[KERNEL] Kernel (deprimed_test) - Break It If You Can

metalspring

Silver Member
Joined
Jan 9, 2011
Messages
2,227
Reaction score
24
Location
hermitage, pa
Interesting, I found a bug or 2 on the cpu by messing with custom speeds... apparently certain speed slots don't read correctly... For example I'm running at 1.265 ghz but the info in setcpu says its 1.2495 ghz or something like that... Also there is a major one with running 1.3 ghz, it somehow reads 2.6 ghz but is super slow and freezes but even changing it to 1.301 works just fine (I doubt anyone really wants to run that constantly though lol)
At any rate I'm loving my linpack scores of 20+ right now :p
 
OP
prime

prime

Kernel Developer
Premium Member
Developer
Joined
Apr 26, 2010
Messages
1,939
Reaction score
70
Location
Florida
Website
github.com
Interesting, I found a bug or 2 on the cpu by messing with custom speeds... apparently certain speed slots don't read correctly... For example I'm running at 1.265 ghz but the info in setcpu says its 1.2495 ghz or something like that... Also there is a major one with running 1.3 ghz, it somehow reads 2.6 ghz but is super slow and freezes but even changing it to 1.301 works just fine (I doubt anyone really wants to run that constantly though lol)
At any rate I'm loving my linpack scores of 20+ right now :p

1.301Ghz on a 600Mhz CPU....priceless

What's the CPU temperature after heavy use? 165° F or more?
 

metalspring

Silver Member
Joined
Jan 9, 2011
Messages
2,227
Reaction score
24
Location
hermitage, pa
1.301Ghz on a 600Mhz CPU....priceless

What's the CPU temperature after heavy use? 165° F or more?

surprisingly it was only about 130-135 at the highest, definitely not stable though, only went 15 minutes before a reboot, 1.275 seems stable enough though
btw what are good load averages?
 
OP
prime

prime

Kernel Developer
Premium Member
Developer
Joined
Apr 26, 2010
Messages
1,939
Reaction score
70
Location
Florida
Website
github.com
These are the vsel values from chevyno1's 1.25ghz ulv kernel

mpu_opps[7] rate=1250000000 opp_id=7 vsel=86
mpu_opps[6] rate=1100000000 opp_id=6 vsel=69
mpu_opps[5] rate=1000000000 opp_id=5 vsel=59
mpu_opps[4] rate=800000000 opp_id=4 vsel=46
mpu_opps[3] rate=550000000 opp_id=3 vsel=34
mpu_opps[2] rate=400000000 opp_id=2 vsel=34
mpu_opps[1] rate=250000000 opp_id=1 vsel=34
 
OP
prime

prime

Kernel Developer
Premium Member
Developer
Joined
Apr 26, 2010
Messages
1,939
Reaction score
70
Location
Florida
Website
github.com
surprisingly it was only about 130-135 at the highest, definitely not stable though, only went 15 minutes before a reboot, 1.275 seems stable enough though
btw what are good load averages?

Good load averages are 0.09 to 0.5 depending on what the phone is doing...you never want to see 8.85 and such as the CPU is way behind.

Here is a sshot I just took:

aefd6043-6a18-bbda.jpg
 
Last edited:

AlphaMack

Member
Joined
Aug 12, 2010
Messages
71
Reaction score
0
Location
California
Checking back into this thread with a (final?) update. I noticed that the latest test kernel from the OP download link is the same (rooted@linbox #7) as the deprimed_base I initially downloaded several months ago and didn't like because of the lag. To avoid any issues I turned off swap, reverted from V6, undid my custom 20overclock, and made sure that the download of the kernel was OK.

Using prime's 1000 script (and eventually copied over to /etc/init.d/20overclock) my D1 couldn't have been more smooth. Android OS usage seems a little high (<20%) but the battery drain is acceptable - seems even better than kernel versions #19 and #20.

For kicks I reintroduced V6 supercharger and I haven't really experienced any problems (yet). The only other things I could think of where my previous experience wasn't working out:

- Swap and V6 not getting along in the same way compcache causes lag when used with V6.
- Something went haywire with the deprimed_base.zip installation.

The jury is still out on the first point because I just turned my swap partition back on and so far so good for now. Of course if the stability goes south I'll immediately know.

So far I'm happy with the results and I think I can finally leave good enough alone until I retire my OG.
 

DaxxTrias

Member
Joined
Jun 28, 2010
Messages
32
Reaction score
0
Hm I didn't notice it til now, but it seems the latest test build does not include the zram module.

I wonder if this breaks compcache?


Edit:

There appears to be 9 module lib files missing as opposed to the last build I used (#19)
 
Last edited:

phonetool

Member
Joined
Mar 22, 2011
Messages
337
Reaction score
0
There's discussion about compcache being broken in the last few pages of this thread. Prime is set to fix it with the next release, which is delayed by his hectic life at the moment.

- onDroid
 

rancur3p1c

Member
Joined
Jun 30, 2010
Messages
66
Reaction score
0
I made some timing changes and such but basically the same as smartass. The latest version of SetCPU exposes the governors functions for easy manipulation.



Have been too busy lately, I haven't abandoned the kernel ;)

In my experience none of the governors actually keep the core in the lowest frequency state...best is on demand which still keeps cpu at full throttle 25% of the time. I eventually gave up on 1.2ghz overclock because of this. Any thoughts?

I have since moved on to Atrix 4g with stock kernel, gets the governing right and almost all time is in lowest state. Moving to custom kernels brings back the same problem...25% of the time cpu is at full throttle, killing battery life.

You might not release another Droid kernel but that might be worth looking at for whatever phone it is you're working on now....
 
OP
prime

prime

Kernel Developer
Premium Member
Developer
Joined
Apr 26, 2010
Messages
1,939
Reaction score
70
Location
Florida
Website
github.com
In my experience none of the governors actually keep the core in the lowest frequency state...best is on demand which still keeps cpu at full throttle 25% of the time. I eventually gave up on 1.2ghz overclock because of this. Any thoughts?

I have since moved on to Atrix 4g with stock kernel, gets the governing right and almost all time is in lowest state. Moving to custom kernels brings back the same problem...25% of the time cpu is at full throttle, killing battery life.

You might not release another Droid kernel but that might be worth looking at for whatever phone it is you're working on now....

It depends on what is going on in the background what time in state shows.

I would need to do some testing on bumblebee to know it isn't behaving as intended, I do know if you change the slots you would need to adjust the bumblebee max frequency while sleeping else the governor ramps up to the top slot.
 
OP
prime

prime

Kernel Developer
Premium Member
Developer
Joined
Apr 26, 2010
Messages
1,939
Reaction score
70
Location
Florida
Website
github.com
I'm sure I will, just want to be certain about the decision ;)
 
Top