Improved CPU/GPU interoperability #5001
Merged
Conversation
This file contains 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
viclafargue
force-pushed
the
cpu-gpu-interop-improvements
branch
from
d8dcaaf to
fa8b051
Compare
raydouglass
pushed a commit
that referenced
this pull request
Dec 15, 2022
**Purpose of this PR** - Provide an implementation of CumlArray that can be backed by either host or device memory - Hook the host-compatible CumlArray into infrastructure for device and memory type selection - Ensure that device-host and host-device transfers are minimized for both GPU and CPU execution of existing CPU/GPU interoperable models **Non-goals of this PR** - This PR is not intended to provide a version of CUML that can be imported or run on a CPU-only system - This PR does not allow CUML to be compiled without nvcc - This PR is not intended to document the CPU/GPU interoperable estimator infrastructure introduced by #4908 and improved in #5001 - This PR is intended *only* to avoid breaking existing functionality. New functionality related to CPU-only execution which is not currently being used in the codebase will be tested in a separate PR **Tests performed on this PR** Besides the obvious (standard test suite), benchmarks have been run against this PR for a variety of conversions to and from the new CumlArray. Additionally, the standard cuML benchmark suite was run against this PR and compared to results from current branch-22.12. Results for both are described below. **Notable features of this PR not immediately related to its purpose** - Implementing the new CumlArray infrastructure required us to break many of the existing circular imports in the codebase - New [utilities](https://github.com/wphicks/cuml/blob/fea-xpu_infra/python/cuml/internals/safe_imports.py) for "safe" imports in the context of GPU-only or CPU-only installs were introduced **Anticipated follow-on PRs** - #5001 - #4970 - PR(s) to make all CPU/GPU-exclusive imports "safe" - PR(s) to add additional testing for CPU-only functionality Authors: - William Hicks (https://github.com/wphicks) - Victor Lafargue (https://github.com/viclafargue) - Corey J. Nolet (https://github.com/cjnolet) Approvers:
dantegd
added
improvement
dantegd
approved these changes
Dec 16, 2022
|
rerun tests |
Codecov Report
Additional details and impacted files@@ Coverage Diff @@
## branch-23.02 #5001 +/- ##
===============================================
Coverage ? 79.24%
===============================================
Files ? 191
Lines ? 12374
Branches ? 0
===============================================
Hits ? 9806
Misses ? 2568
Partials ? 0
Flags with carried forward coverage won't be shown. Click here to find out more. Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here. ☔ View full report at Codecov. |
|
@gpucibot merge |
jakirkham
pushed a commit
to jakirkham/cuml
that referenced
this pull request
Feb 27, 2023
**Purpose of this PR** - Provide an implementation of CumlArray that can be backed by either host or device memory - Hook the host-compatible CumlArray into infrastructure for device and memory type selection - Ensure that device-host and host-device transfers are minimized for both GPU and CPU execution of existing CPU/GPU interoperable models **Non-goals of this PR** - This PR is not intended to provide a version of CUML that can be imported or run on a CPU-only system - This PR does not allow CUML to be compiled without nvcc - This PR is not intended to document the CPU/GPU interoperable estimator infrastructure introduced by rapidsai#4908 and improved in rapidsai#5001 - This PR is intended *only* to avoid breaking existing functionality. New functionality related to CPU-only execution which is not currently being used in the codebase will be tested in a separate PR **Tests performed on this PR** Besides the obvious (standard test suite), benchmarks have been run against this PR for a variety of conversions to and from the new CumlArray. Additionally, the standard cuML benchmark suite was run against this PR and compared to results from current branch-22.12. Results for both are described below. **Notable features of this PR not immediately related to its purpose** - Implementing the new CumlArray infrastructure required us to break many of the existing circular imports in the codebase - New [utilities](https://github.com/wphicks/cuml/blob/fea-xpu_infra/python/cuml/internals/safe_imports.py) for "safe" imports in the context of GPU-only or CPU-only installs were introduced **Anticipated follow-on PRs** - rapidsai#5001 - rapidsai#4970 - PR(s) to make all CPU/GPU-exclusive imports "safe" - PR(s) to add additional testing for CPU-only functionality Authors: - William Hicks (https://github.com/wphicks) - Victor Lafargue (https://github.com/viclafargue) - Corey J. Nolet (https://github.com/cjnolet) Approvers:
jakirkham
pushed a commit
to jakirkham/cuml
that referenced
this pull request
Feb 27, 2023
Authors: - Victor Lafargue (https://github.com/viclafargue) - William Hicks (https://github.com/wphicks) - Corey J. Nolet (https://github.com/cjnolet) - Dante Gama Dessavre (https://github.com/dantegd) - Carl Simon Adorf (https://github.com/csadorf) Approvers: - Dante Gama Dessavre (https://github.com/dantegd) URL: rapidsai#5001
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
No description provided.