Melissa O'Neill wrote:
Clearly, "simplifying" the second version of primes into the first by performing CSE actually makes the code much *worse*. This "CSE- makes-it-worse" property strikes me as "interesting".
So, is it "interesting"...? Has anyone worked on characterizing CSE space leaks (and avoiding CSE in those cases)? FWIW, it looks like others have run into the same problem, since bug #947 in GHC (from October 2006) seems to be along similar lines.
Another must-advertise-work-by-David-Sands incident. Actually, joint work with Joergen Gustavsson: Possibilities and limitations of call-by-need space improvement, ICFP'01. http://doi.acm.org/10.1145/507635.507667 Also references [9] and [10] therein. Ciao, Janis. -- Dr. Janis Voigtlaender http://wwwtcs.inf.tu-dresden.de/~voigt/ mailto:voigt@tcs.inf.tu-dresden.de