Hello,
I played with the dalvik.vm.dexopt-flags in /system/build.prop. Basically you control the class-loading and dalvik-vm there.
streakowner proposed in http://forum.xda-developers.com/showpost.php?p=18111740&postcount=644 the setting v=n,o=v. But disabled this with #, in DSC v1.0 the setting is active.
That means: do not verify and optimize mode verified. Found that info in dexopt/OptMain.c
I did some performance measurement with smartbench 2011 and got the value 553 in the first round, and 465 in the second round. That is very bad for an one ghz powered device. Good would > 1000. /data/dalvik-cache takes 74 MiB with this setting.
At the moment I'm running with
dalvik.vm.dexopt-flags=v=n,o=a,u=y
That means: do not verify, optimize all classes and optimize for uniprocessor.
First I was not impressed, because rebooting with a cleaned dalvik-cache took more than double the time span of the o=v setting. logcat is full off lines telling me what can't be optimized and how much time the load took. But after booting I got the impression the system went faster. With smartbench 2011 I get the value 1036 in the first and 1110 in the second round. /data/dalvik-cache takes 74 MiB with this setting.
Can someone confirm my finding?
I don't say you will double the speed of the streak with this setting. But I think you will notice the change. The trade off is the time for the first boot. But I think that is worth is.
So I propose to use the above setting in all gb-based roms for the streak 5.
Have a nice weekend,
hunderteins
Tested on miui and it's not working i get into a boot loop.
getprop | grep dexopt
[dalvik.vm.dexopt-flags]: [v=n,o=a,u=y]
Well, same ~950 in Quadrant (CPU/mem only)
998400
--
Sergei.
_2bad4u_ said:
Tested on miui and it's not working i get into a boot loop.
Click to expand...
Click to collapse
Ouch! Sorry to hear that. I hope you didn't get in serious trouble.
_n0p_ said:
getprop | grep dexopt
[dalvik.vm.dexopt-flags]: [v=n,o=a,u=y]
Well, same ~950 in Quadrant (CPU/mem only)
998400
Click to expand...
Click to collapse
Quadrant shows 1048 with o=a and 1065 with o=v. Thats not helping either. At least smartbench shows nearly the same "Games Index". But what do they test with ther productivity index? Calculating PI and mandelbrot is so different?
And there is opera. Its definitly faster with o=a. I'll try to find a web page rendering test.
o.k. I get more mixed signals here. Numion tells me that rendering a real world web-page is about 30% faster with o=a (10.5 vs. 14,6 sec). Scragz is 10 % slower with o=a ( 9.7 vs. 10.5 sec).
I'll file it under works for me, but don't try this with miui. Thanks for your help.
Reg. hunderteins
No problem happy to help
hunderteins, this actually lead me to think that you'll try to build highly optimized (armv7, neon, no loop unroll - small cache) libdvm. Am I right?
_2bad4u_ said:
Tested on miui and it's not working i get into a boot loop.
Click to expand...
Click to collapse
It works with MIUI too but you have to delete the another line that contains
dalvik.vm.dexopt-flags=m=y
i guess this lines value is overlapping with the other lines "u=y" value.
---------- Post added at 08:11 AM ---------- Previous post was at 08:03 AM ----------
The result of first test is:
Productivity index=716
Gaming index=981
with smart bench 2012
---------- Post added at 08:20 AM ---------- Previous post was at 08:11 AM ----------
The result of second test is:
Productivity index=1074
Gaming index=1569
with smart bench 2011
DS 5 works faster than ever!
Thank you!
hunderteins said:
Hello,
I played with the dalvik.vm.dexopt-flags in /system/build.prop. Basically you control the class-loading and dalvik-vm there.
streakowner proposed in http://forum.xda-developers.com/showpost.php?p=18111740&postcount=644 the setting v=n,o=v. But disabled this with #, in DSC v1.0 the setting is active.
That means: do not verify and optimize mode verified. Found that info in dexopt/OptMain.c
I did some performance measurement with smartbench 2011 and got the value 553 in the first round, and 465 in the second round. That is very bad for an one ghz powered device. Good would > 1000. /data/dalvik-cache takes 74 MiB with this setting.
At the moment I'm running with
dalvik.vm.dexopt-flags=v=n,o=a,u=y
That means: do not verify, optimize all classes and optimize for uniprocessor.
First I was not impressed, because rebooting with a cleaned dalvik-cache took more than double the time span of the o=v setting. logcat is full off lines telling me what can't be optimized and how much time the load took. But after booting I got the impression the system went faster. With smartbench 2011 I get the value 1036 in the first and 1110 in the second round. /data/dalvik-cache takes 74 MiB with this setting.
Can someone confirm my finding?
I don't say you will double the speed of the streak with this setting. But I think you will notice the change. The trade off is the time for the first boot. But I think that is worth is.
So I propose to use the above setting in all gb-based roms for the streak 5.
Have a nice weekend,
hunderteins
Click to expand...
Click to collapse
Sent from my Dell Streak using Tapatalk 2
Testing this now. Opera Mobile is definitely faster, and I honestly think the start time from a wipe of the dalvik cache wasn't any longer than normal.
Testing this now. Opera Mobile is definitely faster, and I honestly think the start time from a wipe of the dalvik cache wasn't any longer than normal.
Click to expand...
Click to collapse
Second that... wiping Dalvik on the Streak seems to be the same for me
Sent from my cm_tenderloin using Tapatalk 2
Same here. Didn't notice any major difference in boot time then usual.
MIUI seems a bit more responsive after this. And I've always found MIUI to be a bit laggy compared to CM and DSC! Gonna observe it for some more time before I can be 100% sure.
Miui 2.4.20 is not laggy like old versions and it boots very fast.After i edited build.prop it's more responsive and i wrote the results of benches.
Just a warning:
When I was experimenting with this settings, i've got caught in situation when CM7 dvm was trying to use /cache for dalvik-cache and failed.
