[PATCH] Move StructurizeCFG out of R600 to generic Transforms

[Matt Arsenault]

On 06/05/2013 12:55 PM, Owen Anderson wrote:
> Hey,
>
> I have a serious question about this patch that I raised in a prior 
> email, and I'd appreciate it if you'd address it rather than 
> continuing in the opposite direction.
>
> On Jun 5, 2013, at 12:42 PM, Matt Arsenault <Matthew.Arsenault at amd.com 
> <mailto:Matthew.Arsenault at amd.com>> wrote:
>
>>  Move NearestCommonDominator to separate file
>>
>> http://llvm-reviews.chandlerc.com/D906
>>
>> CHANGE SINCE LAST DIFF
>>  http://llvm-reviews.chandlerc.com/D906?vs=2252&id=2279#toc
>>
>> Files:
>>  include/llvm/Analysis/NearestCommonDominator.h
>
> I don't understand why this analysis is necessary.  We already have 
> pair-wise nearestCommonDominator functionality built into DomTree.  If 
> it needs to be extended to support multi-block NCD queries, that 
> should be possible to implement on top of the existing functionality 
> rather than build a new analysis on top of it.
>
I was wondering about that after you asked, but I thought that Christian 
had addressed it earlier today.

> More generally,  I don't understand why it's desirable to move this 
> out of the R600 target.  The comments hand wave about it being useful 
> for other targets, but I don't see anyone actually planning to use it. 
>  If there is interest in it from other targets, then you probably need 
> to /start/ by having a conversation with other target authors about 
> what kind of common canonical form different targets can share rather 
> than simply foisting whatever was convenient for R600 onto them.
>
> --Owen

I'm interested in using it in the AMDIL target, which is why I want to 
do this.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20130605/8190be40/attachment.html>