Hi all ,
CUDA for multi-core CPUs was annonced for last July. I can’t find any news about it :(
Do you know when it’s going to be released ?
Thks,
Larry
Hi all ,
CUDA for multi-core CPUs was annonced for last July. I can’t find any news about it :(
Do you know when it’s going to be released ?
Thks,
Larry
I might be missing something, but what CUDA has to do with multi-core CPU?
I just read in this post that it will come in CUDA 2.1
[url=“http://forums.nvidia.com/index.php?showtopic=75886&pid=431151&mode=threaded&show=&st=&#entry431151”]http://forums.nvidia.com/index.php?showtop...t=&#entry431151[/url]
If you want your program to work both with and without a CUDA device, you would normally have to code your algorithm twice: once in CUDA, and once in C/C++/whatever. The CUDA programming model is restrictive enough that a compiler could directly generate multithreaded, SSE-enabled code in addition to GPU code. Then you only have to write the parallel parts of your application once.
http://www.gigascale.org/pubs/1278.html
Thats the research paper on MCUDA! – I think this is where everything might have started!
seibert,
thanks for explanation.