Discussion of article "OpenCL: The Bridge to Parallel Worlds" - page 2

[Flops]

Please replace"kernel" in the article with "core". It is impossible to read.

[Sceptic Philozoff]

You're the first one to ask. It's not that hard to replace.

Let me make a poll.

I did it right here.

[Vladislav Andruschenko]

The requested page was not found

Check if the path is correct and try again

I can't download files from the article

[Sceptic Philozoff]

Vladon: can't download files from the article

Yes, indeed. The downloaded file turns out to be an html-page, but not an mq5 file.

Thanks, I will inform Service Desk.

[Automated-Trading]

Mathemat:

Yes, indeed. The downloaded file turns out to be an html-page, but not an mq5 file.

Thanks, I will inform Service Desk.

Fixed.

[Vladislav Andruschenko]

Automated-Trading:
Corrected.

thanks

[Alain Verleyen]

Good article. Thanks.

However, I still wonder how opencl can be used with trading.

[Kourosh Hossein Davallou]

Excellent

[intersense]

MetaQuotes:

New article OpenCL: A Bridge to a Parallel World has been released:

By Sceptic Philozoff

OpenCL support is a very good choice, now and in the future of the computing platform heterogeneity is very obvious, but now the same algorithms under the conditions of using OpenCL than CUDA performance is much lower, perhaps CUDA than OpenCL is more underlying, better optimised for its own GPU. NVIDIA's GPUs have better performance, better momentum, and the CUDA compiler has adopted LLVM. NVIDIA's GPU performance is better, the development momentum is better, and CUDA compiler has adopted LLVM, there will be more and more languages to support CUDA, Python can now support, especially CUDA6.0 in the programming ease of use is more prominent, especially Unified Memory technology, in the future, with the CUDA runtime support for automatic data migration is better, the performance of the programme and programming productivity will be better. MQL5 support for OpenCL is a good start, and there may be some things that can be done on CUDA in the future.

[Vladimir Pastushak]

Author or experts answer please:

Where will the code below work faster on the main stone or in the vidicon ? And are there any specifics ?

void OnStart()

  {

   long total= 1000000000;

   for(long i=0;i<total;i++)

      for(long q=0;q<total;q++)

         for(long w=0;w<total;w++)

            for(long e=0;e<total;e++)

               for(long r=0;r<total;r++)

                  for(long t=0;t<total;t++)

                     for(long y=0;y<total;y++)

                        for(long u=0;u<total;u++)

                           func(i,q,w,e,r,t,y,u);

  }