On Tue, Oct 23, 2018 at 04:47:06AM +0000, Mike Larkin wrote:
> On Mon, Oct 22, 2018 at 02:52:42AM -0200, Daniel Bolgheroni wrote:
> > On Fri, Oct 19, 2018 at 04:16:51AM +0000, Mike Larkin wrote:
> > > On Thu, Oct 18, 2018 at 10:34:20PM -0300, Daniel Bolgheroni wrote:
> > > > On Wed, Oct 17, 2018 at 08:42:46PM +0000, Mike Larkin wrote:
> > > > > A 1000Hz host helps here. I get 10.32s real time on sleep 10 with
> > > > > that setting.
> > > > >
> > > > > Note that qemu behaves the same way on OpenBSD.
> > > >
> > > > OK, the output is still slow when on serial, but things improved
> > >
> > > Is the console baudrate 9600 or 115200?
> >
> > It's running at 115200.
> >
> > $ vmctl start 1 -c
> > Connected to /dev/ttyp7 (speed 115200)
>
> ^^^ if this is what you are using to determine that, I'd ask you to ensure
> that you stty com0 115200 in /etc/boot.conf and that the /etc/ttys line
> has 115200 for the console also. The baudrate from the output of the 'cu'
> used by 'vmctl console' always prints 115200 in this case, even if vmd
> is only outputting at 9600.
Yes, it was my assumption. I already had 'stty com0 115200' in /etc/boot.conf
and now adjusted /etc/ttys to 115200 where appropriate. And things improved:
9600 in /etc/ttys:
# time cat /etc/ttys
(...)
1m09.27s real 0m00.00s user 0m00.00s system
#
115200 in /etc/ttys:
# time cat /etc/ttys
0m11.66s real 0m00.00s user 0m00.00s system
#
As for comparison, I made the same test connected to a Orange Pi Zero, also at
115200, using a USB-to-serial converter uftdi(4):
# time cat /etc/ttys
0m04.03s real 0m00.00s user 0m00.00s system
#
I don't know if this is somehow related to interrupts previously discussed and,
in these cases, it's running -current snapshots, e.g. HZ=100.
Thank you.
--
db
