Page MenuHomePhabricator

[Polly] [HACK] [WIP] Add functionality to GPUJIT to track unified memory.

Authored by bollu on Jul 28 2017, 5:25 AM.

Diff Detail


Event Timeline

bollu created this revision.Jul 28 2017, 5:25 AM
bollu added a comment.Jul 28 2017, 5:25 AM

Clearly, this entire patch is a hack. I'd like to know how to add tests for this, and how to ensure that this gets reverted at some point in the future :)

grosser edited edge metadata.Aug 1 2017, 5:33 AM

Hi Siddharth,

I think it is OK to upstream this code, assuming you write and document it in a generic way and explain how it is used.

36 ↗(On Diff #108633)

Why is this needed. Maybe upstream separately?

1213 ↗(On Diff #108633)

There seem to be other situations where a context might already be available. I suggest to upstream this independently.

bollu updated this revision to Diff 109314.Aug 2 2017, 3:48 AM
bollu marked an inline comment as done.
  • Rebase on top of master and take tobias' suggestion into account.
bollu updated this revision to Diff 109315.Aug 2 2017, 3:50 AM

[NFC] upload proper diff.

grosser accepted this revision.Aug 2 2017, 3:54 AM

LGTM in general.

1437 ↗(On Diff #109315)

I don't think it adds any value to discuss COSMO here. Instead, it suggest to just explain what kind of behavior you seeand how this is addressed here.

1439 ↗(On Diff #109315)


This revision is now accepted and ready to land.Aug 2 2017, 3:54 AM
bollu updated this revision to Diff 109317.Aug 2 2017, 4:01 AM
  • [NFC] explain hack better
bollu marked 2 inline comments as done.Aug 2 2017, 4:01 AM
bollu updated this revision to Diff 109318.Aug 2 2017, 4:04 AM
  • [NFC] lint fix.
This revision was automatically updated to reflect the committed changes.
singam-sanjay edited edge metadata.Aug 3 2017, 12:51 PM

This breaks my Julia build.

How are the cudaFree and cudaMallocManaged functions resolved ? Can you add a patch to link GPURuntime to cudart ?