Re: [linux-audio-user] [ANNOUNCE] jack_convolve-0.0.1

Hi, i implemented a small convolution engine for jack.. grab it at http://affenbande.org/~tapas/jack_convolve-0.0.1.tgz untar make ./jack_convolve responsefile.wav It creates as many in/out ports as the response file has channels. Uses fftw3f and libsndfile. Will add libsamplerate support in near future. So for now, make sure the samplerate of jack and the response file match. Here's a ca 1 sec 48khz (resampled from 96khz) stereo response file of a room:

It's from this package:

which has 96khz responses.. Consumes ca. 25-30% cpu load on my 1.2ghz athlon at jack buffer size 2048 [;)] So there's plenty room for optimization (and some return value checking will be added too ;)).. If you know some tricks, let me know.. The sourcecode is pasted below for easier reference. Flo P.S.: thanks to mario lang for collaborating and giving some hints towards using fftw3f instead of fftw and some other optimizations.. P.P.S.: oh yeah, example sound, here you go [hydrogen dry then with output of jack_convolve mixed to it]: http://affenbande.org/~tapas/jack_conv_ex1.ogg And here the convoluted signal alone: http://affenbande.org/~tapas/jack_conv_ex2.ogg P.P.S.: known issues: - won't handle samplerate or buffersize changes gracefully - will bring your machine to a crawl ;) jack_convolve.cc: --------------- /* Copyright (C) 2004 Florian Schmidt This program is free software; you can redistribute it and/or modify it under the terms of the GNU Lesser General Public License as published by the Free Software Foundation; either version 2.1 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU Lesser General Public License for more details. You should have received a copy of the GNU Lesser General Public License along with this program; if not, write to the Free Software Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA. $Id: types_8h-source.html,v 1.1 2004/04/27 18:21:48 joq Exp $ */ #include <jack/jack.h> #include <iostream> #include <sstream> #include <unistd.h> #include <signal.h> #include <stdio.h> #include <sndfile.h> #include <vector> #include <cmath> #include <fftw3.h> jack_client_t *client; std::vector<jack_port_t *> iports; std::vector<jack_port_t *> oports; jack_nframes_t jack_buffer_size; int chunks_per_channel; // channel chunk data std::vector<std::vector <fftwf_complex*> > chunks; // the buffers for the fft float *fft_float; fftwf_complex *fft_complex; // the plan fftwf_plan fft_plan_forward; fftwf_plan fft_plan_backward; float normalize_factor; // per channel we need a ringbuffer holding the fft results of the // audio periods passed to us by jackd. // each needs to be sized jack_buffer_size * chunks_per_channel (see main()) std::vector<fftwf_complex *> ringbuffers; // this gets advanced by jack_buffer_size after each process() callback unsigned int ringbuffer_index = 0; // a vector to hold the jack buffer pointers.. these get resized // during init in main std::vector<jack_default_audio_sample_t *> ibuffers; std::vector<jack_default_audio_sample_t *> obuffers; // channel data std::vector<jack_default_audio_sample_t *>overlaps; int process(jack_nframes_t frames, void *arg) { // std::cout << " " << ringbuffer_index; // get pointer[s] to the buffer[s] int channels = chunks.size(); for (int channel = 0; channel < channels; ++channel) { ibuffers[channel] =

frames)); obuffers[channel] =

frames)); } for (int channel = 0; channel < channels; ++channel) { // copy input buffer to fft buffer for (int frame = 0; frame < jack_buffer_size; ++frame) { fft_float[frame] = (float)(ibuffers[channel][frame]); fft_float[frame+jack_buffer_size] = 0.0; } // fft the input[s] fftwf_execute(fft_plan_forward); // store the new result into the ringbuffer for this channel for (int frame = 0; frame < jack_buffer_size * 2; ++frame) { ringbuffers[channel][ringbuffer_index+frame][0] = fft_complex[frame][0] / normalize_factor; ringbuffers[channel][ringbuffer_index+frame][1] = fft_complex[frame][1] / normalize_factor; } // zero our buffer for the inverse FFT, so we can simply += the // values in the next step. for (int frame = 0; frame < jack_buffer_size * 2; ++frame) { fft_complex[frame][0] = 0; fft_complex[frame][1] = 0; }