[cfe-commits] cfe-commits Digest, Vol 50, Issue 74

[Anton Lokhmotov]

> Updated patch attached.  I'll commit in a week unless there are
> any objections.

Hi Peter,

+    SC_OpenCLWorkGroupLocal,
Do you intend to introduce a storage class specifier for variables declared
in the constant address space?  If yes, I think it would be better to use
SC_OpenCLLocal and SC_OpenCLConstant as the names.  If not, why not?

+  case SC_OpenCLWorkGroupLocal: return "work-group-local"; break;
+  case SC_PrivateExtern:        return "__private_extern__"; break; 
> Because using this "storage class specifier" in source code would be a
> mistake, I deliberately chose a spelling that would not lex to a single
> token, but that would still be useful for AST dumper tools.
Perhaps you could use "__work_group_local__" to be consistent with
SC_PrivateExtern?

+  CGOpenCLRuntime.cpp
Could you please provide an overview of this functionality?  How do you plan
to develop it.  How one would use it?

+// RUN: %clang_cc1 %s -ffake-address-space-map -emit-llvm -o - | FileCheck
%s
Why do you use the fake address map, not the CL one?

Many thanks and best regards,
Anton.

P.S. I nearly missed this patch of yours, I as normally only read cfe-dev.
I appreciate, however, that sending patches to cfe-commits is more
appropriate.  Could you please cc me on submissions to cfe-commits?