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This repository has been archived by the owner on Nov 17, 2023.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

Fix row_sparse_pull with single gpu #10772

Merged
merged 8 commits into from
May 4, 2018

Conversation

leezu
Copy link
Contributor

@leezu leezu commented May 1, 2018

Description

When using the kvstore on a single device, no copy of the source array would be created when pushing to the kvstore. When row sparse pulling from the kvstore, this would lead to a deadlock as both source and target arrays were identical. For now this PR removes the optimization of not creating a copy.

Thanks @eric-haibin-lin for helping to find the cause.

@leezu leezu force-pushed the fixrowsparsepullsinglegpu branch from 47c26c2 to 52377b8 Compare May 1, 2018 22:38
@leezu leezu force-pushed the fixrowsparsepullsinglegpu branch from 52377b8 to b507f78 Compare May 1, 2018 22:39
@eric-haibin-lin eric-haibin-lin self-assigned this May 1, 2018
@leezu leezu force-pushed the fixrowsparsepullsinglegpu branch from b507f78 to c84d427 Compare May 1, 2018 23:20
@eric-haibin-lin eric-haibin-lin mentioned this pull request May 2, 2018
7 tasks
@@ -91,6 +91,21 @@ def check_rsp_pull(kv, count, ctxs, is_same_rowid=False, use_slice=False):
check_rsp_push_pull('device')
check_rsp_push_pull('device', is_push_cpu=False)


def test_row_sparse_pull_single_device():
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The test is in kvstore gpu section but uses local kvstore and cpu context. It would probably make sense to change the test here to device and gpu?

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

You're right. GPU was used previously in this test but these changes were accidentally pushed. Will update the test

@eric-haibin-lin eric-haibin-lin merged commit 7395be8 into apache:master May 4, 2018
anirudh2290 pushed a commit to anirudh2290/mxnet that referenced this pull request May 7, 2018
* Fix row_sparse_pull with single gpu

* Add test

* Fix row_sparse_pull with single gpu

* Add test

* fix sparse retain in comm.h

* remove dedup var

* update test
jinhuang415 pushed a commit to jinhuang415/incubator-mxnet that referenced this pull request May 29, 2018
* Fix row_sparse_pull with single gpu

* Add test

* Fix row_sparse_pull with single gpu

* Add test

* fix sparse retain in comm.h

* remove dedup var

* update test
rahul003 pushed a commit to rahul003/mxnet that referenced this pull request Jun 4, 2018
* Fix row_sparse_pull with single gpu

* Add test

* Fix row_sparse_pull with single gpu

* Add test

* fix sparse retain in comm.h

* remove dedup var

* update test
zheng-da pushed a commit to zheng-da/incubator-mxnet that referenced this pull request Jun 28, 2018
* Fix row_sparse_pull with single gpu

* Add test

* Fix row_sparse_pull with single gpu

* Add test

* fix sparse retain in comm.h

* remove dedup var

* update test
@leezu leezu deleted the fixrowsparsepullsinglegpu branch September 28, 2020 18:33
Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

3 participants