OpenCL: internal implementation tests in MQL5 - page 2

 
Renat:
Yes, we will try to describe the use of OpenCL both in the documentation and in the articles.

Sigh of relief. :) Your site is one of the few where the articles are readable and well understood (digested).

I have this question. Do I understand correctly that this will only be available to users who have the appropriate video card installed that supports this technology? That is, if you write a program using this technology and put it on sale in the Market, then not all users will be able to use this program? For example, my video card, as I understand it is already considered obsolete, since I could not run the script you presented. I guess I will have to upgrade. :)

 
tol64:

Sigh of relief. :) Your site is one of the few where the articles are readable and well understood (digested).

I have this question. Do I understand correctly that this will only be available to users who have the appropriate video card installed that supports this technology? That is, if you write a program using this technology and put it on sale in the Market, then not all users will be able to use this program? For example, my video card, as I understand it is already considered obsolete, since I could not run the script you presented. I guess I will have to upgrade. :)

This script will probably work with the next version - for now I've shown a working result of the internal version under development.

Especially OpenCL in the absence of physical accelerator in operating system (not too old) there is at least one "software device" that can execute the same OpenCL code slowly to the best of its ability.

Use the free utility OpenCL-Z to see your computer's specs:

You can see that there are two devices on the system:

  • GBP: Redwood (Radeon) - graphics card
  • CPU: Intel i7 - software implementation

By default the most powerful device is selected, and in the worst case the slowest CPU.

It means that Expert Advisor with OpenCL will most probably run everywhere unless the computer and operating system are blatantly old (without updates and latest drivers). We will probably make a notice in the Market (it will open on Friday) if the product contains OpenCL and the buyer's computer does not have OpenCL support.

OpenCL-Z
  • 2013.04.22
  • Brought to you by: byleas
  • sourceforge.net
This program was born as a parody of another *-Z utilities like CPU-Z or GPU-Z. It shows some basic information about OpenCL-enabled GPUs and CPUs. You seem to have CSS turned off. Please don't fill out this field. You seem to have CSS turned off. Please don't fill out this field.
 
Renat:

Yes, it will be possible to plug in - that's exactly the aim.

If OpenCL devices are not on the agent, the software implementation will automatically work.

Great! I have MSI's R6970 - 1536 threads (agents) and Gigabyte's HD5870 (1600 processors). In BOINC manager they can be used without crossfire (I just plug one output of the second card or connect one of the outputs of the second monitor. Question: can they both be used without crossfire???
 
Renat:

...

That is, an Expert Advisor with OpenCL will very probably run everywhere unless the computer and the operating system are really old (without updates and latest drivers). We will probably make a notice in the Market (it will open on Friday) if a product contains OpenCL and the computer the buyer is using does not have OpenCL support.

I have a rather old video card(GeForce 9600M GT) on my laptop, I think. The drivers for it don't even seem to be released anymore. :) Downloaded a utility. Nevertheless, this is the result:

---

So, I and many users who haven't upgraded for ~2-3 years, as I understand it, will be able to use this technology. All that remains is to learn how to use it. From those links to articles you provided, everything seems to be clear. What is missing is the most important thing, the examples within the MT5 framework. That is, examples of what may be done using it for trading. As far as I understand it may be even 3-dimensional charts on the price chart, for example. All in all it is very interesting. Thank you for your informative, as always, answers.

 
ilovebtc:
I'm willing to provide my Radeon 6930 if it's more profitable than cryptocurrency. 2-3 c.u. a day would suit me fine on it. If it will be as idle as Core i5 i7 processors, imho not interested.
I put 6970 and 5870 without conditions for the benefit of development.
 

My old lady:

Cor i5-750 processor is not detected.

Will it work on devices with OpenCL 1.0 or do I need 1.1?

Although the scripts from JavaDev that he sent in the summer worked.

 

Like this.... ATI video card - nvidia icon. Maybe I downloaded from the wrong source...?

 
There are clarifications:
1) the number of gpu threads are not separate agents, but are used as mass worker bees directed by swarms to specially parallelised code
2) we support OpenCL 1.1 and higher because it supports double types. The OpenCL 1.0 version can only operate with float whose accuracy is in no way suitable for financial calculations

Try to install new drivers although many cards of previous generations do not support double operations.
Документация по MQL5: Основы языка / Типы данных / Вещественные типы (double, float)
Документация по MQL5: Основы языка / Типы данных / Вещественные типы (double, float)
  • www.mql5.com
Основы языка / Типы данных / Вещественные типы (double, float) - Документация по MQL5
 
Renat:
There are clarifications:
1) the number of gpu threads are not separate agents, but are used as mass worker bees directed by swarms to specially parallelised code
2) we support OpenCL 1.1 and higher because it supports double types. The OpenCL 1.0 version can only operate with float whose accuracy is in no way suitable for financial calculations

Try installing new drivers although many cards of previous generations do not support double operations.
During summer testing ofJavaDev scriptswe faced a problem that my card didn't supportdouble, butfloat worked. Drivers can't fix it, we have to change the card :(
 
Renat:
There are clarifications:
1) the number of gpu threads are not individual agents, but are used as mass worker bees directed by swarms to specially parallelised code
........
If I understand correctly, 1 GPU is one very powerful agent? Can the CPU agents be disabled in that case (due to their low speed relative to the video)? And again: is it possible to have two ATIs without crossfire?