Fibonacci and performance

Fri Apr 27 13:33:00 GMT 2001

Quoth Andrew Haley on Friday, 27 April:
: I was also thinking of some evil way to overwrite the call to
: JvInitClass the first time it was used... :-)

That's how we handled such things in the CM5 (sparc) C* compiler, but
there are all manner of bothersome problems with this for gcc. You've
got to invalidate the i-cache lines manually, for example, on modern
sparcs. I'm guessing there are other architectures where executable
pages are normally mapped read-only.