Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

[CMake] Experimental Jemalloc support. #8874

Merged
merged 2 commits into from Mar 31, 2017
Merged

[CMake] Experimental Jemalloc support. #8874

merged 2 commits into from Mar 31, 2017

Conversation

vit-stepanovs
Copy link

My benchmarks of TF with Jemalloc on Windows had mixed results so far. However, some specific scenarios may benefit from Jemalloc support, so adding it to CMake (OFF by default).

@tensorflow-jenkins
Copy link
Collaborator

Can one of the admins verify this patch?

@vit-stepanovs vit-stepanovs changed the title [CMake] Expiremental Jemalloc support. [CMake] Experimental Jemalloc support. Mar 31, 2017
@yifeif yifeif requested review from mrry and gunan March 31, 2017 18:11
@yifeif yifeif added the stat:awaiting tensorflower Status - Awaiting response from tensorflower label Mar 31, 2017
@gunan gunan requested a review from jhseu March 31, 2017 18:37
@jhseu
Copy link
Contributor

jhseu commented Mar 31, 2017

Jenkins, test this please

@yifeif yifeif merged commit b26c5fb into tensorflow:master Mar 31, 2017
copybara-service bot pushed a commit that referenced this pull request Mar 12, 2024
@trevor-m @tensorflower-gardener
PR #8874: [GPU] Use NCCL user buffers for collective permute and all-…
6eef8fd 
…to-all

Imported from GitHub PR openxla/xla#8874

This PR enables XLA to take advantage of NCCL user buffers for ncclSend/ncclRecv when `--xla_gpu_enable_nccl_user_buffers=true` is used. Requires NCCL 2.20

Copybara import of the project:

--
8de2786d3242c76bed385235b5655156ee187e5f by Trevor Morris <tmorris@nvidia.com>:

Use NCCL user buffers for ncclSend/ncclRecv ops

--
56ceecb1b7fc1606dd00b514bbdb7d039e787b8c by Trevor Morris <tmorris@nvidia.com>:

Include memory space in buffers for collective permute and send/recv

--
64711757e48b619b9e2d322fc49714a94194d8f1 by Trevor Morris <tmorris@nvidia.com>:

Don't offload send, recv

Merging this change closes #8874

FUTURE_COPYBARA_INTEGRATE_REVIEW=openxla/xla#8874 from trevor-m:p2p-user-buffers 64711757e48b619b9e2d322fc49714a94194d8f1
PiperOrigin-RevId: 615104094
@copybara-service copybara-service bot mentioned this pull request Mar 12, 2024
Merged
copybara-service bot pushed a commit that referenced this pull request Mar 12, 2024
@trevor-m @tensorflower-gardener
PR #8874: [GPU] Use NCCL user buffers for collective permute and all-…
2f3118d 
…to-all

Imported from GitHub PR openxla/xla#8874

This PR enables XLA to take advantage of NCCL user buffers for ncclSend/ncclRecv when `--xla_gpu_enable_nccl_user_buffers=true` is used. Requires NCCL 2.20

Copybara import of the project:

--
8de2786d3242c76bed385235b5655156ee187e5f by Trevor Morris <tmorris@nvidia.com>:

Use NCCL user buffers for ncclSend/ncclRecv ops

--
56ceecb1b7fc1606dd00b514bbdb7d039e787b8c by Trevor Morris <tmorris@nvidia.com>:

Include memory space in buffers for collective permute and send/recv

--
64711757e48b619b9e2d322fc49714a94194d8f1 by Trevor Morris <tmorris@nvidia.com>:

Don't offload send, recv

Merging this change closes #8874

FUTURE_COPYBARA_INTEGRATE_REVIEW=openxla/xla#8874 from trevor-m:p2p-user-buffers 64711757e48b619b9e2d322fc49714a94194d8f1
PiperOrigin-RevId: 615104094
copybara-service bot pushed a commit that referenced this pull request Mar 12, 2024
@trevor-m @tensorflower-gardener
PR #8874: [GPU] Use NCCL user buffers for collective permute and all-…
c131d5a 
…to-all

