Addressing External Memory

Is is better to address the "command register" of an external chip as XDATA of 8051 chip(using ALE,Port0,Port2,etc)
This is memory-mapped I/O.
or should I just manually arrange some I/O pins (of 8051) to do that?
This is I/O mapped I/O.
None is "better" than the other, the general consensus is that if P1 and a couple of P3 pins can handle your I/O you go I/O mapped, if not you go memory mapped.
The event of "memory rich" chips such as the p89c668 with 8k RAM and 64k FLASH inside the chip has given the possibility of much more I/O mapped I/O since this frees P0 and P2 for I/O.

Erik

Hello,

Thank You Erik!

I really need help on this
---------------------------------------------------------------
BASE_ADDR = 0x8000
i=1

#define COMMAND(i) (*((volatile char xdata *)(BASE_ADDR +i)))
---------------------------------------------------------------

this code seems to use casting to define COMMAND(i) to point to a specific xdata location

1) Is "COMMAND(i)" a "memory specific pointer"?
2) Does it take up any memory space (eg 2bytes as you would expect of a memory specific pointer)?
3) Is there a better way (or clearer code) to do this?
4) Is the "volatile" really necessary? (or should i do some settings to prevent optimization from affecting code)

Regards,
KC

1) Is "COMMAND(i)" a "memory specific pointer"?
yes
2) Does it take up any memory space (eg 2bytes as you would expect of a memory specific pointer)?
look in the generated assembly
3) Is there a better way (or clearer code) to do this?
many ways, I would name the memory mapped I/Os individually and use
#define SOME_COMMAND_ADDR (BASE_ADDR +nnn)
xdata volatile SOME_COMMAND _at_ SOME_COMMAND_ADDR
or something like that
4) Is the "volatile" really necessary? (or should i do some settings to prevent optimization from affecting code)
Volatile works for this, why reinvent the wheel.

Erik

1) Is "COMMAND(i)" a "memory specific pointer"?

No.
A Pointer is a variable - you are defining an expression

2) Does it take up any memory space (eg 2bytes as you would expect of a memory specific pointer)?

No (apart from the code space necessary to evaluate the expression).
COMMAND is a preprocessor definition - the macro is expanded by the preprocessor before the compiler ever sees it!

3) Is there a better way (or clearer code) to do this?

How about _at_, as Erik suggested?
Or Keil's predefined XBYTE?

4) Is the "volatile" really necessary?

It's your system - only you can tell us that!

...should i do some settings to prevent optimization...

No.
This is precisely the situation for which volatile is provided - use it!