Hi list! I think I ran into a new sort of issue. I thought I'd dabble a little in spatialisation and the technique I blundered into (invented is a rather large word) involves dealing with timing and modulation by generating a rather large amount of shreds that wait for some period of time, then write to a variable, then die. I'd write about it more but right now it's not so interesting because of being untested. Anyway, this made me generate quite a few shreds (some tens of thousands but I'd like to use hundreds of thousands eventually for multiple soundsources and channels and so on....). For this I thought it wise to render in non-real time (cough). Sadly it turns out that ChucK doesn't realy like this. On my laptop it simply crashed the whole computer (which then proceeded to turn itself off) on my GF's (also xp) computer it reports a "out of memory" issue with three rather odd characters (a Y with a dash on it amongst them). After that the computer had to be reset anyway but this one is tempramental in ways we can't blame on ChucK. I could try on my Debian box in a few days. So... now what? Are kilo-shreds simply a area one shouldn't venture into? I think I'm of the opinion that ChucK running in --silent mode should simply allocate more memory and time untill the swap (or one's patience) runs out but frankly I can't imagine that would happen over a test-file of a second and 15,000 or so (according to the mini before it gave up) shreds. Yours, Kas.
Hi Kassen!
I think I ran into a new sort of issue.
Heh, ChucK is certiainly good at generating new sorts of issues!
On my laptop it simply crashed the whole computer (which then proceeded to turn itself off) on my GF's (also xp) computer it reports a "out of memory" issue with three rather odd characters (a Y with a dash on it amongst them). After that the computer had to be reset anyway but this one is tempramental in ways we can't blame on ChucK.
(Are you sure we can't blame ChucK?) To make sure I have facts straight: this is running command line chuck (or miniAudicle) on XP? I postulate this may be the unfinished garbage collector catching up with us. What type of things are happening in these thousands of shreds? New unit generators allocated? String concatenation? new objects?
Are kilo-shreds simply a area one shouldn't venture into?
Kilo-shreds, and even mega/giga/(probably not tera yet)-shreds should be well within the normal chuckian operational range, assuming they don't do things that compell ChucK to leak loads of memory. (In the end, I am fairly certain it *is* ChucK's fault) Can you post a perhaps simplified version of the kilo-shred that is causing this disaster? By the way, we are making progress on garbage collection though nothing is enabled in the release. We'll definite priority boost GC. To sort of quote Samuel L. Jackson: I have had it with this muthachucking garbage in this muthachucking language! Best, Ge!
Ge;
Heh, ChucK is certiainly good at generating new sorts of issues!
Indeed and cheerfully so.
(Are you sure we can't blame ChucK?)
Well, that computer will have issues over nearly anything and needs a reboot or two a day. It needs a re-install, I'll probably set her up with Ubuntu. To make sure I have facts straight:
this is running command line chuck (or miniAudicle) on XP?
Well, both. The Mini has the added advantage of slowly increasing it's shred counter before bailing. I postulate this may be the unfinished garbage collector catching up with
us. What type of things are happening in these thousands of shreds? New unit generators allocated? String concatenation? new objects?
None of those. The only thing those shreds have that would alocate memory is their arguments and the results of calculations being send to ugen members. I was quite carefull there but I might've made a mistake somwhere. It started crashing before I could sort out matters like functionality...
Kilo-shreds, and even mega/giga/(probably not tera yet)-shreds should be well within the normal chuckian operational range, assuming they don't do things that compell ChucK to leak loads of memory.
Right. That's what I thought. (In the end, I am
fairly certain it *is* ChucK's fault) Can you post a perhaps simplified version of the kilo-shred that is causing this disaster?
Yeah, I'll try to nail it down tomorow. To be sure; this initial test *was* ineficient and I was expecting trouble but this was a bit much. By the way, we are making progress on garbage collection though nothing is
enabled in the release. We'll definite priority boost GC. To sort of quote Samuel L. Jackson: I have had it with this muthachucking garbage in this muthachucking language!
I'm fairly sure garbage collection won't help in this particular case. I'm generating two shreds per samp right now so that's a lot of shreds at any given time even before there is a chance for anything to become garbage. I'll try to nail down a simplified yet issue-causing version. Thanks, Kas.
participants (2)
-
Ge Wang -
Kassen