1; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py 2; RUN: llc < %s -mtriple=x86_64-unknown-linux-gnu -mattr=avx512vl,avx512bw,avx512dq | FileCheck %s 3 4define void @EntryModule(i8** %buffer_table) { 5; CHECK-LABEL: EntryModule: 6; CHECK: # %bb.0: # %entry 7; CHECK-NEXT: vxorps %xmm0, %xmm0, %xmm0 8; CHECK-NEXT: movq (%rdi), %rax 9; CHECK-NEXT: movq 24(%rdi), %rcx 10; CHECK-NEXT: vcmpneqps (%rax), %ymm0, %ymm0 11; CHECK-NEXT: vpsrld $31, %xmm0, %xmm1 12; CHECK-NEXT: vpshufd {{.*#+}} xmm2 = xmm1[1,1,1,1] 13; CHECK-NEXT: vpshufd {{.*#+}} xmm3 = xmm1[2,3,2,3] 14; CHECK-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[3,3,3,3] 15; CHECK-NEXT: vpaddd %xmm1, %xmm3, %xmm1 16; CHECK-NEXT: vpsubd %xmm0, %xmm2, %xmm0 17; CHECK-NEXT: vpaddd %xmm1, %xmm0, %xmm0 18; CHECK-NEXT: vmovd %xmm0, (%rcx) 19; CHECK-NEXT: vzeroupper 20; CHECK-NEXT: retq 21entry: 22 %i = bitcast i8** %buffer_table to <8 x float>** 23 %i1 = load <8 x float>*, <8 x float>** %i, align 8 24 %i6 = load <8 x float>, <8 x float>* %i1, align 16 25 %i7 = fcmp une <8 x float> %i6, zeroinitializer 26 %i8 = zext <8 x i1> %i7 to <8 x i32> 27 %i18 = getelementptr inbounds i8*, i8** %buffer_table, i64 3 28 %i19 = load i8*, i8** %i18, align 8 29 %shift = shufflevector <8 x i32> %i8, <8 x i32> undef, <8 x i32> <i32 1, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef> 30 %i20 = add nuw nsw <8 x i32> %shift, %i8 31 %shift13 = shufflevector <8 x i32> %i8, <8 x i32> undef, <8 x i32> <i32 2, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef> 32 %i21 = add nuw nsw <8 x i32> %i20, %shift13 33 %shift14 = shufflevector <8 x i32> %i8, <8 x i32> undef, <8 x i32> <i32 3, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef> 34 %i22 = add nuw nsw <8 x i32> %i21, %shift14 35 %i23 = extractelement <8 x i32> %i22, i32 0 36 %i24 = bitcast i8* %i19 to i32* 37 store i32 %i23, i32* %i24, align 8 38 ret void 39} 40