Hi Hui,
We have a fork of Chapel available on the ROCm github. You do need ROCm and an
AMD APU to run these today. The code is available here:
https://github.com/ROCm-Developer-Tools/chapel/tree/chpl-hsa-master
Code examples which run on the GPU are here:
https://github.com/ROCm-Developer-Tools/chapel/tree/chpl-hsa-master/test/release/examples/hsa
Right now, our code is working on the Chapel-to-C path, so it does not work on
the LLVM path. We’re working on merging our changes back to master.
Mike.
From: Hui Zhang [mailto:[email protected]]
Sent: Tuesday, September 5, 2017 5:09 AM
To: Chapel Sourceforge Developers List <[email protected]>
Subject: [Chapel-developers] Chapel support on GPU
Hello,
I need some information and suggestions about the Chapel's support on GPU. (I
had zero knowledge of GPU)
After doing some research, I only found two separate works related to this
topic:
1. CUDA: performance portability with the Chapel language
http://ieeexplore.ieee.org/document/6267860/?arnumber=6267860&tag=1
2. OpenCL: GPGPU Support in Chapel with the Radeon Open Compute Platform
For the 1st work, I don't find the code in the trunk, how can I play with
that? How does the compilation process actually work? (CUDA ith nvcc and C with
gcc, then how to combine both and generate single binary)
For 2nd, I don't find any detail except the limited information on the slides.
Is that open source and available on Chapel what version? And what I need to
play with it on my own machine? or it has to be on ROCm?
Besides, is there any way for either of the above two to generate LLVM IR for
the GPU-aware Chapel code? (including the host and device code)
Thanks
--
Best regards
Hui Zhang
------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
_______________________________________________
Chapel-developers mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/chapel-developers
