42 lines · plain
1; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py2; RUN: llc -mtriple=x86_64-linux-gnu < %s | FileCheck %s3 4; Ensure vectorized FREMs are not widened/unrolled such that they get lowered5; into libcalls on undef elements.6 7define float @frem(<2 x float> %a0, <2 x float> %a1, <2 x float> %a2, ptr%p3) nounwind {8; CHECK-LABEL: frem:9; CHECK: # %bb.0:10; CHECK-NEXT: pushq %rbx11; CHECK-NEXT: subq $64, %rsp12; CHECK-NEXT: movq %rdi, %rbx13; CHECK-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill14; CHECK-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill15; CHECK-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill16; CHECK-NEXT: callq fmodf@PLT17; CHECK-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill18; CHECK-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload19; CHECK-NEXT: shufps {{.*#+}} xmm0 = xmm0[1,1,1,1]20; CHECK-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload21; CHECK-NEXT: shufps {{.*#+}} xmm1 = xmm1[1,1,1,1]22; CHECK-NEXT: callq fmodf@PLT23; CHECK-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload24; CHECK-NEXT: unpcklps {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]25; CHECK-NEXT: divps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload26; CHECK-NEXT: movaps %xmm1, %xmm027; CHECK-NEXT: shufps {{.*#+}} xmm0 = xmm0[1,1],xmm1[1,1]28; CHECK-NEXT: addss %xmm1, %xmm029; CHECK-NEXT: movlps %xmm1, (%rbx)30; CHECK-NEXT: addq $64, %rsp31; CHECK-NEXT: popq %rbx32; CHECK-NEXT: retq33 %frem = frem <2 x float> %a0, %a134 %fdiv = fdiv <2 x float> %frem, %a235 %ex0 = extractelement <2 x float> %fdiv, i32 036 %ex1 = extractelement <2 x float> %fdiv, i32 137 %res = fadd float %ex0, %ex138 store <2 x float> %fdiv, ptr%p339 ret float %res40}41 42