[clang] [llvm] [Clang][SYCL] Define Intel IGCA preprocessor macro definitions for SYCL during device compilation (PR #227877)

Ian Li via cfe-commits cfe-commits at lists.llvm.org
Thu Nov 5 12:02:38 PST 2026


================
@@ -326,6 +333,16 @@ class LLVM_LIBRARY_VISIBILITY BaseSPIRVTargetInfo : public BaseSPIRTargetInfo {
 
   void getTargetDefines(const LangOptions &Opts,
                         MacroBuilder &Builder) const override;
+
+protected:
+  bool isSupportedOffloadArch(OffloadArch Arch) const override {
+    // Chipstar, vulkan, and amdhsa triples shouldn't accept a -target-cpu.
+    if (getTriple().getOS() != llvm::Triple::UnknownOS)
+      return false;
+
+    // Valid offload archs for SPIRV:
+    return Arch.isGeneric() || Arch.isIntel();
+  }
----------------
ianayl wrote:

Instead of making offload targets, i.e. SPIRV accept all parseable offload architectures, I chose to restrict to only the following offload archs. Please let me know if my logic here is wrong.

https://github.com/llvm/llvm-project/pull/227877


More information about the cfe-commits mailing list