DrPop
11-20-11, 12:09 AM
Here is a link to the article: http://www.tomshardware.com/news/nvidia-cuda-graphics-card-hardware-acceleration,14033.html
Now, what got me wondering was this: "...In June of this year, Nvidia rolled (http://www.prnewswire.com/news-releases/the-portland-group-updates-compilers-to-deliver-nvidia-cuda-architecture-for-x86-platforms-123999879.html) out, with a delay of more than two years, multi-CPU x86 CUDA compilers that runs CUDA code on Intel and AMD processors."
Does that mean in the future a BOINC Project could only have to program ONE version of the code - a "CUDA" version, and then run it through a compiler that would allow it to crunch on multiple CPUs (AMD/INTEL)?
If so, and if it would be efficient enough, that is really cool from a programming perspective. I mean, then a project only needs one Dev team and they've got both GPU & CPU bases covered! :)
Now, what got me wondering was this: "...In June of this year, Nvidia rolled (http://www.prnewswire.com/news-releases/the-portland-group-updates-compilers-to-deliver-nvidia-cuda-architecture-for-x86-platforms-123999879.html) out, with a delay of more than two years, multi-CPU x86 CUDA compilers that runs CUDA code on Intel and AMD processors."
Does that mean in the future a BOINC Project could only have to program ONE version of the code - a "CUDA" version, and then run it through a compiler that would allow it to crunch on multiple CPUs (AMD/INTEL)?
If so, and if it would be efficient enough, that is really cool from a programming perspective. I mean, then a project only needs one Dev team and they've got both GPU & CPU bases covered! :)