Question for developers - using all computational cores during optimisation - page 9

 
Renat Fatkhullin:

We have engaged in a complete rewrite of the tester and optimiser.

We will drastically overhaul and fix the accumulated problems.

When to expect all these solutions?????

especially waiting for:

1. "one core - one job".

2. continued optimisation since stoppage (server crash, reboot, etc.)

 

toRenat Fatkhullin I still insist on an answer ....

When to expect all these solutions?????

1. "one core - one job"

2. continuation of optimization since stoppage (server crash, reboot, etc.)

Renat Fatkhullin - MetaQuotes
  • www.mql5.com
Профиль трейдера
 

1 core counts - 50 idle, beautiful ......

 
Boris Egorov:

1 core counts - 50 idle, beautiful ......

Doesn't show the whole picture of the calculations - how many total jobs, how many completed, how many left...

 
Vladimir Karputov:

It doesn't show the whole picture of the calculations - how many total tasks, how many are completed, how many are left...

is that better?

 

posted in another thread - no effect, no one from the developers has responded

Please fix bug observed in all new versions


2020.07.26 10:54:28.285 PKFDC0:30019 pass (94, 1154) tested with error "OnInit returned non-zero code 1" in 0:00:00.110

2020.07.26 10:54:28.289 Tester expert file ....\MQL5\StochasticCCI.ex5 open error [2]

it's not there and shouldn't be.

the error is given by the network agent after update, before the update it worked fine, on old versions the optimizer does not give errors in the Expert Advisor, the Expert Advisor was not recompiled

completely demolishing and reinstalling the MetaTrader and deleting all data had no effect

the error is observed on two network agents

 
Boris Egorov:

posted in another thread - no effect, no one from the developers has responded

Please fix bug observed in all new versions


2020.07.26 10:54:28.285 PKFDC0:30019 pass (94, 1154) tested with error "OnInit returned non-zero code 1" in 0:00:00.110

2020.07.26 10:54:28.289 Tester expert file ....\MQL5\StochasticCCI.ex5 open error [2]

it's not there and shouldn't be.

the error is given by the network agent after update, before the update it worked fine, on old versions the optimizer does not give errors in the Expert Advisor, the Expert Advisor was not recompiled

completely demolishing and reinstalling the MetaTrader and deleting all data had no effect

the error is observed on two network agents

I have the same problems. I take it that a different indicator is being called in the indicator?

 
Aleksey Vyazmikin:

I'm having the same problem. I take it that another indicator is called in the indicator?

I don't have this, I mean there is no indicator call in the indicator

 

I'm probably going to get banned again, but I'm still posting here solving some problems that MT programmers haven't solved in years, as evidenced by a google search for example

So:

1) Disabling of MT5 logs.
For many years MT programmers were asked to either disable logs or limit their size.
Why did they ask? Because MT is installed on C disk and there is no possibility to specify the other one, and during optimization MT easily fills up with logs several terabytes, i.e. the whole C disk.
Folders
C:\Users\user name \AppData\Roaming\MetaQuotes\Terminal\your long number \logs
C:\Users\user name \AppData\Roaming\MetaQuotes\Terminal\your long number \MQL5\Logs
should be read only.
For this purpose we create another administrator and give him full rights, but diminish his own rights. As a result MT cannot get access to this folder and cannot write logs, the disk will not be clogged with unnecessary logs.

the second variant - does not always work - the essence of a variant in creation of the batch which each certain interval will delete logs ... Really MT can even in one day completely clog up a system disk with logs

2. Deactivation of automatic loading МТ5 updates.
What for? If the optimization is in progress, the network agents become unavailable at the moment of complete loading of the update,
there is a difference between the versions of the main optimizer and its agents. The result is optimization failure.
It is also not good when an update occurs on a battle server during trading, when every millisecond counts.
I think that the updates can be only at the time when the market is not working, i.e. on weekends.
Similarly, the folder
"C:\Documents and Settings\user name\AppData\Roaming\MetaQuotes\Terminal\ .here is your long number.... Make "\liveupdate" read only
by default the folder does not exist, it needs to be created, after installing read only MT sees the updates but cannot download ....
The result - updates do not interfere with optimizer and do not interfere with trading.

3. The problem"one core - one job" has not been solved and isunlikely to be solved.Problem description: I have 60 cores heterogeneous on different servers, when I distribute jobs the optimizer distributes the same number of jobs to all cores, as a result the slowest core takes hours to complete jobs, and 59 other cores stay idle... The problem is not solved since it doesn't bring any profit to MT, but on the contrary takes money from paid cloud agents .... therefore users will continue to be tormented by the optimizer. Pity. The solution - create your own alternative optimizer for example in python. and the funny thing - it will take MT programmers 15 minutes to solve this "problem" ....

 
Boris Egorov:

3. The problem"one core - one job" has not been solved and probablywill not be solved.Problem description: there are 60 heterogeneous cores on different servers, when jobs are distributed, optimizer distributes equal number of jobs to all cores, as a result the slowest core finishes its job for hours, and 59 more cores stay idle... The problem is not solved since it doesn't bring any profit to MT, but on the contrary takes money from paid cloud agents .... therefore users will continue to be tormented by the optimizer. Pity. Exit - create your own alternative optimizer for example in python. and the funny thing - solving this "problem" - it will take MT programmers 15 minutes ....

Something I have a very big doubt that solving such a problem "takes 15 min".

И... I thought they said that the computational kernel is being rewritten. So, "the snail is coming, it's coming". I myself hate that weak computers slow down the whole computation process, and have to not use them at all. But... for now.

About "taking away money" - quite a ridiculous argument. Anyone who owns a farm with a couple or three slow computers has no way to be a potential Cloud customer - if he has money, he won't use slow computers anyway. And if he doesn't have money, he won't buy cloud computing even if he is forced to stop slow computers from running.

Reason: