[clang] [llvm] [AMDGPU] modify named barrier builtins and intrinsics (PR #114550)

Gang Chen via cfe-commits cfe-commits at lists.llvm.org
Fri Nov 1 09:06:17 PDT 2024


================
@@ -920,6 +920,124 @@ class AMDGPULowerModuleLDS {
     return KernelToCreatedDynamicLDS;
   }
 
+  static GlobalVariable *uniquifyGVPerKernel(Module &M, GlobalVariable *GV,
----------------
cmc-rep wrote:

This is where the local pointer address is assigned for a named barrier before going into AMDGPUbackend

https://github.com/llvm/llvm-project/pull/114550


More information about the cfe-commits mailing list