[clang] [SYCL] The sycl_kernel_entry_point attribute. (PR #111389)

Erich Keane via cfe-commits cfe-commits at lists.llvm.org
Mon Oct 7 08:43:21 PDT 2024


================
@@ -455,6 +455,64 @@ The SYCL kernel in the previous code sample meets these expectations.
   }];
 }
 
+def SYCLKernelEntryPointDocs : Documentation {
+  let Category = DocCatFunction;
+  let Content = [{
+The ``sycl_kernel_entry_point`` attribute specifies that a function definition
+defines a pattern for an offload kernel entry point function to be emitted when
+the source code is compiled with ``-fsycl`` for a device target. Such functions
+serve as the execution entry point for a SYCL run-time library to invoke a SYCL
+kernel on a device. The function's parameters define the parameters to the
+offload kernel.
+
+The attribute requires a single type argument that specifies a class type that
+meets the requirements for a SYCL kernel name as described in section 5.2,
+"Naming of kernels", of the SYCL 2020 specification. A unique kernel name type
+is required for each function declared with the attribute. The attribute may
+not first appear on a declaration that follows a definition of the function.
+
+The attribute appertains only to non-member functions and static member
+functions that meet the following requirements:
+
+- Has a ``void`` return type.
+- Is not a variadic function.
+- Is not a coroutine.
+- Is not defined as deleted or as defaulted.
+- Is not declared with the ``constexpr`` or ``consteval`` specifiers.
----------------
erichkeane wrote:

Why no constexpr?  I get no consteval, but constexpr seems reasonable, it should not do offload during constant evaluation (obviously), but that shouldn't prevent being marked as constexpr.  This is particularly problematic in cases of lambdas that end up being implicitly constexpr, which we might want to mark, but it won't be clear this is the case.

https://github.com/llvm/llvm-project/pull/111389


More information about the cfe-commits mailing list