Porting to CUDA was done

[Yung-Yu Chen]

Dear List,

I am pleased to announce that a CUDA-enabled, dual-mode solver hierarchy has been developed and organized into SOLVCON in the name space of solvcon.kerpak.  I have made a note on the porting: http://solvcon.net/yyc/writing/2011/cuda_port.html .  Although there're still things to be done before a meaningful benchmark, for double-precision, the non-memory-optimized CUDA code on Tesla M2050 is about 10 times faster than one core of Xeon 5506.  It doesn't sound too bad for a two-/three-dimensional, mixed-shape CESE code.

The development will be continued immediately after this milestone.

with regards,

Yung-Yu Chen

--
Yung-Yu Chen
PhD candidate of Mechanical Engineering
The Ohio State University, Columbus, Ohio
+1 (614) 859 2436

http://solvcon.net/yyc/

[John Yu]

Yung,
Thank you for your effort. It is a great milestone.
JY

--
You received this message because you are subscribed to the Google Groups "solvcon" group.
To post to this group, send email to sol...@googlegroups.com.
To unsubscribe from this group, send email to solvcon+u...@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/solvcon?hl=en.

--
Sheng-Tao John Yu, Ph.D.
Associate Professor of Mechanical Engineering
The Ohio State University
Room 505, Scott Lab
Columbus, OH 43210
Tel: 614 292 2902
http://cfd.eng.ohio-state.edu/