Discussion of article "How to Install and Use OpenCL for Calculations" - page 2
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
Everything is great: with the step reduced to 0.01, the acceleration is 100 times:
2013.06.27 00:47:02 OpenCL_Sample (EURUSD,M15) Расчеты без OpenCL:
2013.06.27 00:47:07 OpenCL_Sample (EURUSD,M15) Calculating function values = 4828ms
2013.06.27 00:47:07 OpenCL_Sample (EURUSD,M15) Determining the colouring colours = 469 ms
2013.06.27 00:47:07 OpenCL_Sample (EURUSD,M15)
2013.06.27 00:47:07 OpenCL_Sample (EURUSD,M15) Calculations with OpenCL:
2013.06.27 00:47:07 OpenCL_Sample (EURUSD,M15) OpenCL: GPU device 'GeForce GTX 650' selected
2013.06.27 00:47:07 OpenCL_Sample (EURUSD,M15) Calculating function values = 47 ms
2013.06.27 00:47:07 OpenCL_Sample (EURUSD,M15) Determining the colouring colours = 31ms
One thing is missing - practical examples. We have already sounded the starter:
For example, with respect to MQL5, performance increase can be very useful for some script (indicator or Expert Advisor), which performs complex and long analysis of historical data for several symbols and timeframes(here it should be noted that MQL5 program, claiming parallel execution, must be written in a special way using OpenCL API).
But the code example is again purely theoretical (detached from trading).
ps: this is not a complaint, it's a comment )
I tried it in increments of 0.003. It didn't work: GPU memory is too small (HD6870, 1 GB).
But on the CPU it counts itself and does not buzz - even with a step of 0.002 (in the script you need to replace CL_USE_ANY with the number of the OpenCL device corresponding to the CPU):
True, it's not very fast - but it counts! (I can't do it at step 0.001 - now I don't have enough RAM, which I have only 8 GB).
At step 0.001 I can't do it - now I don't have enough RAM, which I have only 8 GB.
At 0.001, either the terminal crashed or some criticalerror popped up. I didn't bother to look into it because it has nothing to do with practice.
ps: memory 16
I asked the questions incorrectly I apologise.....
the questions are based on the picture in the article:
What is that in the second line, Device#2? I have a strong suspicion that it is AMD Radeon HD7970.
And why it has only 2 gigs of memory instead of 3, as it should (all Tahiti have either 3 or 6 gigs of memory).... or it's already a question for developers, how the amount of memory available for OpenCL is determined.
P.S. Personally I'm interested in 7950, because I think it has the best quality/price ratio of the top-end or close to it. 7970, of course, is also good, but they say that its cooling is not very good.
Andrei, it's clear that the practice here will be... well, almost exclusively proprietary. Thanks to the company for giving me a fishing rod, no, dough for a fishing rod! It has provided me with very useful tools, but has not yet told me how to use them. Well, that's not her job. You have to do your own digging.
For OpenCL to be really useful, you need an algorhythm that processes a large amount of information in parallel and produces something useful for trading.
Without these conditions highlighted in blue, OpenCL is absolutely unnecessary. It is necessary to try hard to make it useful. But for some extremely limited users, which are less than 5% 0.3% of the total number, it is possible (I hope).
Andrew, it's clear that the practice here will be... well, almost exclusively proprietary. Thanks to the company for giving me a fishing rod, no, dough for a fishing rod. It has provided me with very useful tools, but has not yet told me how to use them. Well, that's not her job. You have to do your own digging.
For OpenCL to be really useful, you need an algorhythm that processes a large amount of information in parallel and produces something useful for trading.
Without these conditions highlighted in blue, OpenCL is absolutely unnecessary. It is necessary to try hard to make it useful. But for some extremely limited users, which are less than 5% 0.3% of the total number, it is possible (I hope).
Here's a concrete practical problem - https://www.mql5.com/uws. The algorithm of each tool/TF bundle is independent, i.e. the load can be easily parallelised. It is clear that nobody will do it but me =). That's not what I mean.
I would like the examples to move away from tetris and drawing and get closer to trading. What prevents you from making a similar table with multicurrency analysis of all available history?
if metatester agent (cloud) can use this as well
Could anyone attach the OpenCL 1.1 drivers for Nvidia?
It seems the Nvidia has removed the OpenCL support for version 1.1
Thank you.
Could anyone attach the OpenCL 1.1 drivers for Nvidia?
It seems the Nvidia has removed the OpenCL support for version 1.1
Thank you.
And what version of OpenCL does it support then ?
Only version 1.0, If anyone had the file of version 1.1, please attach the file to this discussion.
Thank you.