Autotrade performance

Questions about MultiCharts and user contributed studies.
AAY
Posts: 56
Joined: 29 Nov 2013
Has thanked: 9 times
Been thanked: 30 times

Autotrade performance

Postby AAY » 07 Feb 2014

Hello

I trade multiple strategies on multiple charts, several dozen of them. I do not have any obvious performance issues, but I want to know how to improve performance, just in case.

1. If Multicharts takes 25% CPU on 4-core system, does it mean that it utilizes 100% of one core? Or it uses multiple cores and this is not a concern?

2. Do alerts affect performance? Is there a difference, performance wise, between "on bar close", "every tick" and "once per bar" alert?

3. Any general tips for strategy programming? Do IntrabarPersist variables negatively affect performance and should I avoid them?

Any input is much appreciated.

User avatar
Henry MultiСharts
Posts: 9165
Joined: 25 Aug 2011
Has thanked: 1264 times
Been thanked: 2957 times

Re: Autotrade performance

Postby Henry MultiСharts » 07 Feb 2014

Hello
I trade multiple strategies on multiple charts, several dozen of them. I do not have any obvious performance issues, but I want to know how to improve performance, just in case.
1. If Multicharts takes 25% CPU on 4-core system, does it mean that it utilizes 100% of one core? Or it uses multiple cores and this is not a concern?
Hello AAY,

MultiCharts utilizes the multi-threaded CPU to increase chart indicator calculations by splitting the load between the CPU cores (each chart gets a separate flow). So each core gets its personal load.
2. Do alerts affect performance? Is there a difference, performance wise, between "on bar close", "every tick" and "once per bar" alert?
Each study calculation requires some resources. The more calcualtions you have, the more events happen beacause of these calcualtions - the more resources are utilized.
3. Any general tips for strategy programming? Do IntrabarPersist variables negatively affect performance and should I avoid them?
Any input is much appreciated.
If you need them in your code and you know how to use them - it won't negatively affect the performance.

User avatar
Smoky
Posts: 517
Joined: 03 Dec 2010
Location: Thailand
Has thanked: 99 times
Been thanked: 121 times

Re: Autotrade performance

Postby Smoky » 07 Feb 2014

how many EL codes lines could be run in tick/tick study ? with a powerful computer of course

what append if new tick come before the end of the runing study ?

User avatar
TJ
Posts: 7742
Joined: 29 Aug 2006
Location: Global Citizen
Has thanked: 1033 times
Been thanked: 2222 times

Re: Autotrade performance

Postby TJ » 07 Feb 2014

how many EL codes lines could be run in tick/tick study ? with a powerful computer of course
A typical store bought desktop computer has a quad core CPU running at 2.6 GigaHertz.

That is 2,600,000,000 calculations per second on each core!

Of course that is a simplistic view... most of the operations require more that one clock cycle. And the CPU must also perform housekeeping duties.

But you get the point... today's computers are powerful; even an inexpensive off the shelf run of the mill computer, is fully capable of running MultiCharts with ease.

The best way to assure yourself is to try it out... Go load up MultiCharts with lots of data, add lots of indicators (eg. try multiple copies of MACD and Stochastics), then run some tests. You can try both backtest, or with a demo feed, do some realtime tests.

what append if new tick come before the end of the runing study ?
The ticks are processed FIFO. If there are more ticks than the computer can handle, which can happen during major report release time (eg. FOMC or job report), the ticks are processed one at a time until the cache is cleared.

User avatar
ABC
Posts: 718
Joined: 16 Dec 2006
Location: www.abctradinggroup.com
Has thanked: 125 times
Been thanked: 408 times
Contact:

Re: Autotrade performance

Postby ABC » 11 Feb 2014

how many EL codes lines could be run in tick/tick study ? with a powerful computer of course
I would say that this highly depends on the "quality" of your code as well. Two codes that generate the same outcome can still have dramatically different calculation times.

Regards,
ABC


Return to “MultiCharts”