Re: CPU problems only after hours

IOhannes m zmoelnig via Pd-list <[email protected]> Wed, 15 Apr 2026 10:50:09 +0200

Newsgroups	gmane.comp.multimedia.puredata.general
Message-ID	<[email protected]>

On 4/15/26 10:38, Jakobweb via Pd-list wrote:
> Okay, good to know. I think I dont use externals at the moment. But I will check on it. And you probably mean small numbers on message level, right?

not sure what you mean with "small number on message level".

the problem typically appears in *signals* (because normally, you will 
not have that many messages; at audio rate it is pretty easy to have 
more than 44000 numbers per second; and if all of them a denormals, your 
CPU will have to handle more than 44000 exceptions).

also, when i say "small numbers" i really mean *very small* numbers, as 
in: there absolute value is less than 1.1754943*10^-38

usually, you do not enter such small values directly (not even with 
sensor data). but they might get into the system with filters: if your 
new output includes a slightly diminished last output (e.g. y[n] = 0.99 
* y[n-1]), then this output will start big (e.g. "1") but after a number 
of iterations it will become very very small, entering the realm of 
denormals.

but as i said: i don't think that any built in object suffers from that 
(as the built-ins usually check for such small numbers and set them to 
"0", which prevents the problem)

gfdma
IOhannes

-- 
please do not CC me for list-emails

---
[email protected] - the Pure Data mailinglist
https://lists.iem.at/hyperkitty/list/[email protected]/message/NGVTWOTL4MFZGVVB26KCXRUAAWFFA2Y4/

To unsubscribe send an email to [email protected] mailing list
UNSUBSCRIBE and account-management -> https://lists.iem.at/

OpenPGP_signature.asc (application/pgp-signature, 840 B)