[llvm] [AMDGPU] Enable unaligned scratch accesses (PR #110219)

Matt Arsenault via llvm-commits llvm-commits at lists.llvm.org
Mon Sep 30 05:13:03 PDT 2024


================
@@ -387,8 +387,9 @@ bool GCNTTIImpl::isLegalToVectorizeMemChain(unsigned ChainSizeInBytes,
   // them later if they may access private memory. We don't have enough context
   // here, and legalization can handle it.
   if (AddrSpace == AMDGPUAS::PRIVATE_ADDRESS) {
-    return (Alignment >= 4 || ST->hasUnalignedScratchAccess()) &&
-      ChainSizeInBytes <= ST->getMaxPrivateElementSize();
+    return (Alignment >= 4 || ST->hasUnalignedScratchAccessEnabled() ||
+            ST->enableFlatScratch()) &&
----------------
arsenm wrote:

Same here 

https://github.com/llvm/llvm-project/pull/110219


More information about the llvm-commits mailing list