Works fine with Nexus S dvm (as in DSC v7.0)
And I can confirm higher scores in Smartbench.
More, Smartbench somehow gives results, similar to clock value - ~930 on 998400, ~1340 on 1344000, so overclock does matter in this case.
So do you recommend using it or going back to the original settings?
I sure do.
But just in case keep handy /system/lib/libdvm.so from DSC 7.0 and nandroid backup.
libdvm.so from CM7 has reference to /cache - I wonder if binary change to /data would help.
sinan33 said:
It works with MIUI too but you have to delete the another line that contains
dalvik.vm.dexopt-flags=m=y
i guess this lines value is overlapping with the other lines "u=y" value.
---------- Post added at 08:11 AM ---------- Previous post was at 08:03 AM ----------
The result of first test is:
Productivity index=716
Gaming index=981
with smart bench 2012
---------- Post added at 08:20 AM ---------- Previous post was at 08:11 AM ----------
The result of second test is:
Productivity index=1074
Gaming index=1569
with smart bench 2011
Click to expand...
Click to collapse
Ok so i need to change only the entry dalvik.vm.dexopt-flags. Below is my buildprop part where the dalvik entries are so just confirm as to what all needs to be changed
#streakOwner's recommended
dalvik.vm.startheapsize=48m
dalvik.vm.heapsize=48m
dalvik.vm.execution-mode=int:jit
dalvik.vm.dexopt-flags=v=n,o=v
dalvik.vm.dexopt-data-only=1
dalvik.vm.lockprof.threshold=150
davlik.vm.verify-bytecode=false
ro.android.kernel.checkjni=0
#wifi.supplicant_scan_interval=120
windowsmgr.max_events_per_sec=80
keyguard.no_require_sim=true
Dude_Virus said:
Ok so i need to change only the entry dalvik.vm.dexopt-flags. Below is my buildprop part where the dalvik entries are so just confirm as to what all needs to be changed
#streakOwner's recommended
dalvik.vm.startheapsize=48m
dalvik.vm.heapsize=48m
dalvik.vm.execution-mode=int:jit
dalvik.vm.dexopt-flags=v=n,o=v
dalvik.vm.dexopt-data-only=1
dalvik.vm.lockprof.threshold=150
davlik.vm.verify-bytecode=false
ro.android.kernel.checkjni=0
#wifi.supplicant_scan_interval=120
windowsmgr.max_events_per_sec=80
keyguard.no_require_sim=true
Click to expand...
Click to collapse
Change
dalvik.vm.dexopt-flags=v=n,o=v
to
dalvik.vm.dexopt-flags=v=n,o=a,u=y
or
dalvik.vm.dexopt-flags=v=n,o=a,u=y,m=y (testing now with register optimization)
_n0p_ said:
Change
dalvik.vm.dexopt-flags=v=n,o=v
to
dalvik.vm.dexopt-flags=v=n,o=a,u=y
or
dalvik.vm.dexopt-flags=v=n,o=a,u=y,m=y (testing now with register optimization)
Click to expand...
Click to collapse
Thanks for the values....but dont i need to change the cache size to 74 as mentioned in the original post tht the cache goes to 74 mb
Dude_Virus said:
Thanks for the values....but dont i need to change the cache size to 74 as mentioned in the original post tht the cache goes to 74 mb
Click to expand...
Click to collapse
You've took cache size for heap size. That's different things. hunderteins was referring for catalog with built dalvik cache (/data/dalvik-cache)
Related
I was looking through /system with root explorer to see what I notice that it new from the EVO 4G and I stubbled across a file called adreno_config.txt. I opened it and read through parts of it and it has a lot of values that can be changed. I changed binning from software to hardware and noticed a very slight jump in benchmark scores using antutu benchmark. Mainly I noticed smoother scrolling on homescreens and webpages with a lot of flash content. I was hoping maybe a dev or someone who knows alittle more about these things than I do could look thru this file and see if we could get a performance boost from anything in this file. The exact path to the file is /system/etc/adreno_config.txt. thanks and hopefully something good could come of this. One of the big things I noticed was antialiasing settings that were all turned off by default. If this has already been looked at then I am sorry for the double post and don't mind being told I'm wrong.
I will upload the file if needed, but I figure that anyone who would know anything about this should already have root and s-off so it is really pointless.
Clay333 said:
I was looking through /system with root explorer to see what I notice that it new from the EVO 4G and I stubbled across a file called adreno_config.txt. I opened it and read through parts of it and it has a lot of values that can be changed. I changed binning from software to hardware and noticed a very slight jump in benchmark scores using antutu benchmark. Mainly I noticed smoother scrolling on homescreens and webpages with a lot of flash content. I was hoping maybe a dev or someone who knows alittle more about these things than I do could look thru this file and see if we could get a performance boost from anything in this file. The exact path to the file is /system/etc/adreno_config.txt. thanks and hopefully something good could come of this. One of the big things I noticed was antialiasing settings that were all turned off by default. If this has already been looked at then I am sorry for the double post and don't mind being told I'm wrong.
I will upload the file if needed, but I figure that anyone who would know anything about this should already have root and s-off so it is really pointless.
Click to expand...
Click to collapse
seems like an AWESOME find. subscribed
Don't apologize, sounds like you found something thar we can tinker with. Good post.
did a little bit of searching about this config file..it seems it got monkeyed around with while they were trying to achieve s-off, but I cant find anything showing any performance differnces from changing the values, so hopefully were on to something good
Success100 said:
did a little bit of searching about this config file..it seems it got monkeyed around with while they were trying to achieve s-off, but I cant find anything showing any performance differnces from changing the values, so hopefully were on to something good
Click to expand...
Click to collapse
Could you post it? I'll take a look to see if the file can improve performance.
DDiaz007 said:
Could you post it? I'll take a look to see if the file can improve performance.
Click to expand...
