[m-dev.] for review: MLDS switch optimization

[Fergus Henderson]

On 09-Nov-2000, Tyson Dowd <trd at cs.mu.OZ.AU> wrote:
> On 07-Nov-2000, Fergus Henderson <fjh at cs.mu.OZ.AU> wrote:
> > @@ -1756,6 +1776,9 @@
> >  	=(Info),
> >  	{ ml_gen_info_get_module_info(Info, ModuleInfo) },
> >  	{ module_info_globals(ModuleInfo, Globals) }.
> > +
> > +ml_gen_info_new_label(Label, Info, Info^label := Label) :-
> > +	Label = Info^label + 1.
> >  
> >  ml_gen_info_new_func_label(Label, Info, Info^func_label := Label) :-
> >  	Label = Info^func_label + 1.
> 
> We should really use the counter module in the library to implement
> counters such as this.

Yes.  I just haven't gotten around to that yet.
The surrounding code was written before the existance of the
counter module and for this change I thought it best to
be consistent with the surrounding code.

> > -/******
> > -	% Is it worth including this?  We already have `computed_goto'...
> > +		% This representation for switches is very general:
> > +		% it allows switching on any type, and cases can match
> > +		% on ranges as well as on values.
> > +		% Many target languages only allow switches on ints or enums.
> > +		% Some (e.g. C#) also allow switches on strings.
> > +		% Most target languages only allow matching on values;
> > +		% only some (e.g. GNU C) allow matching on ranges.
> 
> This comment is fine, but it doesn't say what the consequences are.
> Should the backend be responsible for simplifying switches to what is
> available, or should the MLDS code generator look at the backend before
> generating a switch?  (your code does the latter).

OK, good point.  I'll add

		% The MLDS code generator should only generate switches
		% that the target will support.

> > +mlds_output_case_cond(Indent, Context, match_range(Low, High)) -->
> > +	% This uses the GNU C extension `case <Low> ... <High>:'.
> > +	mlds_indent(Context, Indent),
> > +	io__write_string("case "),
> > +	mlds_output_rval(Low),
> > +	io__write_string(" ... "),
> > +	mlds_output_rval(High),
> > +	io__write_string(":\n").
> 
> Isn't this going to cause Pete Ross to thump you when it breaks the
> VC++ backend?  
> 
> Have I missed the bit where you avoid generating this for non GNU C
> compilers?  Or is it not used yet?

Currently this is not used, since the MLDS code generator doesn't
generate switches with `match_range' cases.  I just put it in to
keep the MLDS data structure as general as possible.
(Perhaps I'm overdoing the generality here...)

For the IL back-end, feel free to handle match_range by just calling
error/1.

> > +% This option is not yet documented because it is not yet useful -- currently
> > +% we don't take advantage of GNU C's computed gotos extension.
> > +%		"--no-prefer-switch",
> > +%		"\tGenerate code using computed gotos rather than switches.",
> > +%		"\tThis makes the generated code less readable, but potentially",
> > +%		"\tslightly more efficient.",
> > +%		"\tThis option has no effect unless the `--high-level-code' option",
> > +%		"\tis enabled.  It also has no effect if the `--target' option is",
> > +%		"\tset to `il'.",
> > +
> 
> Perhaps you're over-doing it by adding commented out documentation to
> describe the cases where an unimplemented option has no effect.

The option in question _is_ implemented ;-)
It's just not yet useful.

We do use MLDS computed gotos.  If you use the option, you will see
that we generate different C code as a result.  It's just that in the
mlds_to_c stage we map those to ordinary C switches and gotos, rather
than to GNU C computed gotos, and so the result is no more efficient.
Given that it's no more efficient and that it's less readable, it's
not useful.

-- 
Fergus Henderson <fjh at cs.mu.oz.au>  |  "I have always known that the pursuit
                                    |  of excellence is a lethal habit"
WWW: <http://www.cs.mu.oz.au/~fjh>  |     -- the last words of T. S. Garp.
--------------------------------------------------------------------------
mercury-developers mailing list
Post messages to:       mercury-developers at cs.mu.oz.au
Administrative Queries: owner-mercury-developers at cs.mu.oz.au
Subscriptions:          mercury-developers-request at cs.mu.oz.au
--------------------------------------------------------------------------