[llvm] [AMDGPU] Price only the fmul that fuses into an fadd/fsub as free (PR #226009)

[Shilei Tian via llvm-commits]

================
@@ -0,0 +1,17 @@
+; RUN: opt -passes=slp-vectorizer -mtriple=amdgcn-amd-amdhsa -mcpu=gfx942 -pass-remarks=slp-vectorizer -disable-output < %s 2>&1 | FileCheck %s
+; RUN: opt -passes=slp-vectorizer -mtriple=amdgcn-amd-amdhsa -mcpu=gfx950 -pass-remarks=slp-vectorizer -disable-output < %s 2>&1 | FileCheck %s
----------------
shiltian wrote:

no need for multi-arch here

https://github.com/llvm/llvm-project/pull/226009