Click to collapse
just use root explorer to go in and grab it from your phone. i would definetely make a backup of it on your sd card and nandroid before messing around with it. I dont have s-off just yet (computer is down) or Id mess with it myself
here it is
Thanks I'll take a look to see what he's talking about exactly.
From what it looks, it looks like it is the Adreno 220 graphics configuration settings. It seems it can be tweaked, but I do not know entirely since this is a phone.
Take a look at these lines...
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
; Specify the allocation strategy for multisampling antialiasing buffers.
; Valid values:
; on_demand - allocate MSAA buffer on demand. This is the default value.
; always - always allocate MSAA buffer.
; never - never allocate an MSAA buffer.
;
;MSAABufferAllocation=on_demand
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
; Force the MSAA antialiasing mode to the value in MSAAMode.
; Valid values:
; 0 - (default) do not force the MSAA antialiasing mode.
; 1 - force the MSAA antialiasing mode to the values in MSAAMode and MSAASmoothing.
;
;forceMSAAMode=0
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
; Specifiy the antialiasing mode to use if forceMSAAMode is enabled.
; Valid values:
; 0 - (default) do not antialias.
; 1 - use 2x antialiasing.
; 2 - use 4x antialiasing.
;
;MSAAMode=0
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
Now I don't know if this will fully improve performance, but I do know that when I am tweaking files like this for a PC video card, it'll make all the difference in the world for running games. But these files are within the games directory so this seems to be some sort of universal settings. I don't know entirely since this is a phone. Someone with knowledge in regards to this please chime in.
Yup this has to do with the configuration settings, it is universal. I can see an FPS cap option. Our evos aren't capped by FPS, if anyone would want to know that.
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
; Cap value for frames per second. Do not cap if value < 0 or value > 1000.
; Valid values:
; 0 - (default) do not cap.
; Int - (integer) cap to this number of frames per second if N 1000.
;
;FPSCap=0
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
I'll probably tweak this file to see if it really does anything, but I don't have the time to fiddle with it right now. Anyone who has knowledge in regards to phones, I don't mind if you chime in and educate.
good ****..
i just sent viper a pm and well see what he says about it.
He's one of the more popular devs in the dev section
I pm'ed two Devs... I know files like this are used always in computers for adjusting configurations precisely rather than using presets. I'm just not entirely sure for a phone lol
to get full benefit of these tweaks, you must use chainfire 3D PRO, and enable EGL beta driver when you install CF3D driver. Also, since you have pro, go ahead and hide CF3Ds announce ID. good luck, have fun.
I assume this is what joelZ9614 might have used to "force GPU UI rendering" mentioned in his Shooter Rewind ROM changelog
phatmanxxl said:
I assume this is what joelZ9614 might have used to "force GPU UI rendering" mentioned in his Shooter Rewind ROM changelog
Click to expand...
Click to collapse
I don't believe so, there is another way to do it.
Tilde88 said:
to get full benefit of these tweaks, you must use chainfire 3D PRO, and enable EGL beta driver when you install CF3D driver. Also, since you have pro, go ahead and hide CF3Ds announce ID. good luck, have fun.
Click to expand...
Click to collapse
gotcha. well viper says he's gonna f around with it tonight and report back. so tilde your saying this is only going to benefit gamplay and not system operations?
edit: and here comes viper lol. right on time
phatmanxxl said:
I assume this is what joelZ9614 might have used to "force GPU UI rendering" mentioned in his Shooter Rewind ROM changelog
Click to expand...
Click to collapse
I always wonder what that meant. If anyone uses his ROM, is there a noticeable difference using it?
Success100 said:
gotcha. well viper says he's gonna f around with it tonight and report back. so tilde your saying this is only going to benefit gamplay and not system operations?
Click to expand...
Click to collapse
Most likely gameplay only.
DDiaz007 said:
I always wonder what that meant. If anyone uses his ROM, is there a noticeable difference using it?
Click to expand...
Click to collapse
I've enabled it in my ROM (in my test build) and it seems a little smoother, not amazingly different, but every little bit helps
Joel just told me he's seen the file but hasn't bothered looking into it at first.. he said he's going to though.
Just a few tweaks for those who are rooted and not flashing custom roms. some will already be in there, just edit them.
use root explorer, go to system, click mount r/w hold build.prop till a menu pops up and click the text editor one.
read the entire build.prop to ensure no duplicates
also dont misspell!
If its already in there just edit it. if you dont know what your doing dont bother with this.
render ui with gpu
debug.sf.hw=1
smoother scrolling
windowsmgr.max_events_per_sec=150
better battery by reducing wifi scans
wifi.supplicant_scan_interval = 150
dalvik memory heap size
dalvik.vm.heapsize=48m
screen ppi or font density
ro.sf.lcd_density=235
turn off dithering, not needed anymore in gingerbread
persist.sys.use_dithering=0
more memory
persist.sys.purgeable_assets=1
smoother experience
debug.performance.tuning=1
video.accelerate.hw=1
dalvik.vm.dexopt-flags=m=y
camera/camcorder mods
ro.media.capture.maxres=8m
ro.media.capture.fast.fps=4
ro.media.capture.slow.fps=120
ro.media.capture.flash=led
ro.media.capture.flashMinV=3300000
ro.media.capture.torchIntensity=40
ro.media.capture.flashIntensity=70
ro.media.panorama.defres=3264x1840
ro.media.panorama.frameres=1280x720
ro.camcorder.videoModes=true
better pics
ro.media.dec.jpeg.memcap=8000000
ro.media.enc.hprof.vid.bps=8000000
ro.media.enc.jpeg.quality=100
better playback
media.stagefright.enable-player=true
media.stagefright.enable-meta=true
media.stagefright.enable-scan=true
media.stagefright.enable-http=true
media.stagefright.enable-record=false
i dont use all of these but most. it helps.
Keep in mind in order to update in the future you must keep a back-up of your build.prop to restore if you want to update.
once your done restart
Thanks but do you have a short description of what the tweak will enhance? Some are obvious, but others are not. Thanks again.
