On Thursday 15 May 2008, Bill Hart wrote: > Here is the graph of Magma times: > > http://sage.math.washington.edu/home/wbhart/flint-trunk/graphing/gf2.png > > The crossover is not clear. The change from a smooth curve to a > squiggly line is about 3600. So presumably that is it, but the graph > also seems to change character at about 6200 or 7000 as well. One of > these changes may be cache related. > At 3600, the total data for all three matrices is almost 5mb and the > cache on my machine is 1024kb. But if Magma is using classical > multiplication, then this is pretty much irrelevant anyway, since you > can keep the working data within the cache for quite a while during > the algorithm.
On the other hand: a squiggly line is what one one would expect for Strassen-Winograd due to the extra rows/columns that have to be taken care of. In any case: 3600 seems rather late for 1MB L2 cache. I think Allan Steel once gave a talk about his implementation and stated that they don't use classical block multiplication (I saw some slides with that remark). Martin -- name: Martin Albrecht _pgp: http://pgp.mit.edu:11371/pks/lookup?op=get&search=0x8EF0DC99 _www: http://www.informatik.uni-bremen.de/~malb _jab: [EMAIL PROTECTED] --~--~---------~--~----~------------~-------~--~----~ To post to this group, send email to sage-devel@googlegroups.com To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/sage-devel URLs: http://www.sagemath.org -~----------~----~----~----~------~----~------~--~---
