OpenCl and the tools for it. Reviews and impressions. - page 23

 
Reshetov:

Don't get bent out of shape about the highlighted one, better listen to the developers if you don't believe me and also assume I'm being ugly. See page 20 :

Better yet, try to program some application stuff in OpenCL yourself before you make up your own stuff here.


Better pay attention to the words "regardless of display".

 
IgorM:

Yeah, for that matter, at least they would have surprised us by calculating pi to the millionth digit using OpenCl, maybe such a wonderful innovation would have worked somehow, imho

You could do it on a GPU too, but it's just a case of poor parallelization. GPU is good when you need to do thousands and millions of simple and not very accurate single-type calculations in a short time. According to my estimation, it's the best for autotrading. For example, it is possible to calculate a couple dozens of pairs in a second for all of the available currency pairs. Or the same amount of slightly more sophisticated indicators for 2-4 seconds. Of course, to calculate for thousands of bars, not only current value (for which milliseconds are enough).
 
Mashka for thousands of bars - spicy:) Maybe linear algebra would work?
 
tara:
Mashka for thousands of bars - spicy:) Maybe linear algebra would work?

On a piece of paper, in pencil? :)

 
IgorM: It's more important to get the inputs right, rather than dumping nonsense into the NS...
Now try telling joo that he's "dumping nonsense into NS"...
 
MetaDriver:

On a piece of paper, in pencil? :)


No, on a matrix processor :)
 
IgorM:

I have never even wondered how hard it is to calculate Pi, but it turns out it is not that hard https://ru.wikipedia.org/wiki/Пи_

Can anyone surprise me with code in mql5 and using OpenCl?

Hardly. You need bibles which are able to use super large integers.

Although... I've seen a simple formula somewhere that allows to calculate any sign of pi in hexadecimal, knowing its number after the decimal point (probably, also in hexadecimal). But the calculation itself up to a trillion digits is unlikely to be fast.

But it's a matter of principle here, isn't it? Slow algorithms are purposely taken to illustrate acceleration.

P.S. I'll find the formula and post it.

 
tara:

No, on a matrix processor :)
They're great friends with linear algebra, they're almost in love with it. )) And they're great with trigonometry, too.
 
MetaDriver: They're great friends with linear algebra, they're almost in love with it. )) And they're great with trigonometry, too.
Only without branching.
 
Mathemat:
Well try telling joo now that he's "dumping nonsense into NS"...

no thanks, joo is serious, may not understand that I don't mind OpenCl, but so far I haven't seen the need to use it, except for Mandelbrot fractals, as already mentioned it's for Mandelbrot that you can now not worry about ))))

Reason: