[Scilab-users] [EXTERNAL] Re: issue with parallel_run
Carrico, Paul
paul.carrico at esterline.com
Tue Jun 14 10:48:44 CEST 2016
Thanks Stephane for the help.
It works now
NB: I changed n to 1000 and num to 1E6 (i.e. close to my need) and the speed up is nearly 7 (=7 time faster) :)
Paul
EXPORT CONTROL :
Cet email ne contient pas de données techniques
This email does not contain technical data
De : users [mailto:users-bounces at lists.scilab.org] De la part de Stéphane Mottelet
Envoyé : mardi 14 juin 2016 10:18
À : Users mailing list for Scilab
Objet : [EXTERNAL] Re: [Scilab-users] issue with parallel_run
Hello,
1-parallel_run is designed to cope with input-output data organized in columns
2-you have to precise the size of the output of yout fct() function.
Here is your corrected script (with comments) :
mode(0);
clear;
function random_matrix=fct(n)
M = grand(n,12,'unf',-1,1);
M = gsort(M,"lr","d");
random_matrix = M(1,:)'; // transpose the output
clear M;
endfunction
num = 10;
n = 1000;
// classical calculation
Result = zeros(12,num); // swap the dimensions
tic()
for i = 1 : num
Result(:,i) = fct(n);
end
disp(Result);
time1 = toc(); printf("time 1 = %g\n",time1);
// using parallel_run
Result2 = zeros(num,12);
vect_n = n*ones(1,num);
tic()
Result2 = parallel_run(vect_n, fct,[12,1]); // precise size of output
disp(Result2);
time2 = toc(); printf("time 2 = %g\n",time2);
S.
Le 14/06/2016 à 10:06, Carrico, Paul a écrit :
mode(0);
clear;
function random_matrix=fct(n)
M = grand(n,12,'unf',-1,1);
M = gsort(M,"lr","d");
random_matrix = M(1,:);
clear M;
endfunction
num = 10;
n = 1000;
// classical calculation
Result = zeros(num,12);
tic()
for i = 1 : num
Result(i,:) = fct(n);
end
disp(Result);
time1 = toc(); printf("time 1 = %g\n",time1);
// using parallel_run
Result2 = zeros(num,12);
vect_n = n*ones(1,num);
tic()
Result2 = parallel_run(vect_n, fct);
disp(Result2);
time2 = toc(); rintf("time 2 = %g\n",time2);
--
Département de Génie Informatique
EA 4297 Transformations Intégrées de la Matière Renouvelable
Université de Technologie de Compiègne - CS 60319
60203 Compiègne cedex
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.scilab.org/pipermail/users/attachments/20160614/f4e835b5/attachment.htm>
More information about the users
mailing list