Page MenuHomePhabricator

[OpenMPOpt][SplitMemTransfer] Can move setup instructions upwards?
Needs ReviewPublic

Authored by hamax97 on Sep 7 2020, 6:10 PM.

Details

Summary

Checks if the setup (issue) instructions for __tgt_target_data_begin_mapper can be moved upwards.

Diff Detail

Event Timeline

hamax97 created this revision.Sep 7 2020, 6:10 PM
hamax97 requested review of this revision.Sep 7 2020, 6:10 PM

Tests?

The regression test for this is hide_mem_transfer_latency.ll. But there's still the issue with the AAManager, so the test has the fixmes of where the set up instructions should be moved.

Tests?

The regression test for this is hide_mem_transfer_latency.ll. But there's still the issue with the AAManager, so the test has the fixmes of where the set up instructions should be moved.

But it seems not to include any changes to the test. why not? because this doesn't have an impact without resolving the issue you mentioned?

Tests?

The regression test for this is hide_mem_transfer_latency.ll. But there's still the issue with the AAManager, so the test has the fixmes of where the set up instructions should be moved.

But it seems not to include any changes to the test. why not? because this doesn't have an impact without resolving the issue you mentioned?

Exactly, the mayModify function always returns true at the instruction immediately above the issue, so there's no movement.

hamax97 updated this revision to Diff 292016.Sep 15 2020, 1:46 PM
hamax97 set the repository for this revision to rG LLVM Github Monorepo.
  • Fixed bug when creating the wait runtime call.

tests?

Recall our previous discussion. Given the issue with the AAManager there's no movement in any of the test cases. Anyway, we need first to review D88029 before accepting any of this.