[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [tlaplus] Current status of TLC multi-core scaling
On 10/15/21 8:37 AM, Ognjen Maric wrote:
I'm using TLC to analyze a spec on a machine with 64 cores, and I'm
seeing significantly diminishing returns when adding new cores (a few
percent speed-up going from 10 to 60 workers), and I'd like to know what
the state of the art options for scaling TLC (obviously, I will profile
Searching the mailing list archives, I found a discussion from a couple
of years ago  where it was suggested to use OffHeapDiskFPSet, but the
state queue implementation was still mentioned as a bottleneck. The
1.6.0 release notes also mentioned ByteArrayQueue, switching to which
yielded a 30% speedup with many cores. Is using these two the best I can
do today? Is there anything else on the horizon that might improve scaling.
Also, have you considered documenting the options to switch the
fingerprint/queue implementations in current-tools.pdf? They were not
easy to find otherwise. I'm happy to send a PR...
it is great that ByteArrayQueue combined with OffHeapDiskFPSet improves
scaling for you. This combination is the best you can do as of today,
and it would be greatly appreciated if you could create a PR that
updates current-tools.md .
As to future work on scaling TLC, a design of a scalable state queue
exist, but I halted working on it in favor of other features. If a
volunteer would like to pick up where I left off, I'm more than happy to
You received this message because you are subscribed to the Google Groups "tlaplus" group.
To unsubscribe from this group and stop receiving emails from it, send an email to tlaplus+unsubscribe@xxxxxxxxxxxxxxxx.
To view this discussion on the web visit https://groups.google.com/d/msgid/tlaplus/17d07086-f8d9-41c1-edb9-f177d240c662%40lemmster.de.