[Octave-bug-tracker] [bug #52342] vecnorm command

[Rik]

Follow-up Comment #7, bug #52342 (project octave):

@Marco: Benchmarking is pretty hard to get right.  First, one generally needs
to repeat the test several times.  Second, The way your test was structured on
one line means that the CPU's dynamic frequency scaling was probably not fully
engaged for the first calculation using abs(), but was for the second part of
the test.  That gives an unfair advantage to the vectorized code.

I wrote a small benchmark script (bm_abs.m) which is attached.

 

N = 10;

bm = zeros (N,1);
bm2 = zeros (N,1);

for i = 1:N
  tic;
  (abs(A)).^p;
  bm(i) = toc;
  tic;
  (real(A).^2+imag(A).^2).^(p/2);
  bm2(i) = toc;
endfor

printf ("abs: %f\n", median (bm));
printf ("vec: %f\n", median (bm2));


The results for various p values are shown below:


p = 1

abs: 0.925663
vec: 2.668234

p = 2

abs: 1.033491
vec: 1.123011

p = 3

abs: 0.700889
vec: 2.656083

p = 4

abs: 2.472564
vec: 0.884022

p = 5

abs: 2.525005
vec: 2.689880

p = 6

abs: 2.508996
vec: 0.885091

p = 7

abs: 2.517311
vec: 2.630967

p = 8

abs: 2.498222
vec: 2.637875


Definitely some weirdness associated with p = [4,6].  I don't think those are
common choices, so it may be best to stay with abs(). 



    _______________________________________________________

Reply to this item at:

  <http://savannah.gnu.org/bugs/?52342>

_______________________________________________
  Message sent via/by Savannah
  http://savannah.gnu.org/