Dan Sugalski <[EMAIL PROTECTED]> wrote: > At 12:02 PM +0200 5/5/04, Leopold Toetsch wrote: >>Changing C<add> or any MMDed opcode to look like a branch is a severe >>performance impact for the non-overloaded case.
> If the JIT structure makes it untenable, it doesn't work, and that's > fine. I don't think it has to be quite as bad as it is now, but on > the other hand the performance hit in general needed to make this > work better is probably not worth it. It's not that slow any more. Running overloaded PASM hasn't more overhead then calling a sub. My Pentium 600 runs 1E6 overloaded C<bxor> functions in 1.5 seconds. The overload benchmark is at 3 times the speed of perl 5.8.2 now. The SSE version of memcpy gave it a big boost on Pentiums. And there is still SSE2, which I can't test. > Something to keep in mind once we get more of the base PMC types > implemented, and have more of an idea how much of the MMD code ends > up being bytecode vs C. I don't expect much of the basic functionality being in PASM. leo
