Fibonacci and performance

Alexandre Petit-Bianco
Fri Apr 27 13:42:00 GMT 2001

Tom Tromey writes:

> In this case the gcj assembly output looks very different from the
> g++ output (before the change it was very similar).  For one thing
> the assembly (both at -O2) for the function in question is almost
> twice as long for gcj.

Do you have a working copy of insight? (I don't.) If yes, it might be
interesting to post the interwoven source and assembly dump of both
functions, as I found them very easy to read and interpret. I don't
know how well this functionality can work given that we're using -O2.


More information about the Java mailing list