-psycledelics- - Printable Version: Few 1.8 bugs

-psycledelics- (http://psycle.pastnotecut.org/index.php)
|- TECH BOARD (http://psycle.pastnotecut.org/board.php?boardid=2)
|-- Few 1.8 bugs (http://psycle.pastnotecut.org/threadid.php?boardid=2&threadid=1828)

From Taika-Kim on 30.08.2005, 08:37:

Few 1.8 bugs

I have been doing my new song for some days now with no problems, then today I got a few very strange bugs.

I have a Blitz layer with 3 synths, and I added a fourth (Feedme) to the stack. Suddenly sometimes 2 of the Blitz synths go all the way to 50% on CPU usage instead of the usual less than 5%.

Another one was that one instance on Arguru Crossdelay went totally apeshit and everything that came out sounded like it's sign was switched. (remember the old FT2 function? this sounded very much the same!)

The first one is somewhat repeatable, the second one I've never encountered before.

For now it seems that if I play with the Feedme instance soloed a bit, the Blitzes go crazy on CPU usage. But I'll have to test on this a bit more...

From [JAZ] on 30.08.2005, 21:42:

Please, try with the following exe:

http://psycle.sourceforge.net/psyclebin.zip

It contains an alternate (gotten from musicdsp.org ) method to remove denormals and "bad numbers".
It seems to work better.

__________________
<[JAZ]> Pa pi pa pa pa pi pa.... ;·D

From Angelus on 31.08.2005, 10:02:

Could it be the solution ( ... or maybe the patch) for denormal problems on pentium processors ?

__________________
?( :( :) :] :D .cC( Psycle! )

From Taika-Kim on 31.08.2005, 15:31:

But I have an AMD processor, aren't they free of the denormal-problem?

Anyway, the second bug happened again yesterday, suddenly all Psycle output got totally garbled. I was working intensely so I didn't even check the culprit this time, I just restarted Psycle...

From [JAZ] on 31.08.2005, 21:06:

Denormals is one of the problems, and it is the one that causes high cpu load.

This release also implements dectection about NAN's (Non-Acceptable-Number, or something like that), which causes invalid values, which usually are like +/-inf(inite).

That could cause one of such erroneous outputs.

At last, i think i've identified still one problem with the wire volumes, when exchanging machines, or moving them in general, which could cause bad output aswell.

__________________
<[JAZ]> Pa pi pa pa pa pi pa.... ;·D

From graue on 10.09.2005, 20:57:

NaN means "Not a Number," just for the record.

From DMNXS on 24.09.2005, 01:09:

quote:
Original by graue
NaN means "Not a Number," just for the record.

heheh

...