Re: [hatari-devel] MIDI support for Windows & OSX (with PortMidi)

[Simon Iten <itensimon@xxxxxxxxx>]

> On 4 Oct 2017, at 00:13, Eero Tamminen <oak@xxxxxxxxxxxxxx> wrote:
> 
> Hi,
> 
> On 10/03/2017 11:57 PM, Simon Iten wrote:
>> does or how does portmidi change the way midi is presented to a usb-midi converter?
>> or in other words, why does portmidi on osx behave differently (“better” in my case) then linux with the standard none portmidi method?
> 
> PortMidi sends whole MIDI packages while the other one outputs the data byte-by-byte.
> 
> Which device drivers you use on Linux and which on OSX?
> 
on osx i use the motu drivers from the support page they have. on linux there is no firmware needed, but i seem to remember the device is not fully class compliant. (a hack of the device id is necesarry). but it works out of the box on my linux system. 
> 
>> this is with the same atari software, same synth and same interface. driver issue on linux? (fastlane-usb motu, not fully class compliant)
> 
> Are you using the same version of Hatari?
> 
> You can build Hatari also on Linux with portmidi to compare, whether they still behave the same, i.e. is the issue in the OS / driver level.
> 
> 
i am using a non portmidi version on linux.

does the github version have portmidi for linux already. or do i have to apply the patch?

i will try to test this further.
> 	- Eero
> 
>>> On 3 Oct 2017, at 22:10, Nicolas Pomarède <npomarede@xxxxxxxxxxxx> wrote:
>>> 
>>> Le 03/10/2017 à 21:26, David Savinkoff a écrit :
>>>> ----- Eero Tamminen wrote:
>>>>> PS. Jari has also some questions about jitter he's seeing in MIDI output:
>>>>> 	http://www.atari-forum.com/viewtopic.php?f=51&t=32344
>>>> As a MIDI jitter testing experiment, try the following sound options:
>>>> --sound-buffer-size 10
>>>> --sound-sync true
>>>> These options affect emulation rate, and may improve the jitter.
>>> 
>>> Hi
>>> 
>>> I'm not sure this would help ; for midi output, you'd want as realtime possible behaviour, which I think is better achieved when emulation speed is not audio driven but using usleep in the main loop.
>>> In that case if MFP expects a timer to last 1.39 sec for example under STF, then Hatari should take also exactly 1.39 sec.
>>> 
>>> Plus, MIDI programs don't necessarily use the YM2149 chip, so you would have no audio data to sync with (except playing empty audio buffers).
>>> 
>>> Put another way, I have more trust in usleep precision in that case than in syncing with an audio driver whose latency might vary depending on output sample rate or buffers size :)
>>> 
>>> Nicolas
>>> 
>>> 
> 
>