Mt4 - the Terminal that never sleeps :(
I wanted the EA to coordinate file accesses using GetTickCount(). So I was testing what it can do and got this:
It should sleep for 7 mSec but it never sleeps. Here is my EA, where am I mistaken?? Or is this just the reason why the terminals are so fast - the timer are influenced by the theories of relativity of Einstein?
May be - due to be as fast as possible - the Sleep()-order was put in another thread?
dur is always 0 (on my 10 years old laptop) despite there was in the screen shot the order to sleep for 7 mSec?
beg0 and beg are always the same despite there was in the screen shot the order to sleep for 7 mSec?
This isn't long enough (takes less than a millisecond). It is after the sleep
uint dur,beg=GetTickCount(); // new mSec-Time after Sleep int i = MathRand()%RandVal; while(i-->0) { double x = MathLog(i+2); // just to spend some time here } dur=GetTickCount()-beg; // find the time of calc..
When i is printed, the values are 9, 17 etc.
Maybe use GetMicrosecondCount() or make the loop longer?
Thank you - GetMicrosecondCount() is new to me, but Sleep sleeps for mSecs and both mSec from GetTickcount() before and after the Sleep are the same - how can that be?
The log-calculation is only to spent some time - its not important to me.
The behaviour seems normal to me:
void OnTimer() { uint begin=GetTickCount(); Sleep(500); uint end = GetTickCount(); Comment(StringFormat("Begin: %i\nEnd: %i\nDuration: %i",begin,end,end-begin)); }
Edit: OK, I think I see what you mean.
If you use a large number (like 500), no problem - see above.
But if you use smaller values you end up with either 15 / 16 because of the system timer limitations, or 0.
void OnTimer() { static int sleep=15; static int cnt =0; uint begin=GetTickCount(); Sleep(sleep); uint end = GetTickCount(); printf("Sleep: %i Begin: %i End: %i Dur: %i",sleep,begin,end,end-begin); cnt++; if(cnt>3) { sleep-=2; cnt = 0; } if(sleep<5) EventKillTimer(); }
Hmm,
if you use this version Sleep turns out to be some kind of a random timer.
- Sleep below a certain number is ignored
- and even Sleep(100) is only sometimes 100, sometimes faster, sometimes slower..
Could it be that Windows (below the mt4) tries to be tricky?
//+------------------------------------------------------------------+ //| test_mSec_Sleep.mq4 | //| Copyright 2014, MetaQuotes Software Corp. | //| http://www.mql5.com | //+------------------------------------------------------------------+ #property copyright "Copyright 2014, MetaQuotes Software Corp." #property link "http://www.mql5.com" #property version "1.00" #property strict extern int mSec = 10000; uint maxmSec = 0; //+------------------------------------------------------------------+ //| Expert initialization function | //+------------------------------------------------------------------+ int OnInit() { //--- create timer EventSetMillisecondTimer(mSec); MathSrand(GetTickCount()); return(INIT_SUCCEEDED); } //+------------------------------------------------------------------+ //| Expert deinitialization function | //+------------------------------------------------------------------+ void OnDeinit(const int reason) { //--- destroy timer EventKillTimer(); } //+------------------------------------------------------------------+ //| Timer function | //+------------------------------------------------------------------+ void OnTimer() { //ulong beg0 = GetMicrosecondCount(),begM = beg0%10, slp = begM>3 ? 11-begM : 0; /*L 1*/uint beg0 = GetTickCount(),begM = beg0%10, slp = begM>3 ? 11-begM : 0; if ( begM>3 ) Sleep((int)slp); /*L 3*/uint dur,beg=GetTickCount(); /*L 5*/Sleep(100); /*L 7*/dur=GetTickCount()-beg; maxmSec = fmax((int)dur,maxmSec); // uint beg0 = GetTickCount(),begM = beg0%10, slp = begM>3 ? 11-begM : 0; Comment("OnTime() ",TimeCurrent(),"\nL 1: ",beg0," mSec, ",beg0,"%10 = ",begM," => Sleep : ",slp," mSec \nL 3: ",beg," mSec expected: ",(beg0+slp), "\nL 5: ",dur," mSec Duration after Sleep(100)\nGetTickCount right now ",GetTickCount()); return; }
The Service desk just told me:
"It depends on system clock resolution in the Intel architecture.
There is very known issue with some inaccuracy of system clock - deviation can be up to 15 milliseconds"
Sigh :(
If so even GetMicrosecondCount() does not help!
What a rainy day today.
The Service desk just told me:
"It depends on system clock resolution in the Intel architecture.
There is very known issue with some inaccuracy of system clock - deviation can be up to 15 milliseconds"
Sigh :(
If so even GetMicrosecondCount() does not help!
What a rainy day today.
Have a read of this thread. I found that you could drop the sleep but not the timer.
What I find strange here is the randomness of whether it is 0 or 15ms.
This entire thread is useless. Using sleep (any version) will still result in a race condition between two separate processes.
- If you are coordinating access between to EAs, See my mutex Prevent EA from opening trades at the same time across 2 or more pairs? (Steve) - MQL4 forum
- If it is external, access the file with exclusive read privileges and/or delete the file after read.
race condition between two separate processes.
A thread race is not my problem, but the access of two processes to the same file and my idea instead of looping while h==INVALID_HANDLE I wanted one process to access from 0-4 and the other one from 5-9. Should be easy - theoretically - as both processes´need to open, read/write and close the file always 0 mSec. Well, but there is Intel's minimal resolution of 15 mS.
And "this entire thread is .. " not [ed. note] ".. useless" as at least I learnt about this minimal resolution and I do now understand that many calculations within OnTimer or OnCalculate take either 0 mSec or 15 mSec or 31 MSec ...
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
You agree to website policy and terms of use
I wanted the EA to coordinate file accesses using GetTickCount(). So I was testing what it can do and got this:
It should sleep for 7 mSec but it never sleeps. Here is my EA, where am I mistaken?? Or is this just the reason why the terminals are so fast - the timer are influenced by the theories of relativity of Einstein?
May be - due to be as fast as possible - the Sleep()-order was put in another thread?
dur is always 0 (on my 10 years old laptop) despite there was in the screen shot the order to sleep for 7 mSec?
beg0 and beg are always the same despite there was in the screen shot the order to sleep for 7 mSec?
Even if you set RandVal to 30000 dur stays always 0 and the maxmSec too - wow the terminal never sleeps.
You can even replace int i = MathRand()%RandVal; by int i = 30000; //MathRand()%RandVal; - the terminal never sleeps.
The next update might be able to finish before it was started and the final GetTickCount() is smaller than the initial ;)