This is an archive of the discontinued LLVM Phabricator instance.

enable memory interleaving for X86\SLM arch.
ClosedPublic

Authored by magabari on Jan 11 2017, 1:37 AM.

Diff Detail

Repository
rL LLVM

Event Timeline

magabari updated this revision to Diff 83927.Jan 11 2017, 1:37 AM
magabari retitled this revision from to enable memory interleaving for X86\SLM arch..
magabari updated this object.
RKSimon edited edge metadata.Jan 11 2017, 2:58 AM

Do you have perf stats to confirm that this is safe to do now?

lib/Target/X86/X86TargetTransformInfo.cpp
2084 ↗(On Diff #83927)

I don't suppose you know what the remaining problems are with Atom are?

mkuper added inline comments.Jan 11 2017, 10:08 AM
lib/Target/X86/X86TargetTransformInfo.cpp
2084 ↗(On Diff #83927)

This is my fault - I played it safe, and assumed that since SLM has problematic behavior, so does Atom, since the architectures are similar.
I wasn't aware of any Atom-specific issues when I committed this.
Of course, it's possible Intel have some new data about this, but in its absence, I'd say both are safe to remove.

magabari added inline comments.Jan 24 2017, 1:10 AM
lib/Target/X86/X86TargetTransformInfo.cpp
2084 ↗(On Diff #83927)

Yes it's safe to do that now for SLM.
Regrading Atom, i didn't do any cost changes for Atom so i think this should still be disabled for Atom currently.

Did you manage to do any perf tests on SLM hardware?

Did you manage to do any perf tests on SLM hardware?

Yes. and now it's safe to enable this feature for SLM.

RKSimon accepted this revision.Jan 24 2017, 1:58 PM

OK, LGTM

This revision is now accepted and ready to land.Jan 24 2017, 1:58 PM
This revision was automatically updated to reflect the committed changes.