Very Computer

Please, don't understand me wrong. I don't want to start a war. I
simply want to speed up things on my computer.

My computer is a AMD K6 200Mhz with 48MB (96MB swap) and a on board
SIS 5597 graphic chip.

Performance was okay when running SuSE Linux 5.3. But I didn't like
the devellopment of Linux with complete changes coming quite often
(modules change from 2.0 to 2.2; libc->glibc2.0->glibc2.1 ...). In my
mind an OS should just be installed once, then some patches from time
to time and everything should work for at least 5 years. (I'm Solaris
administrator and am used to keep computers and OS'es for such a long
time) In the last 5 years I had to reinstall Linux 3 times, and now
had to do it again and thought: NO, I'm trying something else.

I heard a lot of good thing about FreeBSD (IDE, SCSI implementation,
TCP/IP perfromance ...) I started installing FreeBSD 4.0 and it was
quite easy to handle (I loved the ppp installation. Never been so
easy!)

But performance is low! Trying nbench, it is 10% less than linux (on
an unloaded machine, which my PC most oftenly is). This didn't bother
me, whats 10%. Disk access is now much faster (DMA wasn't supported
with my linux-system) thats great. Booting time increased with
FreeBSD, thats great to.

But working with it is painful: Though I have the same Xfree version
(3.3.6) X is remarkably slower. I can see windows disappear: first the
become completely gray, than background becomes visible from top to
bottom.*(my mostly used program besides emacs) is a pain
now. System-load increased during Xdvi enormously (up to
80%). Printing (apsfilter) takes horrible long (again, enormous system
load). and it's not only the creation of new fonts.

Maybe, I missconfigured my kernel. Is i586 right for AMD K6. I applied
all points from LINT concerning AMD K6.

What else can I do to speed up things?

Heiko Klein

Quote:>Please, don't understand me wrong. I don't want to start a war. I
>simply want to speed up things on my computer.

>My computer is a AMD K6 200Mhz with 48MB (96MB swap) and a on board
>SIS 5597 graphic chip.

>Performance was okay when running SuSE Linux 5.3. But I didn't like
>the devellopment of Linux with complete changes coming quite often
>(modules change from 2.0 to 2.2; libc->glibc2.0->glibc2.1 ...). In my
>mind an OS should just be installed once, then some patches from time
>to time and everything should work for at least 5 years. (I'm Solaris

This is not the case with FreeBSD either. We had changes just as Linux.
Maybe it was a bit smoother, but probably only a bit.
We changed a.out->ELF, gcc->egcs->gcc, SCSI->CAM (with renaming sd->da),
IDE->ATA (with renaming wd->ad), and certainly more.
Of course, having a running FreeBSD-2.1 does not need to be changed,
but I am sure you would not be happy to install new ports on this
system.

Quote:>administrator and am used to keep computers and OS'es for such a long
>time) In the last 5 years I had to reinstall Linux 3 times, and now
>had to do it again and thought: NO, I'm trying something else.

>But performance is low! Trying nbench, it is 10% less than linux (on
>an unloaded machine, which my PC most oftenly is). This didn't bother

I would not call 10% less significant. This is possibly just an
artefact of the benchmark used. Or is it 10% of Linux's performance?

Quote:>me, whats 10%. Disk access is now much faster (DMA wasn't supported
>with my linux-system) thats great. Booting time increased with
>FreeBSD, thats great to.

>But working with it is painful: Though I have the same Xfree version
>(3.3.6) X is remarkably slower. I can see windows disappear: first the

Did you check the transport? ASFAIS the only serious problem you
are reporting is X11 related. X11 can be transported over local or
over tcp/ip, depending on the DISPLAY env variable.

Quote:>become completely gray, than background becomes visible from top to
>bottom.*(my mostly used program besides emacs) is a pain

As TeX is very compute-bound, this is STRANGE:

Quote:>now. System-load increased during Xdvi enormously (up to
>80%). Printing (apsfilter) takes horrible long (again, enormous system
>load). and it's not only the creation of new fonts.

>Maybe, I missconfigured my kernel. Is i586 right for AMD K6. I applied
>all points from LINT concerning AMD K6.

