Anastasia added inline comments.

================
Comment at: clang/include/clang/Basic/DiagnosticSemaKinds.td:10092
     "use of type 'double' requires %select{cl_khr_fp64|cl_khr_fp64 and 
__opencl_c_fp64}0 support">;
+def err_opencl_3d_image_writes_requires_extensions : Error<
+    "use of type '__write_only image3d_t' requires 
%select{cl_khr_3d_image_writes|cl_khr_3d_image_writes and "
----------------
it seems like we keep expanding similar diagnostics which flows against the 
design policy. Could you try to unify with 
`err_opencl_double_requires_extension` or even with 
`err_opencl_requires_extension`?

Perhaps we could create a mechanism to print custom strings for extensions that 
alias features or it can be done in the follow-up patches too?


================
Comment at: clang/lib/Sema/SemaType.cpp:1732-1734
+    // __opencl_c_3d_image_writes feature. OpenCL C v3.0 API s4.2 - For devices
+    // that support OpenCL 3.0, cl_khr_3d_image_writes must be returned when 
and
+    // only when the optional feature is supported
----------------
azabaznov wrote:
> Note this. Is it OK to refer to API spec here?
I am guessing you want to explain that corresponding features and extensions 
must be in sync?

It makes sense, but it feels like we are missing something in the kernel 
language spec then...


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D106260/new/

https://reviews.llvm.org/D106260

_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

Reply via email to