[PATCH] Always read from /dev/urandom
Jeremy Allison
jra at samba.org
Mon Oct 12 20:22:08 UTC 2015
On Mon, Oct 12, 2015 at 10:18:42PM +0200, Volker Lendecke wrote:
> On Mon, Oct 12, 2015 at 01:06:59PM -0700, Jeremy Allison wrote:
> > I just did the cachegrind tests. Wow. On bin/genrandperf 1000000
> > there's a factor of *120* difference in instructions executed.
>
> Well, that factor is not the whole story. The kernel has to do
> *something*. But I could imagine that the kernel rng has seen more tuning
> than ours. The other factor really is the complexity on our side.
Yeah I know, it's only measuring the userspace, not
kernel space time. But still, 120x less userspace code
is certainly not to be sneezed at :-).
More information about the samba-technical
mailing list