Re: Fat references

Anton Ertl <anton@mips.complang.tuwien.ac.at> wrote:
(snip, someone wrote)


>>> 16 bits: word
>>> 32 bits: double word
>>> 64 bits: quad word


(then I wrote)
>>That is VAX. Everyone else uses 32 bits for a word.


> No, that's the 8008 and its successors. For the PDP-11 and its
> successors (including the VAX) the 32-bit units were called longwords.


Certainly the PDP-11 was described as having a 16 bit word, but I
don't remember double word, or especially quad word being used in
context with the PDP-11. There were no memory operations that would
do that.


The 8008, 8080, 6800, 6502, and others at that time were considered
'eight bit processors' even though the 8080 could do some 16 bit
operations. One might have called 16 bits a word, but it wasn't
commonly needed. (At least much less than for the PDP-11.)


Already being used to a 32 bit word for S/360, I remember being
disappointed at the time that VAX called 16 bits a word. Especially
if one believed that VAX was supposed to be in the 32 bit processor
marketplace.


> "Word" for 32 bits is common for architectures that originated from
> 32-bit architectures that were not sold as extensions of 16-bit or
> 8-bit architectures (e.g., IBM 360, MIPS, SPARC, ARM).


It always made some sense to me. As I remember, in typing a word is
considered to be five characters. The average english word seems
likely to be in the four or five character range, not two.


> How do original 64-bit architectures like the Cray-1 call 16-bit and
> 32-bit units? Do they have any support for that at all?


I believe the Cray-1 does 64 bit memory operations, and that most (if
not all) C compilers for the Cray-1 used a CHAR_BIT of 64. If C char,
short, int, and long are all 64 bits, why have a name other than word
for them?


-- glen

Return to the comp.compilers page.
Search the comp.compilers archives again.