[XLA:GPU][oneAPI] Add support for oneCCL bazel build#42595
Closed
nhatleSummer22 wants to merge 5 commits into
Closed
[XLA:GPU][oneAPI] Add support for oneCCL bazel build#42595nhatleSummer22 wants to merge 5 commits into
nhatleSummer22 wants to merge 5 commits into
Conversation
Contributor
|
Hi, thank you for this PR ! I was trying to use it earlier but ran into problem due to oneCCL "one process per GPU" model (according to Codex), the collectives would fail to initialize. Apparently this is only possible with oneAPI 2026.0. Do you have more info? Also, we are maintaining a fork which now ran llama on a b70 at https://github.com/zml/xla/commits/zml/oneapi2/ We found 2 problems:
Thank you! |
Member
|
@MichaelHudgins Can we simply add a directory in third_party or does this need a bit more work on our side? |
Contributor
Member
Apologies, i missed this one. In general we can, let me message you internally with more specifics. |
dimitar-asenov
approved these changes
Jun 5, 2026
This was referenced Jun 5, 2026
neudinger
added a commit
to zml/xla
that referenced
this pull request
Jun 9, 2026
dimitar-asenov
approved these changes
Jun 9, 2026
copybara-service Bot
pushed a commit
that referenced
this pull request
Jun 9, 2026
Imported from GitHub PR #42595 📝 Summary of Changes This PR enables building oneCCL from source using XLA bazel build system. [oneCCL](https://github.com/uxlfoundation/oneCCL) enables optimized communication pattern on Intel's GPUs. 🎯 Justification This PR is first step to support scale-up functionality on Intel's GPUs. Subsequent PRs will add full support for scale-up. 🚀 Kind of Contribution ✨ New Feature Copybara import of the project: -- 8e59923 by Nhat Le <nhat.le@intel.com>: Add support for oneCCL bazel build -- 6762807 by Nhat Le <nhat.le@intel.com>: Add dependency to trigger oneCCL build -- 644072a by Nhat Le <nhat.le@intel.com>: Fix EOF -- 42901eb by Nhat Le <nhat.le@intel.com>: Make all the necessary headers visible to dependents -- f4c481a by nhatle <nhat.le@intel.com>: Fix for file not found errors in XLA Linux X86 GPU ONEAPI CI Merging this change closes #42595 FUTURE_COPYBARA_INTEGRATE_REVIEW=#42595 from Intel-tensorflow:nhatle/xla_oneccl_bazel_build f4c481a PiperOrigin-RevId: 927245566
copybara-service Bot
pushed a commit
that referenced
this pull request
Jun 9, 2026
Imported from GitHub PR #42595 📝 Summary of Changes This PR enables building oneCCL from source using XLA bazel build system. [oneCCL](https://github.com/uxlfoundation/oneCCL) enables optimized communication pattern on Intel's GPUs. 🎯 Justification This PR is first step to support scale-up functionality on Intel's GPUs. Subsequent PRs will add full support for scale-up. 🚀 Kind of Contribution ✨ New Feature Copybara import of the project: -- 8e59923 by Nhat Le <nhat.le@intel.com>: Add support for oneCCL bazel build -- 6762807 by Nhat Le <nhat.le@intel.com>: Add dependency to trigger oneCCL build -- 644072a by Nhat Le <nhat.le@intel.com>: Fix EOF -- 42901eb by Nhat Le <nhat.le@intel.com>: Make all the necessary headers visible to dependents -- f4c481a by nhatle <nhat.le@intel.com>: Fix for file not found errors in XLA Linux X86 GPU ONEAPI CI Merging this change closes #42595 FUTURE_COPYBARA_INTEGRATE_REVIEW=#42595 from Intel-tensorflow:nhatle/xla_oneccl_bazel_build f4c481a PiperOrigin-RevId: 927245566
copybara-service Bot
pushed a commit
that referenced
this pull request
Jun 9, 2026
Imported from GitHub PR #42595 📝 Summary of Changes This PR enables building oneCCL from source using XLA bazel build system. [oneCCL](https://github.com/uxlfoundation/oneCCL) enables optimized communication pattern on Intel's GPUs. 🎯 Justification This PR is first step to support scale-up functionality on Intel's GPUs. Subsequent PRs will add full support for scale-up. 🚀 Kind of Contribution ✨ New Feature Copybara import of the project: -- 8e59923 by Nhat Le <nhat.le@intel.com>: Add support for oneCCL bazel build -- 6762807 by Nhat Le <nhat.le@intel.com>: Add dependency to trigger oneCCL build -- 644072a by Nhat Le <nhat.le@intel.com>: Fix EOF -- 42901eb by Nhat Le <nhat.le@intel.com>: Make all the necessary headers visible to dependents -- f4c481a by nhatle <nhat.le@intel.com>: Fix for file not found errors in XLA Linux X86 GPU ONEAPI CI Merging this change closes #42595 FUTURE_COPYBARA_INTEGRATE_REVIEW=#42595 from Intel-tensorflow:nhatle/xla_oneccl_bazel_build f4c481a PiperOrigin-RevId: 927245566
