> I noticed that mir cut generation takes considerable time on some > large problems (like rocII-4-11 from miplib). The attached patch makes > two improvements that considerably improve performance in such > instances: > 1. A lot of time was spent on generating a temporary vector in > function aggregate_row. It is a lot faster to reuse an existing > vector. > 2. A search for an element in the same function was done in row order, > where using the elements in the order they are in the column is more > efficient. This changes the generated cuts in some cases, but seems > neutral overall (0.3% less cuts in a test set of 64 miplib instances). >
Patched. Please see the attachment. Andrew Makhorin
glpios06.c.gz
Description: GNU Zip compressed data
_______________________________________________ Help-glpk mailing list [email protected] https://lists.gnu.org/mailman/listinfo/help-glpk
