9 May
2010
9 May
'10
10:15 a.m.
On 5/8/2010 3:06 PM, Tony Duell wrote:
It means we can never come to agreement on the subject. You're a
hardware guy, so naturally you consider the bit-ness of a CPU to be the
size of the data bus. I'm a software guy, so I consider the bit-ness to
be the size of the registers (ALU for the anal among us). So never the
twain shall meet, I guess.
BINGO! You've got it...
The whole point is that it is (IMHO) meaningless to quote a 'number of
bits' for a processor and have it meaningful in all situations.
Sure an 8088 is normally classed as a 16 bit processor, and I normally
class it as one. But when I am wiring up the databus, or when I am
debugging the hardware of my IBM5150 or HP150, then I need to know there
are only 8 data lines. My P850, with its row of 16 toggle swithcs on the
panel, with registers that appear to me 16 bits wide, with 16+16 bit add
instructions, etc is a 16 bit machine most of the time. That is until I
start working inside the CPU and find there's only an 8 bit ALU there.My
TRS-80 Model 4 is an 8-bit machine (Z80A), but I still have 16+16 bit add
instructions. But only an 8 bit data bus.
And so on.
Now, what's a PERQ? THe ALU is 20 bits wide, but the main memory data bus
is only 16 bits wide. The most efficient memory accessses read 64 bits
You're a hardware guy :)
Correct. But I don't see what differece that makes. from 4 memory locations in one go, and then transfer
them to the CPU on 4
consequtive microcode sycles. I normally consider it to be a
20 bit
machine (ALU and procesosr register width), but that doesn't seem to be
the only sensible bit-count.
-tony