local refs

Matthew Fluet fluet@CS.Cornell.EDU
Tue, 11 Dec 2001 15:51:05 -0500 (EST)


> I'm rerunning self-compiles to factor in these other global checks (and
> also to get dominator timing in commonSubexp); 

That was definitely the cuplrit:
            localRef starting
               multi starting
               multi finished in 0.70 + 1.90 (73% GC)
               restore totals 3.03 + 3.26 (52% GC)
                  checkGlobalsViolations totals 1.38 + 2.05 (60% GC)
                  checkViolations totals 0.11 + 0.0 (0.0% GC)
                  dominatorTree totals 0.82 + 1.21 (60% GC)
                  computeDF totals 0.36 + 0.0 (0.0% GC)
                  computeLive totals 0.02 + 0.0 (0.0% GC)
                  computePhi totals 0.09 + 0.0 (0.0% GC)
                  rewrite totals 0.21 + 0.0 (0.0% GC)
               shrink totals 0.41 + 0.0 (0.0% GC)
            localRef finished in 5.47 + 6.54 (54% GC)

Opinions on whether this check is worth keeping in there?

And dominators are a significant portion of commonSubexp:
            commonSubexp starting
               dominatorTree totals 2.52 + 4.40 (64% GC)
            commonSubexp finished in 4.58 + 4.40 (49% GC)