Sent from my MoPho using XDA Premium.
just edited with some descriptions
Thanks for this. I havent unlocked my bootloader yet so this is something I can play around with to see if I can squeze anything more out of this MoPho!
bigbassbrent said:
Thanks for this. I havent unlocked my bootloader yet so this is something I can play around with to see if I can squeze anything more out of this MoPho!
Click to expand...
Click to collapse
Yes its basically making changes to your build.prop no unlocked bootloader needed. An easier way to do this is to pull the build.prop and get it on your comp then make the changes and replace current build.prop. Id make a backup of unmodified build.prop.
So if these tweaks are not in the stock build prop, is there a specific place in the file they go? Or can I just copy paste to anywhere, such as the bottom.
They can go at the end. If my memory serves me correct, the following ones were already in the build.prop and just needed changing. (I didn't do any of the media related ones)
windowsmgr.max_events_per_sec=150 (default = 70)
wifi.supplicant_scan_interval=150 (default = 45)
dalvik.vm.heapsize=48 (default = 40)
ro.sf.lcd_density=235 (default = 240)
I had to add the following lines at the end of the build.prop
persist.sys.use_dithering=0
persist.sys.purgeable_assets=1
debug.performance.tuning=1
video.accelerate.hw=1
dalvik.vm.dexopt-flags=m=y (this one was already present and the same)
I'm just learning, but I remember seeing something about there needing to be a blank line at the end of the file. (hit enter after the last line) Dont know if this is true in this case but I figured I'd put one just in case.
So any that has applied these tweaks which are the best to use for performance enhancement and do you notice the enhancement?
Oh how to make a backup too do I just write it down our what.
Sent from my iphone killer
abrcrmdl23 said:
So any that has applied these tweaks which are the best to use for performance enhancement and do you notice the enhancement?
Oh how to make a backup too do I just write it down our what.
Sent from my iphone killer
Click to expand...
Click to collapse
All of the changes I listed in my last post are aimed at gaining performace except for the ro.sf.lcd_density=235 (default = 240) line which to me just made the fonts smaller. I ended up setting this one back to 240.
As far as performance gains, I didnt record any benchmarks before I made the changes so I don't have a good baseline, but I compared my scores to another unmodified photon.
Smart Bench 2011
Modified scored 3345, 2665
Stock scored 2823, 2656
Quadarant Standard
Modified scored 2624
Stock scored 2743
Based on Quadrant scores, I'm not sure what we have accomplished by making the changes. The Smart Bench scores seemed to be somewhat better after the changes. I'll continue to watch and report any other findings.
Thanks for the info. The only tweak I made so far is the dalvik cache tweak. It made the phone much smoother and faster. Is there any disadvantage in bumping up the dalvik cache up to 55 or 60?
These changes will not show up much in benchmarking, however they will make the device feel smoother and faster. better scrolling and better experience. changing how the device renders the ui wont make your quadrant go up
clifton421 said:
Thanks for the info. The only tweak I made so far is the dalvik cache tweak. It made the phone much smoother and faster. Is there any disadvantage in bumping up the dalvik cache up to 55 or 60?
Click to expand...
Click to collapse
People have mixed opinions about what's the best setting, the higher number allows apps to cache more memory into the ram.
I usually set mine to 64m
if you set it too high each program will use more memory and the available memory for other programs will be reduced
So I was looking through my build prop and has anyone else noticed there are a lot odd Verizon references in there like it says the photon is a Verizon wireless world device, and there is other VZW references it also makes mention
of something that enables lte connectivity and ehspa or something like that
Sent from my iphone killer
abrcrmdl23 said:
So I was looking through my build prop and has anyone else noticed there are a lot odd Verizon references in there like it says the photon is a Verizon wireless world device, and there is other VZW references it also makes mention
of something that enables lte connectivity and ehspa or something like that
Sent from my iphone killer
Click to expand...
Click to collapse
Sprint phones roam on Verizon
so I have never used adb to edit things on my computer do I just open adb up and is there a button to link the 2? i don't really feel like editing these from my phone.
Use root explorer, it makes things soooo much easier.
Is there any change in battery life?
By the way where in the text do i put this syntax's?
Sent from my MB855 using XDA App
Uh what the hell is this shlt. Omg
Sent from my MB855 using XDA App
to pull you have to
Code:
adb pull build.prop/system c:\target folder
Code:
c:\[user_name_\target folder > adb shell
$su
# mount -o remount,rw -t yaffs2 /dev/block/mtdblk3 /system
# exit
$exit
c:\[user_name_\target folder > adb push build.prop /system
c:\[user_name_\target folder > adb shell
$su
# mount -o remount,ro /dev/mtdblock3 /system
#exit
$exit
c:\[user_name_\target folder > adb reboot
Hi everyone! We're back at it again with another DIY guide and this time it's everyone's favorite: Dual Cores.
This guide will assume you don't use any other software that will affect this, such as but not limited to: SetCPU, Virtuous OC.
I'm not sure how Virtuous OC works, as I've never used it, but I included in there just in case. SetCPU has been tested by me to render this useless, as it resets the permissions on the second core.
There are a number of ways to go about doing this:
ramdisk
init.d
init.post_boot.sh
But, the easiest fastest way to get this accomplished is to either upload and post your init.post_boot.sh file, which is located in /system/etc/, OR edit it manually yourself.
There are other ways but this file already exists, and has most of what you need in it and it gets executed at boot anyway.
Take note of the permissions set for this file, because they will need to be set back if you choose to edit outside of Android. You can also edit the file with Android's file editor if you use Root Explorer.