I dont expect big performance hits from adding i686 (which is probably
the correct conf).
--
Dipl.-Math. Wilhelm Bernhard Kloke
Institut fuer Arbeitsphysiologie an der Universitaet Dortmund
Ardeystrasse 67, D-44139 Dortmund, Tel. 0231-1084-257

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

well i have performance problems with DISK I/O on freebsd. i have two
identical machines. one is running linux, second is running
4.1-STABLE. however, linux works with disks about 50% faster. dmesg
reports using udma 66 on fbsd, on linux only udma33...

i tested on partitions mounted async, sync+softupdates,
sync+softupdates+noatime....

with all tests except sync mount, linux is much faster than freebsd.
is ide driver for freebsd slow or i have a problem with myself? ... i
actually hate linux, so don't tell me, that i'm starting holy war.

best regards, brane

-----BEGIN PGP SIGNATURE-----
Version: PGPfreeware 6.5.3 for non-commercial use <http://www.pgp.com>

iQA/AwUBOZKIQe9q7Ci83GJUEQJHygCfV3qmi5oSDf2UI+ygbgEGlqQMf1cAn1OO
gzOPwNfyoK3FrcAVCJkIPN7v
=CHig
-----END PGP SIGNATURE-----

Quote:>-----BEGIN PGP SIGNED MESSAGE-----
>Hash: SHA1

>well i have performance problems with DISK I/O on freebsd. i have two
>identical machines. one is running linux, second is running
>4.1-STABLE. however, linux works with disks about 50% faster. dmesg
>reports using udma 66 on fbsd, on linux only udma33...

>i tested on partitions mounted async, sync+softupdates,
>sync+softupdates+noatime....

>with all tests except sync mount, linux is much faster than freebsd.
>is ide driver for freebsd slow or i have a problem with myself? ... i
>actually hate linux, so don't tell me, that i'm starting holy war.

It is not a good idea to hate Linux. Did you try to disable UDMA66
which may be crappy on your hardware, via sysctl or kernel config?

The original post, though, was mostly satisfied with FreeBSD's disk
performance.
--
Dipl.-Math. Wilhelm Bernhard Kloke
Institut fuer Arbeitsphysiologie an der Universitaet Dortmund
Ardeystrasse 67, D-44139 Dortmund, Tel. 0231-1084-257

Wilhelm B. Kloke <w...@yorikke.arb-phys.uni-dortmund.de> wrote:

Not to mention the recent upgrade of binutils which induced hundreds of
messages on the mailing list!!

X is exactly the same in Linux and FreeBSD, so try to check your XF86config
file with a Linux one.

In my experience FreeBSD has always been faster than Linux (yes i know that
Linux is supposed to have less latency than FreeBSD when forking new
processes, but my experience on normal desktop usage is notably better
responsiveness with FreeBSD). In particular i am here writing a math book on my
laptop. For 378 pages (A4, and with small margins) full of formulas and
figures, it takes 8 seconds compiling. For Xdvi it is instantaneous. The
laptop is a Celeron 400 with 64 Megs memory, and it is heavily (and stupidly)
loaded with Kde and the whole stuff it uses to manage screen, sound and so on.

This is abnormal. The only time i have seen load becoming large is when stupid
kde died living processes floating around eating all CPU. Check carefully if
you don't have netscapes kaudioservers and similar floating around without
control.

I cannot help for specific K6 stuff, but i know a lot of people use them and
are happy with them. Here is my config file, which seems to work OK.

machine         i386
cpu             I686_CPU
ident           ROSE
maxusers        64

