OpenCl and the tools for it. Reviews and impressions. - page 21
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
Mathemat:
There's a simple, very simple loop with a lot of iterations (around one million). It is hoped that there will be a speed increase on OpenCL.
There will only be if:
1. the tester will support OpenCL. Renat has already promised this in the next builds
2. You will learn OpenSL.
3. You will manage to rewrite this heaviest calculation in OpenCL so that the performance from computing to GPU will increase rather than vice versa.
Once all three points are met, then there will be a result. The video card will not do the last 2 points for you.
1. the tester will support OpenCL. Renat has already promised this in the next builds
I wasn't talking about the tester.
Then you get the obvious nonsense. For the sake of not waiting 10 seconds and reducing it to 0.2, you should:
1. Buy a fancy graphics card. It's a waste of money to buy a card and put it in the computer in 0.1 second.
2. Learn OpenSL. Not in 0.1 second.
3. The second, you will try to use mql5 to cram the commands into the GPU in such a way that they save the same 9.8 seconds. It won't work in 0.1 second either.
All in all it will take a lot of time and money. And all this for the sake of coming out of your trousers with joy that some one-time result was achieved with such time and money?
Or maybe it does make sense to wait 9.8 seconds longer?
Assuming that the metaquotes are already in the process of rewriting/porting technical indicators to OpenCl, not a lot of time, so a small pile )))))))
Assuming that the metaquotes are already in the process of rewriting/porting technical indicators to OpenCl, not a lot of time, so a small pile )))))))
So far they say there are some problems and not everything is as good as it seemed at the beginning.
But in general, they have provided an interface directly from MQL5 to OpenCL. Of course it's much more convenient than cutting out tonsils via . dll. And now everyone should decide for himself whether to use it or not.
I suppose that the developers will be able to speed up some MQL5 functions using GPU. And they will do it little by little. But miracles do not happen. GPUs are not CPUs, but rather functionally limited high-speed cores with their own specifics for programming.
Yura, don't be silly. In software firms working with large projects, they put the fastest multicore stones into computers for coders, just to speed up the development process. Even if they develop software which will be single-threaded.
I just gave an example, which seems to be not very good. Besides, you may perform the same calculations written into init() not only there.
Well, I don't know what applications there may be. What is important is that the terminal is slowly but surely turning into a hardware-compatible one.
Yura, don't be foolish.
If only this same terminal were optimized by the genetic algorithm corresponding to autotrading, it would be wonderful. But while MT4 pushes it beyond its direct purpose, while the developers accelerated drawing of Mandelbrot fractals 100 times, so who of us is being foolish?
I was just giving an example, not a very good one it seems.
If only this terminal would be optimized by the genetic algorithm corresponding to autotrading, it would be wonderful. But while MT4 is pushing it beyond its direct purpose, while developers have accelerated drawing Mandelbrot fractals by 100 times, who of us is fooling around?
Right, and I'm being foolish just by asking you to give at least one normal example of an autotrading widget?Yuri, you probably haven't looked at the example code. It is not the video card that draws this fractal. Then the data are saved to bmp and bmp is displayed by standard means of the terminal.
Yuri, you must not have looked at the example code. This fractal is not drawn by video card, only calculations are performed in videokrata. Then data is saved to bmp, and bmp is displayed by standard means of the terminal.
Uh-huh, and the output, based on calculations done in the GPU, is a Malevich drawing. Because I don't know what BMP is?
No one has ever counted the time it takes to render this bmp. Even if they did, it's the same in both cases, doesn't make a difference.