If you choose to edit it yourself and outside of Android, you will need to open the file with Notepad++ (Or whatever qualified editor you use) and add this in anywhere before the line "chown system /sys/devices/system/cpu/cpu0/online"
echo "1" > /sys/devices/system/cpu/cpu1/online
chmod 0444 /sys/devices/system/cpu/cpu1/online
echo "ondemand" > /sys/devices/system/cpu/cpu1/cpufreq/scaling_governor
chmod 0644 /sys/devices/system/cpu/cpu1/cpufreq/scaling_governor
What this does is set the second core's online state to 1 (enabled) and the second line sets this value as read-only, so that the system can't disable its online state when it feels the second core is no longer needed. The third line sets the governor, and I assume you can use any governor that's available in the kernel. However, I'm not sure if both cores can have separate governors. Changing one may change the other. I haven't tested this myself, but sampling frequency and up-threshold works like this, so I assume governors may as well. The last line, if it has anything in common with the third, sets the governor so that you may infact change it with CPU Tuner (since it has the ability to control the second core as well as the first). This line may be obsolete and was only added in as a fail safe.
While you're in there, you can tune your cores even more (assuming you don't use anything like virtuous OC, CPU Tuner, ETC) and set the min and max values by adding this (there should already be something like this for the first core so you can go ahead and add this below that):
echo 192000 > /sys/devices/system/cpu/cpu0/cpufreq/scaling_min_freq
echo 192000 > /sys/devices/system/cpu/cpu1/cpufreq/scaling_min_freq
echo 594000 > /sys/devices/system/cpu/cpu1/cpufreq/scaling_max_freq
As a way to conserve battery life, I set the second core to run at half its max (stock kernel), This works fantastically for CPU Master, which only controls the first core, as you can have all your profiles set, without messing with the second core. (I've found that you can run the first core at 192MHz with the screen off/low battery while still maintaining minimal lag by setting the second core's max to 594MHz!)
Once you've edited the file, you can go ahead and push it back to your device (assuming you've edited it outside of android) and set the permissions back to what they were and reboot.
As always, if any of this is overwhelming, I'll be happy to edit the file for you.
The other methods mentioned above work in pretty much the same fashion (adding those lines in), and if you prefer to use one of those methods, then that's fine as well.
To check whether or not your edits were successful, you can navigate with root explorer to /sys/devices/system/cpu1 and look for "online" which should have permissions of r--r--r-- You can view the file and see if "1" is in there, indicating that the second core is active. You can also use SystemPanel to check your CPU activity, which should show both cores active. As a legend, a grey pie chart means the second core isn't active, but they should both be green and actively being used.
Good luck!
Nice. What sort of difference in user experience have you noticed?
system panel shows second core 100% idle after editing file, also the /sys/devices/system/cpu/cpu1/online file show 0 in view as text instead of 1 any suggestions?
cabbieBot said:
Nice. What sort of difference in user experience have you noticed?
Click to expand...
Click to collapse
much smoother with no lag at all on my end.
rontonomo said:
system panel shows second core 100% idle after editing file, also the /sys/devices/system/cpu/cpu1/online file show 0 in view as text instead of 1 any suggestions?
Click to expand...
Click to collapse
did you reboot after editing? Did you verify that the online file's permissions are r--r--r--? And finally, are you using any cpu controlling software? Setcpu?
what do you recommend besides setcpu to overclock?
cabbieBot said:
what do you recommend besides setcpu to overclock?
Click to expand...
Click to collapse
cpu master or cpu tuner
got it to work by taking out the quotes around the 1
I'm using your latest modified stock kernel. Can you post an additive to add this?
Via My HTC Evo 3D On The Now Network From Sprint.
Rydah805 said:
I'm using your latest modified stock kernel. Can you post an additive to add this?
Via My HTC Evo 3D On The Now Network From Sprint.
Click to expand...
Click to collapse
if you're on the dual core version then this is identical to what's already there. No need to apply.
freeza said:
if you're on the dual core version then this is identical to what's already there. No need to apply.
Click to expand...
Click to collapse
Ahh, just saw it again, dual cores are alive.!!! LOL
Via My HTC Evo 3D On The Now Network From Sprint.
So in theory wouldn't this make ur phone faster? Because after doing some quad testing it seems that my phone is slower. It feels faster but at times the scrolling lags a bit. Anyway to fix the lag? That's if anyone else is having the issue. And I know quads don't mean sh*t but I was curious as to why they would drop.
Sent from my Beastly EVO.. And did I mention it has 3D?
What were the original permission for init.post_boot.sh. I think they were:
RW - R - R
Can someone please confirm?
ericwgarza1 said:
What were the original permission for init.post_boot.sh. I think they were:
RW - R - R
Can someone please confirm?
Click to expand...
Click to collapse
confirmed
youngpro83 said:
So in theory wouldn't this make ur phone faster? Because after doing some quad testing it seems that my phone is slower. It feels faster but at times the scrolling lags a bit. Anyway to fix the lag? That's if anyone else is having the issue. And I know quads don't mean sh*t but I was curious as to why they would drop.
Sent from my Beastly EVO.. And did I mention it has 3D?
Click to expand...
Click to collapse
the reason why there is lag is because your system switches the governor back to userspace after a while... don't know what triggers this, just know that it happens. i have performed this same modification on my phone numerous times and with no success. the phone is smoother managing itself.
i was actually going to ask freeza if he knows of a way to permanently set both cores to ondemand, and permanently have cpu1 online...
that would make it fast. it will stick momentarily, but as soon as that governor is switched back to userspace, you will see lag and a huge performance drop.
---------- Post added at 11:48 AM ---------- Previous post was at 11:36 AM ----------
or maybe i misunderstood the post... and maybe freeze can help me understand here...
so you are saying the lag comes from both cores running at high frequencies? and that if you throttle down the second core, but always have it running, it will improve performance without the lag?
hottubtimemachine said:
the reason why there is lag is because your system switches the governor back to userspace after a while... don't know what triggers this, just know that it happens. i have performed this same modification on my phone numerous times and with no success. the phone is smoother managing itself.
i was actually going to ask freeza if he knows of a way to permanently set both cores to ondemand, and permanently have cpu1 online...
that would make it fast. it will stick momentarily, but as soon as that governor is switched back to userspace, you will see lag and a huge performance drop.
