[m-dev.] LCO

[Fergus Henderson]

On 18-Sep-1998, Erwan Jahier <Erwan.Jahier at irisa.fr> wrote:
> Fergus Henderson wrote:
> > 
> > On 18-Sep-1998, Erwan Jahier <Erwan.Jahier at irisa.fr> wrote:
> > > Zoltan Somogyi wrote:
> > >
> > > > > 2) Is LCO switch off when tracing is on ?
> > > >
> > > > Effectively, yes. The code of the exit event will prevent any kind of last
> > > > call optimization.
> > >
> > > hmm... shouldn't we have a --trace-exit option in order to still be able
> > > to trace programs that makes strong use of LCO ?
> > 
> > Is this a real need, or an imagined need?
> > How common are such programs?
> 
> Well in fact, that is not the reason why I am asking that. 
> The reason is that if we could have a tracer "ligth" enough to allow
> user to always compile their program with trace information on, it would
> be great !

Yes, it would indeed.  I agree that that would certainly be sufficient
justification for a `--trace-exit' option.  Making the tracer sufficiently
light may not be easy, however.

> I've done benchmarks on a program that solve the hanoi tower problem
> with 19 disks, and it happends that the program with tracing on runs
> around 15% slower which is already not too bad (I shall do more
> benchmarks).

Yes, please do, and let us know the results.

-- 
Fergus Henderson <fjh at cs.mu.oz.au>  |  "I have always known that the pursuit
WWW: <http://www.cs.mu.oz.au/~fjh>  |  of excellence is a lethal habit"
PGP: finger fjh at 128.250.37.3        |     -- the last words of T. S. Garp.