> 
> > finally I got burned by the 532's icache :-). A short fix for the problem
> > follows.
> 
> If it's a problem that the cache isn't being invalidated when it
> should be, well, then the cache _should_ be invalidated when it should
> be!  8-)

This is true :-), but I think cache flushing via system call is so
expensive that you don't need to flush the cache anymore once you
got to the point where you could flush the cache...

I think doing it the way I did it (by not invalidating the index)
might be faster then doing an actual cache flush via a trap.

	Matthias

PS: How do you all do this on the i486/i586? The *do* have data/instruction
    caches?

------------------------------------------------------------------------------