Complete.Org: Mailing Lists: Archives: freeciv-dev: May 2005:
[Freeciv-Dev] Re: (PR#12946) FACTOR in consider_settler_action
Home

[Freeciv-Dev] Re: (PR#12946) FACTOR in consider_settler_action

[Top] [All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index] [Thread Index]
To: bdunstan149@xxxxxxxxx
Subject: [Freeciv-Dev] Re: (PR#12946) FACTOR in consider_settler_action
From: "Benoit Hudson" <bh@xxxxxxxxxxxxxxxxxxx>
Date: Sun, 1 May 2005 10:56:33 -0700
Reply-to: bugs@xxxxxxxxxxx

<URL: http://bugs.freeciv.org/Ticket/Display.html?id=12946 >

On Sun, May 01, 2005 at 04:10:42AM -0700, Benedict Adamson wrote:
> Those multipliers are used so we can use integer arithmetic, which in 
> turn is done because integer arithmetic is faster than floating-point 
> arithmetic (right?). But it introduces bugs like this (I have a vague 
> recollection of similar bugs previously). But is this 'optimisation' 
> really worthwhile? Is the performance gain large enough to justify the 
> decreased reliability? I would not be surprised if the gain were small.

There's more than one reason to be wary of floating point, such as:
        1e-8 != (1.0 + 1e-8) - 1.0
Also, Intel boxes have different floating-point (80-bit inside the FPU,
64-bit in memory) than the rest of the world (64-bit everywhere), which
means that client and server may disagree on certain things if they run
on different platforms.

But generally I agree that floating point should be used more often.

        -- Benoît





[Prev in Thread] Current Thread [Next in Thread]