[IPOL discuss] [IPOL tech] Démo de la méthode K_SVD

[Miguel Colom]

Quoting Nicolas Limare <nicolas.limare at cmla.ens-cachan.fr>:
> One cause for large memory footprints can be the parallel code
> sections, if a large array is declared as "private", ie replicated for
> every execution thread (-> x32 on the demo server!). If this data is
> only read or if it can be written without thread conflicts, it should
> be declared as "public", and not replicated.

Yes, asking for memory in parallel code is likely to create an  
explosion on the demand of resources.

Normally we only allocate the needed memory outside the parallel code  
and share it publicly along all the threads.

Even though, and despite all the optimizations we put to save memory  
(sometimes at the expense of increasing the execution time), denoising  
is quite greedy in terms of memory and CPU resources.