Concurrent CSound: Parallel Execution fro High Speed Direct SynthesisSkip other details (including permanent urls, DOI, citation information)
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 3.0 License. Please contact email@example.com to use this work in a way not covered by the license. :
For more information, read Michigan Publishing's access and usage policy.
Page 46 ï~~CONCURRENT PARALLEL EDE' DIRlE HIGH SPEED Nick Bailey Alan Purvis DURHAM MUSIC TECHNOLOGY* School of Engineering & Applied Science University of Durham South Road Ian Pet DU Del Uni Pal Du EN Durham DH1 3LE ENGLAND Michael Clarke, Music Department, Hud ABSTRACT: This group has already a the speed of MIT's CSOUND direct synt] after suitable pre-processing the sco impact of an advanced parallel architect results indicating the speed increases ac demonstration of the system illustrates
Page 47 ï~~The analysis presented here assumes that t elements: useful computation, while the process( directly involved in the production of sound sam processor is busy performing buffering or comm themselves contribute to synthesis; and suspensic the processor awaits host I/O with insufficient b manipulating the program source code and mnput the relative time spent in each of these states. Two scores and orchestras were used to benc example with many short events in the score and inexpensive; programme 2** has fewer events in demads agreat deal of computation. 2.1 Execution Profiling The first experiment attempts to evaluate how and how much is spent in useful computation. processors in four different contexts.
Page 48 ï~~An (c) estimate of the time spent in necessa test but this case, the main program S the sound output buffering software is me communication overhead is retained, but the s moment. From this benchmark, we deduce t and 12% in necessary computation. Put, not compilation of this example: 24% is spent susp 67% is spent in useful computation; only is an encouraging indication of the ability of th operations between concurrent processes withoi throughput. A simulation was devised to estimate the is sorted using the standard utility, but specif' of three. The sorted score is then passed throu lines marked to be routed to non-existent proc allocated sparsely amongst the processors (test a large system. If there is sufficient buffering freely, the execution should take approximatel; c IS 0 0 0 4e fAn errtri, f n~inf
Page 49 ï~~There is a 27% increase in compilatio: required with 32 buffers. This is a signific compared with the performance of some a parallel environment. 2.3 Performance Indices As a final experiment, programme 2 is comp one, two and three processors, and the P processor version of the program runs with expected, as it has a far lower communicatio addition of an extra processor (with the at the performance index to 0.90, but the ad, effect upon the performance index. It is t the addition of an extra processor which in parallel environment. Conclusions cost-effective solution to the problem of synthesis machines of reasonable performaz 1*itn fnrtnr nnw in, ratflif rnfa