---------- Post added at 11:48 AM ---------- Previous post was at 11:36 AM ----------
or maybe i misunderstood the post... and maybe freeze can help me understand here...
so you are saying the lag comes from both cores running at high frequencies? and that if you throttle down the second core, but always have it running, it will improve performance without the lag?
Click to expand...
Click to collapse
I've set my scaling_governor to read-only and so far "ondemand" has stuck since boot yesterday.
hmmm... i wonder what is causing it to switch back and forth me...
I keep triing this but its not working or maybe I'm doing it wrong, so if I post my file could someone edit it for me? THANKS
Sent from my PG86100 using Tapatalk
MPD CPU1 Governor Control
Treve's MPD CPU1 Governor Control
http://forum.xda-developers.com/showpost.php?p=17456447&postcount=7630
working well with this
Freeza, Good stuff man, thanks for posting this! We appreciate you sharing your knowledge!
kr1m3boss said:
I keep triing this but its not working or maybe I'm doing it wrong, so if I post my file could someone edit it for me? THANKS
Sent from my PG86100 using Tapatalk
Click to expand...
Click to collapse
Hi, sorry for the late response. Have been really busy the past few days.
Did you ever get this working?
Well I'm curious about which composition type would make my phone feel the most smooth. I spent about an hour google-ing and found a single webpage that somewhat explains the differences and I would like someone here to tell me which is the best.
# If you turn on debug.sf.hw you can change the composition type
# with this variable.
#
# cpu (default)
# gpu (default when debug.sf.hw = 1)
# mdp if your hardware supports MDPv4
# c2d if you have c2d support
# dyn use copybit
# asn upcoming kernel interface
#debug.composition.type=
Click to expand...
Click to collapse
I also found a thread that about a person using
debug.composition.type=gpu
debug.composition.type=cpu
in his build.prop and apparently that really made his rom fly, although it made mine have touch responsiveness issues
G4 Man2 said:
Well I'm curious about which composition type would make my phone feel the most smooth. I spent about an hour google-ing and found a single webpage that somewhat explains the differences and I would like someone here to tell me which is the best.
I also found a thread that about a person using
debug.composition.type=gpu
debug.composition.type=cpu
in his build.prop and apparently that really made his rom fly, although it made mine have touch responsiveness issues
Click to expand...
Click to collapse
This change fixed the font corrupted and some graphics issues in my phone, don't feel anything like it's smoother
hung.ngomanh said:
This change fixed the font corrupted and some graphics issues in my phone, don't feed anything like it's smoother
Click to expand...
Click to collapse
Lol I'm glad a question I had last year helped you
No improvements on my Note with CM 10.1. Neither in system UI nor in games.
Maybe this tweak just works on CM7 and doesn't work with Cm 9 or higher
xxLeoxx93 said:
No improvements on my Note with CM 10.1. Neither in system UI nor in games.
Maybe this tweak just works on CM7 and doesn't work with Cm 9 or higher
Click to expand...
Click to collapse
To date the build.prop still uses this to determine the composition type even in cm 10.1
{
"lightbox_close": "Close",
"lightbox_next": "Next",
"lightbox_previous": "Previous",
"lightbox_error": "The requested content cannot be loaded. Please try again later.",
"lightbox_start_slideshow": "Start slideshow",
"lightbox_stop_slideshow": "Stop slideshow",
"lightbox_full_screen": "Full screen",
"lightbox_thumbnails": "Thumbnails",
"lightbox_download": "Download",
"lightbox_share": "Share",
"lightbox_zoom": "Zoom",
"lightbox_new_window": "New window",
"lightbox_toggle_sidebar": "Toggle sidebar"
}
This only works on Gingerbread, not anything else. We are still working on finding alternative tweaks for ICS and JBThis is a MOD to boost the performance of the Adreno 205 GPU in the Xperia Play. (This mod should also work on other Adreno devices (Adreno 205+) and whilst some people have had some success I cannot confirm which devices other than the Play it works with)
This mod is a joint collaboration/venture of me and CosmicDan. However, as of version 0.2 CosmicDan has unfortunately left the Xperia 2011 range for better things. I wish him every success in the future.
CosmicDan was able to find a variant of the 'adreno_config.txt' file that contains settings that should work on our device. Since finding this we have found numerous combinations of settings that increase the performance of our chips. See below.
What does it do?
The configuration file consists of multiple options, each which do their very own unique 'tweaking' to the way the GPU performs. One of the most notable options we are using is 'triJuice', an explanation of which is below:
If our phones were to have their own driver application, what this would be doing is essentially moving the 'Quality/Performance' slider towards 'Performance'. This tells the GPU to concentrate on 'Performance' rather than 'Quality'.
See CosmicDan's explanation here:
It's common graphics stuff and the same for PC's, we have told the GPU to put a focus on performance instead of quality. But the quality is not sacrificed much, it's barely noticeable. Maybe some games will look not so well, you'll have to test and see.
It's like we've lowered effect and texture detail on a system-wide level, allowing the CPU and GPU to give more time to work on geometry and frame updates and such. Useful because many games don't have settings for graphics quality.
Click to expand...
Click to collapse
Another notable option is 'forceAutoTextureCompression'. This saves RAM and makes texture rendering faster, but the actual loading of the game might be a tiny bit slower - however once it's loaded it will have faster rendering.
A list of possible settings for this configuration file are below (For detailed descriptions please see attached)
Post 3 lists which settings tend to increase or reduce performance, I will not give instructions on how to modify this however if you to intend on making your own config file please use Post 3 as a guide.
