Compiling custom kernels for P1.1 (with fiasco-gen) - Page 16 - maemo.org

Venemo | # 151 | 2010-04-06, 16:25 | Report

Hi,

I read the first post in this thread, and a few interesting things came into my attention:

Originally Posted by

USB 1.1+2.0+generic drivers,USB hubs,USB video+serial

What is this exactly?
Is it some sort of USB host mode, or what does it do?

egoshin | # 152 | 2010-04-06, 16:38 | Report

Originally Posted by titan View Post

* with 125MHz enabled I noticed that the device spend most of the time in 600MHz

Can you show an output of

cat /sys/devices/system/cpu/cpu0/cpufreq/stats/time_in_state

while kernel has 125MHz capability?

titan | # 153 | 2010-04-06, 16:54 | Report

Originally Posted by Venemo View Post

What is this exactly?
Is it some sort of USB host mode, or what does it do?

check the USB/IP link
http://talk.maemo.org/showpost.php?p...postcount=1259
transparent access to USB devices on another computer via network

The Following 3 Users Say Thank You to titan For This Useful Post:
cproc, OVK, Venemo

hstende | # 154 | 2010-04-06, 17:37 | Report

titan I'm now using your kernel-maemo_2.6.28-maemo20 and use 125-900.
Mine is idling mostly at 125
Nokia-N900-02-8:/sys/devices/system/cpu/cpu0/cpufreq/stats# cat time_in_state
1200000 0
1100000 0
1000000 0
950000 0
900000 17693
850000 0
800000 32
750000 64
700000 175
600000 8204
550000 96
500000 50072
250000 1207
125000 53260

The 500000 was when I had the usb-cable connected,

I did a longer test. Starting with a reboot:

Nokia-N900-02-8:~# uptime
14:32:04 up 15:16, load average: 0.44, 0.14, 0.04

Nokia-N900-02-8:~# cat /sys/devices/system/cpu/cpu0/cpufreq/stats/time_in_state 1200000 0
1100000 0
1000000 0
950000 0
900000 46075
850000 361
800000 256
750000 316
700000 974
600000 36517
550000 826
500000 25800
250000 12983
125000 5377123

This is mostly idling and using wlan checking email little talking.

Last edited by hstende; 2010-04-07 at 12:33.

The Following User Says Thank You to hstende For This Useful Post:
egoshin

clooak | # 155 | 2010-04-06, 19:43 | Report

When you say Nokia pr1.2, are you talking about this sdk package has been precompiled and assembled, ready to use if i go with this kernel?

http://thenokiablog.com/2010/03/22/n...-major-update/

titan | # 156 | 2010-04-06, 21:02 | Report

Originally Posted by egoshin View Post

Can you show an output of
cat /sys/devices/system/cpu/cpu0/cpufreq/stats/time_in_state
while kernel has 125MHz capability?

after an uptime of 30mins (only idling, with Conky and Terminal open, Wifi on, ssh connection)

1200000 0
1100000 0
1000000 0
950000 0
900000 0
850000 0
800000 0
750000 0
700000 0
600000 109312
550000 62
500000 3236
250000 5452
125000 58039

i.e. 600 MHz:62,42%, 550 MHz:0,04%, 500 MHz:1,85%, 250 MHz:3,10%, 125 MHz:32,60% (1265)

egoshin | # 157 | 2010-04-06, 22:24 | Report

Originally Posted by titan View Post

after an uptime of 30mins (only idling, with Conky and Terminal open, Wifi on, ssh connection)

1200000 0
1100000 0
1000000 0
950000 0
900000 0
850000 0
800000 0
750000 0
700000 0
600000 109312
550000 62
500000 3236
250000 5452
125000 58039

i.e. 600 MHz:62,42%, 550 MHz:0,04%, 500 MHz:1,85%, 250 MHz:3,10%, 125 MHz:32,60% (1265)

So, can I suggest you another modification?

Actually - combination of Lehto fix and yours fix:

Have a table of the same size, but remove 550 and add 125

Then - on request to writing to scaling_max_frequency get ready higher frequency, but don't replace 600 yet.

And if screen is ON - replace 600 with higher frequency, if screen is OFF (it can be derived from GPIO module, I think) - put back 600.

Actually, it could have sense to limit frequency even more if screen is OFF and no pulseaudio active but that can be difficult to implement right away.

(I try to avoid writing a new scaling governor... eventually it should be written for mobile devices but today we haven't it).

Alex Atkin UK | # 158 | 2010-04-07, 02:03 | Report

Originally Posted by titan View Post

after an uptime of 30mins (only idling, with Conky and Terminal open, Wifi on, ssh connection)

1200000 0
1100000 0
1000000 0
950000 0
900000 0
850000 0
800000 0
750000 0
700000 0
600000 109312
550000 62
500000 3236
250000 5452
125000 58039

i.e. 600 MHz:62,42%, 550 MHz:0,04%, 500 MHz:1,85%, 250 MHz:3,10%, 125 MHz:32,60% (1265)

Seems you are probably right about the bug at 125Mhz then. Using the extras-devel kernel (250Mhz min, 900Mhz max) uptime 13 hours (most of which I was asleep) I got:

1200000 0
1100000 0
1000000 0
950000 0
900000 60041
850000 337
800000 242
750000 338
700000 1114
600000 253
550000 546
500000 43771
250000 4505106

That looks pretty good to me, shows its idling at 250 properly. That is idling with NOTHING open mind you, Conky I think causes some jumping around of CPU speeds due to probing so much stuff and it being graphical.

Last edited by Alex Atkin UK; 2010-04-07 at 02:07.

nightfire | # 159 | 2010-04-07, 07:27 | Report

Titan, your kernel rocks in so many ways. I was playing around with the overclocking stuff with my own build, but have switched back to yours now.

Why not take some credit with a -titan on the version?

2.6.28.10-rx51-titan has a nice ring to it. As does saying "that's right. I'm running titan."

titan | # 160 | 2010-04-07, 13:26 | Report

Originally Posted by nightfire View Post

Titan, your kernel rocks in so many ways.

thanks. I'm also running it permanently on my device and it works flawlessly
(expect for the bug in the touchscreen calibration program).

Originally Posted by

Why not take some credit with a -titan on the version?

not necessary. If I had severe ego problems I would switch to MeEgo

I just wish more users would vote for it (after testing) so that it can finally go to extras
to be accessible to more people.
Hopefully that would convince Nokia to add some its features to the stock kernel.