Ai Dreams Forum
Member's Experiments & Projects => General Project Discussion => Topic started by: frankinstien on May 31, 2021, 10:03:12 pm
-
I currently have a Vega 56 GPU which is AMD where I'm implementing OpenCL through Cloo and OpenCLTemplate to generalize the ability to work with kernels. So I enhanced an Open Source OpenCL editor for just that, below is an image of it:
Click on image to get a larger view:
(https://i.imgur.com/O0IpKk4.png) (https://i.imgur.com/O0IpKk4.png)
The editor can generate data as well, which makes it very convenient when troubleshooting code. It can store the configuration in a file including the data generated. It can also load from and save to a noSQL database. The noSQL entries are then used to retrieve kernels where they can be compiled and stored in the GPU and launched when needed. There is metadata associated with the kernel parameters so outputs can be configured appropriately.
So, now GPU functions can be associated with concept expressions and are referenced to the noSQL entries, where the KernelEntries have the code, and generalized interface to compile, apply data, and retrieve results.
I will be integrating a debugger, where it can step through the openCL code. It currently will detect compile errors and report them.
-
Thats a pretty good gpu, better than my gtx980. Have u run any raytracing on it? they kick butt at it.
They are also good at Ai, and physics engines, and string matching.
So what stuff have you implemented on it? And whats this program here, I can see its using "i" code, I bet your getting tonnes of output out of it, gpu's are really cool.
Only things I can think of that're cooler are fpga's and quantum computers.
-
So what stuff have you implemented on it?
Currently working with Fourier transform for sound processing and using the gpu for visual processing as well.
-
I don't know why you put that "i" next to the cosine component, seems like worthless notation.
U could compute a GINORMOROUS Fourier spectrum on the gpu, if it was a normal sized window you could probably run at least 10,000 of them simultaneously, quite spectacular. completely superfluous unless u had a reason to. Its good packing the power, why I always keep my desktop there when I need to do something serious that my little ol' lappy cant handle.
-
I don't know why you put that "i" next to the cosine component, seems like worthless notation.
There's a cos(a * k), I don't see a cos(i) or for that matter a variable "i" anywhere in the OpenCL kernel.
-
So it should be, I only like real numbers. ;D