OpenEuphoria

2. Re: Empty procedure optimized out?

• Posted by Tone Škoda <tskoda at email.si> Nov 19, 2004
• 537 views
• Last edited Nov 20, 2004

Tommy Carlier wrote:
> 
> Robert, couldn't you remove calls to empty procedures while parsing
> a program?
...
> And it's used like this:
> }}}
<eucode>
> ignore(someFunction(someParameter1, someParameter2))
> </eucode>
{{{

> The function 'someFunction' should be executed, but the call to the
> empty procedure could be removed. 

It would be much more logical if someFunction would not be executed.

4. Re: Empty procedure optimized out?

• Posted by Robert Craig <rds at RapidEuphoria.com> Nov 19, 2004
• 556 views
• Last edited Nov 20, 2004

Tommy Carlier wrote:
> 
> Robert, couldn't you remove calls to empty procedures while parsing
> a program? I have a procedure to ignore the return value of a function,
> that looks like this:
> }}}
<eucode>
> global procedure ignore(object ignored)
> 	-- ignores its parameter
> end procedure
> </eucode>
{{{

> And it's used like this:
> }}}
<eucode>
> ignore(someFunction(someParameter1, someParameter2))
> </eucode>
{{{

> The function 'someFunction' should be executed, but the call to the
> empty procedure could be removed. That could also optimize this
> situation:
> }}}
<eucode>
> constant DEBUG = 0
> 
> procedure writeToLog(sequence line)
> 	if DEBUG then
> 		puts(logfile, line & '\n')
> 	end if
> end procedure
> </eucode>
{{{

> 
> 'if DEBUG then' can be removed, because you know before running the
> code than it will never be called. That makes the 'writeToLog'-procedure
> an empty procedure that can be optimized out.

Thanks. I'll add it to my list of possible optimizations.

By the way, this was too trivial for the release notes, 
but I took out the warning about parameters declared but not used.

Regards,
   Rob Craig
   Rapid Deployment Software
   http://www.RapidEuphoria.com

6. Re: Empty procedure optimized out?

• Posted by Pete Lomax <petelomax at blueyonder.co.uk> Nov 20, 2004
• 539 views

On Fri, 19 Nov 2004 13:58:55 -0800, Robert Craig
<guest at RapidEuphoria.com> wrote:

>By the way, this was too trivial for the release notes, 
>but I took out the warning about parameters declared but not used.

<shakes head in despair>
You think no-one wanted to know you'd done that!!??
A very simplistic cop-out to deal with the problem, IMNSHO.
BTW, there /are/ some cases when that warning should remain.
</shakes head in despair>

8. Re: Empty procedure optimized out?

• Posted by Patrick Barnes <mrtrick at gmail.com> Nov 20, 2004
• 538 views

On Fri, 19 Nov 2004 18:08:37 -0800, CoJaBo <guest at rapideuphoria.com> wrote:
> > On Fri, 19 Nov 2004 13:58:55 -0800, Robert Craig
> > <guest at RapidEuphoria.com> wrote:
> >
> > >By the way, this was too trivial for the release notes,
> > >but I took out the warning about parameters declared but not used.
> That is probably the most annoying "feature" of Eu.
> I for one am glad at its demise!

But you can turn it off! And unless you're using a library that has
call-backs, if you see the warning, it means that there's something
that you should do about it.

I'd like to have seen the warning remain... but maybe some flag to
stop the warnings appearing in those call_back functions?


In any case, a simple 'without warning' fixed all the remaining
warnings when your code was ready for release.

-- 
MrTrick

10. Re: Empty procedure optimized out?

• Posted by Robert Craig <rds at RapidEuphoria.com> Nov 20, 2004
• 571 views
• Last edited Nov 21, 2004

Juergen Luethje wrote:
> Robert Craig wrote:
> > By the way, this was too trivial for the release notes,
> > but I took out the warning about parameters declared but not used.
> 
> Previously, we could turn it off by using
>    without warning
> if we wanted to do so.
> 
> But now, we can *not* turn it on, if we want to do so.
> May I ask you please to reconsider this decision?

There's a line of code commented out in symtab.e line 510

or equal(vtype, "parameter") -- this is rarely a real problem

If you put that back in, eu.ex will report the warning.

You always have to strike a balance when reporting warnings.
You'd like to catch legitimatate problems, but you don't
want to generate a lot of false alarms. In this case I felt
the false alarm rate was maybe 100x the legitimate problem rate,
so I took it out. For instance, people might be calling a 
set of routines via call_proc() or call_func(), and some of 
those targetted routines might not use all of the arguments that 
are passed to them. 

The situation is different with other private variables. 
If you get a warning that one of those was declared
but not used, it's a simple matter to remove the useless
declaration. Those warnings remain. Most of them are legitimate.

Regards,
   Rob Craig
   Rapid Deployment Software
   http://www.RapidEuphoria.com

12. Re: Empty procedure optimized out?

• Posted by CoJaBo <CoJaBo_EUforum_Address at CJBN.net> Nov 21, 2004
• 551 views

(Take 2, stupid router)
Tony Steward wrote:
> 
> Here Here please reconsider I want the warnings.
Those warnings only affect routine parameters,
not other variables. Those warnings seem to have
no real use.

> 
> regards
> Tony Steward
> 
> 
> On Sat, 20 Nov 2004 12:27:44 +0100, Juergen Luethje <j.lue at gmx.de> wrote:
> > 
> > Robert Craig wrote:
> > 
> > > Tommy Carlier wrote:
> > >>
> > >> Robert, couldn't you remove calls to empty procedures while parsing
> > >> a program?
> > 
> > <snip>
> > 
> > That would be great!
> > 
> > > Thanks. I'll add it to my list of possible optimizations.
> > >
> > > By the way, this was too trivial for the release notes,
> > > but I took out the warning about parameters declared but not used.
> > 
> > Previously, we could turn it off by using
> >    without warning
> > if we wanted to do so.
> > 
> > But now, we can *not* turn it on, if we want to do so.
> > May I ask you please to reconsider this decision?
> > 
> > Regards,
> >    Juergen
> > 
> > 
> -- 
> Regards
> Tony Steward
> www.locksdownunder.com
> 
> IF IT IS TO BE IT IS UP TO ME!
> 
>