Re: [Jack-Devel] JACK on mobile
Le 16 oct. 2013 à 20:17, Christian Schoenebeck <[hidden]> a écrit :
> On Wednesday 16 October 2013 19:07:43 Bradley Justice wrote:
>> I do have a question regarding iOS and audio latency. Does anyone on this
>> list have reliable metrics? I have seen some low figures reported on
>> Internet postings. However, figures I have encountered from individuals who
>> actually measured it seem out of range for pro audio.
>
> JACK iOS (RIP) was configurable down to 2.9ms
>
> http://www.crudebyte.com/jack-ios/jack_ios_shot_3.png
>
> which worked reliable on all iOS devices we tested (while running multiple
> audio apps simultaneously), even on older devices. Probably even lower
> latencies are possible, but we actually never tested going below that buffer
> size (due to other priorities). What I found interesting was that even ancient
> iPhone 1 devices provided reliable low latency.
>
> I have a big sympathy for modern Mach derivative kernels, even though Apple's
> XNU kernel (Mac & iOS) is not top-notch anymore, it still solves things very
> elegant and efficiently. I would love to see some L4 microkernel to enter a
> major distro as alternative to the Linux kernel in future. But due to the old
> problems of i.e. missing drivers, alternative kernels won't have a chance on
> major open source distros that soon I fear ...
>
> CU
> Christian
Well Christian you are speaking of the 2 * buffer size latency here right?
OSX usually decompose I/O latency as = AD conversion latency + input latency offset + 2 * buffer size + output latency offset + DA conversion latency
So I would guess the minimal I/O latency on IOS is more than the 2.9 ms value
.
Stéphane
1381956552.30142_0.ltw:2, <EFC2CE45-433E-4619-8AAE-E97C6EDD4BD4 at grame dot fr>