#makeoptions    DEBUG=-g                #Build kernel with gdb(1) debug symbols

options         INET                    #InterNETworking
options         INET6                   #IPv6 communications protocols
options         FFS                     #Berkeley Fast Filesystem
options         FFS_ROOT                #FFS usable as root device [keep this!]
options         MFS                     #Memory Filesystem
options         PROCFS                  #Process filesystem
options         COMPAT_43               #Compatible with BSD 4.3 [KEEP THIS!]
options         SCSI_DELAY=5000         #Delay (in ms) before probing SCSI
options         UCONSOLE                #Allow users to grab the console
options         USERCONFIG              #boot -c editor
options         VISUAL_USERCONFIG       #visual boot -c editor
options         KTRACE                  #ktrace(1) support
options         SYSVSHM                 #SYSV-style shared memory
options         SYSVMSG                 #SYSV-style message queues
options         SYSVSEM                 #SYSV-style semaphores
options         P1003_1B                #Posix P1003_1B real-time extentions
options         _KPOSIX_PRIORITY_SCHEDULING
options         _KPOSIX_VERSION=199309L
options         ICMP_BANDLIM            #Rate limit bad replies
options         INCLUDE_CONFIG_FILE     # Include this file in kernel
options         SOFTUPDATES
options         "NO_F00F_HACK"
options         USER_LDT                #allow user-level control of i386 ldt
options         PQ_NORMALCACHE          # color for 256k/16k cache

device          isa
device          pci

# Floppy drives
device          fdc0    at isa? port IO_FD1 irq 6 drq 2
device          fd0     at fdc0 drive 0

# ATA and ATAPI devices
device          ata
device          atadisk                 # ATA disk drives
device          atapicd                 # ATAPI CDROM drives
options         ATA_STATIC_ID           #Static device numbering
#options        ATA_ENABLE_ATAPI_DMA    #Enable DMA on ATAPI devices

# SCSI peripherals
device          scbus           # SCSI bus (required)
device          da              # Direct Access (disks)
device          pass            # Passthrough device (direct SCSI access)

# atkbdc0 controls both the keyboard and the PS/2 mouse
device          atkbdc0 at isa? port IO_KBD
device          atkbd0  at atkbdc? irq 1
device          psm0    at atkbdc? irq 12

device          vga0    at isa?

# splash screen/screen saver
pseudo-device   splash

# syscons is the default console driver, resembling an SCO console
device          sc0     at isa?

# Floating point support - do not disable.
device          npx0    at nexus? port IO_NPX irq 13

# Power management support (see LINT for more options)
device          apm0    at nexus? #disable flags 0x20 # Advanced Power Management

# PCCARD (PCMCIA) support
device          card
device          pcic0   at isa? irq 9 port 0x3e0 iomem 0xd4000
#device         pcic0   at isa?

# You may need to reset all pccards after resuming
options         PCIC_RESUME_RESET       # reset after resume

# Serial (COM) ports
device          sio0    at isa? port IO_COM1 flags 0x10 irq 4

# Parallel port
device          ppc0    at isa? irq 7 flags 0x4
device          ppbus           # Parallel port bus (required)
device          lpt             # Printer
device          ppi             # Parallel port interface device

device          ed0     at isa? port 0x280 irq 10 iomem 0xd8000

# Pseudo devices - the number indicates how many units to allocated.
pseudo-device   loop            # Network loopback
pseudo-device   ether           # Ethernet support
pseudo-device   tun             # Packet tunnel.
pseudo-device   pty             # Pseudo-ttys (telnet etc)
pseudo-device   vn              # Vnode driver (turns a file into a device)

# The `bpf' pseudo-device enables the Berkeley Packet Filter.
# Be aware of the administrative consequences of enabling this!
pseudo-device   bpf     3       #Berkeley packet filter

# Sound card
device          pcm0 at isa? irq 5 drq 0 flags 0x11

# USB support

device uhci
device ohci
device usb
device ugen
device uhid
device ukbd
device ums

--
Michel Talon

The SiS 5597 can be a pain sometimes.

Quote:> But working with it is painful: Though I have the same Xfree version
> (3.3.6) X is remarkably slower. I can see windows disappear: first the
> become completely gray, than background becomes visible from top to
> bottom.

I think there was a problem with XFree86 3.3.6, or was it 3.3.7?
Anyways, the SiS never really DID work very well with these versions
of XFree86.  I'd download the latest XFree86 4.0.1 source code, and
compile it.  If you want, I can give you the host.def file I used to
compile it. Beware, though -- I changed the compiler opt. flags to
-march=pentium -O3 -pipe, and -march=pentium may not work on the AMD.

It may be the SiS 5597, as I've pointed out.  I'd upgrade to XFree86
4.0.1, and see if that makes a difference.

