-
Notifications
You must be signed in to change notification settings - Fork 708
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.
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
Use taskflow to create asynchronous tasks. #16976
Conversation
Here's some good news: This tests correctly (with the exception of the one CI check that is down). This means that we have some hope to eventually convert to TaskFlow after all. |
Hi, I'm a deal.II user and I use WorkStream extensively in my project. It has greatly boosted my matrix assembly (with significant speedup up to 4 to 8 cores depending on the size, and gradually saturate beyond that). I'm curious about are we now transitioning from TBB to TaskFlow? What’s the reason for this shift? Efficiency or to prepare for hybrid- and heterogenous-parallel computing? |
@QY-Shi The most notable reason is that TaskFlow is header-only and only relies on standard C++ features, whereas TBB has platform-specific code that is not easy to compile (or cannot be used at all) on some platforms. |
Got it. Thanks again for such a comprehensive and efficient numerical library with the WorkStream functionality. I really enjoyed using it! |
Anyone feels like hitting the green button? I think it's relatively low-risk given that (i) we don't by default use TaskFlow, (ii) that one CI check that does use TaskFlow succeeds. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I do. 😄
This supersedes #15634. It works on my laptop for all tests in
multithreading/
.I will note that the "Jenkins: serial" CI check is configured to use TaskFlow, so this patch is actually tested by the CI infrastructure.