[llvm-branch-commits] [llvm] [AMDGPU] Allocate and reserve registers for address space 13 objects (PR #214248)

Carl Ritson via llvm-branch-commits llvm-branch-commits at lists.llvm.org
Mon Nov 9 02:06:41 PST 2026


================
@@ -1622,9 +1623,12 @@ void AMDGPUPassConfig::addIRPasses() {
 
   addPass(createAtomicExpandLegacyPass());
 
-  if (TM.getOptLevel() > CodeGenOptLevel::None) {
-    addPass(createAMDGPUPromoteAlloca());
+  // With optimizations enabled, do the full promotion of allocas. Without
+  // optimizations, only allocate existing addrspace(13) allocas, which is
----------------
perlfu wrote:

Should probably avoid explicit `addrspace` number as this _could_ change.

https://github.com/llvm/llvm-project/pull/214248


More information about the llvm-branch-commits mailing list