**** BEGIN LOGGING AT Sat Mar 03 03:00:01 2018 Mar 03 07:18:31 ds2 are you there? Can you point me in right direction(research paper/anything) about the offloading algorithms (even for desktop?). It seems to me that the code written for GPU is written differently than for GPU Mar 03 07:20:33 ds2: I think it would be better if we can provide GPGPU on beaglebone using Legacy methods (Using OpenGL shaders to perform computation) but the programmer will still need to wxplicitly offload work to GPU Mar 03 07:21:09 jkrinder: Can you have rma comment on this as well? Mar 03 07:26:34 Can anyone guide me how to submit the Hello world code as mentioned in the ideas page? Mar 03 07:36:21 Anirban: That is mostly an exercise/test kind of stuff .. and pretty simple test to be honest. Mar 03 07:36:38 so, you should be able to figure it out yourself. Mar 03 07:37:13 zeekhuge: How can I find rma Mar 03 07:43:23 Sidharth: You can try to post it on the mailing-list and cc rma along with it ? Mar 03 07:48:36 I guess you want help on the audio platform project ? If that is it, you can even email henrix. Mar 03 07:49:20 And believe me, email might look a bit boring, but it works, ah ... usually. Mar 03 07:50:14 Well I was thinking more on the lines of GPU offload... Mar 03 07:51:16 Yes I'll email them Mar 03 07:51:23 Explore a little more Mar 03 08:05:00 Sidharth: no papers specifically but look up the concept of GPGPU Mar 03 08:05:27 in other words, what people used to do before OpenCL was standardized Mar 03 08:07:40 ds2: So I will provide legacy GPGPU support right? The programmer can explicitly decide whether to run on CPU or GPU right? Mar 03 08:07:51 ds2: Sort of OpenCL but legacy!! Mar 03 08:08:39 ds2: http://www.mathematik.uni-dortmund.de/~goeddeke/gpgpu/tutorial.html Like this tutorial Mar 03 08:12:46 no Mar 03 08:13:12 Sidharth: I'd say you would write examples and document what is practical on the OpenGLES 2.1 we have Mar 03 08:13:34 the CPU->GPU and GPU->CPU path could be too slow for some algs Mar 03 08:17:08 think of it this way - say you have the GPU do MPEG compression but each frame you do on there takes longer to then on the CPU once you add in transfer times Mar 03 08:17:14 that would not be useful Mar 03 08:20:13 So in real life how is it decided on, say Linux or desktops? Mar 03 08:22:00 by the alg Mar 03 08:22:27 the GPU is a small but massively parallel processor(s) Mar 03 08:22:55 the CPU attempts to do this with specialized instructions (i.e. NEON on ARM, MMX on x86) Mar 03 08:23:10 SIMD I think is the term Mar 03 08:24:23 So the driver decides this or the OS...(or should I search more ...) Mar 03 08:27:06 the designer/programmer Mar 03 08:31:10 zeekhuge: It was really a simple test...shouldn't have asked for help...sent the PR Mar 03 08:51:48 /msg NickServ identify 135-246 Mar 03 10:04:54 hello, I want to participate in GSoC-2018, and am finishing the cross compile task, would anyone help me test the binary file? Mar 03 10:05:08 https://github.com/hjhee/gsoc-application/tree/master/ExampleEntryJasonKridner Mar 03 10:55:14 Changed my nickname from Anirban to AnirbanBanik1998 **** ENDING LOGGING AT Sun Mar 04 03:00:01 2018