Account Options

  1. Sign in
The old Google Groups will be going away soon, but your browser is incompatible with the new version.
Google Groups Home
« Groups Home
comp.lang.prolog
+ new post
About this group
Subscribe to this group
This is a Usenet group - learn more
Message from discussion Multi-argument vs. JIT indexing (Was: Destructive assignments (Was: Win from call/n))
The group you are posting to is a Usenet group. Messages posted to this group will make your email address visible to anyone on the Internet.
Your reply message has not been sent.
Your post was successful
 
From:
To:
Cc:
Followup To:
Add Cc | Add Followup-to | Edit Subject
Subject:
Validation:
For verification purposes please type the characters you see in the picture below or the numbers you hear by clicking the accessibility icon. Listen and type the numbers you hear
 
Ulrich Neumerkel  
View profile  
 More options Oct 17 2012, 1:46 pm
Newsgroups: comp.lang.prolog
From: ulr...@mips.complang.tuwien.ac.at (Ulrich Neumerkel)
Date: Wed, 17 Oct 2012 17:46:30 GMT
Local: Wed, Oct 17 2012 1:46 pm
Subject: Re: Multi-argument vs. JIT indexing (Was: Destructive assignments (Was: Win from call/n))
Print | View thread | Show original | Report this message | Find messages by this author

Jan Wielemaker <j...@invalid.invalid> writes:
>On 2012-10-17, Ulrich Neumerkel <ulr...@mips.complang.tuwien.ac.at> wrote:
>> Jan Wielemaker <j...@invalid.invalid> writes:
>>>This is also more or less what SWI-Prolog does.  The only difference
>>>is that it doesn't do anything at compiletime.  Indexing opportunities
>>>are considered at the first call, creating the best single-argument
>>>index for that call.  ...

>> In SWI, indexing is better on the first argument (that is similar
>> to Eclipse) than on the other arguments.  In the first argument,
>> determinism is maintained even if the first clause contains
>> a variable.  On the other arguments this is not the case

>Yes.  Basically, there are linked lists ....

What I hope is that we find a good terminology for all of this
without referring to implementation details. Currently,
multi-argument indexing is a very vague term.

>> On the other hand, YAP is only determinate for
>> known values (thus, the Xs, but never for Ys - regardless
>> of the argument order.)

>> There are many more dimensions that this one, indeed.

>Sure, and there is a lot of room for parameter tuning, both at the
>level of individual predicates as at the global level.  E.g., if a
>predicate often leaves choice points that lead to failure when
>actually tried, it might be an option to create an extra index.  You
>can do hotspot analysis and more carefully examine the index for hot
>predicates ....

It is only relatively recently that toplevels show a bit the
open choicepoints - which has increased interest in this area.
There is also another motivation apart from hotspots:  The
ability to finalize earlier with setup_call_cleanup/3.

 
You must Sign in before you can post messages.
To post a message you must first join this group.
Please update your nickname on the subscription settings page before posting.
You do not have the permission required to post.
Create a group - Google Groups - Google Home - Terms of Service - Privacy Policy
©2013 Google