Re: A92: C Compiler on the TI

[Prev][Next][Index][Thread]

Re: A92: C Compiler on the TI


> Here we are, in our 700MHz, 256 MB of RAM, 10 GB disk world, looking at the 
> TI-92 with its 10MHz processor and 70K of useful RAM (well, more for some 
> of us), and saying "A C compiler is impossible."  However, 70K is nearly 10 
> times as much as PDP-8s had, and they had not only C compilers but 
> UNIX.  (Well, an extremely early UNIX, anyway.)  And that's with a fraction 
> of our processing power.
> Very little is impossible on our calculator, at least in the way of 
> utilities.  (3D simulations will be slow and difficult, but utilities don't 
> require that kind of raw speed.)  It's mostly a matter of thinking from the 
> bottom up, building a compiler for the platform, exploiting its strengths 
> and avoiding its weaknesses, rather than trying to port or copy software 
> from another platform with entirely difficult capabilities.  Sure, we can't 
> port GNU C onto the calculator.  We can't port Visual Basic, either, but 
> that doesn't stop us from writing a BASIC interpreter like the 6K or 
> smaller one on my TRS-80.
> I know it's a cliche, but we need to think outside the box.  Or rather, 
> inside the box--the little plastic box with the keys on the front. :-)

Whoa there... I wasn't bashing all creativity and innovation!  Of course, we
wouldn't even have the programs we have now if it wasn't for David looking at
the calculator and seeing more.  If someone would like to try to write a C
compiler, go for it.  I was wrong saying that it was impossible.  I've never
tried it.

But outside that box is also a little place called reality.  You can't ignore
the limits of the our hardware.  Often the only way to get decent results is
to consider the limits and use those as building blocks for algorithms that
take advantage of the resources more efficiently.  I'm working on a couple of
projects that most would consider "impossible".  In fact, I gave up on one of
them a while back because I deemed it impossible (that and I figured the ti92
would no longer be useful with the 92+ and 89).  But as I see the 92 is still
used, I've decided to try the project again (after I finish my current one).

Both projects will most certainly push the limits of the 92, and I'm willing
to accept the results on both.  Even if that means a lousy couple of frames
per second.  Just to see them work is enough.


====
Aaron Hill (Redmond, Wash.)
Electronic Mail: serac@lightmail.com
IRC Nickname: serac (on EF-Net)
ActiveWorlds Citizenship: serac
References: