Update packages (pybind11 and catch2) and do not use nvidia-docker command #618
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
|
/ok to test |
|
(@gigony, I sent you a slack message with info on how to comply with newer RAPIDS signing requirements, so adding "ok to test" comments won't be required for CI to run) |
Without upgrading Catch2, The following error occurs when building on
Ubuntu 22.04 due to glibc:
cucim/build-debug/_deps/deps-catch2-src/single_include/catch2/
catch.hpp:10830:58: error: call to non-‘constexpr’ function
‘long int sysconf(int)’
10830 | static constexpr std::size_t sigStackSize = 32768 >=
MINSIGSTKSZ ? 32768 : MINSIGSTKSZ;
Signed-off-by: Gigon Bae <gbae@nvidia.com>
This PR updates pybind11 to v2.11.1. Even with the latest version of pybind11, we still have an issue with `pybind11::array_t` when cuCIM is used in multithread without importing numpy in the main thread. pybind/pybind11#4877 Will need to wait for the next release of pybind11. Signed-off-by: Gigon Bae <gbae@nvidia.com>
nvidia-docker binary is not available if user doesn't install nvidia-docker2 package. This change uses runtime option instead of using nvidia-docker command. Signed-off-by: Gigon Bae <gbae@nvidia.com>
gigony
force-pushed
the
update_packages
branch
from
e135ac9
to
4b3d3b8
Compare
This applies the following patches to pybind11: - pybind/pybind11#4857 - pybind/pybind11#4877 to avoid deadlock when using pybind11 without importing numpy in multi-threaded environment.
jakirkham
approved these changes
Oct 30, 2023
|
/merge |
|
Thanks all! 🙏 |
rapids-bot bot
pushed a commit
that referenced
this pull request
Oct 30, 2023
MERGE after #618 ### Fix memory corruption issue in ThreadBatchDataLoader Fix #592 - Return a shared pointer of CuImageIterator, taking ownership of CuImageIterator instance, instead of returning a copy of CuImageIterator - Do not create new instance for CuImageIterator when `__iter__` is called. Instead, return self. - Wait until all threads are finished in ThreadBatchDataLoader destructor ### Fix memory leak in ThreadBatchDataLoader Fix #598 - Throw StopIteration exception early - Do not set nullptr for `raster_data_` in `ThreadBatchDataLoader::next_data()` - Setting nullptr prevents freeing memory in the destructor of ThreadBatchDataLoader ### Fix fatal errors for cache object - GIL is not acquired when using Python object. ### Add iterator-related test cases for memory issue Authors: - Gigon Bae (https://github.com/gigony) Approvers: - Gregory Lee (https://github.com/grlee77) - https://github.com/jakirkham URL: #620
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.
Update Catch2 to v3.4.0
Without upgrading Catch2, The following error occurs when building on
Ubuntu 22.04 due to glibc:
Update pybind11 to v2.11.1
Even with the latest version of pybind11, we still have an issue
with
pybind11::array_twhen cuCIM is used in multithread withoutimporting numpy in the main thread.
See pybind/pybind11#4877
Will need to wait for the next release of pybind11.
Use runtime option instead of using nvidia-docker command
nvidia-docker binary is not available if user doesn't install
nvidia-docker2 package. This change uses runtime option instead
of using nvidia-docker command.
Apply pybind11 patch to avoid deadlock (until new release is available)
This applies the following patches to pybind11:
numpy._coreimports for NumPy 2.0 pybind/pybind11#4857to avoid deadlock when using pybind11 without importing numpy in
multi-threaded environment.