Newsgroups: comp.os.msdos.djgpp
From: tob AT world DOT std DOT com
Subject: Re: Allegro & sprite stretching optimization
Message-ID: <EB7roE.Ko7@world.std.com>
Sender: tob AT world DOT std DOT com (Tom Breton)
Reply-To: tob AT world DOT std DOT com
Organization: BREnterprises
References: <nWvqRIAb1zkzEwAi AT talula DOT demon DOT co DOT uk>
Date: Tue, 3 Jun 1997 19:05:02 GMT
Lines: 41
To: djgpp AT delorie DOT com
DJ-Gateway: from newsgroup comp.os.msdos.djgpp
Precedence: bulk

Shawn Hargreaves <Shawn AT talula DOT demon DOT co DOT uk> writes:
> That would work, but I'm very wary of an API that exposes the internal
> workings of the implementation like that. Designing an interface that is
> dependent on this kind of implementation detail could cause no end of
> problems in the long run, and would restrict the ways in which the
> routine could be developed in the future. It makes me nervous :-)

It's a point. Though as I will elaborate on, the caching strategy still
appears to spin its wheels uncomfortably long even for little sprites.

> It can be done even more simply than that, and there's no need for the
> copy! At the start of the function, push the values of _scratch_mem and
> _scratch_mem_size into some local variables, and reset _scratch_mem to

Thanks, it was easier that way.

I wrote it in C++ (but not tested it yet) the way you suggested, and I
discovered some things:

    There are enough free variables for every stretcher that it still
    will probably take a while to search the cache. I introduced a hash
    for trivial rejections to speed it up, but still it wants to spend a
    while searching. The cache searching could be sped up further with
    heroic measures like a hash table proper instead of hash-rejection,
    but I don't think it's worth it when the user could just handle it.

    The functionality-split that I talked of falls naturally out of
    doing it the cache way. A little tweaking of the interface and it's
    there.

    It wasn't an issue when compilations and uses were 1-to-1, but on
    separating them there are a couple of issues WRT how precise a match
    a user will demand. For example, I can easily imagine that a user,
    stretching a sprite, would not care whether the stretch began on the
    correct fractional pixel. In other cases, I can imagine a user might
    care very much. I coded the cache-test in the most restrictive way,
    but it's another reason why the user might want to handle it
    themselves.

        Tom