2.6 snapshot, sparc + ssh little issues
Diana Eichert
installed fine, that was once I figured out how to make a floppy on my
i386 system that would boot on the sparc. Just dd'ng the floppy image to
a blank floppy didn't work, the sparc kept complaining about the lack of
disklabel, so i disklabled it, no biggy.
i did see one thing already, with the new ssh the "UsePrivilegedPort no"
keyword in the ssh_config errors out the ssh client. You have to remove
it and run ssh with the "-P" switch. Also scp no longer has the ability
to use non-privileged ports as the "-L" switch is not supported.
BTW, just how slow is an LX, what would I relate it to in the world of
Intel processors?
diana
Diana Eichert
VP Technical Services
Nothing in Particular at the Moment, Inc.
deic...@wrench.com
Markus Friedl
> i did see one thing already, with the new ssh the "UsePrivilegedPort no"
> keyword in the ssh_config errors out the ssh client.
i just commited a patch. my typo .
> You have to remove
> it and run ssh with the "-P" switch. Also scp no longer has the ability
> to use non-privileged ports as the "-L" switch is not supported.
workaround:
instead of
% scp -L file foo:/tmp
add an entry to .ssh/config
% cat ~/.ssh/config
host foox
hostname foo
rhostsauthentication no
rhostsRSAauthentication no
% scp file foox:/tmp
this is what 'scp -L' or 'ssh -P' does.
-markus
Chris Cappuccio
on this chip. I think the more accurate result is 33333333 bytes/second
The first time I run md5, I get
dqc$ md5 -t
MD5 time trial. Digesting 10000 10000-byte blocks ... done
Digest = 52e5f9c9e6f656f3e1800dfa5579d089
Time = 3 seconds
Speed = 33333333 bytes/second
The second time I run it,
dqc$ md5 -t
MD5 time trial. Digesting 10000 10000-byte blocks ... done
Digest = 52e5f9c9e6f656f3e1800dfa5579d089
Time = 2 seconds
Speed = 50000000 bytes/second
If I keep running it, I keep getting 50000000 bytes/second (which is what you
report, 47M, or 50000000/1024/1024)
If I run md5 on something else and run md5 again, I get what I got when I
first ran it..
dqc$ md5 -t
MD5 time trial. Digesting 10000 10000-byte blocks ... done
Digest = 52e5f9c9e6f656f3e1800dfa5579d089
Time = 3 seconds
Speed = 33333333 bytes/second
Finally,
If I run md5 -t on a K6-3-400, I get
passiflora$ md5 -t
MD5 time trial. Digesting 10000 10000-byte blocks ... done
Digest = 52e5f9c9e6f656f3e1800dfa5579d089
Time = 3 seconds
Speed = 33333333 bytes/second
on every try....
this k6-3-400 seems to run gcc as fast as my p3-450
On Fri, 15 Oct 1999, J.D. Carlson wrote:
| What I have used for a simple comparison is 'md5 -t'
|
| It is quick, easy, and handy.
|
| On my machines I get these sorts of numbers:
|
| Pentium III-450Mhz 47.3M bytes/sec
| Pentium -200Mhz 15.3M
| Sparc20 10 M
| Sparc4 3 M
| 486 50Mhz 3 M
|
| That is just for processor comparison.
|
| for disk io 'dd if=/dev/zero of=/tmp/junk bs=1024 count=10000' for writes
| (be sure /tmp has enough room)
| and after that 'dd if=/tmp/junk of=/dev/null bs=1024 count-10000' for
| reads
|
| on the same machines: Write Read
| Pentium III-450Mhz 9.2M 12.2M
| Pentium -200Mhz 3.2M 6.2M (scsi disks)
| Sparc20 2.2M 2.2M
|
|
| J.D. Carlson, UCHSC
|
| On Fri, 15 Oct 1999, Diana Eichert wrote:
|
| >
| > BTW, just how slow is an LX, what would I relate it to in the world of
| > Intel processors?
| >
|
|
---
I dress like a pimp
I walk with a limp
I see the Browns for free in
My low-rider blimp.
Chris Cappuccio
between 333333333 and 500000000
the best way to benchmark your boxes is with lmbench!!
it's in the ports collection...
Diana Eichert
diana
Diana Eichert
Diana Eichert
diana
On Fri, 15 Oct 1999, J.D. Carlson wrote:
> What I have used for a simple comparison is 'md5 -t'
>
> It is quick, easy, and handy.
>
> On my machines I get these sorts of numbers:
>
> Pentium III-450Mhz 47.3M bytes/sec
> Pentium -200Mhz 15.3M
> Sparc20 10 M
> Sparc4 3 M
> 486 50Mhz 3 M
>
> That is just for processor comparison.
>
> for disk io 'dd if=/dev/zero of=/tmp/junk bs=1024 count=10000' for writes
> (be sure /tmp has enough room)
> and after that 'dd if=/tmp/junk of=/dev/null bs=1024 count-10000' for
> reads
>
> on the same machines: Write Read
> Pentium III-450Mhz 9.2M 12.2M
> Pentium -200Mhz 3.2M 6.2M (scsi disks)
> Sparc20 2.2M 2.2M
>
>
> J.D. Carlson, UCHSC
>
> On Fri, 15 Oct 1999, Diana Eichert wrote:
>
> >
> > BTW, just how slow is an LX, what would I relate it to in the world of
> > Intel processors?
> >
>
>
Diana Eichert
J.D. Carlson
it computes. Number of bytes divided by whole seconds. I actually
modified md5.c to:
#define TEST_BLOCK_LEN 100000
from
#define TEST_BLOCK_LEN 10000
to allow a longer test with some better resolution. I seem to remember
from the distant past that a valid statistics test required 12 or more
samples, so in a 3 second test it doesn't pass rigour.
In fact I found the modified md5 and ran it again:
the original:
$ /bin/md5 -t
MD5 time trial. Digesting 10000 10000-byte blocks ... done
Digest = 52e5f9c9e6f656f3e1800dfa5579d089
Time = 2 seconds
Speed = 50000000 bytes/second
the modified:
$./md5 -t
MD5 time trial. Digesting 100000 10000-byte blocks ... done
Digest = d48d98b2a2d8bc7619a0045aec866cac
Time = 25 seconds
Speed = 40000000 bytes/second
On a slow 486 it took a lot longer to complete the test. What would have
been better was to modify it so md5 -tt ran it longer and -ttt even
longer. I didn't go that far. Then new snapshots, and I didn't recompile
md5.c to the new value for TEST_BLOCK_LEN. Thanks for pointing that out.
On the other hand, any value on a P3 makes old sparcs look pretty tired.
I ran the old byte benchmarks, haven't tried lmbench, looking around I
noticed one site said the best test is to time kernel compiles. I found
that md5 correlated well with the byte benchmarks for cpu.
Like I said at the outset, quick, easy, handy. True? take it with a grain
of salt.... ;-)
J.D.
> On Fri, 15 Oct 1999, J.D. Carlson wrote:
>
> | What I have used for a simple comparison is 'md5 -t'
> |
> | It is quick, easy, and handy.
> |
> | On my machines I get these sorts of numbers:
> |
> | Pentium III-450Mhz 47.3M bytes/sec
> | Pentium -200Mhz 15.3M
> | Sparc20 10 M
> | Sparc4 3 M
> | 486 50Mhz 3 M
> |
> | That is just for processor comparison.
> |
> | for disk io 'dd if=/dev/zero of=/tmp/junk bs=1024 count=10000' for writes
> | (be sure /tmp has enough room)
> | and after that 'dd if=/tmp/junk of=/dev/null bs=1024 count-10000' for
> | reads
> |
> | on the same machines: Write Read
> | Pentium III-450Mhz 9.2M 12.2M
> | Pentium -200Mhz 3.2M 6.2M (scsi disks)
> | Sparc20 2.2M 2.2M
> |
> |
> | J.D. Carlson, UCHSC
> |
> | On Fri, 15 Oct 1999, Diana Eichert wrote:
> |
> | >
> | > BTW, just how slow is an LX, what would I relate it to in the world of
> | > Intel processors?
> | >
> |
> |
>
J.D. Carlson
Christian Weisgerber
> What I have used for a simple comparison is 'md5 -t'
I'm not sure this has any real world value, but...
> Pentium III-450Mhz 47.3M bytes/sec
> Pentium -200Mhz 15.3M
> Sparc20 10 M
> Sparc4 3 M
> 486 50Mhz 3 M
WinChip W2A-233 11.1M bytes/s (FreeBSD, 100000 blocks)
Sparc IPC 1.1M bytes/s
If I squint real hard I can see the cogwheels turning inside the
IPC...
