Can strategy trader use the GPU for optimizing?

[Amit Suryakant Jamuda]

Hi all,

I have access to GPU, I guess many people does these days. It is something that has become common and usually mid-range laptops and PCs are shipped with one GPU as standard.

Now can MT5 strategy tester use these GPU to run optimization? As of now we can only run optimization in a local PC or one in the network or use the MQL cloud.

Now a days GPUs come with thousand of processors capable of running thousand of threads at once, if strategy tester can use GPUs, won't it be a good idea.

I know and understand, it is easier said than done. However, I think its a capability that's worth every penny. 

[Jose]

AmitJamuda:

Hi all,

I have access to GPU, I guess many people does these days. It is something that has become common and usually mid-range laptops and PCs are shipped with one GPU as standard.

Now can MT5 strategy tester use these GPU to run optimization? As of now we can only run optimization in a local PC or one in the network or use the MQL cloud.

Now a days GPUs come with thousand of processors capable of running thousand of threads at once, if strategy tester can use GPUs, won't it be a good idea.

I know and understand, it is easier said than done. However, I think its a capability that's worth every penny. 

Hi, 

     actually I think it can, through openCl. Have a look at this article  

https://www.mql5.com/en/articles/690

    I suppose you would have to include the parallelization into your experts and, once assigned to a processor, they would make the required calls to the GPU processors. I actually haven't checked it myself yet, but I plan to check its possibilities in the future. I suppose you meant a more direct parallelization approach where each of the processing units  of the GPU was considered as a core of the PC or a Cloud agent, without requiring any special codification...

[Amit Suryakant Jamuda]

jlwarrior:

Hi, 

     actually I think it can, through openCl. Have a look at this article  

https://www.mql5.com/en/articles/690

    I suppose you would have to include the parallelization into your experts and, once assigned to a processor, they would make the required calls to the GPU processors. I actually haven't checked it myself yet, but I plan to check its possibilities in the future. I suppose you meant a more direct parallelization approach where each of the processing units  of the GPU was considered as a core of the PC or a Cloud agent, without requiring any special codification...

Not exactly.

When you use optimization in the Strategy tester, the same EA is run with the same data, however with different inputs. Now these runs happens on different cores of the processor or through multi threading. You can even use PC/Laptop on the home or office network and also the cloud agents. I am asking can these runs happen in the GPU, cause GPUs can run a lot many threads than what processors can as of today. 

[Alain Verleyen]

AmitJamuda:

Not exactly.

When you use optimization in the Strategy tester, the same EA is run with the same data, however with different inputs. Now these runs happens on different cores of the processor or through multi threading. You can even use PC/Laptop on the home or office network and also the cloud agents. I am asking can these runs happen in the GPU, cause GPUs can run a lot many threads than what processors can as of today. 

No it can't.

[Amit Suryakant Jamuda]

angevoyageur:
No it can't.

Any chance of it being in the near future builds?

[Alain Verleyen]

AmitJamuda:
Any chance of it being in the near future builds?

I don't think so but you have to ask directly to Metaquotes to obtain an authoritative answer.

[Amit Suryakant Jamuda]

angevoyageur:
I don't think so but you have to ask directly to Metaquotes to obtain an authoritative answer.

Ok I'll ask the service desk then.

[Maxim Khrolenko]

AmitJamuda:
Ok I'll ask the service desk then.

When you will receive answer, pleaase post it here (if it will be possible). I also want to know.