jBridge "performance" option and Cantabile audio cycles

[TiUser]

Just accidentally I observed that using "performance option" with jBridged plugins ("run isolated") seems to reduce timing load in Cantabile significantly!

The jBridge help page tells this option processes the plugin asynchronuosly for the price of one extra audio buffer. Funny - giving Cantabile one additional buffer in the audio engine settings doesn't even come close to the effect with jBridge... wouldn't it be possible to integrate this processing schema into Cantabile itself?

Don't tell me simply use jBridge... not all plugins run with it. For instance Kontakt4 does not work and crash because it has "64 audio outs" and jBridge can just handle something of about 30+.

Anyone else observed similar effects?

Please try "run isolated" with heavy loading plugins in Cantabile if you have jBridge installed. Free jBridge demo - as I have - does the test trick too. Please report your observations here.

Thanks.

[humphrey]

Hi TiUser,

I can't believe what I observe here: I did some fast tests and what I can see is a decrease in cpu consumption I would call dramatical. In one setup I get about 27% from about 60% befor, in another setup it's 5% compared to 20%.

I think I have to do some further intensive testing - but at first glance this seems to be teriffic.

One observation I made: I tried to decrease the buffer size in the soundcard to compensate the additional buffer. The result was a strong increase in cpu consumption (which was not unexpected). After increasing buffer size to the old value again I didn't get this low consumption again. I had to open the jBridge surface again and declick and click the performance button to get this activated again - strange.

I'll keep you informed if I get further information. Fyi: the plugs I used with jBridge (isolated) were Halion3 and Omnisphere.

Meanwhile many thanks for this hint.

King reagards, humphrey

[TiUser]

Glad to see similar observations with you!

Some other points:

- Did you observe occasional load spiking above the value you had before using jBridge with performance option set?

- Did you notice / feel disadvantages in latency with this "optimization"?

- I noticed that Cantabile's optimized plugin reuse mechanisms kept previously bridged or not bridged plugins in this state during reuse, regardless how these were stored in the next session...

- It looks like jBridge sometimes crashes for reasons I actually do not understand and can not really willingly reproduce.


Also I have no real insight I have my own idea why this difference might be possible. All Audio applications are finally tied to an ASIO driver which determines the clock pulse with which audio buffers have to be calculated. Even with multithreading (and all the balancing issues) finally everything has to be synced in time with ASIO requests. But jBridge is different, it starts separate processes (not threads) for each plugin. My imagination is that with the performance option set in jBridge the strict ASIO sync and the needed realtime response in a single process (= Cantabile application) is loosened and more cpu time can be used to calculate audio. Maybe I am wrong but it's the only difference I can finally see, multiple processes instead of threads in one process.

Wished Brad were here to comment on this idea.

[humphrey]

[quote="TiUser"]Glad to see similar observations with you!

Some other points:

- Did you observe occasional load spiking above the value you had before using jBridge with performance option set?

>>> No, the maximum values showing up in cantabile are much lower than the minimums before changes...

- Did you notice / feel disadvantages in latency with this "optimization"?

>>> No not really. Of course it's always a problem to identify differences when you use low latencies...

- I noticed that Cantabile's optimized plugin reuse mechanisms kept previously bridged or not bridged plugins in this state during reuse, regardless how these were stored in the next session...

>>> yep...

- It looks like jBridge sometimes crashes for reasons I actually do not understand and can not really willingly reproduce.

>>> sorry, only a few tests till now. During this period no crashes observed...

Also I have no real insight I have my own idea why this difference might be possible. All Audio applications are finally tied to an ASIO driver which determines the clock pulse with which audio buffers have to be calculated. Even with multithreading (and all the balancing issues) finally everything has to be synced in time with ASIO requests. But jBridge is different, it starts separate processes (not threads) for each plugin. My imagination is that with the performance option set in jBridge the strict ASIO sync and the needed realtime response in a single process (= Cantabile application) is loosened and more cpu time can be used to calculate audio. Maybe I am wrong but it's the only difference I can finally see, multiple processes instead of threads in one process.

Wished Brad were here to comment on this idea.[/quote]

[TiUser]

I've just joined the jBridge KVR thread.

If anyone is interested in joining this too, jump to:
http://www.kvraudio.com/forum/viewtopic ... 73#4225773