Python Interpreter question.

[Heiko Wundram]

Am Mittwoch, 8. September 2004 06:46 schrieb Terry Reedy:
> Whether the 386 machine
> was ever a real machine or whether it was also microcoded, I do not know.

All CPUs made by Intel up to and including the Pentium (586) were "real" 
machines, which ran the bytecode as it came from RAM, without an intermediate 
layer of microcoding. Only with the wake of the Pentium II (IIRC), the 
concept of microcode with a (somewhat) RISC core became seasonable in the PC 
CPU-producing community.

Also, IIRC, the AMD K6-* were real CPUs in the sense that they didn't have a 
microcode layer between the execution and the runtime environment.

The first AMD CPU to feature microcode (although it wasn't software 
changeable, as it was with all Intel CPUs from the PII on; using a little 
software tool you can change the behaviour of your Pentium to resemble that 
of a Motorola 68000, except for the endianness issue, now isn't that cool?) 
was the Athlon, which used microcode not only for the core instruction set, 
but also to drive its FPU, which is completely different to the standard FPUs 
found in the Intel CPUs (it is a RISC processor in the true sense of the 
word). This is also what made the Athlon so easily extensible, as 
implementing a new set of FPU-instructions (like SSE/MMX/SSE2, etc.) only 
needed some additional microcode on the chip (which is cheap to write), 
something that Intel has much more trouble with, as the FPU is decoupled from 
the CPUs microcode layer, and isn't microcoded at all AFAIK.

Anybody correct me if I'm wrong; this is all that I remember from my first 
semester classes on CPU-design. ;)

Heiko.