[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Java & Win95


I'll have a go tomorrow am to grab those html pages you mentioned.

I've no idea what rules the Java class loader follows.  I think a class
file gets loaded and linked dynamically the first time it's used -- i.e.
if class A uses class B and we recompile class B, we shoudn't need to
recompile class A because the latest version gets bound in at run time.

I've had problems when recompiling something with an error and I don't
notice the error and get the previous (succesfully compiled) class file
loaded in and run ... that gives msyterious results!

> The "warm-up loop" in Peter's program was nice,
> except that sometimes it writes out 1,2,3.....to 16, another time
> it starts at 16 (!!???) and prints out increasing numbers.

I had a commented out print statement both in the warm-up loop and in the
benchmark one.  If these are uncommented, the warm-up loop should print
numbers 0..15 and the benchmark loop from 16 upwards.  Dunno why yours
seem to be out of control!

By the way, you had asked why the warm-up loop.  ComsTime was meant to be
measuring just the context-switch/message-passing overheads.  Having the
warm-up loop ensures that all process have started before we start timing.
Actually, a warm-up loop of length two should be suffucuent to ensure that.