================
@@ -1079,4 +1079,107 @@ bool DiagnoseUnguardedBuiltins::VisitCallExpr(CallExpr
*CE) {
void SemaAMDGPU::DiagnoseUnguardedBuiltinUsage(FunctionDecl *FD) {
DiagnoseUnguardedBuiltins(SemaRef).IssueDiagnostics(FD->getBody());
}
+
+bool SemaAMDGPU::checkAMDGPUTypeSupport(QualType Ty, SourceLocation Loc) {
+ ASTContext &Ctx = getASTContext();
+ llvm::Triple TT = Ctx.getTargetInfo().getTriple();
+ const Type *BaseTy = Ty->getPointeeOrArrayElementType();
+
+ if (Ctx.getTargetInfo().getTriple().isSPIRV()) {
+ // The AMDGPU named barrier type requires special handling in the back-end
+ // and is not supported for SPIR-V
+ if (BaseTy->isAMDGPUNamedBarrierType()) {
----------------
yxsamliu wrote:
Could this also reject named barrier wrappers? The raw type is diagnosed now,
but CodeGen treats wrappers the same way via
`isAMDGPUNamedBarrierTypeOrWrapper()`. A wrapper global on `spirv64-amd-amdhsa`
could still reach the unsupported SPIR-V lowering path.
https://github.com/llvm/llvm-project/pull/207687
_______________________________________________
cfe-commits mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
