[PATCH] D29673: [DebugInfo] Append extended dereferencing mechanism to variables' DIExpression for targets that support more than one address space

[David Blaikie via llvm-commits]

It sounds like there's probably a desire to have a target-specific mapping
from LLVM address space to DWARF address space?

If that's the case, then maybe that target specific hook could also inform
the consumer about the default (ie: instead of adding
the targetSupportsMultipleAddressSpaces hook, have: Optional<unsigned>
getDwarfAddressSpace(unsigned LLVMAddressSpace) - that way the default can
be "None" and any implementation that wants to provide address spaces can
communicate the "default" of its choosing by returning None in that case
and whatever non-None values it wants in other cases))

Matt - is that the kind of flexibility you'd like regarding remapping
address spaces between LLVM and DWARF?

On Tue, Feb 7, 2017 at 2:38 PM Matt Arsenault via Phabricator <
reviews at reviews.llvm.org> wrote:

> arsenm added a comment.
>
> It would also be good if we could leave open the possibility of decoupling
> LLVM address space 0 from OpenCL private address space. I would like to be
> able to someday be able to change that
>
>
> https://reviews.llvm.org/D29673
>
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20170209/76e0c3d5/attachment.html>