[m-dev.] for review: samples/diff update

[Andrew Bromage]

G'day.

You (Fergus Henderson) wrote:

> On 14-Sep-1998, Andrew Bromage <bromage at cs.mu.OZ.AU> wrote:
> > samples/diff/README:
> > samples/diff/TODO:
> > 	Update stuff that's now done, do a couple of minor wording
> > 	changes.
> 
> Have you compared the performance with that of GNU Make?

I ran my test suite on murlibobo, and everything was so fast that,
modulo Mercury startup time, everything happened pretty much at the
same speed.  (This was using -O6 --intermodule-optimization with
array bounds checks disabled.)

Unfortunately I can't test it on other machines just now since
everything that I compile on non-alpha machines (tested: hydra,
mercury, kryten) stack dumps.

> If so, and particularly if you have analysed the reasons for
> whatever differences there are in performance, it might be worth 
> putting something about this in the documentation somewhere.

If the diff algorithm turns out to be the bottleneck, profiling won't
tell you why, unfortunately.  (But I have my suspicions.  See the
documentation in match.m.)

Cheers,
Andrew Bromage