We had a meeting of the Mercury group on Monday, 12 November 2001. Attendees: Zoltan Somogyi (zs) David Overton (dmo) Simon Taylor (stayl) Ralph Becket (rafe) Mark Brown (dougl) Progress reports Ralph Becket (rafe): - Contining to work on the loop invariant hoisting optimization and learning the necessary compiler internals. - Started working on the tutorial book. A new cvs module `books' has been created with a subdirectory called `tutorial' which has been populated with a few draft text files. Updates to this module will not be reported on mercury-reviews. - Has been thinking about putting together a more formal specification than the reference manual of the core language suitable for compiler writers. The idea has received support from the group; hopefully a good deal of the document can be culled near directly from local dissertations. David Overton (dmo): - Has been working on the mode constraints paper. - Has been working on formalizing mode declaration checking. - Has not had time to work on implementation this week. Mark Brown (dougl): - Has been working on getting the debugger retry mechanism to work for all IL types with the .NET back end. - Has set up an automated testing system to support the above work. Zoltan Somogyi (zs): - Has finished debugging the stack slot optimization and obtained some speed-ups. Performance tests were still running during the meeting (seven versions of the compiler with each of -O2 ... -O5!) Some results have shown a 1.3% speed up and others have shown a 7% reduction in stack frame size. Zoltan reported that the optimization was not too costly in terms of compilation time. - Has been working on the stack slot optimization paper which is due in this Friday. - Reports that six summer students will be joining the team, but not including Simon Mei who was with us last year. It was suggested that once they are all here together we give them the one-day introduction to Mercury and a one day introduction to the CVS, compiler and build systems. Simon Taylor (stayl): - Has implemented unification expressions as discussed in last week's meeting. - Has added the option of supplying user-defined operator tables to the parser. - Has fixed up the handling of libraries directly from mmc, adding the option of overriding settings by command line switch rather than several environment variables. There was some discussion on this point, the general concensus being that there needed to be a simplification of the interface to the build system. Simon is working on this issue. Minutes taken by Ralph Becket (rafe@cs.mu.oz.au)