[Scilab-users] parallel_run on linux with 8 cores problem with nb_workers

[lukasz_kocot]

Hi all,

I need help. I was thinging that I understand this, but still not. It not
work as i suppose it should.

Why following example only 3 times call function a(),

-->lines(0);
-->function a=g(arg1)
-->disp("called on "+string(arg1));
-->//for i=1:1000000,
-->  a=arg1*arg1;
-->// end;
-->disp("computed "+string(a));
-->endfunction
-->res=parallel_run(1:24, g, init_param('nb_workers', 8));
 
 called on 1   
 
 computed 1   
 
 called on 2   
 
 computed 4   
 
 called on 3   
 
 computed 9   



...... and following example call it all 24 times:
-->lines(0);
-->function a=g(arg1)
-->disp("called on "+string(arg1));
-->//for i=1:1000000,
-->  a=arg1*arg1;
-->// end;
-->disp("computed "+string(a));
-->endfunction
-->res=parallel_run(1:24, g, init_param('nb_workers', 1));
 
 called on 1   
 
 computed 1   
 
 called on 2   
 
 computed 4   
 
 called on 3   
 
 computed 9   
 
 called on 4   
 
 computed 16   
 
 called on 5   
 
 computed 25   
 
 called on 6   
 
 computed 36   
 
 called on 7   
 
 computed 49   
 
 called on 8   
 
 computed 64   
 
 called on 9   
 
 computed 81   
 
 called on 10   
 
 computed 100   
 
 called on 11   
 
 computed 121   
 
 called on 12   
 
 computed 144   
 
 called on 13   
 
 computed 169   
 
 called on 14   
 
 computed 196   
 
 called on 15   
 
 computed 225   
 
 called on 16   
 
 computed 256   
 
 called on 17   
 
 computed 289   
 
 called on 18   
 
 computed 324   
 
 called on 19   
 
 computed 361   
 
 called on 20   
 
 computed 400   
 
 called on 21   
 
 computed 441   
 
 called on 22   
 
 computed 484   
 
 called on 23   
 
 computed 529   
 
 called on 24   
 
 computed 576  




Please help me understund it and implement it in my projekt.
I have 8 core processor.
 




--
View this message in context: http://mailinglists.scilab.org/parallel-run-on-linux-with-8-cores-problem-with-nb-workers-tp4025610p4025725.html
Sent from the Scilab users - Mailing Lists Archives mailing list archive at Nabble.com.