[llvm] [VectorCombine] foldPermuteOfBinops - support multi-use binary ops and operands in shuffle folding (PR #173153)

[Simon Pilgrim via llvm-commits]

================
@@ -64,14 +64,15 @@ define <4 x float> @fadd_v4f32_mixed_types(<4 x float> %a0) {
   ret <4 x float> %post
 }
 
-; Negative test - multiple use of fadd
 define <4 x double> @fadd_v4f64_multiuse_op(<4 x double> %a, <4 x double> %b) {
 ; CHECK-LABEL: define <4 x double> @fadd_v4f64_multiuse_op(
 ; CHECK-SAME: <4 x double> [[A:%.*]], <4 x double> [[B:%.*]]) #[[ATTR0]] {
 ; CHECK-NEXT:    [[A1:%.*]] = shufflevector <4 x double> [[A]], <4 x double> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
 ; CHECK-NEXT:    [[B1:%.*]] = shufflevector <4 x double> [[B]], <4 x double> poison, <4 x i32> <i32 1, i32 0, i32 1, i32 0>
 ; CHECK-NEXT:    [[OP:%.*]] = fadd <4 x double> [[A1]], [[B1]]
-; CHECK-NEXT:    [[POST:%.*]] = shufflevector <4 x double> [[OP]], <4 x double> poison, <4 x i32> <i32 1, i32 0, i32 3, i32 2>
+; CHECK-NEXT:    [[TMP1:%.*]] = shufflevector <4 x double> [[A]], <4 x double> poison, <4 x i32> <i32 2, i32 3, i32 0, i32 1>
+; CHECK-NEXT:    [[TMP2:%.*]] = shufflevector <4 x double> [[B]], <4 x double> poison, <4 x i32> <i32 0, i32 1, i32 0, i32 1>
+; CHECK-NEXT:    [[POST:%.*]] = fadd <4 x double> [[TMP1]], [[TMP2]]
 ; CHECK-NEXT:    call void @use_v4f64(<4 x double> [[OP]])
----------------
RKSimon wrote:

This wasn't expected - please can you show the cost breakdown from `opt -passes=vector-combine -debug`?

https://github.com/llvm/llvm-project/pull/173153