Code:
; Performance Analysis
performance=normal
disableExtraSwapBlit=0
ignoreGLFlush=0
; Binning
binning=hw
forceGuardband=0
guardbandValue=0
forceGmemSize=0
gmemSize=0
veboSetting=0
veboSetting=0
numBins_weight=80
numGroups_weight=20
; Logging
log.resolves=0
log.pm4=0
log.pm4mem=0
log.shaders=0
log.sc_dev=0
log.sc_dev_shader_name=sc_dev_dump.txt
log.cffdump=0
log.cffdump_with_ifh=0
log.cffdump_no_memzero=0
log.dumpx=0
log.primitives=0
; Debugging
waitForIdleAfterDraw=0
waitForIdleOnSubmit=0
disableSwapTsIdle=0
clockGating=off
useSafeMode=0
redirectDebugMessages=0
forceChipId=Default
; Primitive Conversion.
convertTristrips=default
convertTrifans=default
convertLineloops=default
shader_sub.write=0
shader_sub.read=0
shader_sub.trivialfs=0
; Features and Performance
facenessCulling=default
vboDataAlignment=natural
enableOptimizedTextureUpdates=1
enableOptimizedVboUpdates=1
forceAutoTextureCompression=1
triJuice=1
enableInlineConstantUpdates=1
enableMemoryPool=1
enableFastClears=1
ditherSafeFastClears=0
shadowGmemInAppBuffers=1
textureTiling=0
preserveZStencilOnSwap=0
allowDepthExport=0
untileDynamicTextures=1
fullSurfaceDynamicUpdatePath=1
useGpuTilingHints=1
; MultiSampling Antialiasing (MSAA)
MSAASmoothing=Normal
MSAABufferAllocation=never
forceMSAAMode=0
MSAAMode=0
VAESEnable=0
VAESGenericError=0
VAESFailNth=0
VAESDoNotFailFirstN=0
VAESRandomSeed=0
VAESFailPercent=0
; 2D Settings
2D.HwBlt=1
2D.eglSwapMode=noidle
2D.forceEglSwapInterval=0
2D.eglSwapInterval=0
; LEIA Features
leiaEnableLrzWrites=0
leiaEnableLrzExpansion=0
leiaExportColorForLrzUnresolve=0
leiaEnableFastLrzUnresolves=0
; Oxili settings.
oxiliDisableLazyUpdates=0
oxiliDisableChunkedUpdates=0
oxiliForceShaderDirectUpdates=0
oxiliForceConstantDirectUpdates=0
oxiliForceIstoreCacheMode=1
oxiliForceCstoreSingleBuffer=0
oxiliForceShaderSingleContext=0
oxiliForceSuperthreadMode=1
oxiliForceVsSingleThread=0
oxiliForceFsSingleThread=0
oxiliForceSingleSp=0
oxiliSkipClears=0
oxiliForceSysmemRender=1
; Other settings
FPSCap=60
allowFloatFBOs=1
suppressTimestampInterrupts=0
GPUIdleTimeout=off
GPUIdleTimeoutMsec=0
Please ignore the values of the above settings unless otherwise stated. The majority of these are stock/default values for our device.
Downloads - Official repository
AdrenoBooster v0.7
AdrenoBooster v0.6
AdrenoBooster v0.5 - Quality Edition
AdrenoBooster v0.4.1 - Battery Edition
AdrenoBooster v0.3 - Minimal Edition - This version will give you the best performance - stability ratio. Whilst the other versions may give you better performance but in some rare instances lower quality/artifacts, this version should give you the best of both worlds.
AdrenoBooster v0.1
AdrenoBooster v0.2
Instructions
Download and copy the ZIP to SD Card. Then flash using CWM.
NOTE: Please ensure you reboot your device after the first boot post-installation of the mod or it will not be active.
Requirements
Init.d support
Root
Any Gingerbread ROM
Screenshots
IMPORTANT!! - Please ensure you thank 'CosmicDan' as well for this fantastic mod. A huge portion of the work has been done by him!
(See 5th post if you would like to 'Thank' him)
Troubleshooting
First, give yourself another reboot - Just in case!
If it still doesn't seem to be working for you, check to see if the files have copied to your device. Check the following locations for the following files with any file manager with Root support.
/system/etc/init.d - Filename: 93adreno
/system/etc - Filename: adreno_config.txt
You should also see adreno_config.txt in the following location if the init.d script is working correctly: /data/local/tmp
If this file is not in this location then chances are you do not have init.d support.
Black screen on boot? See here: http://forum.xda-developers.com/showpost.php?p=40013461&postcount=377
Extras
You can assist with the testing of some of these settings by doing the following:
Open a new text file in a standard text editor (Notepad++ or Notepad for Windows)
Choose values from the above post to put into your configuration file. (Use the attached adreno_config.txt file as an indication of what setting does what)
Save this new file as 'adreno_config.txt'
Copy this text file to your phones SD Card.
Open your File Manager on your phone (I use ES File Explorer)
Prepare adreno_config.txt on your SD Card for copying
Navigate to /Data/local/tmp and paste the file there
Reboot.
Please be aware that if you have any cleaner init.d scripts this will NOT work, as when you reboot /data/local/tmp will be deleted.
Devs/Chefs/Tinkerers
If you would like to add this to your ROM please simply drop a short request in this thread or PM.
After which, please ensure proper credit is given.
Current Antutu Highscore
(With the benefits of this mod) - By CrypticRook
Manually Uninstalling the Mod
Navigate to the following locations and remove the files.
/system/etc/init.d - Filename: 93adreno
/system/etc - Filename: adreno_config.txt
/data/local/tmp - Filename: adreno_config.txt
Tested Settings
Untested:
preserveZStencilOnSwap - enabling might improve performance at the increased risk of visual artifacts
Dangerous:
facenessCulling - Turning on causes crash on boot
FPSCap - setting to anything other than 0 (even to 60 or 100) causes unstable 2D rendering
2D.eglSwapMode=interrupt - Causes unstable/looping 2D rendering. Applications fail to initialize.
fullSurfaceDynamicUpdatePath - reduces 3D performance by around 10%.
