I’m hesitating to make this post because I know how vague it is but the fact is that this already happened to me 6 times.
After using Pluto, and happened also once with Franklin, my computer becomes extremely slow. An by extremely I mean > 4 times slower. Windows task manager shows that the clock speed does not go over 0.79 GHz. When this occurs I have to shutdown and restart. A simple reboot is not enough to restore the normal functioning.
Now, the most strange part is that the slowness starts one or two DAYS after having started Pluto. This seems to occur more frequently with Julia 1.7 but last time it occurred was with 1.6.
Tried only once with Neptune and saw no such effect but can’t rule out that possibility. My default browser is Firefox.
I know how crazy this sounds, but 6 times leave me no doubt.
Julia Version 1.7.0-beta2
Commit b570546b68 (2021-06-20 06:31 UTC)
OS: Windows (x86_64-w64-mingw32)
CPU: Intel(R) Core(TM) i7-8750H CPU @ 2.20GHz
LLVM: libLLVM-12.0.0 (ORCJIT, skylake)
JULIA_NUM_THREADS = 6
How much memory usage do you have?
It’s not that. I have 32 GB and usage never got close to it. It’s really the processor clock that gets stuck to no more than 0.79 GHz. Again, a reboot does not solve the problem.
And more, this happens also if I kill the Pluto process. That is, it may happen one day after having finished using Pluto.
(I know this is not good for my reputation but 6 times…)
Please monitor the CPU temperature. The overheating causes the CPU to slow down.
@joa-quim Your reputation is good here. I build supercompercomputers for a living, and I deal with issues with thermal limits all the time. I am dealing with one right now.
As others have said I would install a package which monitors CPU frequence and sensors.
On Gnome / Linux I use OSPower on my laptop.
My advice would be to cap your CPU frequency before starting - keep the CPU within limits you choose.
Thanks guys but it’s not the temperature either. This is a 2 years old laptop and the fan was not even spinning.
I wonder whether the fact that you use Julia with Windows + multiple threads on nightly is what might cause problems (though you say you observe this with 1.6 so idk). Franklin and Pluto (and Weave) run code for you in sandboxed environments, but what happens in the sandbox is up to you, it may be that some of this doesn’t get cleared properly but I don’t know what/why.
It’s going to be a bit of guess work here, maybe a good start would be to try to run things in a session with a single thread and see if you still observe issues. If you don’t anymore, then the culprit is to do with windows+multithreading & I doubt it’s something Pluto/Franklin/Weave could do something about. If you still do then we’re back at square one but given you’re the first to report this (to the best of my knowledge, at least on the Franklin side) unfortunately that’s likely to take time to figure out what the problem really is…
PS: could you maybe comment on what code you were running within Pluto/Franklin when this happened? was it always the same or just random snippets of code with no other package?
I had the impression that this only occurred with Julia 1.7dev but last time I had 1.6 running. But given the time it takes to show up I can’t guarantee I hadn’t run a 1.7 session (I think not but can’t be sure).
When it happened with Pluto when I was using GMT. But with Franklin I just started it with the default example and did nothing else. And in fact I just did it again 1 hour ago to see if it repeats.
Yes, I also think so but this can’t happen only to me.
It did happen again. 3 days after short use of Franklin. Sorry, I already had launched Franklin before seeing the suggestion of using Julia with single thread.
I understand that little can be done with so few information. I think I wont use those programs anymore and I’ll report if this happens again.
Hi, I have also experiences the same time of problems recently. I had to rebook the compute (in my case a Mac running Catalina) or exit pluto (and Mozilla) and start again.