RFC: midlevel RTL

[Jan Hubicka]

> On Sun, Jan 20, 2002 at 06:24:13PM +0100, Jan Hubicka wrote:
> > 	* gensupport.c (init_md_reader): Implicitly inlucde midrtl.md file.
> 
> This I think is a mistake.  I think mid-level rtl and target rtl
> should have completely separate recognizer functions.

As I've described, I want to do that later.  I can't simply change whole code
generation process to not include any lowlevel RTL instruction in one day, as
there is way too many expanders called from everwhere - machine specific
builtins, call convention, string operations, etc.

My plan is to use this as temporary solution - at the moment I can generate
midlevelRTL for arithmetics and move patterns and lowlevel for the rest, where
midlevel gets lowered.  Once all the side cases mentioned are tracked down, we
can easilly modify the generators to generate disjunct recognizer.

Avoiding all the cases require modification in frontends and machine dependent
code.  I hope to do many of these on the branch and once the code is useable I
hope to get it included in mainline in 3.2 development period, so the rest of
changes can be done in mainline tree in small, incremental steps.

Doing all at once looks like way to get completely unmaitainable patch not
included to mailine for a years because it breaks everything else few platforms.

Even in such incomplette form, the midlevel RTL should bring number of benefits
such as the end of libcalls, easier GCSE, strength reduction and similar stuff.

Does that sound sensible?
Honza
> 
> 
> r~