Re: A89: Z80 to 68k??
[Prev][Next][Index][Thread]
Re: A89: Z80 to 68k??
Forgive me, I had no intention of bashing your program. No insult
intended. I was not aware that it was feasible to translate from one
asm to another with not only a different instruction set but an entirely
different architecture; at least not without some insanely complex
program. By monstrosity I meant complexity and coding time, not
bloatware or anything like that.
Cassady Roop
> As this author of this monstrous hoax, I guess I might as well say a
> little bit about it:
>
> The program (Source Guru) is *not* a hoax at all. It is availabe (with
> source code, no less) not only on my web page, but also in the ticalc.org
> archives. It was released several months ago, and was mentioned in a
> ticalc.org news item. The complete source code (in C) is included. It is
> supplied only as a DOS executable but should be compilable on other
> systems.
>
> As far as efficiency, I would hope that you know the 68K calculators have
> faster processors than the Z80 calculators, so it isn't necessary. The
> program is converted using a straightforward algorithm, with only a few
> simple peephole optimzations; the result is code that's about 50% as
> efficient as native 68K code, which just happens to be nicely offset by
> the fact that the 68K is about twice as fast. Testing the programs on the
> calculator shows that this is about true for the TI-89.
>
> It only converts from the TI-85, and has compatibility problems with
> self-modifying code (though a few basic types are handled) and computed
> jumps. Also, variables and interrupts aren't supported. However, the
> following TI-85 games are successfully converted with few problems:
>
> Jezzball 2.2
> Avenger 1.0
> Galaxian 3.0
> Lunoid 7.2
>
> There are also many other programs which come close to working with it,
> and any simple ZShell program with no tricks in it will probably work (as
> without relocation, jump tables/self-modifying code should work
> well). All these converted programs are on ticalc.org (except Avenger,
> due to coyright nonsense). All of these programs worked when I tested
> them, and I just now ran the copy of Lunoid from ticalc.org on my TI-89
> (AMS 1.00, HW 1, Teos 1.02) and it still runs fine.
>
> As for whether it is a "monstrosity" I suppose that judgment will remain a
> matter of opinion. However, the program is not bloated and it does do
> it's intended purpose reasonably well (though there is some work to be
> done, at least implementing the last few instructions, and adding
> variables, interrupts, and a computer jump system would expand its scope).
References: