[Csnd] experimental OpenCL GPU code

[Csnd] experimental OpenCL GPU code

Date	2014-04-19 10:55
From	Victor Lazzarini
Subject	[Csnd] experimental OpenCL GPU code
	In addition to the growing collection of CUDA-based GPU opcodes (which includes phase vocoder analysis, synthesis, additive synthesis, sliding PV pitch shifter, direct convolution and partitioned convolution), I have added now an OpenCL opcode for additive synthesis. This allows people without access to CUDA tools to experiment using their GPUs. I have not yet integrated this into the cmake build, but there is a build.sh with the build command for OSX there (which can be easily modified for linux and Windows). You will find the source, build script and an example CSD in Opcodes/opencl. Note that this is experimental and carries no guarantee of success ;) ======================== Dr Victor Lazzarini Senior Lecturer NUI Maynooth, Ireland victor dot lazzarini at nuim dot ie

Date	2014-04-19 13:04
From	menno
Subject	[Csnd] Re: experimental OpenCL GPU code
	i am not sure i understand. Can i use my videocard to process Csound instructions? And how much faster can it run? Where can i find some info about this? -- View this message in context: http://csound.1045644.n5.nabble.com/experimental-OpenCL-GPU-code-tp5734425p5734427.html Sent from the Csound - General mailing list archive at Nabble.com.

Date	2014-04-19 16:32
From	Askwazzup
Subject	[Csnd] Re: experimental OpenCL GPU code
	menno wrote > i am not sure i understand. > Can i use my videocard to process Csound instructions? And how much faster > can it run? > > Where can i find some info about this? Me too, it would be wonderful to put my 650gtx to some use, as it's just idling away since i got rid of Windows. -- View this message in context: http://csound.1045644.n5.nabble.com/experimental-OpenCL-GPU-code-tp5734425p5734436.html Sent from the Csound - General mailing list archive at Nabble.com.

Date	2014-04-19 17:59
From	Victor Lazzarini
Subject	Re: [Csnd] Re: experimental OpenCL GPU code
	The only info at the moment is the example CSD in the sources. It should be self-explanatory. Here, the additive synthesis with 1024 oscillators runs about twice as fast in the opencl version (and about 3 times in the cuda one) than the serial CPU code. ======================== Dr Victor Lazzarini Senior Lecturer NUI Maynooth, Ireland victor dot lazzarini at nuim dot ie On 19 Apr 2014, at 16:32, Askwazzup wrote: > > menno wrote >> i am not sure i understand. >> Can i use my videocard to process Csound instructions? And how much faster >> can it run? >> >> Where can i find some info about this? > > Me too, it would be wonderful to put my 650gtx to some use, as it's just > idling away since i got rid of Windows. > > > > > -- > View this message in context: http://csound.1045644.n5.nabble.com/experimental-OpenCL-GPU-code-tp5734425p5734436.html > Sent from the Csound - General mailing list archive at Nabble.com. > > > Send bugs reports to > https://github.com/csound/csound/issues > Discussions of bugs and features can be posted here > To unsubscribe, send email sympa@lists.bath.ac.uk with body "unsubscribe csound" > > >

