This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: [GSoC'19, libgomp work-stealing] Task parallelism runtime
> Another option, which I guess starts to go out of scope of your gsoc, is
> parallel depth first (PDF) search (Blelloch 1999) as an alternative to work
> stealing. Here's a presentation about some recent work in this area,
> although for Julia and not OpenMP (no idea if PDF would fit with OpenMP at
> all): https://www.youtube.com/watch?v=YdiZa0Y3F3c
I am actually aware of PDF and the works ongoing on the Julia side.
Also, I think it does not go out of the scope of GSoC,
since the essential goal is to implement a more advanced task parallel scheduler anyway.
> Better cache locality.
Despite previous research results that PDF is better in term of locality,
recently developed advanced work-stealing (WS) schemes improved a lot in terms of data locality.
I think an up-to-date quantitive comparison with SOTA algorithms from both sides is required.
Personally I think the WS framework is more flexible and popular? right now.
I'd like to hear the opinion of others on the subject.
Ray Kim