Pocl 3.0 released with initial Opencl 3.0 support (and LLVM 14)

OpenCL 3.0.12 released (with his SDK)

You could look for online introductory courses in OpenCL. I'm interested in such a course so I can start doing an OpenCL version for another BOINC project. Ralph@Home and Rosetta@Home look like very difficult first projects, though.

The online OpenCL courses I've found so far all appear to either be long and expensive, or be basically addons for someone who is already using OpenCL.

The online OpenCL courses I've found so far all appear to either be long and expensive, or be basically addons for someone who is already using OpenCL.

Today is not necessary to write code "directly" in OpenCl.
You can write c++ code and target OpenCl kernel and device.
For example, you can use oneAPI (it's SYCL) to target Nvidia and AMD gpu.

Today is not necessary to write code "directly" in OpenCl.
You can write c++ code and target OpenCl kernel and device.

An example: GPU performance portability using standard C++ with SYCL

PoCL 4.0 released, with support to OneApi driver (Intel GPU)

You could look for online introductory courses in OpenCL. I'm interested in such a course so I can start doing an OpenCL version for another BOINC project.

This is interesting, but it seems not for entry level user...

Kronos Group released updated documentation about C++ for OpenCL on github

The latest version is 3.0.14

HypSycl/OpenSycl has renamed as AdaptiveCpp

New PoCL 5.0 RC-1 has a new interesting features, OpenCl across networked system (over a LAN)

Have you read the discussions over on Rosetta@home asking for Nvidia GPU support? The first efforts to try it showed that the minirosetta algorithms are too memory-demanding to make much use of enough GPU cores to get very much speedup on Nvidia boards with a currently reasonable amount of video memory. Therefore, such a conversion would require so much of a rewrite that it currently doesn't look reasonable.

This message is 14ys old, but it's true!!
Now we have a cuda/nvidia gpu app (or, better, we have a beta app)


Hope for an opencl/sycl/rocm/whatsoever app for other gpus (Amd, Intel, VideoCore, etc)!

New "SCALE" Software Allows Natively Compiling CUDA Apps For AMD GPUs

SCALE takes CUDA programs as-is and can even handle CUDA programs relying on line NVPTX Assembly. The SCALE compiler also is a drop-in replacement to NVIDIA's nvcc compiler and has a runtime that "impersonates" the NVIDIA CUDA Toolkit.

SCALE has been successfully tested with software like Blender, Llama-cpp, XGboost, FAISS, GOMC, STDGPU, Hashcat, and even NVIDIA Thrust. Spectral Compute has been testing SCALE across RDNA2 and RDNA3 GPUs along with basic testing on RDNA1 while Vega support is still a work-in-progress.

....

At its heart, SCALE is an nvcc-compatible compiler that can compile CUDA code for AMD GPUs, implementations of the CUDA runtime and driver APIs for AMD GPUs, and open-source wrapper libraries that in turn interface with AMD's ROCm libraries.

While ZLUDA for instance was quietly funded by AMD, I'm told by Spectral Compute that they've just been funding this development since 2017 via their consulting business. The only immediate downside seen with SCALE is that it itself is not open-source software but to at least there being a free edition license available for users.

---

Grant
Darwin NT

New "SCALE" Software Allows Natively Compiling CUDA Apps For AMD GPUs
[quote]SCALE takes CUDA programs as-is and can even handle CUDA programs relying on line NVPTX Assembly. The SCALE compiler also is a drop-in replacement to NVIDIA's nvcc compiler and has a runtime that "impersonates" the NVIDIA CUDA Toolkit.

There is A LOT of tools to convert automatically (or semi-automatically) CUDA code to OpenCL/OneApi/SYCL/etc.
From AdaptiveCpp (using C++ as common language between Cuda/OpenCl/Sycl) to SpectralCompute, from HIPIFY to Coriander
But there are 2 problems:
- Nvidia could block your work (as they did with ZLUDA). I know, it's not a good thing for science to block others competitor, but CUDA is closed, so Nvidia can do it.
- Developers have to "try and learn" new sw. And i don't know if they have the desire to....

There is A LOT of tools to convert automatically (or semi-automatically) CUDA code to OpenCL/OneApi/SYCL/etc.
From AdaptiveCpp (using C++ as common language between Cuda/OpenCl/Sycl) to SpectralCompute, from HIPIFY to Coriander

For example, AdaptiveCpp is faster than CUDA on Nvidia gpu!!!

There is A LOT of tools to convert automatically (or semi-automatically) CUDA code to OpenCL/OneApi/SYCL/etc.
From AdaptiveCpp (using C++ as common language between Cuda/OpenCl/Sycl) to SpectralCompute, from HIPIFY to Coriander

For example, AdaptiveCpp is faster than CUDA on Nvidia gpu!!
!

Where's your evidence that the Rosetta programs are written in a way that will allow such a conversion without using only graphics boards with extremely large numbers of GPU cores?

Where's your evidence that the Rosetta programs are written in a way that will allow such a conversion without using only graphics boards with extremely large numbers of GPU cores?

No evidence.
Mine is only a little demonstration that is possible to avoid CUDA, if necessary...