• After 15+ years, we've made a big change: Android Forums is now Early Bird Club. Learn more here.

Root [LINARO] AWEstruck v1.04 12/14/14 | [3.4.107] v1.05b6 5/30/15

vestalpaul said:
Is it possible for two different kernels on the same phone running at the same clock speed to benchmark much differently?
Click to expand...

Absolutely, depends on the kernel source code and how it is compiled. In programming the same goal can be achieved many different ways and also in some cases using different hardware features, each combination with their own efficiency. When compiled it turns into a list of instructions, which due to the previous reasons and also compiler differences can result in different numbers of instructions to the cpu to do the same thing. The clock speed just determines how fast the processor works through the compiled instructions. So a lower clock on a better written and compiled kernel can still be faster than high clocks on an inefficient kernel.
 
spock1104 said:
Absolutely, depends on the kernel source code and how it is compiled. In programming the same goal can be achieved many different ways and also in some cases using different hardware features, each combination with their own efficiency. When compiled it turns into a list of instructions, which due to the previous reasons and also compiler differences can result in different numbers of instructions to the cpu to do the same thing. The clock speed just determines how fast the processor works through the compiled instructions. So a lower clock on a better written and compiled kernel can still be faster than high clocks on an inefficient kernel.
Click to expand...

OK ty! Second question...can the ROM have much effect on your CPU benchmark scores? I've noticed a difference in some things and am becoming curious. Don't want to put anyones hard work down. You guys are all awesome for the work you produce.
Why does my benchmark show CPU poor at times and at other times show CPU general while before it used to show CPU good?
 
vestalpaul said:
OK ty! Second question...can the ROM have much effect on your CPU benchmark scores? I've noticed a difference in some things and am becoming curious. Don't want to put anyones hard work down. You guys are all awesome for the work you produce.
Why does my benchmark show CPU poor at times and at other times show CPU general while before it used to show CPU good?
Click to expand...

I can answer that for you. Yes the rom you're running has a huge impact on bench scores, though its really a combo of kernel+Rom as well as how many resources are being used at the time. In general, the smaller the rom, and the less features it has, the better it will do on a bench. As, less resources are tied up, less memory is being used and the over all strain on the system is lower. The number of third party apps installed, as well as system apps present also have a huge effect on benching. So, ideally, if you wanted to shoot for the highest bench possible, running something like spock's kernel+Slim with the minimum apps possible, would probably get the best results. I wouldnt try this just yet though, as I'm pretty sure the kernel only supports stock builds at the moment. (Dont quote me on that, I need to read back through this thread to catch up on progress)
 
modhatter6811 said:
I can answer that for you. Yes the rom you're running has a huge impact on bench scores, though its really a combo of kernel+Rom as well as how many resources are being used at the time. In general, the smaller the rom, and the less features it has, the better it will do on a bench. As, less resources are tied up, less memory is being used and the over all strain on the system is lower. The number of third party apps installed, as well as system apps present also have a huge effect on benching. So, ideally, if you wanted to shoot for the highest bench possible, running something like spock's kernel+Slim with the minimum apps possible, would probably get the best results. I wouldnt try this just yet though, as I'm pretty sure the kernel only supports stock builds at the moment. (Dont quote me on that, I need to read back through this thread to catch up on progress)
Click to expand...

Thanks for confirming my suspicions. It seems that common sense tends to work here lol.
What I/O scheduler do you guys find works best with this kernel?
 
Just a question, and not necessarily for this kernel, just posting here because I know the kernel devs frequent here....but any chance of ever seeing mV support on a kernel for the awe?
 
vestalpaul said:
Just a question, and not necessarily for this kernel, just posting here because I know the kernel devs frequent here....but any chance of ever seeing mV support on a kernel for the awe?
Click to expand...

Once I work out the KitKat merge issues I will be adding this, rbheromax showed me how
 
spock1104 said:
This is expected if your min clock is 384. They use the same voltage so no point in using 384.
Click to expand...

Actually, I must have been mistaken... It is just in the second core that performance control is showing weird stuff...I had to move the scrubber up in order for all the lower frequencies to then show visible and as soon as I did that... now it just shows 0mhz most of the time until it jumps up to a higher frequency, but I've never seen it even at the lowest frequency I have set.... So the second core is obviously doing something, but most of the time just shows 0mhz, is this just because it's offline? Is it caused by the fact you have enabled mp decision again kernel side?



Also... Have a request for intellidemand and smartAssV2 govs. Of course, only a request.
 
DsyMnapTic said:
Actually, I must have been mistaken... It is just in the second core that performance control is showing weird stuff...I had to move the scrubber up in order for all the lower frequencies to then show visible and as soon as I did that... now it just shows 0mhz most of the time until it jumps up to a higher frequency, but I've never seen it even at the lowest frequency I have set.... So the second core is obviously doing something, but most of the time just shows 0mhz, is this just because it's offline? Is it caused by the fact you have enabled mp decision again kernel side?



Also... Have a request for intellidemand and smartAssV2 govs. Of course, only a request.
Click to expand...

I really don't recommend using performance control with this kernel, it doesn't support configuration of mpdecision or thermal controls. My recommendations are trickster mod, android tuner, or kernel tuner.

Two things can be going on here. One: if you set your clock below 384 you should also change the mpdecision hot plug frequency. The core may be getting shut off before reaching minimum frequency.

