VecMem Update, main branch (2024.10.29.) #757
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
|
|
Great! I will let @stephenswat take a look into this for detailed review |
|
Closes #645. |
stephenswat
approved these changes
Oct 29, 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Outdated suggestions cannot be applied.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.
Updated the project to using vecmem-1.11.0.
This update, most importantly, includes: acts-project/vecmem#299 Which was made by @stephenswat to get the memory usage of our code under control. It also includes acts-project/vecmem#286, which is the reason for all the actual code changes in the PR...
Currently
traccc_throughput_st_cudauses the following amount of memory to reconstruct 100 mu=140 events:So, about 3.85 GB at its peak.
With this update included, this looks like:
About 2.77 GB at its peak.
To complete the picture, this is what I get when using vecmem::pool_memory_resource instead of vecmem::binary_page_memory_resource:
About 3.3 GB at its peak.
And finally, if I turn off memory caching completely, I get:
I.e. 200-700 MB of memory usage.
In all these cases, on the RTX 2060 GPU of my desktop machine, the throughput of reconstructing mu=140 events in a single thread/stream remained in +-10%. Without caching being the slowest of course, and
vecmem::pool_memory_resource(which is based on CUB and std::pmr::unsynchronized_pool_resource, being the fastest. But only by little. Apparently with the full chain running, the caching is not dominating the performance at the moment. 🤔