Re: [Algorithms] Countering FP Voodoo

[Paul at H. <pa...@ru...>]

> What, you mean actually RTFM? Nah, it's never worked before!
Exactly!

> pulls the "good compilers shall not reorder, refactor or transpose FP
I agree that assemblers should do that, but a high-level language compiler 
should do all it can do imo. There's no correlation between input and output 
to maintain.

I actually think the problem here is FP itself tbh, not that it's going to 
go away. (I have an inherent mistrust of anything that can't represent a 
tenth properly but looks like it should). At least with fixed point you know 
where you stand with addition, multiplications, signs, zeroes et al.

In this digital world, it often feels like the only analog thing left around 
is the, er, umm... computer!


Regards,
Paul Johnson.

Managing Director,
Rubicon Mobile, Ltd.
www.rubiconmobile.com

----- Original Message ----- 
From: <rob...@pl...>
To: "Game Development Algorithms" <gda...@li...>
Sent: Monday, April 16, 2007 9:57 PM
Subject: Re: [Algorithms] Countering FP Voodoo


>
>
> Paul at Home wrote:
>>
>> Which I then trump with the "A custom compiler could've sorted that one 
>> out
>> for us" card.
>
> :pulls the "good compilers shall not reorder, refactor or transpose FP
> code and shall have clear rules for constant expressions folding" card
> from his hand and places it on top of the pile.
>
>
>> there's no way in the world that on seeing
>> result = value * 1.0f coming out wrong, I would think to try result = 
>> 1.0f *
>> value instead of filling up my code up with the word epsilon.
>
> What, you mean actually RTFM? Nah, it's never worked before!
>
> Sarcastic comments aside, arguing about FP code issues using examples
> expressed in C/C++ really only highlights C/C++'s crappy support for FP
> IMO, not any weakness of 754 or a particular hardware platform.
>
> - Robin Green.
>
>
> -------------------------------------------------------------------------
> This SF.net email is sponsored by DB2 Express
> Download DB2 Express C - the FREE version of DB2 express and take
> control of your XML. No limits. Just data. Click to get it now.
> http://sourceforge.net/powerbar/db2/
> _______________________________________________
> GDAlgorithms-list mailing list
> GDA...@li...
> https://lists.sourceforge.net/lists/listinfo/gdalgorithms-list
> Archives:
> http://sourceforge.net/mailarchive/forum.php?forum_name=gdalgorithms-list
>
>
>
>