Personally I would use a single [writesf~].
Each [writesf~] has its own background thread where most of the time is being spent waiting for disk, so I'd say there's no point in parallelizing by using several [writesf~] objects. On the contrary, you only end up with more context switches.
Personally, I occasionally experience audio dropouts when using many [readsf~] objects which I attribute to the excessive mutex locking since the objects are not really CPU intensive...
Christof
PS: I think that all [readsf~] resp. [writesf~] objects should really share a single IO thread and use a lockfree fifo instead of mutexes (like in SuperCollider)...
On 28.09.2020 19:51, Fede Camara Halac wrote:
Hi,
What is more efficient for recording, say, a +15 minute 16 channel performance: a single writesf with 16 channels or 8 writesf object with 2 channels each?
Bonus question: What if I place a stereo writesf inside a patch and run 8 separate pd~ objects?
The context: recording network performance using Netty McNetface. This is why I'm trying to reduce as much processing from the main patch while staying within pd-land. (Routing multichannel audio out to another daw would be an option that I'm not considering right now)
Thanks!
f fdch.github.io
Pd-list@lists.iem.at mailing list UNSUBSCRIBE and account-management -> https://lists.puredata.info/listinfo/pd-list