Puppy 4.2 smp

[aragon]

This is a little testbed. It is the standard 4.2 release with an smp-enabled 2.6.25.16 kernel and some bugfixes.

Please look here http://www.murga-linux.com/puppy/viewtopic.php?t=41678 for an updated testversion using kernel 2.6.29.1!!!

NOTES:
- This is no official puppy-release!
- This is no bugfix-release of the official puppy-release! it just includes the bugfixes from the bugfixes-thread!
- This should be seen as an experimental build, as it includes a not heavy tested kernel!
- If you got no dualcore, you won't benefit from the used kernel
- I've tested it (not extensive) from live-cd with my Lenovo Dualcore notebook and everything seems to work.

KERNEL:
The smp-kernel could be found here: http://www.murga-linux.com/puppy/viewtopic.php?p=253482 (thanks to jcoder24)

INCLUDED PATCHES:
from: http://www.murga-linux.com/puppy/viewtopic.php?t=40537):
- network-wizard-apr-1st.tar.gz
- xsane-0.996-NoGimp.pet
- jwmtraypatch_update.pet
- No_bcm43xx_patch.pet
- mountcifs-1.5.pet
- nbtscan-1.5.1a-james-patched.pet
- remax-4.2.pet
- abiword-2.6.3-updated.pet
- rox-docx.pet
- mtpaintsnapshot is changed to the one dejan555 provided (from http://www.murga-linux.com/puppy/viewtopic.php?t=38165) as the (old) included one seems to have some issues.

LOOK:
As nothing is changed in look and feel you can look at the official announcement for screenshots (http://www.murga-linux.com/puppy/viewtopic.php?t=40443)

DOWNLOAD
Iso is 101 mb and hosted by MU (thanks a lot):
URL: http://puppyisos.org/isos/2009-01-to-06 ... mp-aragon/
User: puppy
password:linux

md5-checksum is

Code: Select all

f8a450628d3b38572d0fab459e145a31

COMPILING:
If you want to compile apps and libs (NOT kernel-modules) use the devx file for 4.2 (devx_420.sfs) from here: http://distro.ibiblio.org/pub/linux/dis ... modules-4/

KERNEL-SOURCE:
With some help and interation of/with MU and DC (thank you both) i was able to build a kernel-source package. Download see iso. Size is 131 ~mb

md5-checksum is

Code: Select all

3871b0b1960a741719fa2a8f0e3845f7

Notes on kernel-Src:
- This is NOT the original kernel-source. I've pm-ed jcoder24 for the original one, maybe he could post the original. This one is build with MU's packages and 'make oldconfig && make prepare && make' on his packages.
- Tested only (for now) with vwware-workstation, but successfully. So i would call this a test-release.
- Size is big, i know. But as only this has worked for me on my test (and i've tested some different), i'm uploading this one.

TIPS:
1. Suspend
On my laptop i'm using acpitool with

Code: Select all

acpitool -s

to suspend. Download: http://www.murga-linux.com/puppy/viewtopic.php?t=31753

2. Cool & quiet (frequency-scaling)
With the tips from http://www.murga-linux.com/puppy/viewtopic.php?t=28443 i was able to run puppy with reduced fan-action => reduced noise.

BUGS & FIXES:
1. Nosmp after install
The universal installer ads a 'nosmp'-command to some boot-configurations (syslinux.cfg, extlinux.cfg,...). After installation please check the appropriate file for your installation-type. If there is a 'nosmp' delete 'nosmp', save the file and reboot. Should work then.

or

- download attached puppyinstaller
- backup /usr/sbin/puppyinstaller
- extract attached to /usr/sbin

2. Pwidget Cpu_Ram shows only one core
a. open

Code: Select all

/root/.pwigets/config/Cpu_Ram

b. change line

Code: Select all

color0}C P U ${color3}${cpubar 10,112}

to

Code: Select all

${color0}CPU 1 ${color3} ${cpubar cpu1 10,105}
${color0}CPU 2 ${color3} ${cpubar cpu2 10,105}

c. save
d. restart jwm

3. Loading a wrong puppy_420.sfs
if you have another 420-install on your boot drive, it is possible that the wrong sfs is picked up on boot. you will notice that/if something is wrong by some crazy behaviour. if so, there are some 'fixes/workarounds' for that:
- Rename the second one (temporarely).
- If you got a hd-install, use subdirs for your different installs.

