[llvm-branch-commits] [llvm] [LoopVectorize] Support vectorization of compressing patterns (PR #214491)

Benjamin Maxwell via llvm-branch-commits llvm-branch-commits at lists.llvm.org
Wed Sep 2 09:53:32 PDT 2026


================
@@ -21,6 +21,7 @@
 ; CHECK: VPlan for loop in 'foo' [[BEFORE_OR_AFTER]] VPlanTransforms::handleCountableEarlyExits
 ; CHECK: VPlan for loop in 'foo' [[BEFORE_OR_AFTER]] VPlanTransforms::createLoopRegions
 ; CHECK: VPlan for loop in 'foo' [[BEFORE_OR_AFTER]] VPlanTransforms::introduceMasksAndLinearize
+; CHECK: VPlan for loop in 'foo' [[BEFORE_OR_AFTER]] VPlanTransforms::handleCompressingPatterns
----------------
MacDue wrote:

If the load/store instructions became unmasked, currently we would crash (as a mask is expected). I think handling that in the transform would still be pretty easy (we could reject it and not vectorize, or create and all-true mask). I think both cases are simpler than duplicating the logic to determine in the mask will be the same from the predicator. 

https://github.com/llvm/llvm-project/pull/214491


More information about the llvm-branch-commits mailing list