Better than a xilinx FPGA for oak why not a GPU opencl is a c library that allow

LeafLabs Garden » Support » Maple Hardware Support

Better than a xilinx FPGA for oak why not a GPU opencl is a c library that allow

(4 posts) (4 voices)

Started 5 years ago by josheeg
Latest reply from Adam

josheeg
Member

Better than a xilinx FPGA for oak why not a GPU opencl is a c library that allowsa a GPU to be programmed in c. So maby the maple can load video or EEG or EMG data and a GPU like the one mentioned in the beagle board be used on the oak but one easier to solder. TQFP maby...

The OMAP3530 includes an HD-video capable TMS320C64x+ DSP for accelerated video and audio decoding, and an OpenGL ES 2.0 capable 2D/3D GPU. Video outputs can be provided by the on-board S-video or DVI-D (HDMI connector) outputs.

Posted 5 years ago #
soundcyst
Member

the TMS320C64c+ is a hardware core on the same die as the processor.. it's all one in one chip.

and the best way to program it would just be with regular old c, with a compiler that takes advantage of the accelerator.

and the DSP accelerator isn't the same thing as a parallelized GPU pipeline.. not a bad idea, but a bit half-baked.

Posted 5 years ago #
okie
Administrator

I personally don't know much about exactly what you can do with raw GPUs and OpenCL or other languages and tools that are available regarding, particularly, low-level hardware type things like control over a GPU's pins and communication buses. Out of the box, do the available GPU development tools support using them as stand-alone processors and communicate to whatever you want or do they need to be connected to a processor running an operating system for which there are drivers provided, or what? I've forwarded these questions to someone who will know the answer. Of course, someone else at LeafLabs or on the forums may also know. I think I've slightly rephrased the question.

Posted 5 years ago #
Adam
Member

I think I started this because josh and I were emailing about it off-forum and I suggested using OpenCL on a GPU to accelerate his EEG signal processing.

OpenCL
http://en.wikipedia.org/wiki/OpenCL

okie, yes I think OpenCL needs a host processor to drive the GPU... doesn't work standalone.

josh, I meant use the Maple to digitize EEG sensor data, then send to a laptop running an OpenCL program that uses a GPU to process the EEG digital data. The GPUs excel on decomposable problems like linear algebra and matrix operations.

Do the signal processing in software where it is easier, and use the Maple for what it is really good at, interfacing with sensors.

Posted 5 years ago #

RSS feed for this topic

Reply

You must log in to post.