Mailing list for all users of the OCaml language and system.
 help / color / mirror / Atom feed
From: Pierre Weis <Pierre.Weis@inria.fr>
To: Vincent.Poirriez@univ-valenciennes.fr (Vincent Poirriez)
Cc: Pierre.Weis@inria.fr, zimmer@easynet.fr, caml-list@inria.fr
Subject: Re: Instruction return
Date: Tue, 3 Jun 1997 12:24:38 +0200 (MET DST)	[thread overview]
Message-ID: <199706031024.MAA29332@pauillac.inria.fr> (raw)
In-Reply-To: <3393D7D4.2F33@univ-valenciennes.fr> from Vincent Poirriez at "Jun 3, 97 09:37:40 am"

> Cependant, j'ai rencontré la situation suivante dans un contexte
> où l'efficacité (à l'exécution) est essentielle.
> 
> Trouver l'indice du premier élément d'un tableau qui vérifie
> un prédicat p, s'il y en a un. 
> 
> Ce code étant compile avec l'option -unsafe, je ne peux pas
> compter sur les test de débordement des primitives Array.get ...
> 
> Le choix d'utiliser Array.unsafe_get serait inutile si j'inclus
> ce test de débordement dans la condition 
> du while ou de la fonction récursive ainsi:
> 
> let find p a =
>   let l = Array.length a in
>   let i = ref 0 in
>   while !i<l & not( p (Array.unsafe_get a !i) do
>   incr i
>   done;
>   !i
> 
> J'ai donc prefere le coût d'un try ... with:
> 
> exception Exit_for of int
>  let find p a =
>  let l = Array.length a in
>   try
>    for i = 0 to l-1 do
>     if p (Array.unsafe_get a i) then raise Exit_for i
>    done;
>    l
>   with Exit_for i -> i
[...]

As-tu fait des tests qui prouvent la supe'riorite' d'une version sur
l'autre ?

Comment crois-tu que la boucle for fait le bon nombre de tours puis
s'arre^te ? (En faisant un test a` chaque tour, et pratiquement le
me^me que celui de ta boucle while!).

English short version:

> But, due to runtime efficiency considerations I had to make
> the following choice.
> 
>  I want to find the index of the first item of an array which
> verifies a predicat p, if one exists. 
> 
> It is useless to prefer Array.unsafe_get if I add the test
> in the while (or recursive) condition as below:
> 
> let find p a =
>   let l = Array.length a in
>   let i = ref 0 in
>   while !i<l & not( p (Array.unsafe_get a !i) do
>   incr i
>   done;
>   !i
> 
> 
> I prefer to pay the cost of one try ... with ...
> 
> 
> exception Exit_for of int
>  let find p a =
>  let l = Array.length a in
>   try
>    for i = 0 to l-1 do
>     if p (Array.unsafe_get a i) then raise Exit_for i
>    done;
>    l
>   with Exit_for i -> i

Have you got some runtime figures that prove that this last version is more
efficient than the others ?

By the way, the for loop implies also a test each time its body is
executed (exactly the same as in your while loop) to decide to
continue or stop the loop...

Pierre Weis

INRIA, Projet Cristal, Pierre.Weis@inria.fr, http://pauillac.inria.fr/~weis/







  reply	other threads:[~1997-06-03 10:25 UTC|newest]

Thread overview: 7+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
1997-05-31 17:30 Pascal Zimmer
1997-06-02 14:15 ` Pierre Weis
1997-06-03  8:37   ` Vincent Poirriez
1997-06-03 10:24     ` Pierre Weis [this message]
1997-06-03 14:17       ` Vincent Poirriez
1997-06-02 16:23 Dwight VandenBerghe
1997-06-02 21:41 ` Pierre Weis

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=199706031024.MAA29332@pauillac.inria.fr \
    --to=pierre.weis@inria.fr \
    --cc=Vincent.Poirriez@univ-valenciennes.fr \
    --cc=caml-list@inria.fr \
    --cc=zimmer@easynet.fr \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox