Apple, Google Agree On More SLP Vectorization

[phoronix]

Phoronix: Apple, Google Agree On More SLP Vectorization

After making more widespread use of the Loop Vectorizer, developers at Apple in Google are at least agreeing that LLVM's SLP Vectorizer should be more widely-used as well...

http://www.phoronix.com/vr.php?view=MTQyMjE

[mrugiero]

Chandler is forgetting that's probably harder to read when debugging, that's probably why -O3 and not everywhere. One needs to be able to debug the program, obviously.

[Marc Driftmeyer]

Seriously?

They agree?

Nadev is doing the heavy lifting at Apple. Whether or not Google agrees is irrelevant. Facts bare out and LLVM isn't managed or steered by Google.

[GreatEmerald]

Chandler is forgetting that's probably harder to read when debugging, that's probably why -O3 and not everywhere. One needs to be able to debug the program, obviously.

How about -O0 or -O1?

[mrugiero]

How about -O0 or -O1?

I understand "everywhere" as "every optimization flags or no optimization flags". So, everywhere means for debugging you need to know how to read this instruction sets. Maybe it should be -O3 and -O2, for example, but that's hardly "everywhere".

[GreatEmerald]

I understand "everywhere" as "every optimization flags or no optimization flags". So, everywhere means for debugging you need to know how to read this instruction sets. Maybe it should be -O3 and -O2, for example, but that's hardly "everywhere".

Well, since most people use -O2 when building for release, adding it to -O2 would more or less mean "everywhere". I really don't think they would be crazy enough to add this to -O0.