BrainSimulator Port from CUDA to OpenCL

Hello there,

I saw this very interesting and innovative software, but apparently it is limited to the proprietary nvidia cuda sdk. Porting to OpenCL would enable everyone to run the software, which is very important if you want the software to gain traction and be use by a lot of researchers/users. I know this will involve a lot of work. However, I am convinced using an open standard will benefit the project enormously in the long term. You cannot simply ignore 20% of your potential users.

Mar 09 '16 15:03 cornytrace

What about Linux then?

Dec 09 '16 22:12 philipperemy

What about it? OpenCL is supported on all videocards under Linux as far as I can see.

Dec 11 '16 20:12 cornytrace

Hi, we don't plan to switch to OpenCL or to add support for GNU/Linux in the near future.

Dec 12 '16 10:12 HonzaSinkora

Using a proprietary sdk will only hurt the project in the future, you are basically at the whims of Nvidia. Not to mention that if a lot of software would come to demand CUDA functionality, we would be giving Nvidia a monopoly. This is one of the many reasons why open APIs are better in the long run.

Dec 13 '16 13:12 cornytrace

@cornytrace No one uses Open APIs in Deep Learning. You should ask it to Tensor Flow as well.

Dec 14 '16 17:12 philipperemy