>
> On Sun, 16 Nov 1997, Steven N. Hirsch wrote:
>
> > If there are supposed to be enormous performance improvements, I sure
> > haven't seen them. Midnight Commander, which stats out an entire
> > directory on a chdir, takes about 3x longer to gulp up a large directory
> > under 2.1.63 (on a P150+ 6x86) than does 2.0.32 (on a 486/100, yet). This
> > is not just the first time - but _every_ time. Clearly no win on this
> > point...
>
> i'm using both latest 2.0 and 2.1 kernels, and 2.1.64 wins by a great
> margin. The biggest 'human visible' speedup (for me that is) was
> introduced when the dentries concept was added around 2.1.4x. Eg. startup
> times of 'mc', 'pine', and other tools shortened greatly, things feel much
> smoother.
>
> so if you see a 3 times slowdown, could you try either kernel profiling,
> or ktrace to see where the overhead comes from. Or, much better, produce
> some (short, ugly, doesnt matter) code/script that 'executes much faster
> on 2.0 than on 2.1', so we can reproduce+profile+trace it. It doesnt
> matter how 'corner case' that application is, if it's causing RL problems,
> lets speed it up ;) Thanks,
Relative to my original report above, both machines have 32MB of ram
installed. It looks like using mc's latency in stat-ing a large directory
would be a good test case. I'll run an strace to see what it's doing
system call-wise. The kernel tracing is a bit beyond my experience -
sorry. Once I've analyzed what mc is doing, I'll try to build a test
program to emulate it.
Have _you_ made such a comparison? I have found that /usr/man/man3 is a
great target for the test. Mine is > 70k and has ~2200 files in it.
Steve