Date	2014-04-19 18:21
From	Dominic Melville
Subject	Re: [Csnd] experimental OpenCL GPU code
	Victor, is this available purely on Nvidia cards, utilizing their CUDA package, or will they work on others? Are they totally reliant on CUDA from Nvidia? I've used them successfully on my Nvidia card without problems. I've used other GPU based audio processing outside of Csound on my rpi, so far just some FFT analysis and managed to get BruteFIR running too, although that's a little shakey, solid with some convolution reverbs but awful and crashes with others, need more time and work to find out exactly why. Bit off topic I know but I'm curious to know if the Csound CUDA opcodes could be extended to run on something like the GPU on an rpi, it's a pretty fast chip from my findings and well worth the effort so far. Thanks Dominic On Saturday, April 19, 2014, Victor Lazzarini <Victor.Lazzarini@nuim.ie> wrote: The only info at the moment is the example CSD in the sources. It should be self-explanatory. Here, the additive synthesis with 1024 oscillators runs about twice as fast in the opencl version (and about 3 times in the cuda one) than the serial CPU code. ======================== Dr Victor Lazzarini Senior Lecturer NUI Maynooth, Ireland victor dot lazzarini at nuim dot ie On 19 Apr 2014, at 16:32, Askwazzup <aistiskaikaris@mail.com> wrote: > > menno wrote >> i am not sure i understand. >> Can i use my videocard to process Csound instructions? And how much faster >> can it run? >> >> Where can i find some info about this? > > Me too, it would be wonderful to put my 650gtx to some use, as it's just > idling away since i got rid of Windows. > > > > > -- > View this message in context: http://csound.1045644.n5.nabble.com/experimental-OpenCL-GPU-code-tp5734425p5734436.html > Sent from the Csound - General mailing list archive at Nabble.com. > > > Send bugs reports to > https://github.com/csound/csound/issues > Discussions of bugs and features can be posted here > To unsubscribe, send email sympa@lists.bath.ac.uk with body "unsubscribe csound" > > > Send bugs reports to https://github.com/csound/csound/issues Discussions of bugs and features can be posted here To unsubscribe, send email sympa@lists.bath.ac.uk with body "unsubscribe csound"

Date	2014-04-19 19:19
From	Hlöðver Sigurðsson
Subject	Re: [Csnd] experimental OpenCL GPU code
	Too bad that Nvidia catalyst drivers are shit on Linux, I can't use OpenCl therefore :( I'm just hoping that open source drivers will be able to run OpenCl one day, nvidia CUDA is the way to go clearly when it comes to Linux. 2014-04-19 17:21 GMT+00:00 Dominic Melville <dcamelville@gmail.com>: Victor, is this available purely on Nvidia cards, utilizing their CUDA package, or will they work on others? Are they totally reliant on CUDA from Nvidia? I've used them successfully on my Nvidia card without problems. I've used other GPU based audio processing outside of Csound on my rpi, so far just some FFT analysis and managed to get BruteFIR running too, although that's a little shakey, solid with some convolution reverbs but awful and crashes with others, need more time and work to find out exactly why. Bit off topic I know but I'm curious to know if the Csound CUDA opcodes could be extended to run on something like the GPU on an rpi, it's a pretty fast chip from my findings and well worth the effort so far. Thanks Dominic On Saturday, April 19, 2014, Victor Lazzarini <Victor.Lazzarini@nuim.ie> wrote: The only info at the moment is the example CSD in the sources. It should be self-explanatory. Here, the additive synthesis with 1024 oscillators runs about twice as fast in the opencl version (and about 3 times in the cuda one) than the serial CPU code. ======================== Dr Victor Lazzarini Senior Lecturer NUI Maynooth, Ireland victor dot lazzarini at nuim dot ie On 19 Apr 2014, at 16:32, Askwazzup <aistiskaikaris@mail.com> wrote: > > menno wrote >> i am not sure i understand. >> Can i use my videocard to process Csound instructions? And how much faster >> can it run? >> >> Where can i find some info about this? > > Me too, it would be wonderful to put my 650gtx to some use, as it's just > idling away since i got rid of Windows. > > > > > -- > View this message in context: http://csound.1045644.n5.nabble.com/experimental-OpenCL-GPU-code-tp5734425p5734436.html > Sent from the Csound - General mailing list archive at Nabble.com. > > > Send bugs reports to > https://github.com/csound/csound/issues > Discussions of bugs and features can be posted here > To unsubscribe, send email sympa@lists.bath.ac.uk with body "unsubscribe csound" > > > Send bugs reports to https://github.com/csound/csound/issues Discussions of bugs and features can be posted here To unsubscribe, send email sympa@lists.bath.ac.uk with body "unsubscribe csound"

Date	2014-04-19 19:29
From	Askwazzup
Subject	[Csnd] Re: experimental OpenCL GPU code
	It seems i can't even use Nvidia drivers on linux. Oh well.. -- View this message in context: http://csound.1045644.n5.nabble.com/experimental-OpenCL-GPU-code-tp5734425p5734446.html Sent from the Csound - General mailing list archive at Nabble.com.