-
Notifications
You must be signed in to change notification settings - Fork 2
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
Faster startup #1
Conversation
It's totally useless to open a PR here as these packages are automatically generated, so I'm going to close it. Also, did you actually see a performance improvement? Some time ago I tried this exact change and couldn't see any meaningful speedup. |
Only speedup on Julia 1.5+. [You closed when I was typing, but I confirmed it can be faster, yes, by doing a bit differently, and I was actually thinking of adding to the script that generates after confirming works for this one.] FFMPEG_jll is slowing down startup of Plots, by:
Actually using FFMPEG_jll took a bit more time (I guess random variation, as that part should be less), so I start with adding the optimization here (do both?). I take about half a sec off with only -O1, but the opt I would really want, that's currently only possible this way:
|
Did you actually try your patch? I'm not sure you did. Current % julia-15 --startup-file=no -q
julia> @time using FFMPEG_jll
1.127921 seconds (1.16 M allocations: 70.307 MiB, 1.70% gc time) With your patch: % julia-15 --startup-file=no -q
julia> @time using FFMPEG_jll
1.188155 seconds (1.15 M allocations: 70.166 MiB, 1.66% gc time) |
I always only tried applying the change on a single package, and the difference is really meaningless. I don't consider it a regression, the change is very small. Applying the change on all dependencies to see whether there is a cumulative effect is a task that has been on my backlog for some time. If it is proved that this provides a real speedup this change should be applied in BB, not here anyway |
I've confirmed your non-change in speed by using my change as is (but see below about seemingly otherwise, or why):
but then I get speedup with it if I do:
and try again, so it seems the change isn't actually enforcing With the same startup setting I DO get a speedup if I do:
This is not the full speedup I wanted, but something, and note something is happening as it lowers allocations. [In your case you did get 1.15 M vs 1.16 M allocations before, so something happened.] |
No description provided.