[llvm] [X86] X86FixupInstTuning - fold BLENDPS -> MOVSD (PR #144029)

[Phoebe Wang via llvm-commits]

================
@@ -563,20 +543,25 @@ define <2 x i64> @test21(<2 x i64> %a, <2 x i64> %b) {
 ; bitcast to use the mask-or blend combine.
 
 define <2 x double> @test22(<2 x double> %a0, <2 x double> %a1) {
-; SSE2-LABEL: test22:
-; SSE2:       # %bb.0:
-; SSE2-NEXT:    movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
-; SSE2-NEXT:    retq
+; SSE-LABEL: test22:
+; SSE:       # %bb.0:
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
+; SSE-NEXT:    retq
 ;
-; SSE4-LABEL: test22:
-; SSE4:       # %bb.0:
-; SSE4-NEXT:    blendps {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
-; SSE4-NEXT:    retq
+; AVX1-LABEL: test22:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
+; AVX1-NEXT:    retq
 ;
-; AVX-LABEL: test22:
-; AVX:       # %bb.0:
-; AVX-NEXT:    vblendps {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
-; AVX-NEXT:    retq
+; AVX2-LABEL: test22:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
+; AVX2-NEXT:    retq
+;
+; AVX512-LABEL: test22:
+; AVX512:       # %bb.0:
+; AVX512-NEXT:    vblendps {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
----------------
phoebewang wrote:

I see. I think it's better to keep it as is to show the coverage.

https://github.com/llvm/llvm-project/pull/144029