--
Christian "naddy" Weisgerber na...@mips.rhein-neckar.de
Marc Schneiders
> J.D. Carlson <j...@noc7.uchsc.edu> wrote:
>
> > What I have used for a simple comparison is 'md5 -t'
>
> I'm not sure this has any real world value, but...
>
> > Pentium III-450Mhz 47.3M bytes/sec
> > Pentium -200Mhz 15.3M
> > Sparc20 10 M
> > Sparc4 3 M
> > 486 50Mhz 3 M
>
> WinChip W2A-233 11.1M bytes/s (FreeBSD, 100000 blocks)
> Sparc IPC 1.1M bytes/s
>
> If I squint real hard I can see the cogwheels turning inside the
> IPC...
AMD K6-233 OpenBSD 2.6 beta 25.000.000
Another The same The same
P II 300 OpenBSD 2.5 16.666.666
Dual PPro 180@200 MHz FreeBSD-current 13.157.894 /
13.333.333
386 DX 25 FreeBSD 3.1 1.000.000
486 DX 66 NetBSD 1.4 (snap July) 1.000.000
Sparc IPC OpenBSD 2.6 beta 377.358 /
574.712 /
578.034
Sun 3/80 (diskless) NetBSD 1.4 (snap May) 200.000
--
Marc Schneiders ||
||
ma...@venster.nl || Null message body;
ma...@oldserver.demon.nl || hope that's ok
Markus Friedl
> thanks for the patch and workaraound
hi! i need feedback on ssh. please test and mail any problem!
thx, -markus
chuck
> I just did my first sparc install on an ancient LX and everything
> installed fine, that was once I figured out how to make a floppy on my
> i386 system that would boot on the sparc. Just dd'ng the floppy image to
> a blank floppy didn't work, the sparc kept complaining about the lack of
> disklabel, so i disklabled it, no biggy.
No, no. Ancient is the Sparc 1 with 16MB that I have here. Actually,
it's being salvaged to heal a Sparc 2 that is my firewall.
The LX has, I believe, a 75MHz MicroSparc chip (poor FP perf). It's
certainly quick enough to run as a firewall for T1 and below. It's
an adequate desktop (I used Sparc 2 for that cause I have a 21"
monitor that will ONLY work with a Sun. I'm now down to a 17", but
I get 24bit color.
Truly ancient is the Sun 3 in the garage, or the Motorola "mini" that
has an 18"x18" board holding a whole Meg of RAM!
Ancient is not something from the last 5-6 years. that's just old and
cheap. BTW, if this hasn't been plugged in for a while, expect the NVRAM
chip to lose it's NV-ness soon. I've lost three from 1992 machines that
spent a year in storage.
> BTW, just how slow is an LX, what would I relate it to in the world of
> Intel processors?
I believe it's around 60-70 MIPS, which would put it near a Pentium/75.
The Sparc 2 is 30 MIPS, the Sparc 10/75 kicked up around 80.
FP on SPARC is better than same on CISC chips of same Integer speed.
I/O on the SPARCs was pretty poor then, though SBUS was better than ISA
and you don't have to deal with the brain deadness of interrupts and
such that Intel developed (oh why didn't IBM just wait for the 68k!?)
Segments are for worms.
chuck
jdcar...@gmail.com
> I just did my first sparc install on an ancient LX and everything
> installed fine, that was once I figured out how to make a floppy on my
> i386 system that would boot on the sparc. Just dd'ng the floppy image to
> a blank floppy didn't work, the sparc kept complaining about the lack of
> disklabel, so i disklabled it, no biggy.
>
> to use non-privileged ports as the "-L" switch is not supported.
>
> Intel processors?
>
>
> Diana Eichert
> VP Technical Services
> Nothing in Particular at the Moment, Inc.
> deic...@wrench.com
> I just did my first sparc install on an ancient LX and everything
> installed fine, that was once I figured out how to make a floppy on my
> i386 system that would boot on the sparc. Just dd'ng the floppy image to
> a blank floppy didn't work, the sparc kept complaining about the lack of
> disklabel, so i disklabled it, no biggy.
>
> to use non-privileged ports as the "-L" switch is not supported.
>
> Intel processors?
>