sb]}}p) ## 2-byte Spill
28; CHECK-NEXT:    movl $2, %esi
29; CHECK-NEXT:    movl $8, %edx
30; CHECK-NEXT:    callq _calc_expected_mask_val
31; CHECK-NEXT:    kmovw {{[-0-9]+}}(%r{{[sb]}}p), %k0 ## 2-byte Reload
32; CHECK-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx ## 8-byte Reload
33; CHECK-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %rcx ## 8-byte Reload
34; CHECK-NEXT:    ## kill: def $eax killed $eax killed $rax
35; CHECK-NEXT:    ## kill: def $ax killed $ax killed $eax
36; CHECK-NEXT:    movzwl %ax, %esi
37; CHECK-NEXT:    kmovb %k0, %edi
38; CHECK-NEXT:    callq _check_mask16
39; CHECK-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 ## 16-byte Reload
40; CHECK-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %rdi ## 8-byte Reload
41; CHECK-NEXT:    vpmovd2m %xmm0, %k0
42; CHECK-NEXT:    ## kill: def $k1 killed $k0
43; CHECK-NEXT:    kmovd %k0, %eax
44; CHECK-NEXT:    ## kill: def $al killed $al killed $eax
45; CHECK-NEXT:    movzbl %al, %eax
46; CHECK-NEXT:    ## kill: def $ax killed $ax killed $eax
47; CHECK-NEXT:    movw %ax, {{[-0-9]+}}(%r{{[sb]}}p) ## 2-byte Spill
48; CHECK-NEXT:    movl $4, %edx
49; CHECK-NEXT:    movl %edx, %esi
50; CHECK-NEXT:    callq _calc_expected_mask_val
51; CHECK-NEXT:    movw {{[-0-9]+}}(%r{{[sb]}}p), %si ## 2-byte Reload
52; CHECK-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx ## 8-byte Reload
53; CHECK-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %rcx ## 8-byte Reload
54; CHECK-NEXT:    ## kill: def $ax killed $ax killed $rax
55; CHECK-NEXT:    movzwl %si, %edi
56; CHECK-NEXT:    movzwl %ax, %esi
57; CHECK-NEXT:    callq _check_mask16
58; CHECK-NEXT:    addq $56, %rsp
59; CHECK-NEXT:    retq
60  %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 i16
63  %call9 = call i64 @calc_expected_mask_val(ptr %arraydecay, i32 2, i32 8)
64  %conv10 = trunc i64 %call9 to i16
65  %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 i16
69  %call16 = call i64 @calc_expected_mask_val(ptr %arraydecay, i32 4, i32 4)
70  %conv17 = trunc i64 %call16 to i16
71  %call19 = call i32 @check_mask16(i16 zeroext %conv14, i16 zeroext %conv17, ptr %fname, ptr %arraydecay)
72  ret void
73}
74
75; Function Attrs: nounwind readnone
76declare i8 @llvm.x86.avx512.cvtw2mask.128(<8 x i16>)
77
78; Function Attrs: nounwind readnone
79declare i8 @llvm.x86.avx512.cvtd2mask.128(<4 x i32>)
80
81