[PATCH] D107228: [TTI] Add basic SK_InsertSubvector shuffle mask recognition

[Alexey Bataev via Phabricator via llvm-commits]

ABataev added inline comments.


================
Comment at: llvm/test/Analysis/CostModel/X86/shuffle-insert_subvector.ll:261
+; AVX1-LABEL: 'test_vXf32'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of -1 for instruction: %src64_128 = shufflevector <2 x float> %src64, <2 x float> undef, <4 x i32> <i32 0, i32 1, i32 undef, i32 undef>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of -1 for instruction: %src64_256 = shufflevector <2 x float> %src64, <2 x float> undef, <8 x i32> <i32 0, i32 1, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
----------------
RKSimon wrote:
> ABataev wrote:
> > -1?
> This is an example of subvector widening that I was wondering whether we should just set cost = 0 - at the moment the shuffle mask matcher doesn't recognise it at all (hence the -1).
Yes, I think better to make it 0.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D107228/new/

https://reviews.llvm.org/D107228