https://github.com/eholk/harlan

Used for Astro ?

@+
*_*

---

https://github.com/eholk/harlan

Used for Astro ?

@+
*_*

Do you really think the project would rewrite their current apps because a new language came out?

From what I understand the project had some C or C++ code and modified it to work with BOINC.

They are having enough trouble getting an OpenCL app going. Kyong is probably having to learn OpenCL and somehow get the current app to work with it, in between running the project. OpenCL is supported on the 3 gpu types that BOINC currently knows of (ATI/AMD, Nvidia and Intel). Cuda only works on Nvidia.

[edit]it seems it sits on top of OpenCL according to the website linked[/edit]

---

BOINC blog

App is in C. But some parts of code must be changed prior to think about gpu with CUDA.

I've been searching against Harlan and while there are an appreciable number of hits there is neither the volume or type of hits that suggest to me Harlan is mature, stable or popular.

On the other hand, if I were coding a GPU app, transforming/porting/translating an existing serial app into a GPU app or whatever it is they're doing, and a new language/compiler came along and claimed to make that job easier I would probably have a look at it and evaluate. I believe it is possible (in some situations) to avoid 6 months additional work by abandoning 2 months of existing work and starting over with new/different/better tool(s).

Zarck posts the same messages on all BOINC projects fora. Not very constructive as he is well aware CUDA and OpenCL are used by projects that have support for GPUs. I don't think he understands the difficulty in writing a BOINC app for the CPU, let alone GPU apps. GPU apps are much more difficult, because of:
1. programming languages (CUDA/OpenCL)are less frequently used than C/C++
2. parallelism of GPUs
3. different GPU architectures and their optimizations
4. missing and/or broken libraries (AMDs FFT library for instance)
5. lack of support by AMD/nVidia

Whatever Zarck's intentions and "abilities", I don't think it hurts to spread the word and help potential users become aware of emerging tools they might find helpful in their quest.