[PGO][Offload] Allow PGO flags to be used on GPU targets (#94268)

This pull request is the third part of an ongoing effort to extends PGO
instrumentation to GPU device code and depends on
https://github.com/llvm/llvm-project/pull/93365. This PR makes the
following changes:

- Allows PGO flags to be supplied to GPU targets
- Pulls version global from device
- Modifies `__llvm_write_custom_profile` and `lprofWriteDataImpl` to
allow the PGO version to be overridden

1 files changed, 3 insertions, 0 deletions

diff --git a/clang/lib/CodeGen/CodeGenPGO.cpp b/clang/lib/CodeGen/CodeGenPGO.cpp
index 7923738..ff8b133 100644
--- a/clang/lib/CodeGen/CodeGenPGO.cpp
+++ b/clang/lib/CodeGen/CodeGenPGO.cpp
@@ -1357,6 +1357,9 @@ void CodeGenPGO::setProfileVersion(llvm::Module &M) {
 
     IRLevelVersionVariable->setVisibility(llvm::GlobalValue::HiddenVisibility);
     llvm::Triple TT(M.getTargetTriple());
+    if (TT.isAMDGPU() || TT.isNVPTX())
+      IRLevelVersionVariable->setVisibility(
+          llvm::GlobalValue::ProtectedVisibility);
     if (TT.supportsCOMDAT()) {
       IRLevelVersionVariable->setLinkage(llvm::GlobalValue::ExternalLinkage);
       IRLevelVersionVariable->setComdat(M.getOrInsertComdat(VarName));