This was referenced Jun 10, 2026
copybara-service Bot
pushed a commit
that referenced
this pull request
Jun 10, 2026
Imported from GitHub PR #42595 📝 Summary of Changes This PR enables building oneCCL from source using XLA bazel build system. [oneCCL](https://github.com/uxlfoundation/oneCCL) enables optimized communication pattern on Intel's GPUs. 🎯 Justification This PR is first step to support scale-up functionality on Intel's GPUs. Subsequent PRs will add full support for scale-up. 🚀 Kind of Contribution ✨ New Feature Copybara import of the project: -- 8e59923 by Nhat Le <nhat.le@intel.com>: Add support for oneCCL bazel build -- 6762807 by Nhat Le <nhat.le@intel.com>: Add dependency to trigger oneCCL build -- 644072a by Nhat Le <nhat.le@intel.com>: Fix EOF -- 42901eb by Nhat Le <nhat.le@intel.com>: Make all the necessary headers visible to dependents -- f4c481a by nhatle <nhat.le@intel.com>: Fix for file not found errors in XLA Linux X86 GPU ONEAPI CI Merging this change closes #42595 FUTURE_COPYBARA_INTEGRATE_REVIEW=#42595 from Intel-tensorflow:nhatle/xla_oneccl_bazel_build f4c481a PiperOrigin-RevId: 929784497
copybara-service Bot
pushed a commit
to tensorflow/tensorflow
that referenced
this pull request
Jun 10, 2026
Imported from GitHub PR openxla/xla#42595 📝 Summary of Changes This PR enables building oneCCL from source using XLA bazel build system. [oneCCL](https://github.com/uxlfoundation/oneCCL) enables optimized communication pattern on Intel's GPUs. 🎯 Justification This PR is first step to support scale-up functionality on Intel's GPUs. Subsequent PRs will add full support for scale-up. 🚀 Kind of Contribution ✨ New Feature Copybara import of the project: -- 8e599233097ee4da17aea66b2ddf844ee160c471 by Nhat Le <nhat.le@intel.com>: Add support for oneCCL bazel build -- 6762807a4d7b0b6392fdfa493cdb3270ea50cefa by Nhat Le <nhat.le@intel.com>: Add dependency to trigger oneCCL build -- 644072af588c031f96342cf15cf6e65967225660 by Nhat Le <nhat.le@intel.com>: Fix EOF -- 42901eb57429c14c416636f7b5a68ce87f9c895b by Nhat Le <nhat.le@intel.com>: Make all the necessary headers visible to dependents -- f4c481a006fbd38f8eca26e04d0f75b108bb8618 by nhatle <nhat.le@intel.com>: Fix for file not found errors in XLA Linux X86 GPU ONEAPI CI Merging this change closes #42595 Reverts changelist 929420822 FUTURE_COPYBARA_INTEGRATE_REVIEW=openxla/xla#42595 from Intel-tensorflow:nhatle/xla_oneccl_bazel_build f4c481a006fbd38f8eca26e04d0f75b108bb8618 PiperOrigin-RevId: 929784497
copybara-service Bot
pushed a commit
that referenced
this pull request
Jun 10, 2026
Imported from GitHub PR #42595 📝 Summary of Changes This PR enables building oneCCL from source using XLA bazel build system. [oneCCL](https://github.com/uxlfoundation/oneCCL) enables optimized communication pattern on Intel's GPUs. 🎯 Justification This PR is first step to support scale-up functionality on Intel's GPUs. Subsequent PRs will add full support for scale-up. 🚀 Kind of Contribution ✨ New Feature Copybara import of the project: -- 8e59923 by Nhat Le <nhat.le@intel.com>: Add support for oneCCL bazel build -- 6762807 by Nhat Le <nhat.le@intel.com>: Add dependency to trigger oneCCL build -- 644072a by Nhat Le <nhat.le@intel.com>: Fix EOF -- 42901eb by Nhat Le <nhat.le@intel.com>: Make all the necessary headers visible to dependents -- f4c481a by nhatle <nhat.le@intel.com>: Fix for file not found errors in XLA Linux X86 GPU ONEAPI CI Merging this change closes #42595 FUTURE_COPYBARA_INTEGRATE_REVIEW=#42595 from Intel-tensorflow:nhatle/xla_oneccl_bazel_build f4c481a PiperOrigin-RevId: 929784497
copybara-service Bot
pushed a commit
that referenced
this pull request
Jun 11, 2026
