Skip to content
This repository was archived by the owner on Nov 17, 2023. It is now read-only.

Revert PR 17767 for fixing GPU memory usage regression #18283

Merged
leezu merged 2 commits into from
May 13, 2020
Merged

Revert PR 17767 for fixing GPU memory usage regression #18283

leezu merged 2 commits into from
May 13, 2020

Conversation

@rondogency
Copy link
Contributor

@rondogency rondogency commented May 12, 2020

Revert PR #17767 for fixing #18280 GPU memory usage regression

Tested locally and observed GPU memory decreasing after reverting

Resolved revert conflict in elemwise_binrary_ops.h and test_operators.py

@mxnet-bot
Copy link

mxnet-bot commented May 12, 2020

Hey @rondogency , Thanks for submitting the PR
All tests are already queued to run once. If tests fail, you can trigger one or more tests again with the following commands:

  • To trigger all jobs: @mxnet-bot run ci [all]
  • To trigger specific jobs: @mxnet-bot run ci [job1, job2]

CI supported jobs: [sanity, centos-cpu, edge, miscellaneous, unix-gpu, clang, website, windows-gpu, windows-cpu, unix-cpu, centos-gpu]

Note:
Only following 3 categories can trigger CI :PR Author, MXNet Committer, Jenkins Admin.
All CI tests must pass before the PR can be merged.

@apeforest
Copy link
Contributor

apeforest commented May 12, 2020 •
edited
Loading

@rondogency could you change the description to "... fixes #18280". Only if you use the magic word "fix" or "fixes", it will auto close the issue. Thanks

@apeforest apeforest requested a review from ptrendx May 12, 2020 19:06
@apeforest
Copy link
Contributor

apeforest commented May 12, 2020

@ptrendx Are you okay with this revert? Thanks!

@ptrendx
Copy link
Member

ptrendx commented May 12, 2020

Yes, I'm fine with it, although I would not want this PR to close that issue, as there is discussion happening there on an actual fix.

@rondogency
Copy link
Contributor Author

rondogency commented May 12, 2020

I agree with @ptrendx that we should not close the issue, also I noticed that this one is picked into 1.7 release, we should revert it also

@rondogency
Copy link
Contributor Author

rondogency commented May 12, 2020

@mxnet-bot run ci [all]

@mxnet-bot
Copy link

mxnet-bot commented May 12, 2020

Jenkins CI successfully triggered : [windows-cpu, centos-gpu, edge, sanity, miscellaneous, unix-cpu, website, unix-gpu, windows-gpu, centos-cpu, clang]

@rondogency
Copy link
Contributor Author

rondogency commented May 12, 2020

@mxnet-bot run ci [centos-cpu, miscellaneous, unix-gpu, windows-gpu, windows-cpu, unix-cpu, centos-gpu]

@mxnet-bot
Copy link

mxnet-bot commented May 12, 2020

Jenkins CI successfully triggered : [miscellaneous, unix-gpu, centos-cpu, unix-cpu, windows-gpu, centos-gpu, windows-cpu]

@josephevans
Copy link
Contributor

josephevans commented May 12, 2020

@mxnet-bot run ci [centos-gpu]

@mxnet-bot
Copy link

mxnet-bot commented May 12, 2020

Unauthorized access detected.
Only following 3 categories can trigger CI :
PR Author, MXNet Committer, Jenkins Admin.

@leezu
Copy link
Contributor

leezu commented May 12, 2020

As @ptrendx targets having runtime compilation for cuda kernels in 2.0, I wonder if this should be reverted on 1.x branch only?

@rondogency
Copy link
Contributor Author

rondogency commented May 12, 2020

@mxnet-bot run ci [centos-gpu, unix-cpu, unix-gpu, windows-cpu]

@mxnet-bot
Copy link

mxnet-bot commented May 12, 2020

Jenkins CI successfully triggered : [windows-cpu, unix-cpu, centos-gpu, unix-gpu]

@rondogency
Copy link
Contributor Author

rondogency commented May 13, 2020

@mxnet-bot run ci [centos-cpu]

@mxnet-bot
Copy link

mxnet-bot commented May 13, 2020

Jenkins CI successfully triggered : [centos-cpu]

@rondogency
Copy link
Contributor Author

rondogency commented May 13, 2020

@leezu we need to resolve the current regression first and ptrendx would work on a fix

@rondogency
Copy link
Contributor Author

rondogency commented May 13, 2020

@ptrendx can you help to merge it, thanks!

@leezu leezu merged commit 47a38d1 into apache:master May 13, 2020
@rondogency rondogency deleted the revert-17767 branch May 13, 2020 20:53
@rondogency
Copy link
Contributor Author

rondogency commented May 13, 2020

@ciyongch Can you help to cherry-pick this into 1.x and 1.7 release, as this addresses a memory regression fix

rondogency added a commit to rondogency/incubator-mxnet that referenced this pull request May 14, 2020
@rondogency
Revert PR 17767 for fixing GPU memory usage regression (apache#18283)
ac4fce1 
* Revert "Fix and optimize handling of vectorized memory accesses (apache#17767)"

This reverts commit 5542d03.

* add license to reverted file
@ciyongch
Copy link
Contributor

ciyongch commented May 14, 2020

Thanks @rondogency , do you mind to create a PR to 1.7 as well, as it's already merge/created in master/v1.x now :)

@rondogency
Copy link
Contributor Author

rondogency commented May 14, 2020

@ciyongch yes I will do it shortly

rondogency added a commit to rondogency/incubator-mxnet that referenced this pull request May 14, 2020
@rondogency
Revert PR 17767 for fixing GPU memory usage regression (apache#18283)
ea301d3 
* Revert "Fix and optimize handling of vectorized memory accesses (apache#17767)"

This reverts commit 5542d03.

* add license to reverted file
ptrendx pushed a commit that referenced this pull request May 27, 2020
@rondogency
Revert PR 17767 for fixing GPU memory usage regression (#18283) (#18311)
4033a85 
* Revert "Fix and optimize handling of vectorized memory accesses (#17767)"

This reverts commit 5542d03.

* add license to reverted file
eric-haibin-lin pushed a commit that referenced this pull request May 29, 2020
@rondogency
Revert PR 17767 for fixing GPU memory usage regression (#18283) (#18309)
d621e50 
* Revert "Fix and optimize handling of vectorized memory accesses (#17767)"

This reverts commit 5542d03.

* add license to reverted file
AntiZpvoh pushed a commit to AntiZpvoh/incubator-mxnet that referenced this pull request Jul 6, 2020
@rondogency @AntiZpvoh
Revert PR 17767 for fixing GPU memory usage regression (apache#18283)
f603590 
* Revert "Fix and optimize handling of vectorized memory accesses (apache#17767)"

This reverts commit 5542d03.

* add license to reverted file
ChaiBapchya pushed a commit to ChaiBapchya/mxnet that referenced this pull request Aug 15, 2020
@rondogency @ChaiBapchya
Revert PR 17767 for fixing GPU memory usage regression (apache#18283) (…
47ea684 
…apache#18309)

* Revert "Fix and optimize handling of vectorized memory accesses (apache#17767)"

This reverts commit 5542d03.

* add license to reverted file
Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.

Reviewers

2 more reviewers

@apeforest apeforest apeforest approved these changes

@ptrendx ptrendx ptrendx approved these changes

Reviewers whose approvals may not affect merge requirements

Assignees

No one assigned

Labels

None yet

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

7 participants

@rondogency @mxnet-bot @apeforest @ptrendx @josephevans @leezu @ciyongch