Imported from GitHub PR openxla/xla#8874

This PR enables XLA to take advantage of NCCL user buffers for ncclSend/ncclRecv when `--xla_gpu_enable_nccl_user_buffers=true` is used. Requires NCCL 2.20

Copybara import of the project:

--
8de2786d3242c76bed385235b5655156ee187e5f by Trevor Morris <tmorris@nvidia.com>:

Use NCCL user buffers for ncclSend/ncclRecv ops

--
56ceecb1b7fc1606dd00b514bbdb7d039e787b8c by Trevor Morris <tmorris@nvidia.com>:

Include memory space in buffers for collective permute and send/recv

--
64711757e48b619b9e2d322fc49714a94194d8f1 by Trevor Morris <tmorris@nvidia.com>:

Don't offload send, recv

Merging this change closes #8874

FUTURE_COPYBARA_INTEGRATE_REVIEW=openxla/xla#8874 from trevor-m:p2p-user-buffers 64711757e48b619b9e2d322fc49714a94194d8f1
PiperOrigin-RevId: 615104094
copybara-service bot pushed a commit that referenced this pull request Mar 15, 2024
@trevor-m @tensorflower-gardener
PR #8874: [GPU] Use NCCL user buffers for collective permute and all-…
57a77df 
…to-all

Imported from GitHub PR openxla/xla#8874

This PR enables XLA to take advantage of NCCL user buffers for ncclSend/ncclRecv when `--xla_gpu_enable_nccl_user_buffers=true` is used. Requires NCCL 2.20

Copybara import of the project:

--
8de2786d3242c76bed385235b5655156ee187e5f by Trevor Morris <tmorris@nvidia.com>:

Use NCCL user buffers for ncclSend/ncclRecv ops

--
56ceecb1b7fc1606dd00b514bbdb7d039e787b8c by Trevor Morris <tmorris@nvidia.com>:

Include memory space in buffers for collective permute and send/recv

--
b3e776cb8486f2952dcb60a753dcea3c11da4d87 by Trevor Morris <tmorris@nvidia.com>:

Don't offload send, recv

Merging this change closes #8874

FUTURE_COPYBARA_INTEGRATE_REVIEW=openxla/xla#8874 from trevor-m:p2p-user-buffers b3e776cb8486f2952dcb60a753dcea3c11da4d87
PiperOrigin-RevId: 615104094
copybara-service bot pushed a commit that referenced this pull request Mar 15, 2024
@trevor-m @tensorflower-gardener
PR #8874: [GPU] Use NCCL user buffers for collective permute and all-…
c2aa79b 
…to-all

Imported from GitHub PR openxla/xla#8874

This PR enables XLA to take advantage of NCCL user buffers for ncclSend/ncclRecv when `--xla_gpu_enable_nccl_user_buffers=true` is used. Requires NCCL 2.20

Copybara import of the project:

--
8de2786d3242c76bed385235b5655156ee187e5f by Trevor Morris <tmorris@nvidia.com>:

Use NCCL user buffers for ncclSend/ncclRecv ops

--
56ceecb1b7fc1606dd00b514bbdb7d039e787b8c by Trevor Morris <tmorris@nvidia.com>:

Include memory space in buffers for collective permute and send/recv

--
b3e776cb8486f2952dcb60a753dcea3c11da4d87 by Trevor Morris <tmorris@nvidia.com>:

Don't offload send, recv

Merging this change closes #8874

FUTURE_COPYBARA_INTEGRATE_REVIEW=openxla/xla#8874 from trevor-m:p2p-user-buffers b3e776cb8486f2952dcb60a753dcea3c11da4d87
PiperOrigin-RevId: 615104094
copybara-service bot pushed a commit that referenced this pull request Mar 18, 2024
@trevor-m @tensorflower-gardener
PR #8874: [GPU] Use NCCL user buffers for collective permute and all-…
6d82b56 
…to-all

Imported from GitHub PR openxla/xla#8874

This PR enables XLA to take advantage of NCCL user buffers for ncclSend/ncclRecv when `--xla_gpu_enable_nccl_user_buffers=true` is used. Requires NCCL 2.20

Copybara import of the project:

--
98acdf27d4eba6b19652a76d3f7dcd6630349fc5 by Trevor Morris <tmorris@nvidia.com>:

Use NCCL user buffers for ncclSend/ncclRecv ops

--
bcc289b49bcf2086b50a86a2381ea1b80acd3dd2 by Trevor Morris <tmorris@nvidia.com>:

Include memory space in buffers for collective permute and send/recv

--
4a83d8906b6b5e305dad23fc1d8b9a5069637279 by Trevor Morris <tmorris@nvidia.com>:

Don't offload send, recv

--
0083a418c4ab119ed5a0eb061113104980476943 by Trevor Morris <tmorris@nvidia.com>:

Fix conditional

Merging this change closes #8874

FUTURE_COPYBARA_INTEGRATE_REVIEW=openxla/xla#8874 from trevor-m:p2p-user-buffers 0083a418c4ab119ed5a0eb061113104980476943
PiperOrigin-RevId: 615104094
copybara-service bot pushed a commit that referenced this pull request Mar 18, 2024
@trevor-m @tensorflower-gardener
PR #8874: [GPU] Use NCCL user buffers for collective permute and all-…
55c8a38 
…to-all

Imported from GitHub PR openxla/xla#8874

This PR enables XLA to take advantage of NCCL user buffers for ncclSend/ncclRecv when `--xla_gpu_enable_nccl_user_buffers=true` is used. Requires NCCL 2.20

Copybara import of the project:

--
98acdf27d4eba6b19652a76d3f7dcd6630349fc5 by Trevor Morris <tmorris@nvidia.com>:

Use NCCL user buffers for ncclSend/ncclRecv ops

--
bcc289b49bcf2086b50a86a2381ea1b80acd3dd2 by Trevor Morris <tmorris@nvidia.com>:

Include memory space in buffers for collective permute and send/recv

--
4a83d8906b6b5e305dad23fc1d8b9a5069637279 by Trevor Morris <tmorris@nvidia.com>:

Don't offload send, recv

--
0083a418c4ab119ed5a0eb061113104980476943 by Trevor Morris <tmorris@nvidia.com>:

Fix conditional

Merging this change closes #8874

FUTURE_COPYBARA_INTEGRATE_REVIEW=openxla/xla#8874 from trevor-m:p2p-user-buffers 0083a418c4ab119ed5a0eb061113104980476943
PiperOrigin-RevId: 615104094
copybara-service bot pushed a commit that referenced this pull request Mar 18, 2024
@trevor-m @tensorflower-gardener
PR #8874: [GPU] Use NCCL user buffers for collective permute and all-…
5cd046b 
…to-all

Imported from GitHub PR openxla/xla#8874

This PR enables XLA to take advantage of NCCL user buffers for ncclSend/ncclRecv when `--xla_gpu_enable_nccl_user_buffers=true` is used. Requires NCCL 2.20

Copybara import of the project:

--
98acdf27d4eba6b19652a76d3f7dcd6630349fc5 by Trevor Morris <tmorris@nvidia.com>:

Use NCCL user buffers for ncclSend/ncclRecv ops

--
bcc289b49bcf2086b50a86a2381ea1b80acd3dd2 by Trevor Morris <tmorris@nvidia.com>:

Include memory space in buffers for collective permute and send/recv

--
4a83d8906b6b5e305dad23fc1d8b9a5069637279 by Trevor Morris <tmorris@nvidia.com>:

Don't offload send, recv

--
0083a418c4ab119ed5a0eb061113104980476943 by Trevor Morris <tmorris@nvidia.com>:

Fix conditional

Merging this change closes #8874

FUTURE_COPYBARA_INTEGRATE_REVIEW=openxla/xla#8874 from trevor-m:p2p-user-buffers 0083a418c4ab119ed5a0eb061113104980476943
PiperOrigin-RevId: 615104094
copybara-service bot pushed a commit that referenced this pull request Mar 19, 2024
@trevor-m @tensorflower-gardener
PR #8874: [GPU] Use NCCL user buffers for collective permute and all-…
c0b3c14 
…to-all

Imported from GitHub PR openxla/xla#8874

This PR enables XLA to take advantage of NCCL user buffers for ncclSend/ncclRecv when `--xla_gpu_enable_nccl_user_buffers=true` is used. Requires NCCL 2.20

