Why don't I read the articles? - page 3

 
Reshetov: It is, IMHO, a thankless task to learn a left-handed language that has such limitations and subtleties. Moreover, the gain in multitasking in OpenCL is lost in the losses in data transfer from one hardware to another.

Ask MetaDriver about it.

He probably knows more about OpenCL than I do. He will tell you whether it is a good idea or not.

 
Mathemat:
Ask MetaDriver about it. He probably understands more about OpenCL than I do.
Do I need it?
 
Reshetov: Do I need it?

The limitations and subtleties stem from the hardware, not from the C99 language itself, which is quite standard (only the vector additions to it are non-standard).

It is OpenCL that is supposed to remove these limitations.

The hardware produced by NVidia, AMD, Intel and the rest is going in exactly the same direction - towards the possibility of easy use of multitasking. There is no getting away from it.

 
Reshetov:
Do I need it?
Yuri, I've wanted to ask you for a long time (but don't be too offended). Do you use your neural networks in real trading? Do they bring you money? I am referring to this article 6 years ago, it must have been a lot of water since 6 years ago.
Как найти прибыльную торговую стратегию - Статьи по MQL4
  • www.mql5.com
Как найти прибыльную торговую стратегию - Статьи по MQL4: торговые системы
 
Mathemat:
The limitations and subtleties come from the hardware, not from the C99 language itself. OpenCL is designed to remove these limitations.
This, by the way, is another drawback of OpenCL, i.e. being tied to the hardware. Because if you create an OpenCL-linked application, only users possessing the hardware you need can get the same or better results. Universality is lost and so is interest.
 
DmitriyN:
Yuri, I've wanted to ask you for a long time (but don't be too offended). Do you use your neural networks in real trading? Do they bring you money? I'm referring to this article 6 years ago, for 6 years, probably, a lot of water has flowed.
I have already told many people and I will tell them again: I hate those who like to count money in other people's pockets.
 
Reshetov: This, by the way, is another drawback of OpenCL, i.e. being tied to hardware.

No, Yura, there is no hardware binding. OpenCL is just hardware-independent, not hardware-binding.

The same code in OpenCL (not taking into account the directives on hardware vendors) is executed on any hardware. It is another matter that it is not executed in the same way. But it is executed!

 
Reshetov:
I've told many people before and I'll say it again: I hate people who like to count money in other people's pockets.
I'm not going to count your money, believe me. You didn't answer the question. Does neuro work in your talented hands or not? Yes or no? In real life.
 
Mathemat:

No, Yura, there is no link to hardware, that's your speculation. OpenCL is hardware-independent, not tied to hardware.

The same code in OpenCL (not taking into account the directives that apply to hardware vendors) is executed on any hardware. Another thing is that it is not the same.

And my point is that if you close your eyes to one and the other, it seems to be the same, but if you look at performance, the difference is significant.
 
Mathemat:

Ready to see any criticism of my last two articles here (on OpenCL). Links are one and two.

What are they missing, is the level of presentation too high, etc.

I myself expected more interest - especially due to the interest in the OpenCL thread. But there was nothing from anyone on the subject.

Really cool articles I've read with great pleasure. I would like to continue the started topic of hardware-accelerated optimization.

1 Since C++AMP is coming, OpenGL topic is not disclosed yet, it is very interesting to speed up calculations in Metatrader.

2 I'd love to read an article on "Accelerating cloud network calculations using a graphics card".

Can we expect something similar from you in the near future ? The topic is very interesting and fascinating.

Reason: