File: luflop.m

package info (click to toggle)
freemat 4.0-5
  • links: PTS, VCS
  • area: main
  • in suites: jessie, jessie-kfreebsd, wheezy
  • size: 174,736 kB
  • ctags: 67,053
  • sloc: cpp: 351,060; ansic: 255,892; sh: 40,590; makefile: 4,323; perl: 4,058; asm: 3,313; pascal: 2,718; fortran: 1,722; ada: 1,681; ml: 1,360; cs: 879; csh: 795; python: 430; sed: 162; lisp: 160; awk: 5
file content (26 lines) | stat: -rw-r--r-- 813 bytes parent folder | download | duplicates (8) 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
 
function f = luflop (L, U)						    %#ok
%LUFLOP given L and U, computes # of flops required to compute them
%
% Example:
% f = luflop (L, U)
%
% Given an LU factorization, compute how many flops took to compute it.  This
% is the same as (assuming U has a zero-free diagonal):
%
%   Lnz = full (sum (spones (L))) - 1 ;
%   Unz = full (sum (spones (U')))' - 1 ;
%   f = 2*Lnz*Unz + sum (Lnz) ;
%
% except that no extra workspace is allocated for spones (L) and spones (U).
% L and U must be sparse.
%
% Note: the above expression has a subtle undercount when exact numerical
% cancelation occurs.  Try [L,U,P] = lu (sparse (ones (10))) and then
% luflop (L,U).
%
% See also LU

% Copyright 1995-2007 by Timothy A. Davis.

help luflop
error ('luflop mexFunction not found!  Use umfpack_make to compile luflop.') ;