Copybara import of the project:

--
98acdf27d4eba6b19652a76d3f7dcd6630349fc5 by Trevor Morris <tmorris@nvidia.com>:

Use NCCL user buffers for ncclSend/ncclRecv ops

--
bcc289b49bcf2086b50a86a2381ea1b80acd3dd2 by Trevor Morris <tmorris@nvidia.com>:

Include memory space in buffers for collective permute and send/recv

--
4a83d8906b6b5e305dad23fc1d8b9a5069637279 by Trevor Morris <tmorris@nvidia.com>:

Don't offload send, recv

--
0083a418c4ab119ed5a0eb061113104980476943 by Trevor Morris <tmorris@nvidia.com>:

Fix conditional

Merging this change closes #8874

FUTURE_COPYBARA_INTEGRATE_REVIEW=openxla/xla#8874 from trevor-m:p2p-user-buffers 0083a418c4ab119ed5a0eb061113104980476943
PiperOrigin-RevId: 615104094
copybara-service bot pushed a commit that referenced this pull request Mar 19, 2024
@trevor-m @tensorflower-gardener
PR #8874: [GPU] Use NCCL user buffers for collective permute and all-…
393389d 
…to-all

Imported from GitHub PR openxla/xla#8874

This PR enables XLA to take advantage of NCCL user buffers for ncclSend/ncclRecv when `--xla_gpu_enable_nccl_user_buffers=true` is used. Requires NCCL 2.20

Copybara import of the project:

--
98acdf27d4eba6b19652a76d3f7dcd6630349fc5 by Trevor Morris <tmorris@nvidia.com>:

Use NCCL user buffers for ncclSend/ncclRecv ops

--
bcc289b49bcf2086b50a86a2381ea1b80acd3dd2 by Trevor Morris <tmorris@nvidia.com>:

Include memory space in buffers for collective permute and send/recv

--
4a83d8906b6b5e305dad23fc1d8b9a5069637279 by Trevor Morris <tmorris@nvidia.com>:

Don't offload send, recv

--
0083a418c4ab119ed5a0eb061113104980476943 by Trevor Morris <tmorris@nvidia.com>:

Fix conditional

Merging this change closes #8874

FUTURE_COPYBARA_INTEGRATE_REVIEW=openxla/xla#8874 from trevor-m:p2p-user-buffers 0083a418c4ab119ed5a0eb061113104980476943
PiperOrigin-RevId: 615104094
copybara-service bot pushed a commit that referenced this pull request Mar 19, 2024
@trevor-m @tensorflower-gardener
PR #8874: [GPU] Use NCCL user buffers for collective permute and all-…
dd25232 
…to-all

Imported from GitHub PR openxla/xla#8874

This PR enables XLA to take advantage of NCCL user buffers for ncclSend/ncclRecv when `--xla_gpu_enable_nccl_user_buffers=true` is used. Requires NCCL 2.20

Copybara import of the project:

--
98acdf27d4eba6b19652a76d3f7dcd6630349fc5 by Trevor Morris <tmorris@nvidia.com>:

Use NCCL user buffers for ncclSend/ncclRecv ops

--
bcc289b49bcf2086b50a86a2381ea1b80acd3dd2 by Trevor Morris <tmorris@nvidia.com>:

Include memory space in buffers for collective permute and send/recv

--
4a83d8906b6b5e305dad23fc1d8b9a5069637279 by Trevor Morris <tmorris@nvidia.com>:

Don't offload send, recv

--
0083a418c4ab119ed5a0eb061113104980476943 by Trevor Morris <tmorris@nvidia.com>:

Fix conditional

Merging this change closes #8874

FUTURE_COPYBARA_INTEGRATE_REVIEW=openxla/xla#8874 from trevor-m:p2p-user-buffers 0083a418c4ab119ed5a0eb061113104980476943
PiperOrigin-RevId: 615104094
copybara-service bot pushed a commit that referenced this pull request Mar 19, 2024
@trevor-m @tensorflower-gardener
PR #8874: [GPU] Use NCCL user buffers for collective permute and all-…
6385b9b 
…to-all

