MIDI-OX User Forum - Print Page


    
      
        MIDI-OX User Forum
        (http://www.midiox.com/cgi-bin/yabb/YaBB.pl)
      

        MIDI-OX >> Scripting Questions >> Strange performance problem
        
(Message started by: chkk on Aug 28^th, 2001, 6:42am)

Title: Strange performance problem
Post by chkk on Aug 28^th, 2001, 6:42am

Hi Jamie,

as you might know by my last posts, I have been putting the MIDI-OX COM interface to a hard test in the last days ...

I discovered a very strange - but reproducible - performance problem.

Take a very simple MIDI-OX COM client which simple forwards the event it gets from MIDI-OX to an other output, no processing at all.

When you quickly play a lot of notes on the keyboard (using all 10 fingers helps ...) for some minutes, suddenly the notes will come out with a very heavy delay (more than a second), sounds as if MIDI-OX is close to breaking down ... only way to stop it is to stop playing for some seconds.

Watching the Windows 2000 task manager reveals, that the CPU load for MIDI-OX increases from about 5 percent (when everything is OK) to about 40 percent (when the notes get delayed). There is a basic workload of about 45 percent from Cubase with some VST instruments, this workload does not change.

If I log the time required for the processing in the event handler for the MIDI-OX Com object, there is no increase at the moment the problems start, processing time is 0-2 ms.

There is no paging (576MB, about 400 free ...), I disabled the input/output display in MIDI-OX to save some CPU cycles.

Any ideas ?

Greetings,
Christian

Title: Re: Strange performance problem
Post by JamieOConnell on Aug 28^th, 2001, 7:54am

Are you using MIDI Yoke NT when this happens? There is a known occasional latency problem with the NT version of MIDI Yoke because of it's implementation. I have never seen any latency problem using only MIDI-OX.

The task manager will attribute the problem to one application or another because MIDI Yoke NT creates worker threads on behalf of attached applications.

Title: Re: Strange performance problem
Post by chkk on Aug 28^th, 2001, 3:38pm

Yes, I use MIDI Yoke NT (Version 1.70 on Windows 2000 Professional SP2).

The task manager shows the CPU load peak in MIDI-OX. If I don't use any MIDI Yoke ports and only send data from my "real" MIDI in to my "real" MIDI out using the MIDI-OX COM interface, I get the problem as well (but I didn't uninstall MIDI Yoke before the test).

I migrated a simple MIDI thru test program from MIDI-OX to another MIDI COM component (MIDI-IO from Mabry Software), so far I have not seen/heard the problem, even in combination with MIDI Yoke.

But as MIDI-IO has only the basic receive / send functionality compared to the wealth of features in MIDI-OX, I am still investigating ...

Title: Re: Strange performance problem
Post by JamieOConnell on Aug 28^th, 2001, 10:49pm

on 08/28/01 at 15:38:36, chkk wrote:

If I don't use any MIDI Yoke ports and only send data from my "real" MIDI in to my "real" MIDI out using the MIDI-OX COM interface, I get the problem as well (but I didn't uninstall MIDI Yoke before the test).

I see. That's good information. I didn't read the first message carefully enough to realize that this was only happening when a script was active. Is this script using a connection point sink, or is it using MIDI polling? We're pretty much at COM's mercy on this either way, but I'm still curious.

Title: Re: Strange performance problem
Post by chkk on Aug 29^th, 2001, 12:29am

on 08/28/01 at 22:49:35, JamieOConnell wrote:

Is this script using a connection point sink, or is it using MIDI polling?

It is using a connection point sink.