Imported from GitHub PR #42595 📝 Summary of Changes This PR enables building oneCCL from source using XLA bazel build system. [oneCCL](https://github.com/uxlfoundation/oneCCL) enables optimized communication pattern on Intel's GPUs. 🎯 Justification This PR is first step to support scale-up functionality on Intel's GPUs. Subsequent PRs will add full support for scale-up. 🚀 Kind of Contribution ✨ New Feature Copybara import of the project: -- 8e59923 by Nhat Le <nhat.le@intel.com>: Add support for oneCCL bazel build -- 6762807 by Nhat Le <nhat.le@intel.com>: Add dependency to trigger oneCCL build -- 644072a by Nhat Le <nhat.le@intel.com>: Fix EOF -- 42901eb by Nhat Le <nhat.le@intel.com>: Make all the necessary headers visible to dependents -- f4c481a by nhatle <nhat.le@intel.com>: Fix for file not found errors in XLA Linux X86 GPU ONEAPI CI Merging this change closes #42595 FUTURE_COPYBARA_INTEGRATE_REVIEW=#42595 from Intel-tensorflow:nhatle/xla_oneccl_bazel_build f4c481a PiperOrigin-RevId: 929784497
copybara-service Bot
pushed a commit
that referenced
this pull request
Jun 11, 2026
Imported from GitHub PR #42595 📝 Summary of Changes This PR enables building oneCCL from source using XLA bazel build system. [oneCCL](https://github.com/uxlfoundation/oneCCL) enables optimized communication pattern on Intel's GPUs. 🎯 Justification This PR is first step to support scale-up functionality on Intel's GPUs. Subsequent PRs will add full support for scale-up. 🚀 Kind of Contribution ✨ New Feature Copybara import of the project: -- 8e59923 by Nhat Le <nhat.le@intel.com>: Add support for oneCCL bazel build -- 6762807 by Nhat Le <nhat.le@intel.com>: Add dependency to trigger oneCCL build -- 644072a by Nhat Le <nhat.le@intel.com>: Fix EOF -- 42901eb by Nhat Le <nhat.le@intel.com>: Make all the necessary headers visible to dependents -- f4c481a by nhatle <nhat.le@intel.com>: Fix for file not found errors in XLA Linux X86 GPU ONEAPI CI Merging this change closes #42595 FUTURE_COPYBARA_INTEGRATE_REVIEW=#42595 from Intel-tensorflow:nhatle/xla_oneccl_bazel_build f4c481a PiperOrigin-RevId: 929784497
copybara-service Bot
pushed a commit
to tensorflow/tensorflow
that referenced
this pull request
Jun 11, 2026
Imported from GitHub PR openxla/xla#42595 📝 Summary of Changes This PR enables building oneCCL from source using XLA bazel build system. [oneCCL](https://github.com/uxlfoundation/oneCCL) enables optimized communication pattern on Intel's GPUs. 🎯 Justification This PR is first step to support scale-up functionality on Intel's GPUs. Subsequent PRs will add full support for scale-up. 🚀 Kind of Contribution ✨ New Feature Copybara import of the project: -- 8e599233097ee4da17aea66b2ddf844ee160c471 by Nhat Le <nhat.le@intel.com>: Add support for oneCCL bazel build -- 6762807a4d7b0b6392fdfa493cdb3270ea50cefa by Nhat Le <nhat.le@intel.com>: Add dependency to trigger oneCCL build -- 644072af588c031f96342cf15cf6e65967225660 by Nhat Le <nhat.le@intel.com>: Fix EOF -- 42901eb57429c14c416636f7b5a68ce87f9c895b by Nhat Le <nhat.le@intel.com>: Make all the necessary headers visible to dependents -- f4c481a006fbd38f8eca26e04d0f75b108bb8618 by nhatle <nhat.le@intel.com>: Fix for file not found errors in XLA Linux X86 GPU ONEAPI CI Merging this change closes #42595 Reverts 028c67b FUTURE_COPYBARA_INTEGRATE_REVIEW=openxla/xla#42595 from Intel-tensorflow:nhatle/xla_oneccl_bazel_build f4c481a006fbd38f8eca26e04d0f75b108bb8618 PiperOrigin-RevId: 929784497
Member
|
Hi @nhatleSummer22 . We are merging this without the change to |
copybara-service Bot
pushed a commit
to tensorflow/tensorflow
that referenced
this pull request
Jun 11, 2026
Reverts 028c67b FUTURE_COPYBARA_INTEGRATE_REVIEW=openxla/xla#42595 from Intel-tensorflow:nhatle/xla_oneccl_bazel_build f4c481a006fbd38f8eca26e04d0f75b108bb8618 PiperOrigin-RevId: 930185853
copybara-service Bot
pushed a commit
to tensorflow/tensorflow
that referenced
this pull request
Jun 11, 2026
To add support for other operations like all-gather/collective dots to the collective kernel thunk we should make it agnostic to custom kernels first. Reverts 028c67b FUTURE_COPYBARA_INTEGRATE_REVIEW=openxla/xla#42595 from Intel-tensorflow:nhatle/xla_oneccl_bazel_build f4c481a006fbd38f8eca26e04d0f75b108bb8618 PiperOrigin-RevId: 930464574
This was referenced Jun 11, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
5 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
📝 Summary of Changes
This PR enables building oneCCL from source using XLA bazel build system. oneCCL enables optimized communication pattern on Intel's GPUs.
🎯 Justification
This PR is first step to support scale-up functionality on Intel's GPUs. Subsequent PRs will add full support for scale-up.
🚀 Kind of Contribution
✨ New Feature