[Patch][Review Request][llvm-link] Fix One Performance Bug

[Manman Ren]

This patch can actually speed up our LTO compilation. It reduces the execution time of ModuleLinker::computeTypeMapping, which is used during LTO.
It changed from O(N^2) to O(N) where N is number of modules.

Thanks for working on this,
Manman

On Apr 29, 2013, at 7:43 AM, Wan, Xiaofei wrote:

> +1.  I don't think that we need to change Module.  LTO compilation of xalan does not take 30 minutes, so I imagine that this is a problem with llvm-link. 
>  
> For xalan, the linking time is about 30 seconds, not 30 minutes, after this fix, it is 2 seconds
>  
> For sure, the issue exists in llvm-link; I don’t see llvm-link is often used, but this is often used in our business.
>  
>  
> Thanks
> Wan Xiaofei
> From: Nadav Rotem [mailto:nrotem at apple.com] 
> Sent: Monday, April 29, 2013 11:01 AM
> To: Du Toit, Stefanus
> Cc: Wan, Xiaofei; llvm-commits at cs.uiuc.edu
> Subject: Re: [Patch][Review Request][llvm-link] Fix One Performance Bug
>  
>  
> On Apr 28, 2013, at 6:44 PM, "Du Toit, Stefanus" <stefanus.du.toit at intel.com> wrote:
> 
> 
> To be more specific, I think the allocation of this data structure belongs
> in llvm-link itself.
>  
> +1.  I don't think that we need to change Module.  LTO compilation of xalan does not take 30 minutes, so I imagine that this is a problem with llvm-link. 
> _______________________________________________
> 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/20130503/92cda467/attachment.html>