[clang] [SYCL] Basic code generation for SYCL kernel caller offload entry point functions. (PR #133030)

[Erich Keane via cfe-commits]

================
@@ -14794,9 +14803,36 @@ void ASTContext::getFunctionFeatureMap(llvm::StringMap<bool> &FeatureMap,
   }
 }
 
-static SYCLKernelInfo BuildSYCLKernelInfo(CanQualType KernelNameType,
+static SYCLKernelInfo BuildSYCLKernelInfo(ASTContext &Context,
+                                          CanQualType KernelNameType,
                                           const FunctionDecl *FD) {
-  return {KernelNameType, FD};
+  // Host and device compilation may use different ABIs and different ABIs
+  // may allocate name mangling discriminators differently. A discriminator
+  // override is used to ensure consistent discriminator allocation across
+  // host and device compilation.
+  auto DeviceDiscriminatorOverrider =
+      [](ASTContext &Ctx, const NamedDecl *ND) -> std::optional<unsigned> {
+    if (const auto *RD = dyn_cast<CXXRecordDecl>(ND))
+      if (RD->isLambda())
----------------
erichkeane wrote:

My concern here is: What happens if a `genericLambda` gets here, which I believe will have an unset 'mangling-id'.   I'd VERY much like to not let this get pushed to the next one, in part since the ways forward aren't a particularly sizable amount of work.

I see a few ways forward:
1- Create an example that shows the mangler properly handling the generic-lambda.  Note you might have to do some work to get here, as getting generic-lambdas mangled often requires some surprising things to make happen.
2- Prove in some-way that generic-lambda cannot get here (which, would actually take quite a bit of convincing, dependent types find their way to sneak into the mangler all the time unfortunately, particularly with SFINAE and `diagnose_if`/etc).
2- 'assert(!isGenericLambda() && "dont' know how to handle generic lambdas yet");`



https://github.com/llvm/llvm-project/pull/133030