Author: Dave Gomboc
Date: 15:01:49 10/04/01
Go up one level in this thread
On October 04, 2001 at 00:44:55, Uri Blass wrote: >On October 04, 2001 at 00:28:35, Dave Gomboc wrote: > >>On October 04, 2001 at 00:11:37, Uri Blass wrote: >> >>>On October 03, 2001 at 18:53:31, Dave Gomboc wrote: >>> >>>>Observing that the parallel algorithm was super-linearly quicker immediately >>>>gives rise to a superior sequential algorithm: alternate checking from the front >>>>of the array and the back of the array for the zero. It would quickly be >>>>observed that the parallel algorithm does not have super-linear speedup over >>>>this superior sequential algorithm. In addition, the sequential algorithm may >>>>be further optimized by ignoring the front of the array altogether, which would >>>>reduce the parallel speedup further. >>>> >>>>To ignore the superior sequential algorithm and instead publish results >>>>comparing only the original, poor sequential algorithm with the parallel >>>>algorithm is not merely disingenuous but simply bad science. >>>> >>>>Dave >>> >>>The question here is a practical question. >>> >>>It is possible that the original sequential algorithm that chess programs use is >>>a poor algorithm and this is the point of Bruce. >>> >>>Uri ><snipped> >>And that point was already superseded weeks ago. Both Bob and I have clearly >>stated earlier that comparing a parallel algorithm against a _bad_ sequential >>algorithm is useless. > >It is not useless because you may learn from it to improve the sequential >algorithm. > >It is also possible that the programmer only cares about using more than one >processor and in that case (s)he is not going to waste time about improving the >sequential search. > >Uri It is useless because you can learn from the parallel algorithm to improve the sequential algorithm and there is no need to waste time reporting results against a sequential algorithm that is known to be bad. Dave
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.