Hello,
First thanks for gnucap.
You told in a previous
topic:(https://lists.gnu.org/archive/html/help-gnucap/2012-12/msg00008.html)
"The code is designed with parallel in mind, but I have not had the time to
follow through.Two key places .. one is parallel model
evaluation, the other is to split the matrix into sections and
solve the sections in parallel.The split already happens."
I start a work to parellelize(with openmp, and then maybe cuda) the LU
decomposition in gnucap.
I see your topic about parallelism, and i have some time to do it.
Futhermore, we definitly need faster simulation result for our application.
What kind of solution did you have in mind:
-Is it the "section" you design with row, diagonal, and column? In this
case did you want to use , the fact that if all the section beetween
_lownode[mm] and mm are calculated we could computed the element.
In this case we could have a dependence graph(or tree) applied to your
storage matrix section, mostlyy used to parrallelize Gilbert-Peierls
Algorithm .
-Is it an iterative method,with the problem that the convergence could
take theorically an infinite number of operation.(so maybe not a good way)
-Is it parallelize only the map(multiplicaton beetwen element) of dot
product, and then maybe parallelize the reduction(addition beetween
elements).
-Is it a totally another way?
The other question i have :
-Did you had in mind to apply permutation matrix to ease implementation
of parrallelism, or directly doing the best matrix in evalution of netlist.
Regards,
Beranger Six
_______________________________________________
Gnucap-devel mailing list
[email protected]
https://lists.gnu.org/mailman/listinfo/gnucap-devel
