[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [computer-go] Hardware-Instruction.

Don Dailey wrote:
I don't know of anyone who has ever compared a pure serial program to
a pure parallel program. Despite the claims, the only way this can
happen is if someone writes the best serial program they can without
giving ANY consideration to a parallel implementation, then converting
it in a 100% faithful way.
I don't know what "pure parallel" is supposed to mean, but I originally wrote my program for single CPU only and only did the conversion afterwards. This is extremely painful and bug-prone because the program was never designed with it in mind - I can understand that someone more experienced in the field would have avoided that route at any cost. At least valuable debugging & design lessons were learned, so it was still a gain, if only on a personal level.

For serious academic papers, it  is important to make comparisons with
apples to apples.  You want to understand the behavior of the parallel
algorithm and you don't want to start with completely different pieces
of code.  The papers you  read were quite honest in acknowledging this
fact and  pointing out that a  pure serial version might  be written a
little  bit differently.   On  the  other hand  Vincent  claims a  40X
slowdown, and you realize that he is exaggerating.  You are being kind
when  you say he  "exaggerates" in  this case  because we  are talking
about less than a 10% difference, not a 4000% difference.
Without supporting data or a way to reproduce the results such claims are useless. The paper I referred to was talking about a 60% improvement. Now it's only 10%? Which number are we supposed to believe?

Quite frankly I've stopped believing all the ones that can't be independently reproduced - it just too funny how the ones that can be reproduced are always so much worse.

--
GCP
_______________________________________________
computer-go mailing list
computer-go@xxxxxxxxxxxxxxxxx
http://www.computer-go.org/mailman/listinfo/computer-go/