[PATCH] D125712: [SLP][X86] Improve reordering to consider alternate instruction bundles

[Simon Pilgrim via Phabricator via llvm-commits]

RKSimon added a comment.

Don't we need to account for cost somehow?



================
Comment at: llvm/lib/Target/X86/X86TargetTransformInfo.cpp:5331
+    return (NumElements == 4 && ST->hasSSE3()) ||
+           (NumElements == 8 && ST->hasAVX());
+  if (ElemTy->isDoubleTy())
----------------
The AVX case is unnecessary as any NumElements % 4 == 0 case will be supported as multiple ADDSUB ops.


================
Comment at: llvm/lib/Target/X86/X86TargetTransformInfo.cpp:5334
+    return (NumElements == 2 && ST->hasSSE3()) ||
+           (NumElements == 4 && ST->hasAVX());
+  return false;
----------------
Same here NumElements % 2 == 0 is enough


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D125712/new/

https://reviews.llvm.org/D125712