Imported from GitHub PR openxla/xla#8874

This PR enables XLA to take advantage of NCCL user buffers for ncclSend/ncclRecv when `--xla_gpu_enable_nccl_user_buffers=true` is used. Requires NCCL 2.20

Copybara import of the project:

--
98acdf27d4eba6b19652a76d3f7dcd6630349fc5 by Trevor Morris <tmorris@nvidia.com>:

Use NCCL user buffers for ncclSend/ncclRecv ops

--
bcc289b49bcf2086b50a86a2381ea1b80acd3dd2 by Trevor Morris <tmorris@nvidia.com>:

Include memory space in buffers for collective permute and send/recv

--
4a83d8906b6b5e305dad23fc1d8b9a5069637279 by Trevor Morris <tmorris@nvidia.com>:

Don't offload send, recv

--
0083a418c4ab119ed5a0eb061113104980476943 by Trevor Morris <tmorris@nvidia.com>:

Fix conditional

Merging this change closes #8874

FUTURE_COPYBARA_INTEGRATE_REVIEW=openxla/xla#8874 from trevor-m:p2p-user-buffers 0083a418c4ab119ed5a0eb061113104980476943
PiperOrigin-RevId: 615104094
copybara-service bot pushed a commit that referenced this pull request Mar 19, 2024
@trevor-m @tensorflower-gardener
PR #8874: [GPU] Use NCCL user buffers for collective permute and all-…
ac23618 
…to-all

Imported from GitHub PR openxla/xla#8874

This PR enables XLA to take advantage of NCCL user buffers for ncclSend/ncclRecv when `--xla_gpu_enable_nccl_user_buffers=true` is used. Requires NCCL 2.20

Copybara import of the project:

--
98acdf27d4eba6b19652a76d3f7dcd6630349fc5 by Trevor Morris <tmorris@nvidia.com>:

Use NCCL user buffers for ncclSend/ncclRecv ops

--
bcc289b49bcf2086b50a86a2381ea1b80acd3dd2 by Trevor Morris <tmorris@nvidia.com>:

Include memory space in buffers for collective permute and send/recv

--
4a83d8906b6b5e305dad23fc1d8b9a5069637279 by Trevor Morris <tmorris@nvidia.com>:

Don't offload send, recv

--
0083a418c4ab119ed5a0eb061113104980476943 by Trevor Morris <tmorris@nvidia.com>:

Fix conditional

Merging this change closes #8874

FUTURE_COPYBARA_INTEGRATE_REVIEW=openxla/xla#8874 from trevor-m:p2p-user-buffers 0083a418c4ab119ed5a0eb061113104980476943
PiperOrigin-RevId: 615104094
copybara-service bot pushed a commit that referenced this pull request Mar 19, 2024
@trevor-m @tensorflower-gardener
PR #8874: [GPU] Use NCCL user buffers for collective permute and all-…
f5b39c6 
…to-all

Imported from GitHub PR openxla/xla#8874

This PR enables XLA to take advantage of NCCL user buffers for ncclSend/ncclRecv when `--xla_gpu_enable_nccl_user_buffers=true` is used. Requires NCCL 2.20

Copybara import of the project:

--
98acdf27d4eba6b19652a76d3f7dcd6630349fc5 by Trevor Morris <tmorris@nvidia.com>:

Use NCCL user buffers for ncclSend/ncclRecv ops

--
bcc289b49bcf2086b50a86a2381ea1b80acd3dd2 by Trevor Morris <tmorris@nvidia.com>:

Include memory space in buffers for collective permute and send/recv

--
4a83d8906b6b5e305dad23fc1d8b9a5069637279 by Trevor Morris <tmorris@nvidia.com>:

Don't offload send, recv

--
0083a418c4ab119ed5a0eb061113104980476943 by Trevor Morris <tmorris@nvidia.com>:

Fix conditional

Merging this change closes #8874

PiperOrigin-RevId: 617140675
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@mrry mrry mrry approved these changes

@gunan gunan Awaiting requested review from gunan

Assignees
No one assigned
Labels
cla: yes stat:awaiting tensorflower Status - Awaiting response from tensorflower
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.

None yet
6 participants
@vit-stepanovs @tensorflow-jenkins @jhseu @mrry @yifeif @googlebot