Re: namd-cuda-intel vs. namd-intel

From: Nicholas M Glykos (glykos_at_mbg.duth.gr)
Date: Wed May 13 2009 - 13:21:23 CDT

Dear All,

I do not mean to hijack the thread, but still, I thought I should
speak-up, so here it goes: I have been reading a bit about cuda, and even
started a toy-project of mine as an excuse for gaining some first-hand
experience as to how it looks like writing C for a GPU. The feeling I got
was a very strong deja-vu of writing assembly for the Z80 (old guy): with
all memory details exposed, and with very strigent requirements for
efficient memory accesses, the real question is not why the speedup is
only x4 or x5, but how on earth the NAMD developers managed to get any
speedup with such a complex program and in such a short time of developing
the cuda code. From my point-of-view, hats off to the programmers.

My twopence,
Nicholas

ps. Maybe the 'zero-copy accesses' in 2.2 will improve 'speed-ups' ?

-- 
          Dr Nicholas M. Glykos, Department of Molecular Biology
     and Genetics, Democritus University of Thrace, University Campus,
  Dragana, 68100 Alexandroupolis, Greece, Tel/Fax (office) +302551030620,
    Ext.77620, Tel (lab) +302551030615, http://utopia.duth.gr/~glykos/

This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:52:47 CST