From: Craig Chambers (chambers@cs.washington.edu)
Date: Sat Feb 01 2003 - 15:30:01 PST
I'm publishing some sample Whirlwind optimizations (for 501 purposes,
and as additional test cases for the DFA framework in whirlwind).
Both work solely over the CFG; this was intended to make them easier
to understand and write, but I'm not sure this is true, given
Whirlwind's dataflow-node-oriented design. One optimization is
constant propagation & folding, originally written by Sorin then
modified by me. The other is CSE. The CSE pass is a sample solution
to the next 501 project assignment, so Andrei don't look! I'm also
publishing an additional test file to exercise CSE. To help the CSE
implementation, I added a copy_mutable operation to sets (and
reindented some of the code to make it more readable in the current
Cecil style). I successfully used the constant prop pass over my test
suite and towers, and I used CSE over the test suite and it's now
running through towers.
There seems to be a problem in Whirlwind with writing optimizations
whose replacement graphs have just the CFG, but whose enclosing graph
also has a DFG. I don't know what the problem is. I made some
changes that should have helped, but it's not fully working yet. So
beware.
Sorin, please check out my sample solution, to see what you think.
-- Craig
_______________________________________________
Cecil mailing list
Cecil@cs.washington.edu
http://mailman.cs.washington.edu/mailman/listinfo/cecil
This archive was generated by hypermail 2.1.5 : Sat Feb 01 2003 - 15:28:38 PST