[llvm-dev] [RFC] New pass: LoopExitValues

[Mehdi Amini via llvm-dev]

Hi,

> On Sep 23, 2015, at 12:09 PM, Steve King via llvm-dev <llvm-dev at lists.llvm.org> wrote:
> 
> On Wed, Sep 23, 2015 at 12:00 PM, Hal Finkel <hfinkel at anl.gov> wrote:
>>> 
>>> Should we try the patch in it's current location, namely after LSR?
>> 
>> Sure; post the patch as you have it so we can look at what's going on.
>> 
> 
> http://reviews.llvm.org/D12494
> 
> One particular point: The algorithm checks that SCEV's are equal when
> their raw pointers are equal.  Is that a future-proof feature of
> SCEVs?


I don’t know much about this optimization or its implementation, but I just ran it on our test-suite (both the public one and some internal benchmarks) on OS X (X86) and I didn’t observe much changes:

- 3.37% regression on SingleSource/Benchmarks/Shootout/matrix
- 3.87% and 2.46% improvements on two SPEC benchmarks.

(Patch applied on top of r248302).

— 
Mehdi