| Igor Breger | a2f8ca9 | 2016-09-05 08:26:51 +0000 | [diff] [blame] | 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' |
| 5 | source_filename = "mask_set.c" |
| 6 | target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128" |
| 7 | target triple = "x86_64-unknown-linux-gnu" |
| 8 | |
| 9 | declare void @llvm.dbg.declare(metadata, metadata, metadata) |
| 10 | |
| 11 | ; Function Attrs: nounwind uwtable |
| 12 | declare i64 @calc_expected_mask_val(i8* %valp, i32 %el_size, i32 %length) |
| 13 | ; Function Attrs: nounwind uwtable |
| 14 | declare i32 @check_mask16(i16 zeroext %res_mask, i16 zeroext %exp_mask, i8* %fname, i8* %input) |
| 15 | |
| 16 | ; Function Attrs: nounwind uwtable |
| 17 | define 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: callq _calc_expected_mask_val |
| 21 | ; CHECK-NEXT: movl %eax, %edx |
| 22 | ; CHECK-NEXT: movw %dx, %r9w |
| 23 | ; CHECK-NEXT: movzwl %r9w, %esi |
| 24 | ; CHECK-NEXT: kmovw {{[0-9]+}}(%rsp), %k0 ## 2-byte Reload |
| 25 | ; CHECK-NEXT: kmovb %k0, %edi |
| 26 | ; CHECK-NEXT: movq {{[0-9]+}}(%rsp), %rdx ## 8-byte Reload |
| 27 | ; CHECK-NEXT: movq {{[0-9]+}}(%rsp), %rcx ## 8-byte Reload |
| 28 | ; CHECK-NEXT: callq _check_mask16 |
| 29 | %d2 = bitcast <2 x i64> %a to <8 x i16> |
| 30 | %m2 = call i8 @llvm.x86.avx512.cvtw2mask.128(<8 x i16> %d2) |
| 31 | %conv7 = zext i8 %m2 to i16 |
| 32 | %call9 = call i64 @calc_expected_mask_val(i8* %arraydecay, i32 2, i32 8) |
| 33 | %conv10 = trunc i64 %call9 to i16 |
| 34 | %call12 = call i32 @check_mask16(i16 zeroext %conv7, i16 zeroext %conv10, i8* %fname, i8* %arraydecay) |
| 35 | %d3 = bitcast <2 x i64> %a to <4 x i32> |
| 36 | %m3 = call i8 @llvm.x86.avx512.cvtd2mask.128(<4 x i32> %d3) |
| 37 | %conv14 = zext i8 %m3 to i16 |
| 38 | %call16 = call i64 @calc_expected_mask_val(i8* %arraydecay, i32 4, i32 4) |
| 39 | %conv17 = trunc i64 %call16 to i16 |
| 40 | %call19 = call i32 @check_mask16(i16 zeroext %conv14, i16 zeroext %conv17, i8* %fname, i8* %arraydecay) |
| 41 | ret void |
| 42 | } |
| 43 | |
| 44 | ; Function Attrs: nounwind readnone |
| 45 | declare i8 @llvm.x86.avx512.cvtw2mask.128(<8 x i16>) |
| 46 | |
| 47 | ; Function Attrs: nounwind readnone |
| 48 | declare i8 @llvm.x86.avx512.cvtd2mask.128(<4 x i32>) |
| 49 | |