"David S. Miller" writes:
>    > Indeed nice.  On the SS10 with Happy Meal ethernet cards the code path
>    > at each end works out to something like:
>    This is *not* a microoptimization. Its the sort of real optimization I
>    keep mentioning. Eliminating copies is significant in networking
>    stacks. It isn't the same thing as making getpid() run two million
>    times a second.
> 
> I would not be even close to the numbers I get if:
> 
> 	a) my task switch wasn't twice as fast as Solaris on Sparc
> 	b) my syscal entry/exit was not as fast
> 	c) my interrupt latency was not so low

Thats silly. Lets say we are on an FDDI ring. Nice 4k MTU. Are you
going to tell me, with a straight face, that a 4k copy isn't going to
blow the syscall delay into the small fractions of a percent? Hell,
are you going to tell me that a 1.5k copy and checksum isn't going to
blow it to hell?  Eliminating the unneeded COPY is the win. The
micro-optimizations are down in the noise.

I'm not arguing this any longer. Go on and continue microoptimizing. I
have no interest in spending my own life that way.

Perry