cory@three.mv.com (Cory Kempf) (09/03/90)
chewy@apple.com (Paul Snively) writes: > High-Level Events will have a higher priority than user-initiated >events, so that AppleEvent throughput will be at an acceptable level). Uh, have you lookede at the actual throughput rate of AppleEvents? I was running some experiments the other day with AppleEvents... I wrote a simple program that just fed Apple Events to itself. I ran this on a MacIIx with a couple of background tasks running in the background (GLA, CD Remote, Moire, and waiting). I noticed that I was giving up the CPU *FAR* too much... According to the timing that I did, each call to WNE took on the order of 6 ticks. I tried stacking the queue (preposting 5 AE... it actually took longer in the loop). I also tried GNE instead of WNE: no effect. I got rid of GLA, waiting, and CD Remote: WNE now took only 2 ticks. For kicks, I tried removing the call to WNE (leaving in the call to post the AE). Each post took about .4 ticks (still too long in my book). I also tried removing the call for posting events... the entire loop (1000 cycles) completed within 1 tick (the program said that it took zero). This is on an unloaded (e.g. no serious background processing) MacIIx and system 7a9. Now grant, this *IS* and alpha system release, and there is supposed to be some code in there for debugging. I think that these numbers need to be improved by at least two orders of magnitude before the AE system can be used for internal program communications (which is a real pisser: I was planning on using it for my interobject communications to build a multithreaded application. Now, I either must special case or use a new system). I know the system 7 development team is planning on speeding up the system, but I think that a two order of magnitude improvement is a bit much to expect. (I would really love to be proven wrong though!!) I really do not think that AE throughput is at an "Acceptable Level" +C -- Cory Kempf I do speak for the company (sometimes). The EnigamI Co. 603 883 2474 email: cory@three.mv.com, harvard!zinn!three!cory
chewy@apple.com (Paul Snively) (09/08/90)
In article <459@three.mv.com> cory@three.mv.com (Cory Kempf) writes: > > High-Level Events will have a higher priority than user-initiated > >events, so that AppleEvent throughput will be at an acceptable level). > > Uh, have you lookede at the actual throughput rate of AppleEvents? Yes. Of course, we have the advantage of having more recent software than you do. > This is on an unloaded (e.g. no serious background processing) MacIIx > and system 7a9. Therein lies the catch. The performance problem (I'd go so far as to call it a bug in the Process Manager's scheduling) certainly hadn't been fixed as of a9. > I think that these > numbers need to be improved by at least two orders of magnitude before > the AE system can be used for internal program communications (which > is a real pisser: I was planning on using it for my interobject > communications to build a multithreaded application. Now, I either > must special case or use a new system). I agree with your assessment of a9, and I'm quite certain that you'll be pleased with the beta release. > I know the system 7 development > team is planning on speeding up the system, but I think that a two > order of magnitude improvement is a bit much to expect. (I would really > love to be proven wrong though!!) Well, in the case of AppleEvents, that speed-up has already been around for some time--it just hasn't gone out to "the world at large" yet. Again, people will be very very surprised by the beta release (by way of comparison to a9 in particular). > I really do not think that AE throughput is at an "Acceptable Level" You're quite right. The comment wasn't intended to refer to the version that you have; it was intended to refer to versions that I have. __________________________________________________________________________ Paul Snively Macintosh Developer Technical Support Apple Computer, Inc. chewy@apple.com Just because I work for Apple Computer, Inc. doesn't mean that I believe what they believe, or vice-versa. __________________________________________________________________________