Hi,
For what it's worth, I tested this patch with pocl and it made
no difference to the test suite (as expected), because we #define
the OpenCL address spaces to fixed id ones before clang at
the moment.
BR,
Pekka
On 02/05/2013 08:17 PM, Tanya Lattner wrote:
Ping.
Anyone opposed to this?
-Tanya
On Jan 30, 2013, at 2:41 PM, Tanya Lattner <[email protected]> wrote:
Attached is a patch to change how address spaces are mangled.
Currently, if you specify local, global, or constant OpenCL address spaces you
get some strange numbers printed in the mangled name which correspond to the
values in LangAs for opencl_constant, opencl_local, and opencl_global. I
believe it makes more sense to mangle based upon the real address space values
that are defined by the target. It makes it easier to understand what the
address spaces of the arguments to a function really are. I've included a test
case for both the OpenCL case and if someone used the attribute in C++ as well.
Please review.
Thanks,
Tanya
<addrSpaceMangle.patch>_______________________________________________
cfe-commits mailing list
[email protected]
http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
_______________________________________________
cfe-commits mailing list
[email protected]
http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
_______________________________________________
cfe-commits mailing list
[email protected]
http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