4. Printing Problems
Some 4.2-user have printing-problems. DaveS provided this link that helped him to solve his printing-problems http://www.murga-linux.com/puppy/viewtopic.php?t=40665

-----------
aragon

[Lobster]

Thanks very much Aragon.

I have a dual core AMD.
No idea what to expect will be faster?

Anyway downloaded OK and am running 4..2 smp . . .
Rather than use 'puppy pfix=ram' on bootup (recommended)
I went straight into boot up using the previous 4.2 sfs
- no problems

Would this be a step towards a 64bit Puppy - isn't that something also enabled in the kernel compiled or am I confused.com?

Thanks again Aragon

Puppy runs on
AMD Athlon(tm) 64 X2 Dual Core Processor 5000+ 2593.69MHz

- of course he does

[Caneri]

Hi aragon,

I wonder if this SMP enabled version and updates can find it's way into the Git build tree...thanks, as I look for the dual core enabled versions to take advantage of dual core on my kit.

Eric.

[aragon]

hi lobster,

what should be expected?

- better performance in running multiple programs at a time
- better performance in running smp-optimized programs
- ...??

an example for the second would be the creation of large sfs-files, decoding + encoding media.

Regarding 64 bit
if you're back from confused.com you can visit me at the isle of the clueless

aragon

ps: maybe try
http://en.wikipedia.org/wiki/Symmetric_multiprocessing
http://en.wikipedia.org/wiki/64bit

[aragon]

hi eric,

for this one, i don't think that we should do that. it's just an unofficial quick shot. 4.2 is closed and i don't know much about the kernel-package i used (beside of, that i'm using it for about half a year at home).

there are actually some discussions, plans and actions in the kennel - 5 thread http://www.murga-linux.com/puppy/viewtopic.php?t=40488 to build a 'newer', 'more modern' configured standard-kernel for the/a future puppy-version. big_bass is the man who has accepted the challenge, so he would be the one to contact.

generally we should use the git-rep for future releases.

aragon

[MU]

with 2 cores, you can feel 2 advantages.

1.) apps that support it, like "mksquashfs".
This is seldom.
Building a huge sfs with this tool can take an hour or so.
With the seond core, it needs just 30 minutes.

2.) no more "blocking".
The kernel itself "manages" the applications.
If you run mplayer to watch a video on only 1 core, all other applications might react very slow, as mplayer will use a lot of the cpu resources.
If you have 2 cores, mplayer will use the resource of only one core, so that the other applications can use the second one.
Like this they still react snappy.

Mark

[Béèm]

Could the benchmark tools in hardinfo hardware information be a valid means of verifying if smp is working and to what extend it is effective?

[Lobster]

Thanks Mark,

That was simple and helpful info
The second blocking point will make a difference and I will try and be aware of it and stick with an smp version of Puppy if available . . .

With vala / genie if you 'multithread' will this be handled by separate processors or is this another issue?

[MU]

Threads in C are usually active as one process.
As vala is converted to C during compilation, it should inherit the thread model.

2 "programs" run beside each other, but they share a memory segment.
For example Gtkbasic internally is based on threads.
One thread runs the Basic Interpreter (that does the "actions", and a second one the Gtk mainloop (that manages the window and related events like buttonclicks).
Both can write to some of the internal C variables of Gtkbasic, so that they can exchange values.
If a button is clicked, a variable containing the button-name is set (so a value is written to a common memory segment).
This variable is read some milliseconds later by the Basic interpreter, so that it can perform the desired action.
So despite both "programs" run "side by side" internally, they are not completely independent from each other, otherwise they could not comunicate together.

So I am not certain, that the kernel could spread them to several cores.
But I am not certain here.

