[llvm-commits] [llvm] r46959 - in /llvm/trunk: lib/Analysis/ScalarEvolution.cpp test/Analysis/ScalarEvolution/2008-02-11-ReversedCondition.ll

[Wojciech Matyjewicz]

Hi,

> I'm going to have to back out this patch temporarily. 

Okay.

> This is causing  
> performance regression in an important benchmark. I don't think this  
> patch is really at fault. But I need time to investigate.

Which benchmark is it? Probably, an expression for the iteration count
can be now created for a loop that wasn't previously analyzable. It may
be the case that -indvars currently introduces code to compute the
iteration count before the loop, or to compute the exit value for some
register in the loop after the loop. Sometimes this kind of code can be
really long...

Wojtek