[llvm-commits] [llvm-gcc-4.2] r63123 - in /llvm-gcc-4.2/branches/Apple/Dib/gcc: cgraphunit.c ipa-inline.c passes.c tree-pass.h

[Dale Johannesen]

On Jan 27, 2009, at 11:51 AMPST, Evan Cheng wrote:
> On Jan 27, 2009, at 11:40 AM, Dale Johannesen wrote:
>> On Jan 27, 2009, at 11:36 AMPST, Evan Cheng wrote:
>>
>>> Author: evancheng
>>> Date: Tue Jan 27 13:36:13 2009
>>> New Revision: 63123
>>>
>>> URL: http://llvm.org/viewvc/llvm-project?rev=63123&view=rev
>>> Log:
>>> Restore gcc inliner in teh Dib branch for now. It's exposing some  
>>> performance issues.
>>
>> Is this some general performance issue, or just that some specific  
>> code that's been extensively tuned to work well with gcc's inliner  
>> doesn't work as well now?  The latter would be expected
>
> I considered it a general performance issue that just happen to be  
> exposed by a particular test case.

Do you have evidence for this?   I've looked at it quite a bit, and  
concluded that llvm-gcc's inliner does less inlining on average than  
gcc's, but that this is not, in general, a problem:  in other words,  
benchmarks that are better and worse are about equal in number.
(And performance being roughly equal, smaller code size is to be  
preferred.)

I understand that some benchmarks are more important than others in  
particular situations, and I'm not objecting to the patch, but I think  
you are overgeneralizing what the problem is.