[m-rev.] for review: minor parser optimisation

[Julien Fischer]

On Tue, 27 May 2008, Peter Wang wrote:

> Branches: main
>
> A small optimisation for the parser.  The speedup on tools/speedtest is about
> 0.78%
>
> library/parser.m:
> 	Add a mode declaration for `parser_get_token' that takes the expected
> 	token as input, so that the expected and actual tokens can be unified
> 	within the procedure.  If they don't match, it avoids creating a new
> 	parser state that would be discarded by the caller.  The call in
> 	`check_for_higher_order_term' is very likely to fail.
>
>
> PS. The problem I'm trying to solve is the reading of .analysis files.  When
> making an .analysis file, reading in other analysis files can take nearly half
> of the running time, mainly due to large structure sharing/reuse results.
> Suggestions would be appreciated.
>
>
> Index: library/parser.m
> ===================================================================
> RCS file: /home/mercury/mercury1/repository/mercury/library/parser.m,v
> retrieving revision 1.59
> diff -u -r1.59 parser.m
> --- library/parser.m	8 May 2008 00:57:31 -0000	1.59
> +++ library/parser.m	27 May 2008 04:08:01 -0000
> @@ -1018,14 +1018,17 @@
>
> %-----------------------------------------------------------------------------%
>
> -:- pred parser_get_token(token::out,
> -    state(Ops, T)::in, state(Ops, T)::out) is semidet.
> +:- pred parser_get_token(token, state(Ops, T), state(Ops, T)).
> +:- mode parser_get_token(in, in, out) is semidet.
> +:- mode parser_get_token(out, in, out) is semidet.
>
> parser_get_token(Token, !PS) :-
>     parser_get_token_context(Token, _Context, !PS).
>
> -:- pred parser_get_token_context(token::out, token_context::out,
> -    state(Ops, T)::in, state(Ops, T)::out) is semidet.
> +:- pred parser_get_token_context(token, token_context,
> +    state(Ops, T), state(Ops, T)).
> +:- mode parser_get_token_context(in, out, in, out) is semidet.
> +:- mode parser_get_token_context(out, out, in, out) is semidet.
>
> parser_get_token_context(Token, Context, ParserState0, ParserState) :-
>     Tokens0 = parser_state_get_tokens_left(ParserState0),

I suggest that you add a comment stating that the implied mode of these
predicates is intentional (and why, i.e. what the log message says).

Julien.
--------------------------------------------------------------------------
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
--------------------------------------------------------------------------