OpenCL address space and mangling

[Tanya Lattner]

It doesn't make sense to me to not translate the address space to something meaningful (something other than a random number). Targets should define the address space map if they are supported.

What problem are you trying to solve?

-Tanya 


On Jul 18, 2013, at 10:11 AM, Michele Scandale <michele.scandale at gmail.com> wrote:

> Hello to everybody,
> 
> I've noticed a problem in the mangling of function names where address spaces
> are used. The problem is related to opencl and cuda address spaces: the target
> address space map is used to translate address space, but this conversion can
> produce the same mangling from different address space (on X86 all opencl
> address spaces are mapped to the address space zero). See attached example.
> 
> Commit r174688 introduced the usage of the target translation map. Reverting it
> seems be ok as a solution. Do you agree with this solution?
> 
> Thanks in advance.
> 
> Best regards.
> 
> Michele Scandale
> 
> 
> <test.cl>
> _______________________________________________
> cfe-commits mailing list
> cfe-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits