Re: [Valgrind-users] cpu cycles

[Nicholas N. <nj...@ca...>]

On Mon, 28 Jul 2003, Joerg Beyer wrote:

> but after all, the instruction counter still is a resonable
> value if you want to verify that some source code optimization
> works as you expected, right?

Most of the time, probably, but fewer instructions doesn't always mean
fewer cycles.  I would use just 'time' or cycles (eg. counted with
performance counters, via Rabbit or similar tool) to evaluate an
optimisation.  Although that only gives you results on your dev machine.
That's one of the biggest problems with all this complicated hardware.

N