OpenCL: internal implementation tests in MQL5 - page 67
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
Thank you.
The articles are very interesting and very detailed.
I am interested in the subtleties of using data and programming, but unfortunately I don't have time to cover such huge volumes to understand everything described in these articles.
As far as I understand optimization is possible, but only in a custom way.
To connect an additional agent (agents) in MT5 to the tester that will be used for calculations of variant optimization of Expert Advisors (in addition to the agents that display CPU cores) is impossible now and it is not implemented programmatically?
...
To connect additional agent(s) to the tester in MT5, which will be used for calculations of variants during optimization of Expert Advisors (in addition to agents, which display CPU cores), now there is no possibility and it is not implemented programmatically ?
Good afternoon.
Thanks tol64, I understood in which direction I should look. I have studied the articles glancing through them, my knowledge is not enough to understand the code on the fly, and no matter how much I read, it's difficult to understand all the details.
I looked through the Help section but didn't get the information I needed.
I have read all 70 pages of the topics it took me a whole day to read, but overall, the picture is getting clearer.
I wish I could use the processing power of my video card in the strategy tester, when optimizing an EA on my PC. I have 4 agents (4 cores - 2500K), it is much better than on my laptop, but it is not enough to optimize EAs.
I have checked the script for OpenCL calculations and attached the results.
p.s. I haven't installed any additional libraries and drivers, four months ago I've installed AMD SDK for mining, I think it was enough.
The video may seem strange - it's 6950 AMD, with flashed 6970 bios, so all 1536 conveyors are open and frequency is up to 825.
Where's everybody gone?! Strange, not a single new comment in a few days.
Judging from what I've read, I understood that the code text can use OpenCL functions and this part of the program (ekspert or otherwise) will be calculated on GPU.
Therefore I have a question: if such EA is tested on the tester, the GPU will be involved in calculation of each paramentary and therefore it may speed up the overall process?
So I have a question: if such an EA is tested on the tester, then in the calculation with each paramentarians the GPU will be connected and hence, for a large number of passes this may speed up the overall process ?
So again, let's include permission to use OpenCL in cludes, although it is clear that there is no Tesla en masse (cost of cards from 1500 euros and up). But let the functionality be for the future - the topic is going to evolve all over the place anyway.
We have done tests on Nvidia Tesla cards and it turns out that their drivers allow them to work in service processes and under remote desktop.
So again, let us enable OpenCL in cludes, although it is clear that there is no massively available Tesla (cards cost from 1,500 euros and up). But let the functionality be for the future - the topic will be developing everywhere anyway.
Of course I can't see the whole "iceberg", but maybe if you dig deeper into the drivers of other cards, there is a possibility for them as well?
If it's not possible at this stage, maybe allow the user (optionally) to add a new user for this or something similar, which would allow others to surrender the card to the cloud too.
This results in the fastest device being selected by default in multi-device configurations.
It would be great if this changed, but it's not up to us.