PortableCL 1.2 Still Coming While POCL 1.3 Will Further Improve Open-Source OpenCL

[phoronix]

Phoronix: PortableCL 1.2 Still Coming While POCL 1.3 Will Further Improve Open-Source OpenCL

It's been a number of months since last having any major news to report on POCL, the "PortableCL" project providing a portable OpenCL/compute implementation that can run on CPUs, select GPUs, and other accelerators...

http://www.phoronix.com/scan.php?page=news_item&px=POCL-1.2-RC-1.3-Dev

[cesarcafe]

This looks really nice, and very useful!! If there was a similar effort for CUDA, it would really make GPGPU development sweeter than it is today. Anyway, PortableCL is going to make my GPGPU life easier, but if there was a CUDA counterpart, then it would be perfect.

[oleid]

If there was a similar effort for CUDA, it would really make GPGPU development sweeter than it is today.

There is - sort of: https://github.com/ROCm-Developer-Tools/HIP

[mzs.112000]

Okay, can we get OpenCL to actually work on Quadro FX 770M or similar old GPU's? They are still useful for light tasks today(arguably better than some new CPU's), if they had functional OpenCL or a CUDA implementation that still worked(the Ubuntu package and the Arch package both want to install a new version of CUDA that doesn't support old GPU's).

[cesarcafe]

There is - sort of: https://github.com/ROCm-Developer-Tools/HIP

Thanks for telling. I searched for CUDA alternative implementations in the past, and there are some attempts, but either unfinished or unmaintained/abandoned. I didn't know of HIP (which, although not being a CUDA implementation, could be quite useful in my GPGPU work). However, it seems to be Linux-only, but I target Linux, MacOS, and Windows, so I believe there's still no adequate solution to a CUDA alternative implementation. In any case, I didn't know of PortableCL, and it's going to be really useful to me (on the OpenCL side of things).