Un-stressing Tight/Recursive Loops + Other questions

[rcmaehl]

Okay. I've read a few threads that suggest using

Sleep(10)

in long Do...Until Loops to lower CPU usage. Is that like the official amount for best Performance/Load or something? Also, a semi-related question, How long does Sleep(0) sleep for? It has to sleep for some amount of time (probably less than a millisecond, and in the nanosecond range) because it gets executed or do all sleep(0)s get removed upon compiling?

EDIT: Removed grammatical error.

Edited December 6, 2011 by rcmaehl

[water]

Give this little script a try. Run it from SciTe, run it compiled and compare the displayed time.

$a = TimerInit()
Sleep(0)
$b = TimerDiff($a)
MsgBox(0,"", $b)

As compiling means to put the script and everything needed for execution into an exe file your script is still interpreted when you run the exe.

So I guess you won't see a difference because sleep(0) will not be removed.

Edited December 6, 2011 by water

[rcmaehl]

Give this little script a try. Run it from SciTe, run it compiled and compare the displayed time.

$a = TimerInit()
Sleep(0)
$b = TimerDiff($a)
MsgBox(0,"", $b)

As compiling means to put the script and everything needed for execution into an exe file your script is still interpreted when you run the exe.

So I guess you won't see a difference because sleep(0) will not be removed.

Wouldn't running that in a finite loop about 1000 or so times and then calculating the average give a better estimation?

[water]

But you add the overhead of the loop. And as Sleep(0) needs very little processing time you mainly measure the loop.

See the difference:

$iTimerA = TimerInit()
$iTimerB = TimerInit()
For $i = 1 To 1000
    Sleep(0)
Next
$IResultA = TimerDiff($iTimerA)
For $i = 1 To 1000
Next
$IResultB = TimerDiff($iTimerB)
MsgBox(0,"", "1000 loops with Sleep(0) take: " & $IResultA & " milliseconds" & @CRLF & _
    "1000 loops without Sleep(0) takes: " & $IResultB - $IResultA & " milliseconds")

Edited December 6, 2011 by water

[Melba23]

rcmaehl,

Soemwhere here there is a thread where Jon explained all this, but can I find it.....

I seem to remember that Sleep(10) is the lowest value that Sleep will accept - anything lower still gives you a Sleep(10) - and that this is the result of a Windows limitation rather than anything in AutoIt. I believe it was reduced from Sleep(15) at some time in AutoIt's past when the Windows API changed internally. By the way monoceres produced a UDF which had nanosecond accuracy and he found that 100ns was quite sufficient to offload his CPU - so Sleep(10) should be more than adequate. And you do realise that you do not need a Sleep in a GUIGetMsg loop as the function does its own idle?

As far as I know Sleep(0) is a special case which means that you give up the timeslice allocated by Windows - quite what effect that will have on your script I have no idea.

I hope this helps. I will keep looking for that thread....

M23

[BrewManNH]

Putting a sleep into a Do...Until loop will only help with CPU load depending upon what you're doing inside that loop. If it's running a For...Next loop inside that loop 10000 times, every time through the loop, that reads files into arrays it probably won't have any effect on CPU load. If all you're doing is looking for a pixel color to appear it might help somewhat.