Testing the new MQL5 compiler for x64 platforms - 2 to 10 times faster calculations! - page 4
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
The benchmark is not so easy to take, it is licensed from both sides.
Nonsense, I have worked for various foreign firms in the area of embeddings. Not a single firm (and there were 500 programmers each) invented compilers, not even C, let alone C++. All took free gcc and improved it for their own processors and it took not so much effort, there were special teams of programmers who knew how to do it. And there were non-x86 processors, VLIV + DSP, all sorts of them, even such feeble ones that couldn't multiply 2 int. And nobody invented bicycles. But MQ are proud)).
What do we have as a result? It all works fine on MT4/5 and it is impossible to run the strategies on android and eplet because the compiler is designed for x86/64 only.
Nonsense, I have worked for various foreign firms in the area of embeddings. Not a single firm (and there were 500 programmers each) invented compilers, not even C, let alone C++. All took free gcc and improved it for their own processors and it took not so much effort, there were special teams of programmers who knew how to do it. And there were non-x86 processors, VLIV + DSP, all sorts of them, even such feeble ones that couldn't multiply 2 int. And nobody invented bicycles. But MQ are proud)).
What do we have as a result? The result is that everything works fine on MT4/5 and it is impossible to run the strategy on android and Apple because the compiler is designed for x86/64 only.
I do not understand many decisions.
Instead of wasting time and human resources on reinventing another bicycle, and then the whole forum for years to catch bugs, glitches, etc., take ready-made solutions and begin to modify them to fit your needs, to focus efforts on speed and quality of data receipt, to work out execution libraries (their speed and reliability), to create connectors to different exchanges, make decisions that simplify the connection of mathematical packages for use in trading (the same MathLab).
But it seems that MQ has its own way, the main thing is to protect sellers of black boxes (code that their next grails) could not hack, while other pluses in this solution I do not see. Only whether it was worth the trouble...
But it seems that MQ has its own way, the main thing is to protect black box sellers (code so that their next grails) could not be hacked, while other pluses in this solution I do not see. Only whether it was worth the trouble...
Has anyone checked the indicators? If the table is to be believed, they should really get a lot faster.
But because of the bubble test, somehow the confidence in all the results drops. 25% gain over c++ on simple array traversal? I don't believe it (c)
Once again, for those who didn't get it right away: array operations, mathematics, complex class hierarchy and the like are very well accelerated many times over. This means that it is possible to do calculations many times faster. This is a huge advantage.
All code for reproduction is attached and the task of public proofing is respected.
Very few people understand why our own secure language is important in our case. But it is a critical component.
For those who don't know - MQL4/MQL5 contains the most indicators and Expert Advisors for trading platforms in the world, while MQL4/MQL5 itself appeared in the ranks of programming languages.
The next changes in Metatrader 5 will bring a large set of new features in MQL5 that, together with its current capabilities, will allow to write powerful analytical systems directly in the terminal. However much detractors may laugh, but in MQL5 you can even write a computer game inside the terminal. And not even one, and run them in parallel.
Where is the code attached?
Sorry, it was unexpectedly deleted when editing the post later on the tablet.
Posted it again.
Well, of course, it wasn't you who commented the test C-4 of his code of 20 000 lines. In his code, I think, not only "made only CopyXXX and a couple of lines from if + OrderSend". And the acceleration is only 15 - 20%. Where is this speedup by times?
I assume that my opponent has common sense in logical calculations.
Unfortunately, I am wrong. Even detailed explanations with clarifications do not help to fight the urge to criticize on the spot.
If the speedup in the code for 20 000 lines is 15-20%, it means that the amount of own calculations is small and most of the Expert Advisor's time is spent on system calls which are already optimized when written in pure C++.
Nonsense, I have worked for various foreign firms in the area of embeddings. Not a single firm (and there were 500 programmers each) invented compilers, not even C, let alone C++. All took free gcc and improved it for their own processors and it took not so much effort, there were special teams of programmers who knew how to do it. And there were non-x86 processors, VLIV + DSP, all sorts of them, even such feeble ones that couldn't multiply 2 int. And nobody invented bicycles. But MQ are proud))
You're confusing the scope.
We have a language for creating secure, safe and alienable programs for mass distribution and not "let's create a buggy and disposable compiler for internal consumption". A hardware is likely to morally die before a compiler update is released.