[clang] [compiler-rt] [llvm] [PGO][AMDGPU] Add basic HIP offload PGO support (PR #177665)

[Yaxun Liu via cfe-commits]

================
@@ -2179,3 +2217,26 @@ void createProfileSamplingVar(Module &M) {
   appendToCompilerUsed(M, SamplingVar);
 }
 } // namespace llvm
+
+namespace {
+
+// For GPU targets: Allocate contiguous arrays for all profile data.
+// This solves the linker reordering problem by using ONE symbol per section
+// type, so there's nothing for the linker to reorder.
+StructType *InstrLowerer::getProfileDataTy() {
----------------
yxsamliu wrote:

will do

https://github.com/llvm/llvm-project/pull/177665