Hello,
I would like to parallel my code but I am facing some issues. So far I have
just used the decorator @parallel. Some parts of my code needs to access at
some sets/lists stored in the memory and the refresh them.
*Q1-* Is there any method that let me to share memory between process?
Other blocks of my code do not need to access to the memory so I manage to
parallelised them but I have some timing issues: now the code is slower!
I looked at the timing of each step (writing the list of inputs, the
parallelised block, access to the outputs) and the last one (access to the
outputs) consists roughly in the 99% of the total time. The function which
I am referring to is the following
def parallelcount(vector,paralleloutput):
for i,j in paralleloutput:
vector=[x + y for x, y in zip(vector, j)]
return vector
*Q2- *Is there any faster/smart way to access at the paralleloutput?
I got that the output of a parallel process is a generator so, since I am
using it inside a quite big loop, it could be not the best efficient way to
speed up the code. My feeling is that at each loop's step sage reads my
parallel block, builds a generator from it but it doesn't compute anything,
once the generator is called by "parallelcount" it starts to compute
everything (this should explain why all the time is used just for accessing
to the outputs). If it works in this way building the generators should be
the main reason of the slowness, indeed my loop has roughly 10^5 steps but
the parallel count receives just 10 inputs, so probably is not worth
building 10^5 generators.
*Q3- *Am I right? If not why is it slower then the not parallelised code?
If yes may force sage to do not build a generator each time but just
compute the outputs (since they are just 10)?
A positive answer to Q1 may let me avoid to build so many generators
(indeed I may parallelise the above loop instead of the inner one).
Cheers,
Marco
--
You received this message because you are subscribed to the Google Groups
"sage-support" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
To post to this group, send email to [email protected].
Visit this group at https://groups.google.com/group/sage-support.
For more options, visit https://groups.google.com/d/optout.
