[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [sc-dev] Ringz+Dust issue

Josh,

Yes, this is like what I was thinking of. Two thoughts:

* The mFilterRemain should also be explicitly handled - if blocksize
changes, we can't guarantee it stays at 1
* It's probably worth putting this in a macro, since many other
filters use this loop syntax and so probably need the same fix...? (In
fact, lots of filters seem to init by using ClearUnitOutputs, which
probably means they're vulnerable to the Impulse.ar gotcha?)

Something like

#define PUSH_LOOPVALS \
         int tmp_floops = unit->mRate->mFilterLoops; \
         int tmp_fremain = unit->mRate->mFilterRemain; \
         unit->mRate->mFilterLoops = 0; \
         unit->mRate->mFilterRemain = 1;
#define POP_LOOPVALS \
         unit->mRate->mFilterLoops = tmp_floops; \
         unit->mRate->mFilterRemain = tmp_fremain;

Seem OK?
Dan


2008/12/15, Josh Parmenter <josh@xxxxxxxxxxxxxxxxx>:
> Ooops... left out a line:
>
>         int tmp = unit->mRate->mFilterLoops;
>         unit->mRate->mFilterLoops = 0;
>         Ringz_next(unit, 1);
>         unit->mRate->mFilterLoops = tmp;
>
>
>  Josh
>
>  On Dec 14, 2008, at 2:42 PM, Dan Stowell wrote:
>
>
> > Ah, I see, the old Impulse.ar(1) gotcha!
> >
> >
> > > it should probably do everything next does, except output
> > >
> >
> > Creating output is not a problem, in fact I think it *should* create
> > output (so that things further down the chain initialise properly). A
> > workaround/solution could be to *temporarily* change the mFilterLoops
> > and mFilterRemain values in the unit->mRate struct to represent a
> > 1-sample blocksize, which would allow the calculations in Ringz_next()
> > all to drop through correctly without requiring large amounts of
> > extra/duplicated code.
> >
> > I'm going to bed now but can look at this in the morning, that is if
> > someone else doesn't get to it first ;)
> >
> > Dan
> >
> >
> > 2008/12/14, Josh Parmenter <josh@xxxxxxxxxxxxxxxxx>:
> >
> > > I am remembering parts of this now... Below creates the problem where
> > > Impulse.ar(1) misses that first sample.
> > >
> > > Dan - I think the solution may be to create another next (or 'first')
> > > function that calculates the first sample. Basically, it should probably
> do
> > > everything next does, except output. This way, the initial sample and
> filter
> > > state is still calculated, but the mFilterRemain is avoided. My guess
> is,
> > > when called with one sample (after the Ctor) the pointers are accessing
> junk
> > > in the server memory.
> > >
> > >
> > > Josh
> > >
> > >
> > > On Dec 14, 2008, at 1:53 PM, Dan Stowell wrote:
> > >
> > >
> > >
> > > >
> > > > 2008/12/14, Josh Parmenter <josh@xxxxxxxxxxxxxxxxx>:
> > > >
> > > >
> > > > > On Dec 14, 2008, at 1:26 PM, Batuhan Bozkurt wrote:
> > > > >
> > > > >
> > > > > > Somehow Ringz.ar remembers its past state...
> > > > > >
> > > > > >
> > > > > >
> > > > >
> > > > > Like I said, this would indicate to me that some memory isn't being
> > > > >
> > > >
> > > init'd
> > >
> > > >
> > > > > correctly.
> > > > >
> > > > > I'll look it it later.
> > > > >
> > > > >
> > > >
> > > > Josh, I had the same feeling as you, but as JH points out Ringz
> > > > doesn't use RTAlloc. It correctly initialises all its variables, as
> > > > far as I can tell.
> > > >
> > > > I think I have found the problem. When the Ringz ugen is created, it
> > > > calls Ringz_next() but with a blocksize of 1 (this is standard ugen
> > > > init procedure). However, the iterations which look like
> > > >
> > > > LOOP(unit->mRate->mFilterLoops,
> > > > ...
> > > > );
> > > > LOOP(unit->mRate->mFilterRemain,
> > > > ...
> > > > );
> > > >
> > > > assume that the blocksize is what it normally is (i.e. 64, for most
> > > > people). This causes the looping to write outside the memory that it
> > > > should do (and probably therefore leaves garbage around to "remember"
> > > > inappropriately).
> > > >
> > > > I'm not sure how best to fix this. One way is that the constructor
> > > > could call ClearUnitOutputs() rather than trying to run Ringz_next(),
> > > > which from my testing does prevent the nasties. Not sure if this
> > > > change would have bad side-effects...? What do you think?
> > > >
> > > > Dan
> > > >
> > > > _______________________________________________
> > > > sc-dev mailing list
> > > >
> > > > info (subscription, etc.):
> > > >
> > >
> http://www.beast.bham.ac.uk/research/sc_mailing_lists.shtml
> > >
> > > > archive:
> https://listarc.bham.ac.uk/marchives/sc-dev/
> > > > search:
> > > >
> > > https://listarc.bham.ac.uk/lists/sc-dev/search/
> > >
> > > >
> > > >
> > >
> > > ******************************************
> > > /* Joshua D. Parmenter
> > > http://www.realizedsound.net/josh/
> > >
> > > "Every composer – at all times and in all cases – gives his own
> > > interpretation of how modern society is structured: whether actively or
> > > passively, consciously or unconsciously, he makes choices in this
> regard. He
> > > may be conservative or he may subject himself to continual renewal; or
> he
> > > may strive for a revolutionary, historical or social palingenesis." -
> Luigi
> > > Nono
> > > */
> > >
> > >
> > > _______________________________________________
> > > sc-dev mailing list
> > >
> > > info (subscription, etc.):
> > >
> http://www.beast.bham.ac.uk/research/sc_mailing_lists.shtml
> > > archive:
> https://listarc.bham.ac.uk/marchives/sc-dev/
> > > search:
> https://listarc.bham.ac.uk/lists/sc-dev/search/
> > >
> > >
> >
> >
> > --
> > http://www.mcld.co.uk
> >
> > _______________________________________________
> > sc-dev mailing list
> >
> > info (subscription, etc.):
> http://www.beast.bham.ac.uk/research/sc_mailing_lists.shtml
> > archive: https://listarc.bham.ac.uk/marchives/sc-dev/
> > search:
> https://listarc.bham.ac.uk/lists/sc-dev/search/
> >
>
>  ******************************************
>  /* Joshua D. Parmenter
>  http://www.realizedsound.net/josh/
>
>  "Every composer – at all times and in all cases – gives his own
> interpretation of how modern society is structured: whether actively or
> passively, consciously or unconsciously, he makes choices in this regard. He
> may be conservative or he may subject himself to continual renewal; or he
> may strive for a revolutionary, historical or social palingenesis." - Luigi
> Nono
>  */
>
>
>  _______________________________________________
>  sc-dev mailing list
>
>  info (subscription, etc.):
> http://www.beast.bham.ac.uk/research/sc_mailing_lists.shtml
>  archive: https://listarc.bham.ac.uk/marchives/sc-dev/
>  search: https://listarc.bham.ac.uk/lists/sc-dev/search/
>


-- 
http://www.mcld.co.uk

_______________________________________________
sc-dev mailing list

info (subscription, etc.): http://www.beast.bham.ac.uk/research/sc_mailing_lists.shtml
archive: https://listarc.bham.ac.uk/marchives/sc-dev/
search: https://listarc.bham.ac.uk/lists/sc-dev/search/