Re: [CSOUND-DEV] CSOUND-DEV] Simple Midi Synth - multi core mutes/cuts voices
Date | 2015-10-19 19:28 |
From | moguillansky |
Subject | Re: [CSOUND-DEV] CSOUND-DEV] Simple Midi Synth - multi core mutes/cuts voices |
I am also in the need of spreading load across cores. Working with global audio variables to communicate with fx instrs, there is no way known to me to specify that an instrument is only a producer or a consumer of a specific variable, and I can confirm that only one fx seems to serialize all "parallel" producers. The Mixer* opcodes would solve the specification problem, but are they used for the analysis to determine which instruments can be run in parallel? My own results don't seem to show this. Is there a way to output the dependency graph? Eduardo Moguillansky ?? On Sat, Oct 17, 2015 at 1:41 PM, Victor Lazzarini [via Csound] <[hidden email]> wrote: I am not sure there will be an instance for each voice, but the dependency will mean that it will wait for all voices to run before it runs, and that creates a bottleneck. One solution that it might be tried, for effects, is to use another Csound process for effects and connect it via Jack. You can run that process in a single thread. Victor Lazzarini Dean of Arts, Celtic Studies, and Philosophy Maynooth University |
Date | 2015-10-19 19:39 |
From | jpff |
Subject | Re: [CSOUND-DEV] CSOUND-DEV] Simple Midi Synth - multi core mutes/cuts voices |
the function to print semantic information is csp_orc_sa_print_list whic is commented out in Engine/new_orc_parser.c ad in OOPs/compile.c I guess it could be made available under a -m flag, but if you have sources just uncomment these lines ./Engine/new_orc_parser.c:137: //csp_orc_sa_print_list(csound); ./OOps/compile_ops.c:73: //csp_orc_sa_print_list(csound); On Mon, 19 Oct 2015, moguillansky wrote: > I am also in the need of spreading load across cores. Working with global > audio variables to communicate with fx instrs, there is no way known to me to > specify that an instrument is only a producer or a consumer of a specific > variable, and I can confirm that only one fx seems to serialize all "parallel" > producers. The Mixer* opcodes would solve the specification problem, but are > they used for the analysis to determine which instruments can be run in > parallel? My own results don't seem to show this. Is there a way to output the > dependency graph? > > Eduardo Moguillansky > > ?? |