memory management in PIC

Read a book on basic computer concepts. Memory can be any width in bits - and has been in the past.

1. They are all different memory spaces, not shared like in some Von Neumann machines you may be familiar with. Code memory is usually flash, the others are static RAM technology. PICs are all Harvard architecture, though some of the newer ones map stuff around a bit to blur things.
2. different PICs have different instruction widths-- 12, 14, 16, or 24 bits, depending on family. If the width of this one is 16 bits, then 64k x 16 (not 16k) means that 65536 instructions can be stored (minus any tables and other fixed data).
3. The hardware stack contains return addresses for interrupts and calls. Obviously this poses a limit for reentrancy unless you create a software stack.
4. The RAM ("Register Files") is just for data storage. The memory is usually banked (you have part of the address, say 8 bits, in the instruction and another part in another register). This memory space is shared with SFRs (special function registers), of which there can be quite a few. "Access RAM" is another twist to make the ugly banking less inefficient- allowing a single bank containing a bit of RAM and all the SFRs to be accessed directly.
5. There might be EEPROM data space in addition, but it's not accessed directly.

Have a look at the memory maps in the data sheets and reference manuals.

Best regards, Spehro Pefhany

Colin =^.^=

I think you mean 64K*16 program memory.

The first number is how many words of storage. The second number is the number of bits per word.