DyC Progress Log

7/3/97:
BKG: The one-unit (no branches) test case looks good. Ready for run-time testing once merger changes are made. The framework for multi-unit regions is there and ready for eye-ball testing.
4/9/97:
BKG: I have replaced static operands in the template code with RTCONSTs and inserted store ops in the setup code corresponding to those uses. The RTCONSTs go through for my small test program and print in the format Matthai requested.
3/20/97:
BKG: Separation of setup and template using BTA information works. Now to test RTCONSTs...
3/18/97:
Charlie and Matthai have written a Perl script that merges setup and template blocks. Markus has added many fixes to the BTA (e.g., killing all derived static temps at a makeDynamic). Brian has tested p2topo and confirmed that it works. A boilerplate has been added to dyc2c, and it is successfully excised from the graph. The compiler is set up to perform a pseudo-separation of setup and template code by simply copying the dynamic region, inserting an IGOTO, and putting in the per-block tags that the back-end script needs.
9/25/96:
BKG: Merged with Markus's BTA changes. Successfully compiled my version of the code after solving problems with compiling functions using varargs.
9/25/96:
MM: BTA with reachability conditions seems to work.
9/17/96:
BKG: Added several more routines to the GRAPH ADT. Wrote the tail-merging code.
9/13/96:
BKG: Added traversal routines to the GRAPH ADT. Wrote the unit-splitting code. Discovered problem with the way Multiflow orders links to predecessor and successor blocks.
9/10/96:
BKG: Finished first cut of the GRAPH ADT.
9/10/96:
BKG: Changed dyc2c to use symbolic names as function arguments, rather than strings or integers. Finished STACK and LIST ADTs.
9/6/96:
BKG: Wrote code to convert TEMPs to RTCONSTs, partition basic blocks, and assign blocks to specialization units.
8/28/96:
BKG: Finished first pass on adding the VALUE_RTCONST value type.
8/27/96:
BKG/MM: Agreed on formats of RTCONSTs and BTA information.
8/23/96:
BKG: Added p2topo to compute topological information.
8/22/96:
BKG: Linked in p2dyc phase and added a command-line-option template.
8/21/96:
MM: Checked in original version of Multiflow, with build fixes.
8/19/96:
BKG: Fixed a number of build problems so that we can use the pmakefiles and do_make.
8/16/96:
BKG: Completed intra-procedural dyc2c.

Last updated March 18, 1997.
Brian Grant (grant@cs.washington.edu)