12/19: Need testers for kernel that can simultaneously battery save and overclock

[arrrghhh]

set ramsize 0x10000000
set ramaddr 0x10000000
set mtype 2292
set KERNEL zImage
set initrd initrd.gz

set cmdline "lcd.density=240 7x00.a11=500 msmvkeyb_toggle=off pmem.extra=1 gsensor_axis=2,1,3 rel_path=andboot hw3d.force=1 acpuclock.oc_freq_khz=726000 physkeyboard=rhod500 force_cdma=1"

Just to let you know, there were a couple of deprecated commands in your startup. They're in bold/red. They shouldn't hurt anything, but there's no need for them either.

[bzo]

I have a fix for using sleep mode of 1 while overclocked. I haven't tested 0, but it should work if 0 works normally. I was under the impression that the sleep mode of 0 was broken.

Test kernel attached. I'll get the patch submitted as soon as I get at least a few reports of success.

[manekineko]

Originally Posted by bzo

I have a fix for using sleep mode of 1 while overclocked. I haven't tested 0, but it should work if 0 works normally. I was under the impression that the sleep mode of 0 was broken.

Test kernel attached. I'll get the patch submitted as soon as I get at least a few reports of success.

I can confirm that at least initially, your fix appears to work.

I tested this by adding overclock, and setting sleep mode to 1 on a regular kernel. I then booted my phone 3 times and put it to sleep, waiting until the green light came on (this actually took quite a few more than 3 boots, because most of the time the green light would never come on). After the green light came on, I attempted to wake the phone with the power button. In all 3 cases, the green light changed orange but the phone did not wake.

I then installed your kernel, while keeping everything else the same. I booted, and then put the phone to sleep. Green light came on, I successfully woke the phone.

What was the cause of the issue?

I'm going to update first post and ask for more testers.

[bzo]

Originally Posted by manekineko

What was the cause of the issue?

Normally, the PLL2 clock is setup to run at 1056mhz and divided by 2 for the regular 528mhz top speed of the cpu. The max speed of PLL2 is 1200mhz, so if a 2 divider is used, the max frequency possible is 600mhz. So for overclocking, we skip the divider and use the frequency directly so that frequencies above 600mhz are possible.

It appears that in the deeper sleep modes, the clocks are automatically being shut down and powered up again on wake up. The wakeup assumes that PLL2 should be brought back up to 1056mhz and does not reset the divider back to 2, so the cpu is trying to run at 1056mhz which will obviously not work.

The fix checks that PLL2 is being run at the frequency we expect at wakeup or whenever the cpu frequency is changed, and corrects if necessary.

[docnas]

maybe im missing something im using the startup overclocking and when i let it sleep the green indicator goes on meaning the phone sleeps properly or am i missing something? I havent used your kernel yet as im trying out tiads package right now to see how it works but just wanted to know if im mistaken about my phone actually sleeping when the light was green.

[manekineko]

Originally Posted by docnas

maybe im missing something im using the startup overclocking and when i let it sleep the green indicator goes on meaning the phone sleeps properly or am i missing something? I havent used your kernel yet as im trying out tiads package right now to see how it works but just wanted to know if im mistaken about my phone actually sleeping when the light was green.

Yeah, the green indicator means your phone is sleeping properly. Overclocked phones have no problem sleeping.

What this kernel enables you to do is set the battery saving pm.sleep_mode=1 command which is normally incompatible with overclock. Without this kernel, after an overclocked phone goes to sleep, when it wakes up again it crashes.

pm.sleep_mode=1 about doubled my battery life without overclock, I haven't had a chance to observe battery life with overclock and pm.sleep_mode=1.

[docnas]

Thanks for the info. well lemme run a few more tests on tiads package ( so far i got a quadrant score of 596 or was it 594 neocore was 18.1 so it looks good to me) and then il try out this kernel. will report my findings when i test it out and let it run for a bit.

[fishingmedic]

Been running this kernel since last night, no issues at all with it waking up, goes to sleep nice, not sure on batt life with it as I was up late and didn't give it a full charge after loading up the kernel, but so far so good on stability. 20.9 neocore w/o sound, 632 quadrant, no issues found knock on wood...

[arrrghhh]

Originally Posted by bzo

Normally, the PLL2 clock is setup to run at 1056mhz and divided by 2 for the regular 528mhz top speed of the cpu. The max speed of PLL2 is 1200mhz, so if a 2 divider is used, the max frequency possible is 600mhz. So for overclocking, we skip the divider and use the frequency directly so that frequencies above 600mhz are possible.

It appears that in the deeper sleep modes, the clocks are automatically being shut down and powered up again on wake up. The wakeup assumes that PLL2 should be brought back up to 1056mhz and does not reset the divider back to 2, so the cpu is trying to run at 1056mhz which will obviously not work.

The fix checks that PLL2 is being run at the frequency we expect at wakeup or whenever the cpu frequency is changed, and corrects if necessary.

As always, awesome explanation bzo. Thanks

I need to test this kernel! Oy, work is killing me...

[ryanmv213]

Originally Posted by fishingmedic

Been running this kernel since last night, no issues at all with it waking up, goes to sleep nice, not sure on batt life with it as I was up late and didn't give it a full charge after loading up the kernel, but so far so good on stability. 20.9 neocore w/o sound, 632 quadrant, no issues found knock on wood...

do you mind posting your sprint startup.txt you are currently using with this kernel? Also are you changing anything for those scores via tiad8's methods in his thread? What differences have you noticed between ACL's 12.18/19 and this? Also what class sd card are you using? Thanks for the link earlier.