[PATCH] D101555: [SLP]Improve handling of compensate external uses cost.

[Simon Pilgrim via Phabricator via llvm-commits]

RKSimon added inline comments.


================
Comment at: llvm/test/Transforms/SLPVectorizer/X86/alternate-cast-inseltpoison.ll:72
 ; SSE-NEXT:    [[AB6:%.*]] = fptoui float [[A6]] to i32
 ; SSE-NEXT:    [[AB7:%.*]] = fptoui float [[A7]] to i32
 ; SSE-NEXT:    [[R31:%.*]] = shufflevector <4 x i32> [[TMP2]], <4 x i32> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef>
----------------
RKSimon wrote:
> RKSimon wrote:
> > Still not performing fptoui on the entire <4 x i32>? 
> This is purely a cost-model issue - fptoui for 2f32 is 8 but 4f32 is 18 (looks like the model assumes they scalarize which they don't) - these are really wrong, but shouldn't stop this patch.
@ABataev If you rebase this should be fixed after rGeb6429d0fb94fd467e03d229177ae6ff3a44e3cc + rG3ae7f7ae0a33961be48948205981aea91920d3aa


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D101555/new/

https://reviews.llvm.org/D101555