Question for developers - using all computational cores during optimisation - page 2

 
++, for the efficient use of resources.
 

I would like to remind the developers once again

the situation is horrible right now, 16 cores on the main server are fulfilling 20 jobs and 40 more are idle on the network, please solve the problem of making the most of all cores .... this is just catastrophically slowing down the optimization process

 
Boris Egorov:

Why during the optimization metatrader 5 (latest versions) does not use all available computational cores? I marked in green the cores that perform calculations, the rest are idle.

They are saving your resources ))))

Seriously though, I'm restraining myself with an inordinate amount of force so as not to swear at MQ.

Everywhere you touch it gets worse.

Thought they already fixed it, but no. In the new builds some of the agents go into finish mode. You can resuscitate these agents if one of them(s) turn off and on again.

Then GA does not wait, as Slava wrote 85 tasks, and each agent gets a new task immediately after preceeding calculation.

Developers, turn GA back to version 2000! Everything was great there!

 
Boris Egorov:
Sergey Chalyshev:

If you really want to change and not just grumble (like me), why not give developers a reproducible example, where kernels are disabled/standby?

You can based on standard EA (if it's reproducible with it too), but with as much detail as possible, so they can reproduce the behaviour for themselves.

 
Andrey Khatimlianskii:

If you really want a change and not just a grumble (like me), why don't you give the developers a reproducible example where the cores are shutdown/standby?

What kind of example ... there is no one that optimizes for many parameters, for example from 30 different parameters, i posted screenshots, i can do it again

I posted the screenshots, I can share them again,

the problem is that the optimizer gives new jobs to a small number of local agents as soon as they finish counting them out, and the free network agents, which by definition are much more numerous, keep on idling, it's idiotic

... this slows down optimization by at least 10 times, which is extremely important, either calculate in 1-2 hours or like now - 10 hours ..... what was previously 3 hours is now 11 and a half hours ... developers are so fucking optimized

the developers have also imposed a restriction on network agents, only physical cores are allowed, this is a global bug,

this is a global bug - it is not the programmers of metatrader who should decide - but the user, for example my servers are all ok, so i really need all the threads - thisrestriction on physical cores only for networking agents is stupid

Go back to the way it was before

Optimiser task allocation error
 
Boris Egorov:

What kind of example ... Any example that optimises for multiple parameters, e.g. for 30 different parameters, I've posted screenshots, I can do it again.

Any example that can be reproduced.

Advisor, optimisation parameters, advisor parameters, number of local and remote cores.

To be able to take, run, and verify that there is a problem (or explain why this behaviour is chosen and will not change).

 
Rebuilding the tester is now a priority for us. A lot of things will be rewritten.

The problem of a rational task manager will be solved.
 
Andrey Khatimlianskii:

If you really want to change and not just grumble (like me), why not give the developers a reproducible example where kernels are disabled/standby?

It could be based on standard EA (if it's reproducible with it too), but with as much detail as possible, so they can reproduce the behaviour for themselves.

Everyone knows about this effect, and the developers have seen these claims many times. A few months ago they even improved the situation slightly (yes, it was even worse).

 

Many times I wrote about this problem, but I was sent to read how the genetic algorithm works. I do know how it works, and in my 4th year at university I even implemented it myself as a lab.

My situation was even worse, here is a screenshot:


With version 2286 it got better, there is no such an obvious bug anymore, but periodically half of the agents still fail for good. I know how to fix it, but it's a pain.

 
Maksim Emeliashin:

My situation was even worse, here's a screenshot:




It's exactly my situation
Reason: