[llvm-commits] [llvm] r48725 - in /llvm/trunk: include/llvm/Transforms/IPO/InlinerPass.h include/llvm/Transforms/Utils/InlineCost.h lib/Transforms/IPO/InlineSimple.cpp lib/Transforms/IPO/Inliner.cpp lib/Transforms/Utils/InlineCost.cpp

Evan Cheng evan.cheng at apple.com
Mon Mar 24 16:48:41 PDT 2008


I considered removing the BB penalty. It's definitely a possibility  
since it doesn't make that much sense to me either.

But we do need to bump up the limit in addition to other tweaks, 200  
is just much too conservative. gcc default is 600. It's kinda  
comparing apples to oranges, but still. I am not seeing ill effects  
from the limit increase, so it's a good sign.

BTW, my comment was wrong, it's increased to 400.

Evan

On Mar 24, 2008, at 1:41 PM, Chris Lattner wrote:

> On Mar 23, 2008, at 11:37 PM, Evan Cheng wrote:
>> URL: http://llvm.org/viewvc/llvm-project?rev=48725&view=rev
>> Log:
>> Increasing the inline limit from (overly conservative) 200 to 300.
>> Given each BB costs 20 and each instruction costs 5, 200 means a 4
>> BB function + 24 instructions (actually less because caller's size
>> also contributes to it).
>
> Hi Evan,
>
> Instead of bumping up the threshold, maybe it would make sense to
> reduce or eliminate the per-bb penalty?  BB's by themselves don't make
> code bigger.
>
> -Chris
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits




More information about the llvm-commits mailing list