--
- Donn

-----= Posted via Newsfeeds.Com, Uncensored Usenet News =-----
http://www.newsfeeds.com - The #1 Newsgroup Service in the World!
-----==  Over 80,000 Newsgroups - 16 Different Servers! =-----

do you have the pcm0: pci error msg and udma write error msg with FreeBSD 4? I don't
know how to fix this
problem.
I am using gigabyte bx2000+ wiht Pentinum III cpu 667.

Sam

I can confirm what D. Miller says. I have helped a friend to install
RedHat Linux (6.1) on a box with a SiS (6326) graphic chip, and X plainly did not
work. We had to upgrade to a more recent version of X than what shipped in the
distro. If you want a good well supported card under X, take a Matrox.

--
Michel Talon

Quote:> This is not the case with FreeBSD either. We had changes just as Linux.
> Maybe it was a bit smoother, but probably only a bit.
> We changed a.out->ELF, gcc->egcs->gcc, SCSI->CAM (with renaming sd->da),
> IDE->ATA (with renaming wd->ad), and certainly more.
> Of course, having a running FreeBSD-2.1 does not need to be changed,
> but I am sure you would not be happy to install new ports on this
> system.

Hm, that doesn't sound excellent, but I still will go on trying.

Quote:> >But performance is low! Trying nbench, it is 10% less than linux (on
> >an unloaded machine, which my PC most oftenly is). This didn't bother

> I would not call 10% less significant. This is possibly just an
> artefact of the benchmark used. Or is it 10% of Linux's performance?

Maybe it is the benchmark, comparing a UltraSparc166/SunOS with a
PentiumPro200/Linux, the Linux machine always seemed to be faster. (So
it wasn't, considering several other programs)

Quote:> >But working with it is painful: Though I have the same Xfree version
> >(3.3.6) X is remarkably slower. I can see windows disappear: first the

> Did you check the transport? ASFAIS the only serious problem you
> are reporting is X11 related. X11 can be transported over local or
> over tcp/ip, depending on the DISPLAY env variable.

The DISPLAY environment variable must be set by XFree itself. I'm
using fvwm and I'm starting emacs with fvwm. I never set the DISPLAY.
Maybe you can point out where to look?

XFree configuration is the same as under Linux. There aren't a lot of
parameters I could include for my graphics adapter, and I included all
the same. Resolution, colordepth etc. is the same, too.

Some told me to try XFree86 4.0.1. But it doesn't support the SiS5597!

Thanks for your help,

Heiko Klein

Hi,

I think I could fix the problem:

At first I tried to compile the kernel for my AMD K6 with i686
support, a hint I got here. This enable me to learn a lot about the
booting process of FreeBSD ;-) , because I had to load the generic
kernel again.

Then I compiled it using i486 586 and 686 without any special
AMD-options. And that speed up a lot of things (previous compilation
of kernel used ~ 2000s, now only 1000s) X running fast again.

Then I even tried to speed things up even more: using optimization
(-O3 and -m486) which improved FP benchmarks about 20%.

Then I switched of the posix-real-time scheduling (what is it good
for?). Which gave another percent (though now improvements are less
then variations of the benchmark-tests).

But at least, I'm at the point where I wanted to get: A system as fast
as it was before (with linux) (plus DMA suport). That means I will
stick some more time with *BSD just to learn something different.

Heiko

PS: Sometimes, system-load is quite high (for example when compiling
the kernel: 750s user, 250s system). This is quite a lot, though
Solaris has a high system-load, too. System load of Linux (everything
as shown by top) never raises higher than 5%.

On 14 Aug 2000 10:42:42 +0200, Heiko Klein

I think posix-real-time scheduling is a proces that schedules the
processes (which process is allowed to use the cpu at which time
and for how long); the schedule algorithm is posix 'standard' and
made for real time systems, for example think about an ABS system
in a car which has to be real time (otherwise you will get a really
strange car accident ;-)

Hope i'm right ;-)

Martin

--
Linux Gebruikers Handleiding v1.2 : http://2mypage.cjb.net
Linux RedHat 6.1 Kernel 2.2.16  Toshiba P233 MHz, 32 Mb RAM
2:40pm up 34 min, 3 users, load average: 0.87, 0.38, 0.18
Western Civilization, that would be a good idea!