ALSA folks can you help us to fix Chris' problems ?
for the ALSA folks that that are not yet aware about,
a few days ago I released an experimental harddisk-based realtime sampler
(which streams the files directly from disk with precaching of the initial
parts)
see the story here: http://eca.cx/lad/2000/Jul/0149.html
download the sampler here:
http://www.gardena.net/benno/linux/linuxsampler.tar.gz
anyway I am using OSS for my tests on an AWE64 with 3x128 fragments
and using the RAW midi device ( /dev/midi00) to process input
(I open the device in blocking mode and
then sit in a loop which does
while(1) {
read(midifd,buf,1);
process_midi_byte() ;
}
Chris gets "Resource temporarily unavailable" errors
within the midi loop.
Is that because ALSA has problems with blocking MIDI input
(I recall discussion about blocking/non blocking MIDI , but I can't remember
the conclusion) ?
Under OSS/Free blocking read()s work nicely and return as soon midi bytes are
available.
now to the other stuff
On Fri, 14 Jul 2000, Chris Baugher wrote:
> Hello,
>
> I was just playing with your disk-sampler and I've run into two problems.
> First, mlockall() returns an error. Second, the program immediately exits
> after I hit the first note. Here is the output:
(before starting playing, wait 1-2 secs until all threads get fired up , or you
will get a segfault sometimes when pressing the first key because all
messages queues are not yet initialized)
the instrument requires about 35MB available which wants to get mlocked()
I exaggerated with the allocateded diskstreams (100) , so simply
in types.h set MAX_INPUT_STREAMS to 40
and MAX_AUDIO_VOICES to 32
and the mlock should be ok
>
>
> fluke:/home/cygnus/linuxsampler# ./disksamp ./sine.sam
> main start ....
> Instrument::load fd=3
> header res=16400
> LOAD: num_waves=61
> WARNING, can't mlockall() memory ! mlockall: Cannot allocate memory
> midi: start ....
> WARNING, can't mlockall() memory ! mlockall: Cannot allocate memory
> AUDIO: using STEREO output
> using ALSA audio output ....
> ALSA: fragmentsize=128
> ALSA: numfragments=3
> WARNING, can't mlockall() memory ! mlockall: Cannot allocate memory
> ERROR in read_midi_bytes: read: Resource temporarily unavailable
> fluke:/home/cygnus/linuxsampler#
>
>
> The "ERROR in read_midi_bytes:" only appears when the first note is hit.
> I tried putting a usleep(200000) in the MidiIn main loop. The error
> doesn't apear and sometimes it makes noise but not always. Is it possible
> this is an ALSA problem?
Probably, especially the ES1371 (I have one too seems to show up a weird
behavior when using very small fragments ( try fragmentsizes of 256 and 512
using the commandline and see if it helps)
What version are you using?
For now OSS here , because I am currently at home on my old ALSA-less PII400
:-)
> I tried the test program rawmidi.c in alsa-lib/test and it produces notes
> so it seems midi is working. aplay works fine as well.
> This is what my system looks like if it helps:
>
> Pentium 250 (overclocked)
> 64MB ram
> 3.2 gig EIDE disk
> SB AudioPCI (ens1371)
> crappy old Yamaha keyboard
> 2.2.14 kernel patched for low-latency
> ALSA 0.5.8b
>
> Ideas?
Regarding the MIDI issue, I am waiting for an answer from the ALSA folks since
I haven't do much midi IO on my other box which has ALSA installed plus
I have not access to this box currently.
As for ALSA audio output, I am doing nothing magic: I simply setup
the card in BLOCK mode and then do simple pcm_write() of fragments
in a loop.
>
> BTW, I think this is a very cool project. I would someday like to get a
> midi drumkit(I am a drummer and YES drummers are very latency sensitive:)
> and if I could use my PC as the sound module that would be fantastic! I
> don't know how much I could contribute to this project but I'll see what I
> can do.
That would be cool, plus think about the fact the having virtually unlimited
disk space, will allow you to sample every fine nuance of the drumkit
( eg. multiple samples per different velocities etc)
Let us know if you are able to solve your problems.
Benno.
>
> Chris|
