RE: [MI400] MI as the MI language of the as400

["Steve Richter" <srichter@xxxxxxxxxxxxx>]

-----Original Message-----
From: mi400-admin@midrange.com [mailto:mi400-admin@midrange.com]On
Behalf Of Jon Paris

>Another aspect that played into the whole ILE issue was the desire to get
>away from generating MI.  The W-code gives far greater scope for
>optimization than MI or so I'm told.  MI is just too "lumpy".
>to a cross platform intermediate, optimization techniques can be shared and
>the way is open to using cross-compilers.  For example the new C and C++
>compilers.  Even plain ole RPG programs can benefit significantly from the
>new optimization techniques.


The bullshit alarm just went off.

Please point me to the technical papers written by the cross compiler team
that explain why you cant optimize "lumpy" code and please provide some
examples of lumpy vs non lumpy code.

To optimize at compile time you examine the instruction stream to see what
instructions can be run in parallel, what conditional branches can be tested
ahead of time. Based on those findings you add instructions to the risc
output to maximize the parallelism and pipeline efficiency that the powerpc
chip will do on its own without any help from the people in Toronto. The
requirements of a debugger probably complicate things a lot so maybe
rollback code is needed.  This is all my guess, and I welcome the experts
coming on this list and educating me on the subject.

WCode is not published. MI is. < unproven and pointless accusations as to
why and who deleted. ( hint: who benefits from the crime ) > WCode and ILE
work together. If an alternative, non ile process is allowed on the iSeries,
wcode is not needed and I would be very suprised if the resulting code was
not fully optimizable.

respectfully,

Steve Richter