1 files changed, 62 insertions, 7 deletions
diff --git a/llvm/test/CodeGen/X86/extractelement-load.ll b/llvm/test/CodeGen/X86/extractelement-load.ll
index f2d07b1ee1a..76c563e728e 100644
--- a/llvm/test/CodeGen/X86/extractelement-load.ll
+++ b/llvm/test/CodeGen/X86/extractelement-load.ll
@@ -9,13 +9,20 @@ define i32 @t(<2 x i64>* %val) nounwind  {
 ; X32-SSE2-LABEL: t:
 ; X32-SSE2:       # %bb.0:
 ; X32-SSE2-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; X32-SSE2-NEXT:    movl 8(%eax), %eax
+; X32-SSE2-NEXT:    pshufd {{.*#+}} xmm0 = mem[2,3,0,1]
+; X32-SSE2-NEXT:    movd %xmm0, %eax
 ; X32-SSE2-NEXT:    retl
 ;
-; X64-LABEL: t:
-; X64:       # %bb.0:
-; X64-NEXT:    movl 8(%rdi), %eax
-; X64-NEXT:    retq
+; X64-SSSE3-LABEL: t:
+; X64-SSSE3:       # %bb.0:
+; X64-SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = mem[2,3,0,1]
+; X64-SSSE3-NEXT:    movd %xmm0, %eax
+; X64-SSSE3-NEXT:    retq
+;
+; X64-AVX-LABEL: t:
+; X64-AVX:       # %bb.0:
+; X64-AVX-NEXT:    movl 8(%rdi), %eax
+; X64-AVX-NEXT:    retq
   %tmp2 = load <2 x i64>, <2 x i64>* %val, align 16		; <<2 x i64>> [#uses=1]
   %tmp3 = bitcast <2 x i64> %tmp2 to <4 x i32>		; <<4 x i32>> [#uses=1]
   %tmp4 = extractelement <4 x i32> %tmp3, i32 2		; <i32> [#uses=1]
@@ -127,7 +134,8 @@ define float @t6(<8 x float> *%a0) {
 ; X32-SSE2-NEXT:    pushl %eax
 ; X32-SSE2-NEXT:    .cfi_def_cfa_offset 8
 ; X32-SSE2-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; X32-SSE2-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X32-SSE2-NEXT:    movaps (%eax), %xmm0
+; X32-SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1,2,3]
 ; X32-SSE2-NEXT:    xorps %xmm1, %xmm1
 ; X32-SSE2-NEXT:    cmpeqss %xmm0, %xmm1
 ; X32-SSE2-NEXT:    movss {{.*#+}} xmm2 = mem[0],zero,zero,zero
@@ -142,7 +150,7 @@ define float @t6(<8 x float> *%a0) {
 ;
 ; X64-SSSE3-LABEL: t6:
 ; X64-SSSE3:       # %bb.0:
-; X64-SSSE3-NEXT:    movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
+; X64-SSSE3-NEXT:    movshdup {{.*#+}} xmm1 = mem[1,1,3,3]
 ; X64-SSSE3-NEXT:    xorps %xmm0, %xmm0
 ; X64-SSSE3-NEXT:    cmpeqss %xmm1, %xmm0
 ; X64-SSSE3-NEXT:    movss {{.*#+}} xmm2 = mem[0],zero,zero,zero
@@ -165,3 +173,50 @@ define float @t6(<8 x float> *%a0) {
   %cond = select i1 %cmp, float 1.000000e+00, float %vecext
   ret float %cond
 }
+
+define void @PR43971(<8 x float> *%a0, float *%a1) {
+; X32-SSE2-LABEL: PR43971:
+; X32-SSE2:       # %bb.0: # %entry
+; X32-SSE2-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X32-SSE2-NEXT:    movl {{[0-9]+}}(%esp), %ecx
+; X32-SSE2-NEXT:    movaps 16(%ecx), %xmm0
+; X32-SSE2-NEXT:    movhlps {{.*#+}} xmm0 = xmm0[1,1]
+; X32-SSE2-NEXT:    xorps %xmm1, %xmm1
+; X32-SSE2-NEXT:    cmpltss %xmm0, %xmm1
+; X32-SSE2-NEXT:    movss {{.*#+}} xmm2 = mem[0],zero,zero,zero
+; X32-SSE2-NEXT:    andps %xmm1, %xmm2
+; X32-SSE2-NEXT:    andnps %xmm0, %xmm1
+; X32-SSE2-NEXT:    orps %xmm2, %xmm1
+; X32-SSE2-NEXT:    movss %xmm1, (%eax)
+; X32-SSE2-NEXT:    retl
+;
+; X64-SSSE3-LABEL: PR43971:
+; X64-SSSE3:       # %bb.0: # %entry
+; X64-SSSE3-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X64-SSSE3-NEXT:    xorps %xmm1, %xmm1
+; X64-SSSE3-NEXT:    cmpltss %xmm0, %xmm1
+; X64-SSSE3-NEXT:    movss {{.*#+}} xmm2 = mem[0],zero,zero,zero
+; X64-SSSE3-NEXT:    andps %xmm1, %xmm2
+; X64-SSSE3-NEXT:    andnps %xmm0, %xmm1
+; X64-SSSE3-NEXT:    orps %xmm2, %xmm1
+; X64-SSSE3-NEXT:    movss %xmm1, (%rsi)
+; X64-SSSE3-NEXT:    retq
+;
+; X64-AVX-LABEL: PR43971:
+; X64-AVX:       # %bb.0: # %entry
+; X64-AVX-NEXT:    vpermilpd {{.*#+}} xmm0 = mem[1,0]
+; X64-AVX-NEXT:    vxorps %xmm1, %xmm1, %xmm1
+; X64-AVX-NEXT:    vcmpltss %xmm0, %xmm1, %xmm1
+; X64-AVX-NEXT:    vmovss {{.*#+}} xmm2 = mem[0],zero,zero,zero
+; X64-AVX-NEXT:    vblendvps %xmm1, %xmm2, %xmm0, %xmm0
+; X64-AVX-NEXT:    vmovss %xmm0, (%rsi)
+; X64-AVX-NEXT:    retq
+entry:
+  %0 = load <8 x float>, <8 x float>* %a0, align 32
+  %vecext = extractelement <8 x float> %0, i32 6
+  %cmp = fcmp ogt float %vecext, 0.000000e+00
+  %1 = load float, float* %a1, align 4
+  %cond = select i1 %cmp, float %1, float %vecext
+  store float %cond, float* %a1, align 4
+  ret void
+}