I think that we can not parallelize the outermost loop because we
noticed that most elements are dependant on the previous computations.
(a block is dependent to all block beetwen its lownode(bn) and its
row_value(mm) ) To bypass this issue we thought about using a
dependence tree. This tree would be created when we get the system and
updated each time the lownodes change. not making this dependence tree
time consuming. Please refer to the PDF attached to this mail
explaining it with some figures. Please keep in mind this document is
a draft for now..
The PDF link :
https://www.dropbox.com/s/ogmazfro3wpbajo/answer_gnucap_digest_88-1.pdf
--
Beranger Six,
CCamy système
06 33 16 10 17
_______________________________________________
Gnucap-devel mailing list
[email protected]
https://lists.gnu.org/mailman/listinfo/gnucap-devel
