[PATCH] D109594: [AMDGPU] Initialize LDS pointers after alloca, but before call.

[Matt Arsenault via Phabricator via llvm-commits]

arsenm added inline comments.


================
Comment at: llvm/lib/Target/AMDGPU/AMDGPUReplaceLDSUseWithPointer.cpp:195
-    // Split entry basic block of kernel K.
-    auto *EI = &(*(K->getEntryBlock().getFirstInsertionPt()));
-    IRBuilder<> Builder(EI);
----------------
All this patch should do is skip the insertion point past the allocas clustered at the start of the entry block. If there are further allocas in the program which may be broken, you should not be concerning yourself with them. I don't want to add a workaround for non-entry allocas here


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D109594/new/

https://reviews.llvm.org/D109594