Re: An introduction to MI by example II (part B)...

Excerpts from mi400: 16-Nov-99 RE: An introduction to MI b..
Jon.Paris@halinfo.it (1773*) 


> [MI] was never the native language for anything. Just something the old
> OPM compilers generate as an intermediate language.  

Well, arguably MI *is* the pre-ILE "assembler" language on the AS/400,
provided you accept the definition of assembler language as: "The
lowest-level language useful to an applications programmer."  Of course,
Original MI is very much a 'storage-to-storage' model, which hobbles
advanced compiler optimizations.  W-code and the "new" MI are based on a
stack model, and were designed as such primarily to address this
shortcoming. 

> TIMI does stand for Technology Independent Machine Interface but that is
> _not_ MI.  

Mmm, well...  These terms are probably close enough for government work.
 *8-)  The TIMI is effectively *composed* of MI instructions (and their
associated AS/400 "objects"), eh? 

> As Hans Boldt pointed out on Midrange-L the MI that you are programming
> in here is further away from the system in some ways than programming in
> C.  

Further away from any chance at modular, optimizable, bindable code for
certain.  There have been no enhancements to the MI Transformer (other
than minor bug fixes and support for new/modified instructions) since we
wrote it in V3R6, and optimization of OPM (Original Program Model)
programs has never resurfaced as a key requirement, (particularly after
the nice boost from the RISC hardware). 

> I don't want to discourage people from using MI (well in some ways I do
> <grin>) but don't use it thinking you're getting closer to the metal -
> you're not. You can just do things that can't be done in RPG - unless of
> course you use the MI functions of the C library. This has the added
> advantage of being a bound call whereas the calls to MI programs are
> dynamic. 

Absolutely agreed. 

> All new OS/400 function tends to be developed in C and C++ and not in
> PL/MI (which itself is not MI) - IBM never (or hardly ever) wrote in MI.
>  The current IBM compilers generate W-code (a sort of itty bitty
> assembler) which is transformed into MI-prime I think, for translation
> purposes.  But MI-prime is almost identical to W-code [...] 

W-code is more platform-neutral than MI-prime, but they are very close
indeed -- close enough to where the transliteration is a relatively
miniscule portion of overall program-creation processing. 

>  [W-code] in turn is very close to the Risc instruction set.  

I beg to differ, but W-code is so far above the RISC instruction set
that we've got satellites in low-earth orbit closer to the RISC
instruction set than W-code is...  W-code is an intermediate language
only -- it's a whole lotta work (done by tens of thousands of lines of
code) to turn that into RISC instructions. 

> That is why the C compiler and its built-ins can do a better job than MI
> for the RISC boxes. 

+---
| This is the MI Programmers Mailing List!
| To submit a new message, send your mail to MI400@midrange.com.
| To subscribe to this list send email to MI400-SUB@midrange.com.
| To unsubscribe from this list send email to MI400-UNSUB@midrange.com.
| Questions should be directed to the list owner/operator: dr2@cssas400.com
+---