Delay problem

[SeregaZ]

i have one small problem with Delay. i try to read midi file and play it from array. in this array stores "time" of midi. it is a very big array some times. so when play this array - i used delay, but some times Delay(1) is too big. it is in millisec - but i need microsec. have some one same like problem and how you solved it?

[RSBasic]

http://www.purebasic.fr/german/viewtopi ... 40#p322040

[SeregaZ]

thanks, will try it

[SeregaZ]

Code: Select all

Global QueryPerformanceFrequency.q, QueryPerformanceFactor.f
QueryPerformanceFrequency_(@QueryPerformanceFrequency)
QueryPerformanceFactor = 1e6/QueryPerformanceFrequency

Procedure DelayMicroSeconds(Time.q)
   Protected StartTime.q, EndTime.q
   QueryPerformanceCounter_(@StartTime)
   Repeat
      Delay(0)
      QueryPerformanceCounter_(@EndTime)
   Until (EndTime-StartTime)*QueryPerformanceFactor > Time
EndProcedure



a = ElapsedMilliseconds()
For n = 1 To 1000
   DelayMicroSeconds(300)
Next
Debug ElapsedMilliseconds()-a

this code from german forum is super, but it can be changed to more small pause? i need 1/44100 some kind of:
0. 000 022 675s

[Danilo]

You should remove the Delay(0) from the loop.

A delay of 0 doesn't make sense. Because of that, at least in Windows
a delay of 0 has a special meaning, see Sleep function.
Sleep() was used for Delay on Windows, but I have something in mind
that it was changed some versions ago. Not sure how PB handles a Delay of 0 now.
Maybe it just does nothing, so a delay of 0 wouldn't make sense.

Anyway, with so high precision you don't want any delay.

Under heavy OS load (many running applications and all cores and CPU's busy),
you will probably get timing problems because of task switching.
Setting a higher priority for the thread that runs the timing and MIDI output code
could at least help a little bit. See also this discussion at StackOverflow
and Microsecond Resolution Time Services for Windows.

[infratec]

Hi,

I optimized it a bit (I hope):

Code: Select all

Procedure DelayMicroSeconds(Time.q)
  
  Protected StartTime.q, EndTime.q
  
  QueryPerformanceCounter_(@StartTime)
  Time / QueryPerformanceFactor
  Repeat
    QueryPerformanceCounter_(@EndTime)
  Until EndTime - StartTime >= Time
  
EndProcedure

Bernd

[Danilo]

By the way, does anybody know if calling QueryPerformanceFrequency() only once (at application start)
could make problems?
Modern processors often provide a way to change CPU frequency while running, AFAIK.
Lower Frequency for power saving in Notebooks, or higher frequency for a limited time to boost performance.

Could this influence high precision timer codes?

[CELTIC88]

try this

Code: Select all

Procedure DelayEx(timeMicroSeconds.q)
  ; Name...........: DelayEx()
  ; Description ...: Sleeps down to 0.1 microseconds
  ; Parameters ....:  timeMicroSeconds      - Amount of microseconds to sleep
  ; Author ........: Celtic88
  timeMicroSeconds=-1*(timeMicroSeconds*10)
  ProcedureReturn ZwDelayExecution_(0,@timeMicroSeconds)
EndProcedure

Test= ElapsedMilliseconds()
DelayEx(1000000)
Debug Str(ElapsedMilliseconds()-Test) + " Ms"

[Dude]

Time / QueryPerformanceFactor

Division by zero error.

[Helle]

@Danilo: For modern CPUs: Invariant TSC/TscInvariant!
Intel wrote:
CPUID 80000007H, EDX Bit8 Invariant TSC
The invariant TSC will run at a constant rate in all ACPI P-, C-, and T-states. This is the architectural behavior
moving forward. On processors with invariant TSC support, the OS may use the TSC for wall clock timer services
(instead of ACPI or HPET timers). TSC reads are much more efficient and do not incur the overhead associated with
a ring transition or access to a platform resource.
AMD wrote:
CPUID 80000007H, EDX Bit8 TscInvariant
TscInvariant: TSC invariant. The TSC rate is ensured to be invariant across all P-States, C-States,
and stop grant transitions (such as STPCLK Throttling); therefore the TSC is suitable for use as a
source of time.

[Danilo]

Thanks, Helle!

- Time Stamp Counter

[SeregaZ]

thanks for a lot answers i have no succes...
0. 000 022 675s
delay millisec that DelayMicroSeconds proc microsec even dont know

problem is - when i play sound - tempo is drunk. delay pause count is a little wrong. i make attach to illustrate this drunk tempo. you will need to fix path to Array.txt and choice one of three method of pause by comment and uncomment in code: delay(1), infratec and CELTIC88 (how work CELTIC88's variant i am even dont understand)

for compare with correct tempo i add winamp plugin and that original melody.

https://www.dropbox.com/s/sdvel9hsu6yq9 ... r.zip?dl=1

can you help?

[infratec]

Time / QueryPerformanceFactor
Division by zero error.

Try to replace QueryPerformanceFactor.f with
QueryPerformanceFactor.d

Bernd

[infratec]

And

Code: Select all

DelayMicroSeconds(MusArr(i)\pause * 22.675)

Bernd

[SeregaZ]

and with .f and with .d and with 22.675 and with 22 and with 23 - same drunk tempo ((