Two: The CPU stats seem to get thrown off every time mpdecision shuts off the second core. If changing the hot plug frequency doesn't work try disabling mpdecision, reboot, and see if the stats look okay again. I think that the stats system just doesn't know how to deal with the hot plug and the CPUs are still operating properly. If that is the case, it's up to you if you want to deal with odd stats or keep mpdecision off. Also if verified I can add stat correction to my to-do list.
 
Okay well I opened up trickster mod, just to see what it was saying about my already set kernel settings with performance control. I noticed that in mp decision control the idle frequency is set to 486... Maybe I broke it when I moved the slider around on the second core, or otherwise it's just not reporting the settings correctly, which the latter is what I'm guessing since the second CPU core shows as offline most of the time in trickster mod.



*Also, there is no 384mhz... Which is why I felt the need to go down to 270 in the first place. I just felt that 486 was an unnecessary rise to the minimum default frequent



... And won't performance control take priority over any third party app I use, since it's built into the ROM... could I even just remove it with titanium without breaking anything on this ROM? (Brittnearl's "L" awe v1.0)

**personally I don't like performance control and think it should be removed**
 
DsyMnapTic said:
Okay well I opened up trickster mod, just to see what it was saying about my already set kernel settings with performance control. I noticed that in mp decision control the idle frequency is set to 486... Maybe I broke it when I moved the slider around on the second core, or otherwise it's just not reporting the settings correctly, which the latter is what I'm guessing since the second CPU core shows as offline most of the time in trickster mod.



*Also, there is no 384mhz... Which is why I felt the need to go down to 270 in the first place. I just felt that 486 was an unnecessary rise to the minimum default frequent



... And won't performance control take priority over any third party app I use, since it's built into the ROM... could I even just remove it with titanium without breaking anything on this ROM? (Brittnearl's "L" awe v1.0)

**personally I don't like performance control and think it should be removed**
Click to expand...
As far as removing it, It will only break the option in settings... As far as it not working right, It works great with stock kernel. allows control over everything... I think the real cause is incorrect kernel code which is y the scubbers are funky with spocks kernel, he left stock default clock speeds but changed frequency table to one without them... Also u get random freeze reboots
 
I've never gotten random freeze reboots with it, however I have figured out a method to keep both cores running and doing what I want.... In the mp decision control settings I just set min_CPUs and max_CPUs both to 2 and now they run fine.

Do you know what scroff_single_core is though?
 
brittnearl said:
As far as removing it, It will only break the option in settings... As far as it not working right, It works great with stock kernel. allows control over everything... I think the real cause is incorrect kernel code which is y the scubbers are funky with spocks kernel, he left stock default clock speeds but changed frequency table to one without them... Also u get random freeze reboots
Click to expand...

Its not incorrect kernel code, I set it so 384 isn't used for scaling, meaning that 486 would be preferred, since their voltages are the same. Its a well documented piece of the Qualcomm CPU table. That's not the only frequency I did this to, there are several others. It would be poor code for me to just remove them, especially considering that the settings change depending on what bin CPU you have.
 
spock1104 said:
Its not incorrect kernel code, I set it so 384 isn't used for scaling, meaning that 486 would be preferred, since their voltages are the same. Its a well documented piece of the Qualcomm CPU table. That's not the only frequency I did this to, there are several others. It would be poor code for me to just remove them, especially considering that the settings change depending on what bin CPU you have.
Click to expand...
Ok well im no rocket scientist but if pcontrol works with stock kernel it should work with urs... Its common sense that if u change freq scale but dont change default freqs that it will cause issues
 
brittnearl said:
Ok well im no rocket scientist but if pcontrol works with stock kernel it should work with urs... Its common sense that if u change freq scale but dont change default freqs that it will cause issues
Click to expand...

Alright now.... Go easy, Lol... I'm pretty sure Spock has good reasons for what he has done. I mean after all he has tweaked the crap out of this stubborn little kernel and I don't think there's any need to get hasty...I just don't have the knowledge of how kernels are supposed to work so I'm poking around trying to figure things out (though not at a development level by any means, that's beyond me)
 
DsyMnapTic said:
Alright now.... Go easy, Lol... I'm pretty sure Spock has good reasons for what he has done. I mean after all he has tweaked the crap out of this stubborn little kernel and I don't think there's any need to get hasty...I just don't have the knowledge of how kernels are supposed to work so I'm poking around trying to figure things out (though not at a development level by any means, that's beyond me)
Click to expand...
Spocks kernel is the most stable of any oc for this device :) i just dont get how these issues are treated as "normal" :confused: when the stock kernel has none of them... Im not to good with source code though i did make an app for ps2 modding a long time ago or i would look into this myself... Give me an rom and i can make it a beast lol in conclusion i dont get everyones disregard to stability being first priority :confused:
 
brittnearl said:
Spocks kernel is the most stable of any oc for this device :) i just dont get how these issues are treated as "normal" :confused: when the stock kernel has none of them... Im not to good with source code though i did make an app for ps2 modding a long time ago or i would look into this myself... Give me an rom and i can make it a beast lol in conclusion i dont get everyones disregard to stability being first priority :confused:
Click to expand...

Maybe stability is an ongoing process of testing to see if this or that works and then adding something that breaks something else.... More features are probably all "in between" types of things, to give people more options while in fact the subtle stability tweaks and bug squashing is a continuous process that goes on behind the scenes.... Weren't all the attempts and constant kernel releases, about getting things to operate stably in the first place?
 
You must log in or register to reply here.
Back
Top Bottom