Enhance device context pool #9293
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
dzhwinter
reviewed
Mar 21, 2018
| namespace paddle { | ||
| namespace platform { | ||
|
|
||
| DeviceContextPool* DeviceContextPool::pool = nullptr; | ||
|
|
||
| const platform::DeviceContext* DeviceContextPool::Get( | ||
| const platform::Place& place) { | ||
| platform::DeviceContext* DeviceContextPool::Get(const platform::Place& place) { |
There was a problem hiding this comment.
platform::DeviceContext* DeviceContextPool::Get(const platform::Place& place) const| @@ -65,6 +65,18 @@ bool is_cpu_place(const Place &); | |||
| bool places_are_same_class(const Place &, const Place &); | |||
| bool is_same_place(const Place &, const Place &); | |||
|
|
|||
| struct PlaceHash { | |||
| std::size_t operator()(const Place &p) const { | |||
| constexpr size_t num_dev_bits = 4; | |||
There was a problem hiding this comment.
4 bit is not enough, the GPU box product has 32 cards in one node. Then will lead to an overlap of dev_id << num_dev_bits | p.which()
| @@ -159,7 +160,7 @@ class DeviceContextPool { | |||
| } | |||
|
|
|||
| /*! \brief Return handle of single device context. */ | |||
| const platform::DeviceContext* Get(const platform::Place& place); | |||
| platform::DeviceContext* Get(const platform::Place& place); | |||
There was a problem hiding this comment.
should add const suffix?
| @@ -159,7 +160,7 @@ class DeviceContextPool { | |||
| } | |||
|
|
|||
| /*! \brief Return handle of single device context. */ | |||
| const platform::DeviceContext* Get(const platform::Place& place); | |||
| platform::DeviceContext* Get(const platform::Place& place); | |||
|
|
|||
| template <typename Place> | |||
| const typename DefaultDeviceContextType<Place>::TYPE* GetByPlace( | |||
mikeseven
added a commit
to mikeseven/Paddle
that referenced
this pull request
Mar 22, 2018
* commit '9c35b0dc1ba0ace5acf721685802a21045ea1249': (36 commits) Fix dist compile error (PaddlePaddle#9320) Fix bug for backward tanspiler when using parallel_do operator. (PaddlePaddle#9282) update fix transpiler bug Update index_en.rst (PaddlePaddle#9286) "fix mixed_vector bug" (PaddlePaddle#9319) Update index_en.rst (PaddlePaddle#9280) Adjust some contents in write_docs_en.rst for Contribue Documentation (PaddlePaddle#9147) CMake refine for HIP support. Fix CI. Reuduce memory copy when communication between trainer and pserver. (PaddlePaddle#9271) Modified build.sh and remove build_doc.sh fix doc Enhance device context pool (PaddlePaddle#9293) Device blobs are created only in training. Added testing attribute Shrink batch_norm_grad's inputs updates prepare and create op before run wip small fix initial commit ... # Conflicts: # cmake/external/eigen.cmake
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.