parallelization extremely slow on i7

Good Morning/Afernoon/Evening depending where you are ;)

I am just checking out the parallelization features of ML (R2012b) and I don't really see any advantages so far, though I am only executing sample code to be honest.

I have 4 cores on an Intel Core i7 and I use 4 workers to run the simple magic-square sample code. A run takes approximately 40 seconds while simply looping it without parallel code takes about 0.24s .

Can someone briefly comment on the matlabpool features and traps, most likely I am just in a regime of work where parallelization is useless.

code 1

matlabpool(4)
spmd
  q = magic(labindex + 2);
end
for ii=1:length(q)
  figure, imagesc(q{ii});
end
matlabpool close

code 2

for i=1:1:4
    q=magic(i+2);
    figure, imagesc(q);    
end

Answer 1

[Matt J]

Where is this "sample" to be found? The computation looks way too small for parallelization to be fruitful. Try something like,

q = magic(labindex+2000);