Re: Ada bench

["David Sauvage" <pariakaman@gmail.com>]

Hi,
thanks for your advised comments,

"Matthew Heaney" <mhea...@on2.com> wrote
> How do you know from where the improvement came -- the change in
> data structure or the change in read mechanism?

please not i run the program just a few times to get an average value.
(System config : GNU/Debian Sid kernel 2.6.10, AMD 1.2 Ghz)
- Values are in millisecond, N=1
- for the GNAT.Spitbol.Table method, i just instantiate the generic
with 40_000 as discriminant (same value as the C spellcheck
implementation)
- for the AI302.Containers.Indefinite_Hashed_Maps method i used the
default AI302.Strings.Hash as hash function
- When using the Ada.Streams.Stream_IO method, i used a 4_096 buffer.

GNAT.Spitbol.Table + Ada.Streams.Stream_IO    (*1)              : 145
GNAT.Spitbol.Table + Ada.Text_IO                                : 175
AI302.Containers.Indefinite_Hashed_Maps + Ada.Streams.Stream_IO : 278
AI302.Containers.Indefinite_Hashed_Maps + Ada.Text_IO           : 310

"Matthew Heaney" <mhea...@on2.com> wrote
> Have you tried using the C++ hash function with an Ada program that
> uses Ada.Containers.Hashed_Maps?

nop, excellent i will try this soon.

(*1) this implementaton can be also find here
http://psycose.apinc.org/ada_shootout/

--
David