[llvm-commits] [llvm] r94615 - in /llvm/trunk: include/llvm/Analysis/InlineCost.h lib/Analysis/InlineCost.cpp

[Jakob Stoklund Olesen]

On Jan 26, 2010, at 4:35 PM, Eric Christopher wrote:

> Actually it seemed pretty reasonable, I was more worried about the ever present performance problems with Os.  The size wins seem pretty nice.

Yes, there were some unreasonable inlinings for -Os before.

>> It would make sense to bump the -Os threshold from 50 to 75, I'll try to measure the performance impact.
> 
> This would be awesome. A bit frustrating that size isn't the only thing that matters with Os, but...
> 
> I've probably got a couple of testcases if you're interested :)

If you have test that are sensitive to the amount of inlining under -Os, I would like to see them.

/jakob