EA optimization using OpenCL


My Multi Symbol Multi Period Scanner EA performs 6400 calculation per second, and consuming 80% CPU and 3GB Ram

My CPU is Core i7 6700 and GPU is Nvidia GTX 970

The calculation scans 14+ buffers on 11 timeframe on 130+ symbols

I am thinking to convert my EA calculation to OpenCL

Here is results for test for CPU vs GPU

Can you advise if i convert, calculation may be faster?

In sequence Bitonicsort, MatrixMault, Wavelet

double test

float test


It can become faster on OPEN depending on how you divide the workload.

since your EA is multi-symbol, if you are able to divide your symbols across the various GPU cores, then you will definitely get faster results. 

I have developed stand alone application using CUDA and OpenCL for GPUs achieving extremely high speedup. 

But the whole division of workload across CPU and GPUs has to be well planned.