[m-dev.] new instmap implementation

[Andrew Bromage]

G'day all.

You (Simon Taylor) wrote:

> Hi Andrew (and everyone else),
> 
> I noticed in the CVS log:
> > Log Message:
> > Update of alias branch to use new instmap implementation.
> 
> Thanks!

No prob.

> Just a few questions:
> 
> - This seems to be a fairly major change, are we going to see a diff?

I hadn't planned on it since everyone needed it urgently.  If you'd
all like one I can provide, but it's probably just as easy to put it
in with the "merge with the main branch" diff (which I'll be no doubt
doing anyway).

> - What impact did this change have on the performance of the compiler?

Didn't check.  It did make the compiler stop producing an exponential
number of inst_keys for large term constructions, so the worst case is
certainly improved. :-)

> - What are you using to compile the alias branch?

Whatever was installed on cyclone yesterday.

> 	 - mercury-latest doesn't compile the alias branch library.

No, it hasn't for some time now.

David and I spoke about this a little while ago after he had some weird
problems with this.  His cure was simple and straightforward: Since there
are three people doing major development work on the alias branch, we
can afford to keep it up-to-date with the main branch as much as possible,
particularly in the case of binary compatibility.  Since we're only
working in the compiler directory, you may consider the runtime and
library parts of the alias branch closed.  (Does anyone know how to
signal this using CVS?)  Use the main branch runtime and library instead.
They work fine.

Cheers,
Andrew Bromage