58 lines · plain
1; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py2; RUN: llc < %s -mtriple=i386-unknown -mattr=+sse2,-sse4.1 | FileCheck %s --check-prefix=X863; RUN: llc < %s -mtriple=x86_64-unknown -mattr=+sse2,-sse4.1 | FileCheck %s --check-prefix=X644 5define void @test(ptr %b, i16 %a0, i16 %a1, i16 %a2, i16 %a3, i16 %a4, i16 %a5, i16 %a6, i16 %a7) nounwind {6; X86-LABEL: test:7; X86: # %bb.0:8; X86-NEXT: movl {{[0-9]+}}(%esp), %eax9; X86-NEXT: movd {{.*#+}} xmm0 = mem[0],zero,zero,zero10; X86-NEXT: movd {{.*#+}} xmm1 = mem[0],zero,zero,zero11; X86-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]12; X86-NEXT: movd {{.*#+}} xmm0 = mem[0],zero,zero,zero13; X86-NEXT: movd {{.*#+}} xmm2 = mem[0],zero,zero,zero14; X86-NEXT: punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3]15; X86-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]16; X86-NEXT: movd {{.*#+}} xmm0 = mem[0],zero,zero,zero17; X86-NEXT: movd {{.*#+}} xmm1 = mem[0],zero,zero,zero18; X86-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]19; X86-NEXT: movd {{.*#+}} xmm0 = mem[0],zero,zero,zero20; X86-NEXT: movd {{.*#+}} xmm3 = mem[0],zero,zero,zero21; X86-NEXT: punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1],xmm3[2],xmm0[2],xmm3[3],xmm0[3]22; X86-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]23; X86-NEXT: punpcklqdq {{.*#+}} xmm3 = xmm3[0],xmm2[0]24; X86-NEXT: movdqa %xmm3, (%eax)25; X86-NEXT: retl26;27; X64-LABEL: test:28; X64: # %bb.0:29; X64-NEXT: movd {{.*#+}} xmm0 = mem[0],zero,zero,zero30; X64-NEXT: movd {{.*#+}} xmm1 = mem[0],zero,zero,zero31; X64-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]32; X64-NEXT: movd %r9d, %xmm033; X64-NEXT: movd {{.*#+}} xmm2 = mem[0],zero,zero,zero34; X64-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3]35; X64-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]36; X64-NEXT: movd %r8d, %xmm137; X64-NEXT: movd %ecx, %xmm238; X64-NEXT: punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]39; X64-NEXT: movd %edx, %xmm140; X64-NEXT: movd %esi, %xmm341; X64-NEXT: punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1],xmm3[2],xmm1[2],xmm3[3],xmm1[3]42; X64-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]43; X64-NEXT: punpcklqdq {{.*#+}} xmm3 = xmm3[0],xmm0[0]44; X64-NEXT: movdqa %xmm3, (%rdi)45; X64-NEXT: retq46 %tmp = insertelement <8 x i16> zeroinitializer, i16 %a0, i32 047 %tmp2 = insertelement <8 x i16> %tmp, i16 %a1, i32 148 %tmp4 = insertelement <8 x i16> %tmp2, i16 %a2, i32 249 %tmp6 = insertelement <8 x i16> %tmp4, i16 %a3, i32 350 %tmp8 = insertelement <8 x i16> %tmp6, i16 %a4, i32 451 %tmp10 = insertelement <8 x i16> %tmp8, i16 %a5, i32 552 %tmp12 = insertelement <8 x i16> %tmp10, i16 %a6, i32 653 %tmp14 = insertelement <8 x i16> %tmp12, i16 %a7, i32 754 store <8 x i16> %tmp14, ptr %b55 ret void56}57 58