Jon Harrop wrote: >> Actually, since I was in Redmond last January for three days talking >> personally to those guys at Lang.NET, I feel safe to say they have >> shelved the JIT for several years, and their optimizations have not >> kept pace with those in the JVM. (See my blog entry previously >> mentioned.) > > My benchmark results disproved your belief.
for what it's worth, in PyPy we discovered that hotspot produces much better code than the CLR when the bytecode doesn't follow the standard pattern produced by the java/c# compilers. In particular, we heavily use exceptions to model control flow in our RPython program (e.g., every "for" loop needs to catch StopIteration), but the CLR JIT it not able to optimize such a case, and thus the first versions of the CLI backend produced very slow code; to have reasonable performances, we rely on our own inliner/malloc removal/exception inliner, which gave a speedup of something like 30x, IIRC. On the other hand, hotspot produces much better code[1], and moreover we get faster code if we *disable* our own optimizations, since if we use them it results in more code to analyze because of the inlining. [1] http://blogs.sun.com/jrose/entry/longjumps_considered_inexpensive ciao, Anto --~--~---------~--~----~------------~-------~--~----~ You received this message because you are subscribed to the Google Groups "JVM Languages" group. To post to this group, send email to [email protected] To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/jvm-languages?hl=en -~----------~----~----~----~------~----~------~--~---
