[llvm] [VPlan] Introduce explicit ExtractFromEnd recipes for live-outs. (PR #100658)

via llvm-commits llvm-commits at lists.llvm.org
Wed Aug 21 02:25:52 PDT 2024


================
@@ -8538,9 +8540,8 @@ static void addUsersInExitBlock(
   // and there is nothing to fix from vector loop; phis should have incoming
   // from scalar loop only.
   if (MiddleVPBB->getNumSuccessors() != 2)
----------------
ayalz wrote:

So non-empty exitingValuesToFix implies two successors; having exit block as the only successor is not considered.

Looking forward following our roadmap, initial VPlan should probably always start with both successors, and optimized later to have a single successor if scalar epilog is required or redundant.

Does empty exitingValuesToFix imply no edge from middle block to exit block, i.e., can such an edge exist w/o any lcssa phi's to fix?

https://github.com/llvm/llvm-project/pull/100658


More information about the llvm-commits mailing list