Address spaces in MC are completely untested

[Pete Cooper]

Address spaces are also used for the x86 segment registers (fs, gs), but like Tom said for R600, i’m not sure if x86 actually needs any MC level features to implement this.

Thanks,
Pete
On Jun 27, 2013, at 3:10 PM, Tom Stellard <tom at stellard.net> wrote:

> On Thu, Jun 27, 2013 at 09:43:43AM -0400, Rafael Espíndola wrote:
>> Reviewing http://llvm-reviews.chandlerc.com/D1043 I noticed that
>> implementation of address spaces in MC was a bit messy. Parts of some
>> streamers support it, others parts ignore it and yet others assert
>> that it is 0.
>> 
>> I wanted to refactor it, but there is a problem: We don't have *any*
>> tests. Is this used only by out of tree targets? What about R600,
>> doesn't it use address spaces? Only for dynamic memory?
>> 
> 
> R600 uses address spaces, but I don't think we currently use any MC
> features that need to be aware of them.  We only emit code and not much
> else.
> 
> -Tom
> 
>> Another option is that all data in object files is in the default
>> address space for all targets and this patch removes dead code.
>> Without tests it is hard to tell.
>> 
>> If you have an out of tree target that uses address spaces, please try
>> this patch and let me know if it breaks something for you.
>> 
>> This was added back in r63377 by Sanjiv Gupta from microchip, so it is
>> possible that this was only used by PIC and is really dead code.
>> 
>> Cheers,
>> Rafael
> 
> 
>> _______________________________________________
>> llvm-commits mailing list
>> llvm-commits at cs.uiuc.edu
>> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
> 
> 
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20130627/bb27c88f/attachment.html>