[Mesa-users] Technical issue: calculations going incredibly slow

[Warrick Ball]

> This is indeed a good hint. I mean, I don't know which thermal zones represent, but some are indeed > 85 °C. And considering I have a HP Omen, which is famous for overheating, this is surely something I need to consider.

To be honest, I don't know what the thermal zones are either but it's a handy way of quickly getting an idea of the temperature without having to install any software.

> What makes me wonder is how it is possible that only MESA is slowed by the CPU overheating and not all the other processes. Is it something possible?

I don't think I've ever actually hit an overheating problem, so I don't have an intuition for how application performance "feels" when the system decides to slow down the CPU to avoid overheating.  I would only expect a similar performance hit in applications that, like MESA, use nearly all of your cores nearly continuously for long enough to heat up the CPU to a potentially damaging temperature.

A potential workaround might be to not run MESA on all available cores and see if you can strike a balance between how hot the CPU gets and how long it can hold out before overheating (or, ideally, if it can reach an equilibrium where it never gets dangerously hot).

Anyway, CPU overheating isn't really a MESA issue.  If anyone else on the user list has ideas, they might chip in, but you'll probably have more luck with a forum for your hardware.

Cheers,
Warrick


___________

Warrick Ball
Postdoc, School of Physics and Astronomy
University of Birmingham, Edgbaston, Birmingham B15 2TT
W.H.Ball at bham.ac.uk
+44 (0)121 414 4552

On Tue, 4 May 2021, amedeoromagnolo at gmail.com wrote:

> Hi Warrick,
> 
> The evolutionary phase is completely random. The slowing down is uniquely dependent on when I run it the first time on my pc.
>
>       ask these questions because I have two ideas.  First, you might be filling enough RAM that the various microphysics tables start using the hard drive page file, which could be slow.  If this might be the case, try adding
>
>            num_steps_for_garbage_collection = 1000 ! or something greater than default 0
>
>       to the `&star_job` part of your inlist.
> 
> 
> I checked it and it wasn't the problem. Even because I have a 50 GB RAM and I would be very worried if MESA took so much space :)
>
>       My second (probably wrong) idea is that your CPU is overheating and the system is slowing it down.  I'm not sure about this because it isn't clear that rebooting would give it enough time to cool off.  There are many tools available but
>       on Linux you can try [1]
>
>            cat /sys/class/thermal/thermal_zone*/temp
>
>       for a quick look (output is in "millicelsius").  If this is getting to ~80-85°C, this might be your problem.
> 
> 
> This is indeed a good hint. I mean, I don't know which thermal zones represent, but some are indeed > 85 °C. And considering I have a HP Omen, which is famous for overheating, this is surely something I need to consider. What makes me wonder is how
> it is possible that only MESA is slowed by the CPU overheating and not all the other processes. Is it something possible?
> 
> Amedeo
> 
> 
> 
> On Tue, 4 May 2021 at 11:20, Warrick Ball <W.H.Ball at bham.ac.uk> wrote:
>       Hi Amedeo,
>
>       I haven't tried your inlists (thanks for attaching them!) but a few more details would help.  What phase of evolution is the star in when it starts slowing down?  If you reboot and run `./rn`, does it always start slowing down at the same
>       timestep?  Or is it always after about 10 minutes, and not necessarily at the same timestep?
>
>       I ask these questions because I have two ideas.  First, you might be filling enough RAM that the various microphysics tables start using the hard drive page file, which could be slow.  If this might be the case, try adding
>
>            num_steps_for_garbage_collection = 1000 ! or something greater than default 0
>
>       to the `&star_job` part of your inlist.
>
>       My second (probably wrong) idea is that your CPU is overheating and the system is slowing it down.  I'm not sure about this because it isn't clear that rebooting would give it enough time to cool off.  There are many tools available but
>       on Linux you can try [1]
>
>            cat /sys/class/thermal/thermal_zone*/temp
>
>       for a quick look (output is in "millicelsius").  If this is getting to ~80-85°C, this might be your problem.
>
>       Cheers,
>       Warrick
>
>       [1] https://askubuntu.com/questions/15832/how-do-i-get-the-cpu-temperature
>
>       ___________
>
>       Warrick Ball
>       Postdoc, School of Physics and Astronomy
>       University of Birmingham, Edgbaston, Birmingham B15 2TT
>       W.H.Ball at bham.ac.uk
>       +44 (0)121 414 4552
>
>       On Tue, 4 May 2021, mesa-users at lists.mesastar.org wrote:
>
>       > Hi all,
>       >
>       > here is my problem. I have set a pretty vanilla simulation on MESA (inlists attached), but because of some weird black magic my calculations go fast enough over the first ten minutes, and then they dramatically slow down. And I'm not
>       talking about
>       > timesteps, but about the actual calculation time for each timestep.
>       >
>       > Even more weirdly, if I reset my PC and restart the simulations (./re) the code goes fast for another ~10 minutes and then slows again.
>       >
>       > I'm using Ubuntu 20.04, and I made sure no other process was running at the same time. And yes: I set the right number of cores.
>       >
>       > Did this happen to anyone else? Do you have any idea on how to solve this?
>       >
>       > Cheers,
>       >
>       > Amedeo
>       >
>       >
> 
> 
>