IBM

I don't get how the machine-interface (IBM calls it "MI") works. They say the AS/400 is somewhat unusual because there are two machine-levels that play a role in practice. The upper one is called the MI-level and the lower one is the CISC/RISC platform, therefore MI, the Machine Interface, is in a very real sense what makes your machine an AS/400 rather than just a souped-up PowerPC. AS/400 is made in layers, and the operating system does not depend on the lowest layer (the hardware), therefore you can change processors (e.g. from CISC to Power) without the need of touching OS/400, and applications (the highest layer) are also independent.

How is it possible? Doesn't it sound like "java"? And exactly, what is the Machine-Level? And what is above and below the MI?

Its a binary layer, as in when you write applications for IBM i or OS/400 you don't actually compile to the underlying CPU's code, afaik. Compilers produce a CPU independent intermediate code which the OS changes to PPC, x86 etc code as needed. It's a neat concept.

Heh, I've toyed with that idea in my various notes towards hobbyist OS development. Never knew that it saw production use in a professional OS.

Doesn't it sound like "java"?

Java, especially AOT compilation, does indeed work on a similar scheme.

So, is it similar to byte-code?

Y888099 wrote: So, is it similar to byte-code?

"Bytecode" is just a generic term for the machine language of a virtual-machine architecture, so, yes.

I'm curious, does it always do the translation dynamically on loading, or does it cache it in the filesystem somewhere?

I believe it's cached, but this (like many things in System i) is abstracted away.

There is a really good book that explains everything called "Inside The AS/400". I've been meaning to pick up a copy. When you see what IBM moved into silicon you'll wonder why we still use the architecture we do.