[llvm-commits] [llvm] r167286 - in /llvm/trunk: lib/Transforms/Scalar/LoopIdiomRecognize.cpp test/Transforms/LoopIdiom/basic.ll test/Transforms/LoopIdiom/memmove.ll test/Transforms/LoopIdiom/memset_noidiom.ll test/Transforms/LoopIdiom/multi-dimen

[Chandler Carruth]

On Fri, Nov 2, 2012 at 1:33 PM, Chris Lattner <clattner at apple.com> wrote:
>
> On Nov 2, 2012, at 1:33 AM, Chandler Carruth <chandlerc at gmail.com> wrote:
>
>> Author: chandlerc
>> Date: Fri Nov  2 03:33:25 2012
>> New Revision: 167286
>>
>> URL: http://llvm.org/viewvc/llvm-project?rev=167286&view=rev
>> Log:
>> Revert the switch of loop-idiom to use the new dependence analysis.
>>
>> The new analysis is not yet ready for prime time. It has a *critical*
>> flawed assumption, and some troubling shortages of testing. Until it's
>> been hammered into better shape, let's stick with the working code. This
>> should be easy to revert itself when the analysis is ready.
>
> Thanks.  More generally, I don't think anything in the default compiler flow should switch to it until a (LLVM) loop optimization expert reviews it in detail.

Yea, Dan has educated me on many of these issues. I'm likely going to
re-instate this patch without the need for DA soon, and fix up a few
other things.