Octave 6.1 issue

classic Classic list List threaded Threaded
3 messages Options
Reply | Threaded
Open this post in threaded view
|

Octave 6.1 issue

Rory Conolly
Hi, I've been working on a large optimization project (several hundred lines of code in multiple files spread over a main  directory  (folder) and several sub-directories). Have been making good progress with the work in Octave 5.2. Typical optimization runs take 6 to 12 hr on a Windows 10 computer with an AMD Ryzen 9 3900X processor and plenty of RAM, with each iteration of the optimization taking about 25 - 30 seconds. A typical optimization run can involve 300 to over 1000 iterations of the main program.

I recently installed Octave 6.1 on the same computer. The optimization runs identically, initially. But with time, the duration of each iteration starts to increase. After a couple of hundred iterations, each iteration is taking over a minute. This does not happen in Octave 5.2 - the iteration time stays the same until the optimization run completes.

I'm using fminsearch, but I have adapted fminsearcbnd from the MATLAB user library. This allows me to specify bounds for the parameters being varied, which is important for my problem. So I call fminsearchbnd, which in turn calls fminsearch.

The code is too large to post here but I thought I should bring this to the attention of the developers. Happy to help if there is something I can do.

- Rory Conolly


Reply | Threaded
Open this post in threaded view
|

Re: Octave 6.1 issue

Andreas Weber-6
Am 29.12.20 um 17:09 schrieb Rory Conolly:
> ...
> I recently installed Octave 6.1 on the same computer. The optimization
> runs identically, initially. But with time, the duration of each
> iteration starts to increase. After a couple of hundred iterations, each
> iteration is taking over a minute.

Hi Rory, do you see an increase of RAM usage while running? Is it
possible that you system starts swapping?

-- Andy


Reply | Threaded
Open this post in threaded view
|

Re: Octave 6.1 issue

yamane
In reply to this post by Rory Conolly
Hi,

There is a bug on Octave 6.1 reducing the performance a lot: https://savannah.gnu.org/bugs/?59711

Kind regards,
Renato



Em ter, 29 de dez de 2020 16:10, Rory Conolly <[hidden email]> escreveu:
Hi, I've been working on a large optimization project (several hundred lines of code in multiple files spread over a main  directory  (folder) and several sub-directories). Have been making good progress with the work in Octave 5.2. Typical optimization runs take 6 to 12 hr on a Windows 10 computer with an AMD Ryzen 9 3900X processor and plenty of RAM, with each iteration of the optimization taking about 25 - 30 seconds. A typical optimization run can involve 300 to over 1000 iterations of the main program.

I recently installed Octave 6.1 on the same computer. The optimization runs identically, initially. But with time, the duration of each iteration starts to increase. After a couple of hundred iterations, each iteration is taking over a minute. This does not happen in Octave 5.2 - the iteration time stays the same until the optimization run completes.

I'm using fminsearch, but I have adapted fminsearcbnd from the MATLAB user library. This allows me to specify bounds for the parameters being varied, which is important for my problem. So I call fminsearchbnd, which in turn calls fminsearch.

The code is too large to post here but I thought I should bring this to the attention of the developers. Happy to help if there is something I can do.

- Rory Conolly