Optimize TensorList resizing. #5638
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
Signed-off-by: Michal Zientkiewicz <[email protected]>
Signed-off-by: Michał Zientkiewicz <[email protected]>
|
CI MESSAGE: [18510898]: BUILD STARTED |
JanuszL
reviewed
Sep 18, 2024
JanuszL
approved these changes
Sep 18, 2024
|
CI MESSAGE: [18510898]: BUILD PASSED |
mzient
commented
Sep 18, 2024
| template <typename Backend> | ||
| void TensorList<Backend>::recreate_views() { | ||
| // precondition: type, shape are configured | ||
| uint8_t *sample_ptr = static_cast<uint8_t *>(contiguous_buffer_.raw_mutable_data()); | ||
| int64_t num_samples = shape().num_samples(); | ||
| auto &data_ptr = contiguous_buffer_.get_data_ptr(); |
There was a problem hiding this comment.
Hoisting this line was perhaps the biggest saving here.
mzient
commented
Sep 18, 2024
| for (int64_t i = 0; i < num_samples; i++) { | ||
| // or any other way | ||
| auto tensor_size = shape().tensor_size(i); | ||
|
|
||
| std::shared_ptr<void> sample_alias(contiguous_buffer_.get_data_ptr(), sample_ptr); | ||
| tensors_[i].ShareData(sample_alias, tensor_size * type_info().size(), is_pinned(), shape()[i], | ||
| tensors_[i].ShareData(std::shared_ptr<void>(data_ptr, sample_ptr), |
There was a problem hiding this comment.
Having an intermediate variable and moving it was noticeably slower (but still noticeably faster than passing by const-ref and copying).
awolant
approved these changes
Sep 19, 2024
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Outdated suggestions cannot be applied.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.
Category:
Refactoring (Redesign of existing code that doesn't affect functionality)
Description:
This change optimizes the performance of
TensorList::Resizeshared_ptrin ShareData is now passed by value, allowing move semantics and reducing the number of atomic operationsAdditional information:
Many of the changes were tuned experimentally. Don't hesitate to ask if you see something not obvious or outright weird.
Affected modules and functionalities:
Buffer, Tensor, TensorList
Key points relevant for the review:
Tests:
No new functionality or functional changes - all existing tests apply
Checklist
Documentation
DALI team only
Requirements
REQ IDs: N/A
JIRA TASK: N/A