Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Fri, 22 Jun 2012 04:19:23 +0400
From: Solar Designer <>
Subject: Re: relbench and changed format names

On Fri, Jun 22, 2012 at 01:57:25AM +0200, magnum wrote:
> I think labeling the NT format (in n incarnations) as n different 
> formats will introduce a lot of user confusion.

Yes, but we already had two different names (and labels) for it in

> relbench could use the fastest found.
> But no, I don't have any specific suggestion (unless "relbench could use 
> the fastest found" is a suggestion. Maybe it is).

It is a specific suggestion, and a reasonable one.  However, it would be
bad for some use cases of relbench - e.g., when trying to see the effect
of compiler options.

> I just really dislike 
> the idea of putting implementation-specific stuff in the format label. 
> They do not belong there.

But if we put these implementation IDs in the ALGORITHM_NAME field, that
mixes them with build-specific detail that relbench should not (and does
not) consider.  So we'd need yet another field for the implementation ID,
or actually it may just as well be a string formatting convention.

My " (label X)" suggestion is precisely such a convention.

I have no better idea right now.  I think we should just do that, with
the label in FORMAT_NAME where needed, formatted as suggested above.


Powered by blists - more mailing lists

Confused about mailing lists and their use? Read about mailing lists on Wikipedia and check out these guidelines on proper formatting of your messages.