[PATCH] D57779: [SLP] Add support for throttling.

[Dinar Temirbulatov via Phabricator via llvm-commits]

dtemirbulatov marked 2 inline comments as done.
dtemirbulatov added inline comments.


================
Comment at: test/Transforms/SLPVectorizer/AArch64/horizontal.ll:58
+; CHECK-NEXT:    [[TMP12:%.*]] = insertelement <2 x i32> [[TMP11]], i32 1, i32 1
+; CHECK-NEXT:    [[TMP13]] = add nsw <2 x i32> [[TMP12]], [[TMP0]]
+; CHECK-NEXT:    [[TMP14:%.*]] = extractelement <2 x i32> [[TMP13]], i32 1
----------------
ABataev wrote:
> Does not seem to me we should vectorize this. 2 add ops have different attributes: one has just `nsw`, another one `nuw nsw`.
> Plus, 2 inserts+vec add+2 extracts does not look more beneficial than 2 scalar adds
This is correct,  the vector add should simply take the smallest common subset of flags. And the test is used with "-slp-threshold=-6" flag that it why we vectorized those two add operations.


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D57779/new/

https://reviews.llvm.org/D57779