[m-rev.] for review: improve hash tables

[Peter Wang]

On 2010-07-19, Julien Fischer <juliensf at csse.unimelb.edu.au> wrote:
> >>
> >>>Branches: main
> >>>
> >>>Speed up hash tables using a "fat" list representation for bucket chains,
> >>>instead of assoc_list.  The benefits are less type_info constructions (an
> >>>assoc_list requires a type_info for list and a type_info for pair) and,
> >>>likely, better locality.  This change speeds up hash_table_test.m by 25%,
> >>>on an x86-64 machine.
> >>
> >>How did you test for that?
> >
> >Test for what?
> 
> Test that the speedup was 25%.  (I guess I actually mean how did you measure that?)

I installed the library with and without the change, and for each,
compiled tests/hard_coded/hash_table_test.m and ran "time ./hash_table_test"
a few times.  This was in asm_fast.gc at -O5 with transitive intermodule
optimisation.

Peter
--------------------------------------------------------------------------
mercury-reviews mailing list
Post messages to:       mercury-reviews at csse.unimelb.edu.au
Administrative Queries: owner-mercury-reviews at csse.unimelb.edu.au
Subscriptions:          mercury-reviews-request at csse.unimelb.edu.au
--------------------------------------------------------------------------