[PATCH] D101477: [SLP]Fix the crash on cost calculation if non-compatible vectors shuffled.

Sanjay Patel via Phabricator via llvm-commits llvm-commits at lists.llvm.org
Fri Apr 30 08:02:06 PDT 2021


spatel added inline comments.


================
Comment at: llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp:3600-3601
+        } else {
+          auto *SubVT =
+              FixedVectorType::get(VecTy->getElementType(), EENumElts - Idx);
+          Cost +=
----------------
This could use a comment or a code example to explain. I can't tell just from looking at this what the new SubVT represents.


================
Comment at: llvm/test/Transforms/SLPVectorizer/AMDGPU/add_sub_sat.ll:226
 
-define <3 x i16> @uadd_sat_v3i16(<3 x i16> %arg0, <3 x i16> %arg1) {
-; GFX7-LABEL: @uadd_sat_v3i16(
+define <2 x i16> @uadd_sat_v9i16_combine_vi16(<9 x i16> %arg0, <9 x i16> %arg1) {
+; GFX7-LABEL: @uadd_sat_v9i16_combine_vi16(
----------------
Add a new test instead of changing the existing test?
IIUC, it's not the "GFX7" run that is crashing, but we don't have any check lines for "GFX8"? It would be better to just create a new file without "-instcombine", so we can see exactly what SLP will create on this test.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D101477/new/

https://reviews.llvm.org/D101477



More information about the llvm-commits mailing list