OpenCl and the tools for it. Reviews and impressions. - page 14
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
Unlike you, I personally test cludes. And I have also been developing, testing and planning this project for many years.
So leave the statements about my ignorance in this area alone.
Man, I do look like a clown. But I don't have that many agents in the same Agents tab.
Maybe you as a developer and project manager can somehow explain where my agents go? The internet seems to be one for all. Maybe they all ran away to you? What do you use to lure them?
Let me send you the logs and we'll figure it out. I want to get high performance too, not wait until optimization is barely finished. Or do you seriously believe that I'm interested in berating developers here? I berate crutches which seem to give out performance, but in reality I don't see it.
I've already added one more agent per core on a local, in order to improve performance somehow. At least for 3D rendering two processes per core produce high speed, that's why I've already gone that extra mile.
And apparently others have the same problems:
Yura. You are right. It's the way it is so far.
Man, I do look like a clown. But I don't have that many agents in the same Agents tab.
Really? Although the direction of estimation is correct.
Just a little cheating on your part, that's all. All's fair for proof, isn't it?
It's not for nothing that you nicked the right part of the screen where the status of the network "connected or not" is shown - on purpose you showed the test mode in inactive state (I should know the peculiarities of small nuances of each line state in the above screenshot). Therefore the conclusion is unambiguous - blatant cheating.
Clean test requirements:
The "agents available" metrics are instantly deflated as soon as the network load goes down, as the network tries to shut down (send to sleep) as many agents as possible to conserve resources. Agents can only be accounted for when the load is on.
Running a task with 500 passes and then showing a low network load is cheating.
So there you go. You can argue with me, but to make such weak arguments in the face of my repeated corrections is inexcusable.
The root of your problem is the lack of using the MQL5 Cloud Network in practice. For they are not even aware of accounting for agents and warming up the network - they saw the counters of a sleeping network and forward to the forums to talk nonsense. Anyone who has spent an hour checking the cloud will understand the scheme of work.
What do you think?
Just a little cheating on your part, that's all. All means are good for proof, aren't they?
Renat, I am not going to argue and cheat. I saw your screenshot and made sure that both the number and quality of agents on your network is really high to have a normal performance.
I am not able to achieve such a result. That is why I say, let us not blame each other, but let us deal with the problem. I will now run the tests again and take new screenshots to show that the number of passes I have exceeds 10000. I will show in the screenshots how the network works in different modes, because in forward testing mode more than four agents, i.e. one from each network could not be attracted and the whole farm is barely moving.
Here is the first screenshot, of the current state without running optimisation. I don't cut anything out, i.e. you can see the date and time at the bottom. There are prices in the market overview. There is no flapper.
I have 100% serious use of the cloud network.
I specifically re-run the tests right now with forward passes - about 6,500 agents turned on at the same time.
Here is the second screenshot, the optimisation is running. Passes are over 10,000
Here's the third screenshot. The agents have warmed up and are running
Here's another screenshot. The network is silent.
Here's the third screenshot. The agents have warmed up and started
Here we go again with the cheating.
You have to prove through screenshots if words do not work.
Don't you know the peculiarities of genetics and the concept of population size? Enabling genetics mode and generating duplicate tasks of 64 to 256 is great for finding thousands of agents.
Take a look at the article " MQL5 Cloud Network speeds up calculations " and discussion of the article "MQL5 Cloud Network accelerates calculations" - the testing mode of genetics in the cloud has been analyzed there.
I have a 100% serious use of the claud network.