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

Register the memory mapped buffer in datasource to improve H2D throughput #13814

Merged
merged 17 commits into from
Aug 16, 2023
Merged

Register the memory mapped buffer in datasource to improve H2D throughput #13814

merged 17 commits into from
Aug 16, 2023

Conversation

vuule
Copy link
Contributor

@vuule vuule commented Aug 3, 2023 •
edited
Loading

Description

On systems where pageable memory uses host page tables, cudaHostRegister is very cheap. Since host buffer registration can improve throughput, datasource now registers the entire memory mapped buffer when host page tables are used.

This mainly impacts the CSV reader, which reads input files using a host_read call.

Checklist

  • I am familiar with the Contributing Guidelines.
  • New or existing tests cover these changes.
  • The documentation is up to date with these changes.

@vuule vuule added feature request New feature or request cuIO cuIO issue Performance Performance related issue labels Aug 3, 2023
@vuule vuule self-assigned this Aug 3, 2023
@vuule vuule added the non-breaking Non-breaking change label Aug 3, 2023
@github-actions github-actions bot added the libcudf Affects libcudf (C++/CUDA) code. label Aug 3, 2023
@vuule vuule changed the title Register memory mapped buffer in datasource to improve H2D throughput Register the memory mapped buffer in datasource to improve H2D throughput Aug 7, 2023
@vuule vuule marked this pull request as ready for review August 7, 2023 20:05
@vuule vuule requested a review from a team as a code owner August 7, 2023 20:05
davidwendt
davidwendt requested changes Aug 10, 2023
View reviewed changes
cpp/src/io/utilities/datasource.cpp Outdated Show resolved Hide resolved
@vuule @davidwendt
Add CUDF_CUDA_TRYs
cd025b8 
Co-authored-by: David Wendt <[email protected]>
bdice
bdice approved these changes Aug 16, 2023
View reviewed changes
Copy link
Contributor

@bdice bdice left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Looks fine!

@vuule vuule added the 5 - Ready to Merge Testing and reviews complete, ready to merge label Aug 16, 2023
@jlowe jlowe mentioned this pull request Aug 16, 2023
Open
@vuule
Copy link
Contributor Author

vuule commented Aug 16, 2023

/merge

@rapids-bot rapids-bot bot merged commit 3f99569 into rapidsai:branch-23.10 Aug 16, 2023
@vuule vuule deleted the perf-read_csv-register-mmap branch August 16, 2023 21:46
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@davidwendt davidwendt davidwendt approved these changes

@bdice bdice bdice approved these changes

@karthikeyann karthikeyann Awaiting requested review from karthikeyann karthikeyann is a code owner automatically assigned from rapidsai/cudf-cpp-codeowners

Assignees

@vuule vuule

Labels
5 - Ready to Merge Testing and reviews complete, ready to merge cuIO cuIO issue feature request New feature or request libcudf Affects libcudf (C++/CUDA) code. non-breaking Non-breaking change Performance Performance related issue
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
3 participants
@vuule @bdice @davidwendt