[PATCH] D61726: [Pass Pipeline] Run another round of reassociation after loop pipeline

Wed May 15 06:06:33 PDT 2019

spatel added a comment.

In D61726#1502041 <https://reviews.llvm.org/D61726#1502041>, @nemanjai wrote:

> > It would be interesting to see how that result translates on a more typical x86 build machine. Either way, I suspect we'll get different opinions about whether a 0.3% time increase is minimal and whether that cost is worth paying for the runtime perf gains. This might be a case for differentiating between -O2 and -O3?
>
> I don't really have access to a typical x86 server. I can get the numbers on my laptop but I'm not sure how typical that is. Would that suffice?

IMO, it's not required for you to gather more data, but some form of x86 is the common case, so that would be a better data point for most people. If we don't get that experiment pre-commit, then I'd expect some x86 bot to flag this change if it's a problem.

> Also, I am happy to guard this with an `-O3` requirement.

That would remove potential controversy (again, just my opinion) because that's how we limited 'AggressiveInstCombine', but let's see if anyone else (@efriedma @echristo ?) has a different idea.

Repository:
  rL LLVM

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D61726/new/

https://reviews.llvm.org/D61726