Return Styles: Pseud0ch, Terminal, Valhalla, NES, Geocities, Blue Moon.

Pages: 1-

An O(n^2) matrix multiplication algorithm

Name: Anonymous 2009-03-21 17:01

Will one ever be discovered?
Will it be practical?

Name: Anonymous 2009-03-21 17:29

Will one ever be discovered?
yes.

Will it be practical?
only for extremely large matrices.

Name: Anonymous 2009-03-21 17:44

No.
No.

Name: Anonymous 2009-03-21 17:47

Forget it, it's NP-complete

Name: Anonymous 2009-03-21 17:51

Forget it, we have multi-threading anyways.

Name: Anonymous 2009-03-21 18:22

Matrix multiplication for matrices equal to or less than 4x4 in size are effectively O(n^2) on platforms with floating-point SIMD/vector instructions, essentially a form of parallel computing.

>> 5 has it right. Later next year, Intel will be releasing their 32-core x86 gpu/computation platform, larrabee.

Name: Anonymous 2009-03-21 18:53

>>6
floating point numbers have extremely limited range.

Name: Anonymous 2009-03-21 20:14

>>7
Relative to what?

Name: Anonymous 2009-03-21 20:50

>>6 is stupid I bet you think if you unroll n^3 loops of a normal matrices multiplication algorithm you will get O(1) algorithm lol

Name: Anonymous 2009-03-21 21:32

>>9
It's not unrolled, the inner third loop is executed in parallel, nigger.

Name: Anonymous 2009-03-21 21:42

>>9
Yeah. Nigger.

Name: Anonymous 2009-03-21 21:52

>>10-11
yHBT

Name: Anonymous 2009-03-21 21:54

>>12
You have been niggered.

Name: Anonymous 2009-03-21 22:37

>>8
arbitrary precision integers.
or this: http://darcs.augustsson.net/Darcs/CReal/CRealI.hs

Name: Anonymous 2010-12-08 22:55

Name: Anonymous 2011-02-04 12:10

Name: Anonymous 2011-02-04 17:46

Don't change these.
Name: Email:
Entire Thread Thread List