Re: SD: Let me propose something then...
[Prev][Next][Index][Thread]
Re: SD: Let me propose something then...
The problem with that is that the address of the memory allocated to the
program is variable, which will slow down the program, sort of like using
PROGRAM_ADDR. So if you are going to implement something like this, using
usgard like relocation for some or all RAM adresses might be a better idea.
Also if you are planning to make it possible for a program to allocate
memory while it is running, remember that you need some where to make sure
that all other programs update their pointers to much the new memory layout.
Wont the new way make it less obvious to the programer what is actually
happening? As long as you are using a table like ZShell, there will be a lot
of addresses which it wont be possible to call, so the programmer will need
to know what addresses are supported anyway. If the programmer needs to know
about the ROM calls anyway then using ROM calls directly seems just as easy.
Another solution would be to include functions like ROM_CALL and FPROM_CALL
and make the assembler automatically detect which one fits the current call.
That would make it easier to use the ROM, but on the TI85 that would require
a lot of space.
Dines
----- Original Message -----
From: <Farid76@aol.com>
To: <shell-developers@lists.ticalc.org>
Sent: 6. januar 2000 20:00
Subject: Re: SD: Let me propose something then...
>
> In fact, what I'm going to do is not to restrict memory access but to
> allocate memory in a temporary variable when programs need it, so each
> program would have a pointer to a safe memory area and won't corrupt
memory
> used by others programs. Obviously, this method will only work with
> shell-specific programs.
>
> In order to make easier ROM access, I will write my own asm compiler which
> will check for ROM calls, then if the address is not the same in all ROM
> versions, it will change them into calls to shell functions which use a
ROM
> table (like in ZShell).
>
> > Making it impossible not to corrupt memory is going to be very hard when
no
> > buld in support for that kind of a function is not build in ot the CPU.
On
> > the TI85 you would have to check all instructions before they where
> > executed, which is going to make everything extremely slow. If it was
> > possible to do this kind of a thing it would be a great feature in a
shell
> > though as many of the problems with badly written programs could be
> avoided.
> >
> > Making multitaksing work would really require a new language, but you
would
> > nned some speical functions in your shell which the program could call
to
> > take advantage of the multitasking feature. Making it easier to access
the
> > ROM is hard to, a lot of people have thought about this, and no one
seems
> to
> > have a better solution than the one currently used. You have to
consider
> the
> > differences between ROM version, which makes this kind of thing hard to
> > implement.
> >
> > Dines
>
References: