Dear list,

after analyzing test data, we had to update our calculations on the table
parameters.
Right now, after 270 million chains have been generated, for every
new unique chains produced, one merging chain is also produced.
It makes sense to stop at this point where the ratio is 50/50.
The new policy is to generate 270 million chains per table, which
will contain 190 million unique chains (=2^27.5).
Because of that 2^9.5 instead of 2^8.5 table must be generated and
the lookup complexity also doubles.
This table parameter is the recommended value. The effect of producing
a table with more chains is that it takes longer to generate it not
only because more chains are needed, but also because the ratio of
chains to merges is lower. The advantage of having a table with 2^28.5
unique chains is that it takes half the time to do a lookup with it
than with two tables of size 2^27.5
So in a sense we are trading lookup complexity for generation complexity
here. The lookup performance can also be improved later when all tables
are generated by appending new chains to the existing tables.
This will be an option if there is a need to perfect the tables and can
be decided then.
Instructions on the handling of finished tables will be posted when
time permits. Until then you can just let the data lay around.

_______________________________________________
A51 mailing list
[email protected]
http://lists.lists.reflextor.com/cgi-bin/mailman/listinfo/a51

Reply via email to