Mark

[MU]

Beem,

type:
cat /proc/cpuinfo

I get here:

Code: Select all

sh-3.00# cat /proc/cpuinfo 
processor	: 0
vendor_id	: AuthenticAMD
cpu family	: 15
model		: 107
model name	: AMD Athlon(tm) Dual Core Processor 5050e
stepping	: 2
cpu MHz		: 2593.906
cache size	: 512 KB
physical id	: 0
siblings	: 2
core id		: 0
cpu cores	: 2
apicid		: 0
initial apicid	: 0
fdiv_bug	: no
hlt_bug		: no
f00f_bug	: no
coma_bug	: no
fpu		: yes
fpu_exception	: yes
cpuid level	: 1
wp		: yes
flags		: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt rdtscp lm 3dnowext 3dnow pni cx16 lahf_lm cmp_legacy svm extapic cr8_legacy 3dnowprefetch
bogomips	: 5189.97
clflush size	: 64
power management: ts fid vid ttp tm stc 100mhzsteps

processor	: 1
vendor_id	: AuthenticAMD
cpu family	: 15
model		: 107
model name	: AMD Athlon(tm) Dual Core Processor 5050e
stepping	: 2
cpu MHz		: 2593.906
cache size	: 512 KB
physical id	: 0
siblings	: 2
core id		: 1
cpu cores	: 2
apicid		: 1
initial apicid	: 1
fdiv_bug	: no
hlt_bug		: no
f00f_bug	: no
coma_bug	: no
fpu		: yes
fpu_exception	: yes
cpuid level	: 1
wp		: yes
flags		: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt rdtscp lm 3dnowext 3dnow pni cx16 lahf_lm cmp_legacy svm extapic cr8_legacy 3dnowprefetch
bogomips	: 5188.00
clflush size	: 64
power management: ts fid vid ttp tm stc 100mhzsteps

So infos for both cores are displayed.


If you want to check the usage, extract the attached archive to /usr/bin/.
Then type in a console:
cpuinfo

You should see such self updating values:

Code: Select all

sh-3.00# cpuinfo
CPU: 5%   14%

I modified it to display info for 2 cores, for 4 cores, it had to be extended more.

Mark

[ttuuxxx]

This is a little testbed.

aragon

Congrads aragon looks like there will be some puplets based on this.
Great idea and Glad it works
ttuuxxx

[aragon]

hi guys,

simplest way to test is to run htop, it's in 4.2.

if you get e pic like below (see 1 + 2) both cores are running.

aragon

[Béèm]

Great idea, Aragon.
I see my two processors as well.

[Caneri]

Hi Aragon,

I installed this over G_Pup but the dual core is not active after installing to a usb stick.

What do I modprobe to try to load the smp enabled stuff?..yes, very technical term that.

Htop shows one core...grrr...any ideas.

It runs fine off the cd.

TY,
Eric

[Christopher]

Thank you
very cool

[aragon]

@ eric

i had to consult Wolf Pup, as he had the same with G_pup and solved it in 1.2.

Will report back.

aragon

[aragon]

@ eric

please control you syslinux.cfg. the universanl-installer seems to write 'nosmp' statement. if you delete 'nosmp' (please backup first) problem might be solved.

please report if this works.

i found this in the unnamed-pup thread.

aragon

[Caneri]

Hi Aragon,

Yes Wolf Pup had the the usb problem in G_Pup.

Here's my cfg..

default puppy
display boot.msg
prompt 1
timeout 80

F1 boot.msg
F2 help.msg

label puppy
kernel vmlinuz

There is no mention of "no smp" in /usr/lib/syslinux/files/syslinux.cfg

I'll try a clean usb if I can wipe one here...Thanks,
Eric

[Caneri]

OK..I found this in extlinux.conf.

default puppy
display boot.msg
prompt 1
timeout 80

F1 boot.msg
F2 help.msg

label puppy
kernel vmlinuz
append initrd=initrd.gz pmedia=usbflash nosmp

I'll edit this and reboot....Eric

[Caneri]

Ok..edited the extlinux.conf

Now it says...

default puppy
display boot.msg
prompt 1
timeout 80

F1 boot.msg
F2 help.msg

label puppy
kernel vmlinuz
append initrd=initrd.gz pmedia=usbflash


Htop shows 2 cores but pwidgets only shows one core.

Hooray...I'm getting there...I liked G_Pup but I found I like the full Puppy and the bling

Eric