Re: [AL] ANN: FBlend - Fast color blenders

[<xav...@ho...>]

I was wondering, why don't you just make add those optimizations "inside"
allegro? ;)

----- Original Message -----
From: "Bob" <oha...@vi...>
To: <al...@ca...>
Cc: <ag...@ca...>
Sent: Tuesday, October 02, 2001 12:37 AM
Subject: [AL] ANN: FBlend - Fast color blenders


>
> FBlend v0.1 has been released!
>
> What is FBlend?
> ---------------
> Do you think the Allegro blenders are too slow? Do you want to do special
> effects in 16-bpp but can't? Then FBlend is for you! FBlend is
> a series of special routines that do color blending -fast-.
>
> FBlend is also functionaly equivalent to Allegro's blenders, so you can
just
> make slight changes in your code to benefit from the extra speed!
>
>
> Mini benchmarks(*):
>
> The SSE color add code is 7.63x faster then Allegro's!
> The MMX color add code is 5.01x faster than Allegro's.
> The  C  color add code is 2.5x faster than Allegro's.
>
> The SSE color trans code is 3.85x faster than Allegro's.
> The MMX color trans code is 2.74x faster than Allegro's.
> The  C  color trans code is 1.97x faster than Allegro's.
>
>
> * These numbers come from my computer and are dependent on my hardware
>    and software setup. Your Mileage May Vary.
>
>
> Where do I get it?
> ------------------
>
> Web Page: http://sf.net/projects/fblend/
>
> Direct links to download:
> http://prdownloads.sourceforge.net/fblend/fblend-0.1.zip
> http://prdownloads.sourceforge.net/fblend/fblend_docs-0.1.zip
>
>
> Supported platforms:
> --------------------
>
> FBlend comes with makefiles for MSVC and Mingw.
> It requires the same setup as Allegro.
>
> Makefiles for DJGPP, Linux, BeOS, etc are welcomed!
>
>
>
> --
> - Robert J Ohannessian
> "Microsoft code is probably O(n^20)" (my CS prof)
> http://pages.infinit.net/voidstar/
>
>
>From <all...@ca...> Mon Oct 01 21:40:12 2001
Received: from VL-MS-MR003.sc1.videotron.ca [24.201.245.36] by canvaslink.com with ESMTP
  (SMTPD32-6.00) id A41219AA03E4; Mon, 01 Oct 2001 21:10:42 -0400
Received: from videotron.ca ([24.202.202.88]) by
          VL-MS-MR003.sc1.videotron.ca (Netscape Messaging Server 4.15
          MR003 Jul 24 2001 16:23:26) with ESMTP id GKK0EB03.6OT for
          <al...@ca...>; Mon, 1 Oct 2001 21:06:11 -0400 
Message-ID: <3BB...@vi...>
Date: Mon, 01 Oct 2001 21:06:24 -0400
From: Bob <oha...@vi...>
User-Agent: Mozilla/5.0 (Windows; U; Windows NT 5.0; en-US; rv:0.9.4) Gecko/20010913
X-Accept-Language: en-us
MIME-Version: 1.0
To: al...@ca...
References: <3BB...@vi...> <OE3...@ho...>
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: 8bit
Subject: Re: [AL] ANN: FBlend - Fast color blenders
Precedence: bulk
Sender: all...@ca...
Reply-To: al...@ca...
X-UIDL: 304633504
Status: O
Content-Length: 915
Lines: 26

Javier González wrote:

> I was wondering, why don't you just make add those optimizations "inside"
> allegro? ;)



I tried to, but there's a problem: The Allegro blenders are all pixel based 
(they take on 2 values, mix them, and return the result). My blenders are 
line based (rect based even), so including them in Allegro would mean that 
all the primitive functions need a rewrite to accomodate the change. Not 
only that, but it'll need support of every blender, in every possible mode, 
for every primitive type, with blenders being either coded in C or assembly, 
and the caller in either assembler or C.

This makes quite a mess.

Perhaps for Allegro 4.1 when/if I get time to do all of the above, but in 
the mean time, I'll just be making a couple of blenders and perhaps other 
effects.


-- 
- Robert J Ohannessian
"Microsoft code is probably O(n^20)" (my CS prof)
http://pages.infinit.net/voidstar/