Print

Print


First, let me caveat that I'm a colleague and friend of the inventor of co-arrays - Bob Numrich.

However, just to point out the cloud in the silver-lining ...  I do have some fears that the elegance of co-arrays will breed bad software practices as developers start inserting communication willy-nilly into their code.  At least at the moment, I've partially convinced many developers of the value of encapsulating common code fragments within subroutines, but now perhaps they'll just put the halo-exchanges right in the loop that needs them ...   (Sounds good until you need to switch to a deeper halo throughout 1 million lines of code ...)

But even acknowledging such possible abuse, the merits of co-array are so strong that I heavily endorse whatever actions promote this to being standardized and more importantly widely implemented.

Cheers,

- Tom




On Aug 26, 2008, at 3:11 PM, Hedley, Bruce wrote:

Having had experience with both PVM and MPI, I can only say:

"Bring on the Co-arrays !" 

They will, as Van clearly outlines, save people time and headaches.

Bruce Hedley
Atomic Energy of Canada Ltd

Thomas Clune
Acting Head of the Software Integration and Visualization Office
NASA GSFC (610.3)     
301-286-4635 (W) 240-266-0400 (F)