We had a meeting of the Mercury group on Fri March 9, 2001.

Attendees:
	Tyson Dowd (trd)
	Zoltan Somogyi (zs)
	Thomas Conway (conway)
	David Overton (dmo)
	Fergus Henderson (fjh)
	Mark Brown (dougl)
	Adrian Pellas-Rice (apell)
	Simon Taylor (stayl)

1. General Business

	- Zoltan has submitted the final version of the ARC discovery
	  project grant application.

	- Zoltan, Fergus and Tyson will be in Seattle from 17-26
	  March.

	- Zoltan suggested that the pseudoknot benchmark (Fergus has a
	  Mercury implementation) and the Mercury version of 'error'
	  should be added to the distribution.

	- There was a discussion of the need for a Windows machine for
	  testing the Windows and .Net distributions.

	  	- VMWare considered a bit inconvenient because of its
		  memory usage.

		- Dual booting a machine also inconvenient.

		- Decided that Zoltan's old laptop, Traveller, will be
		  chained to a desk and run Windows so that anyone who
		  needs it can use it.

	- Tom asked for a status report on the accurate garbage
	  collector.  Tyson and Fergus responded:

	  	- It needs a lot of testing.

		- Does not yet handle foreign code which handles
		  Mercury data structures.

		- Has suffered software rot with respect to changes in
		  the RTTI data structures.

		- Does not yet handle closures properly.

		- Does not yet handle existential types properly.

		- Will need performance tuning and perhaps even a
		  redesign of the algorithm.

2. Progress reports

	We went around the table, with each person in turn discussing what
	they had been working on recently and/or what they were planning
	to work on.

	Tyson Dowd (trd)

		- Arrived home from India without having contracted
		  cholera.

		- In addition to travelling to Seattle this month, he
		  will be in Cambridge from 17/4 to 11/5 and may then
		  spend some time in Belgium visiting Mission
		  Critical.

		- Built a Debian package of 0.10.0.

	Zoltan Somogyi (zs)

		- Found a bug in deep profiling after 10 days of
		  debugging.

		- Two new optimisations implemted for the LLDS back
		  end which will hopefully replace value numbering.
		  Still needs to benchmark to work out when these
		  optimisations should be turned on.

	Thomas Conway (conway)

		- Has been tracking down the deep profiling bug with
		  Zoltan.

		- Has been writing up his thesis.

		- Has been benchmarking independent 'and' parallelism
		  (&).

		- Produced a histogram of sizes of allocated cells in
		  the compiler.

		- From next week will be back at MDS Mondays to
		  Thursdays.  Will be writing up his thesis on Fridays
		  -- mostly at home, but may come in to the office
		  occasionally.

	David Overton (dmo)

		- Has been debugging the new mode checker to get a set
		  of benchmarks working.

	Fergus Henderson (fjh)

		- Has been attempting to move expansion of equivalence
		  types to a later stage of compilation, but has been
		  having problems since some phases need to know the
		  original type and others need to know the expanded
		  type.  Zoltan suggested doing the expansion early,
		  but keeping both types around so that different
		  passes can use whichever they need.

		- Has been fixing a reported bug where the compiler
		  aborts with a software error if a program contains a
		  higher-order term without mode and determinism
		  declarations.  Will change the code to give a nice
		  error message instead.

		- Found a bug in the Component Pascal type system
		  which means they will need to do a bit of a redesign
		  of their language to make it verifiably safe in
		  .Net.

	Mark Brown (dougl)

		- Has started his new top secret job and has been
		  reading the CLI spec from Microsoft.

		- Has setup CVS pserver on Venus.

	Adrian Pellas-Rice (apell)

		- Has been debugging his work on sequence
		  quantification.

	Simon Taylor (stayl)

		- Has been fixing bugs and reading background
		  material.


Minutes taken by David Overton (dmo@cs.mu.oz.au).