Hallo Tim!
building this is much easier than building devel_0_38 from cvs and it's hopefully quite stable at the moment ...
Is it already possible to use threaded sf and fftw?
with
./configure --enable-jack --disable-alsa --enable-simd --enable-optimize --enable-threadedsf --enable-gathreadlocks --enable-newhash --enable-lockfree I get the following linking error (something with midi):
ss.o -ljack -lrt -ltk8.4 -lX11 -lpthread -lm -L/usr/X11R6/lib -ltcl8.4
pd-s_midi.o(.text+0x4da): In function glob_midi_dialog': : undefined reference to
midi_oss_init'
pd-s_midi.o(.text+0xc83): In function sys_open_midi': : undefined reference to
midi_oss_init'
pd-s_midi.o(.text+0xd67): In function sys_reopen_midi': : undefined reference to
midi_oss_init'
s_midi_alsa.o(.text+0x4e): In function sys_do_open_midi': : undefined reference to
snd_seq_open'
s_midi_alsa.o(.text+0xa8): In function sys_do_open_midi': : undefined reference to
snd_seq_create_simple_port'
s_midi_alsa.o(.text+0x108): In function sys_do_open_midi': : undefined reference to
snd_seq_create_simple_port'
s_midi_alsa.o(.text+0x13b): In function sys_do_open_midi': : undefined reference to
snd_seq_client_info_malloc'
s_midi_alsa.o(.text+0x150): In function sys_do_open_midi': : undefined reference to
snd_seq_get_client_info'
s_midi_alsa.o(.text+0x164): In function sys_do_open_midi': : undefined reference to
snd_seq_client_info_set_name'
s_midi_alsa.o(.text+0x16f): In function sys_do_open_midi': : undefined reference to
snd_seq_client_info_get_client'
s_midi_alsa.o(.text+0x186): In function sys_do_open_midi': : undefined reference to
snd_seq_set_client_info'
s_midi_alsa.o(.text+0x191): In function sys_do_open_midi': : undefined reference to
snd_seq_client_info_free'
s_midi_alsa.o(.text+0x1d0): In function sys_do_open_midi': : undefined reference to
snd_midi_event_new'
s_midi_alsa.o(.text+0x2e5): In function sys_putmidimess': : undefined reference to
snd_seq_event_output_direct'
s_midi_alsa.o(.text+0x404): In function sys_poll_midi': : undefined reference to
snd_midi_event_init'
s_midi_alsa.o(.text+0x42e): In function sys_poll_midi': : undefined reference to
snd_seq_event_input_pending'
s_midi_alsa.o(.text+0x46e): In function sys_poll_midi': : undefined reference to
snd_midi_event_decode'
s_midi_alsa.o(.text+0x4b9): In function sys_poll_midi': : undefined reference to
snd_seq_event_input'
s_midi_alsa.o(.text+0x4f0): In function sys_close_midi': : undefined reference to
snd_seq_close'
s_midi_alsa.o(.text+0x4fe): In function sys_close_midi': : undefined reference to
snd_midi_event_free'
collect2: ld returned 1 exit status
make[1]: *** [pd] Fehler 1
make[1]: Leaving directory `/home/holzi/pd-0.38-4-devel-1/src'
make: *** [all-recursive] Fehler 1
LG Georg
building this is much easier than building devel_0_38 from cvs and it's hopefully quite stable at the moment ...
- Is it already possible to use threaded sf and fftw?
i don't think so ... this would require some changes in pd's thread handling that are not really minor ... since these are most likely changes, that miller won't accept i don't think that i'll work on this in the near future ... at the moment you can either gain speed or click-free loading, but not both ...
- with
./configure --enable-jack --disable-alsa --enable-simd --enable-optimize --enable-threadedsf --enable-gathreadlocks --enable-newhash --enable-lockfree
hm ... disabling alsa will break it, due to the alsa midi code ... maybe --disable-alsa should enable oss midi ...
thanks for the feedback ... tim
hi
just reporting a working version of devel 0.38.4 (stable release) with jack -cb_scheduler (no glitches).
with ./configure --enable-lockfree --enable-newhash --enable-jack --enable-fftw --enable-simd --enable-optimize=pentium4 --enable-threadedsf --enable-atomic
** fftw AND threadedsf = wrong?
starting pd -rt -jack -cb_scheduler
** should i add something else (devel specific)?
thank you tim, pat
** fftw AND threadedsf = wrong?
fftw: resorting the dsp chain is not threadsafe soundfiler: resorting the dsp chain after loading data to table threaded soundfiler: resorting takes place in the helper thread ... BOOM
starting pd -rt -jack -cb_scheduler
** should i add something else (devel specific)?
basically fine ... still, the cb_scheduler may result in lower latencies, but it's rather untested and thus might be not that stable ... on the other hand ... what happens to an untested piece of software, that no one dares to test :-)
cheers ... tim