> > On Mon, 28 Jul 1997, Chris Evans wrote:
> >
> > > I just compared 2.1.46 vs. 2.1.46+pentium memcpy patch, and interestingly
> > > enough found that the UNIX byte benchmarks tended to _drop_ a fair bit,
> > > with the exception of process creation and execl throughput. (Note that I
> > > only ran the basic 'system' tests - TCP bandwidth etc. to be determined
> > > when I find the newer benchmarks)
> > >
> > > This is most interesting since I used to swear by the patch.
> I'll rerun the benchmark tonight and post the results. Can anyone point me
> to the most comprehensive suite that contains all the TCP
> bandwidths/latencies etc. as well as pipes + process creation + file IO?
ok, one of the possible reasons might be that memcpy() is not inlined
anymore ... the current (bit slower) string-operation based memcpy()
inlines _very_ well. For a hot path with many get_user()'s this might show
up ... although i do not know what the real reason is, this is just a
guess.
-- mingo