Performance Boosts:
forceAutoTextureCompression - Turning this on seems to help a LOT with 3D performance. Might increase load times by a a tiny amount.
triJuice - setting it to the max value of 3 increases particle/shader/lighting performance a LOT with a minor loss in quality
forceMSAAMode - enabling this will force no anti-aliasing as long as MSAAMode is left at 0. Could increase performance on some things but make them look very chunky
2D.HwBlt - Enabling this should enhance GPU hardware acceleration in gingerbread. It says default is enabled but I've set it to 1 anyway.
oxiliForceVsSingleThread=1 - Must be enabled with oxiliForceFsSingleThread=1 for performance increase. Prolonged usage has negative impact (Needs more testing).
oxiliForceFsSingleThread=1 - Must be enabled with oxiliForceSingleSp=1 for performance increase. Prolonged usage has negative impact (Needs more testing).
oxiliForceSingleSp=1 - Must be enabled with oxiliForceVsSingleThread=1 for performance increase. Prolonged usage has negative impact (Needs more testing).
clockGating - Turning this on might save power consumption (Currently being tested more to confirm). However there is no performance drop by having this enabled
Seemed to hurt performance: (I did not test these much, could do with more tests one-by-one)
leiaEnableLrzExpansion - enabling might.... do something.
leiaEnableFastLrzUnresolves - enabling might improve performance
oxiliDisableChunkedUpdates - enabling might improve performance
oxiliForceShaderDirectUpdates - enabling might improve performance
oxiliForceConstantDirectUpdates - enabling might improve or reduce performance
oxiliForceShaderSingleContext - enabling might improve performance at cost of quality
oxiliForceSuperthreadMode - enabling might improve performance or reduce it. Probably conflicts with above one.
suppressTimestampInterrupts - enabling might improve or reduce performance
Null/No Difference - These options made no effect on performance from their default settings
shadowGmemInAppBuffers - no effect on performance or quality
Great! Really looking forward to seeing what you can do
Thanks for letting me know you opened a topic
I've been doing some research and here's what ideas I've found so far:
clockGating - Turning this on might save power consumption. Will need to test if it has a performance hit.
facenessCulling - Turning on causes crash on boot
forceAutoTextureCompression - Turning this on MIGHT save RAM and/or MIGHT increase load times/CPU usage.
triJuice - raising this value might increase performance but decrease quality of mipmapping.
shadowGmemInAppBuffers - disabling might improve performance but break some things
preserveZStencilOnSwap - enabling might improve performance at the increased risk of visual artifacts
fullSurfaceDynamicUpdatePath - enabling might improve performance at the increased risk of visual artifacts
forceMSAAMode - enabling this will force no anti-aliasing as long as MSAAMode is left at 0. Could increase performance on some things but make them look very chunky
2D.HwBlt - Enabling this might enhance GPU hardware acceleration in gingerbread. It says default is enabled but I've set it to 1 anyway.
2D.eglSwapMode - changing to interrupt mode might.... do something. LETS TRY IT!
leiaEnableLrzExpansion - enabling might.... do something.
leiaEnableFastLrzUnresolves - enabling might improve performance
oxiliDisableChunkedUpdates - enabling might improve performance
oxiliForceShaderDirectUpdates - enabling might improve performance
oxiliForceConstantDirectUpdates - enabling might improve or reduce performance
oxiliForceShaderSingleContext - enabling might improve performance at cost of quality
oxiliForceSuperthreadMode - enabling might improve performance or reduce it. Probably conflicts with above one.
oxiliForceVsSingleThread - enabling might improve or reduce performance
oxiliForceFsSingleThread - enabling might improve or reduce performance
oxiliForceSingleSp - enabling might improve or reduce performance
FPSCap - setting to 30 or 60 might help with all-round performance and reduce lag spikes. Maybe.
suppressTimestampInterrupts - enabling might improve or reduce performance
Right, that's a lot of things for me to try one by one. If anyone else wants to, go for it. Just remember it could completely break your boot
Re: [WIP] Adreno Configuration Settings [Improved Performance?]
I did lol.. In the adreno thread xD
Ill make it more clear next time.
I've just tried a few things... Managed to make Antutu crash lot. Just finally completed a full benchmark but then the OS died. Going well so far. Cya in about 7 hours!
Edit: it looks like you can copy it to data/local/temp and the settings will apply for next boot. After which it is then removed from the directory. Which for me makes things a bit easier!
Sent from my R800i using xda app-developers app
Oh by the way, copying the config file to /system/lib/egl/ definitely does nothing. I tested by setting FPS cap to 5, no effect. File needs to be at /data/local/tmp/ - it does *not* get wiped on reboot so its OK.
EDIT: THe file isn't removed from /data/local/tmp/ for me =\ maybe you have an init.d script that erases it or something.
Re: [WIP] Adreno Configuration Settings [Improved Performance?]
Possibly. Ill check.
Btw, I just managed 13fps on the OpenGL ES2.0 test on Antutu. It was hilarious, so many artifacts... But it passed it!
...Until it died on the SD card test. If you use Antutu I suggest doing custom tests and taking SD out. Or reducing OCs. I think my 1.6ghz may be the problem.
Or perhaps we should use stock clock a to base our tests on?
Edit: seems like almost every setting you try also kills the bootanimation lol
Sent from my R800i using xda app-developers app
Spizzy01 said:
Possibly. Ill check.
Btw, I just managed 13fps on the OpenGL ES2.0 test on Antutu. It was hilarious, so many artifacts... But it passed it!
...Until it died on the SD card test. If you use Antutu I suggest doing custom tests and taking SD out. Or reducing OCs. I think my 1.6ghz may be the problem.
Or perhaps we should use stock clock a to base our tests on?
Edit: seems like almost every setting you try also kills the bootanimation lol
Sent from my R800i using xda app-developers app
Click to expand...
Click to collapse
I can't get past the 2D/sprite test in Antutu (the little Androids) it freezes at the end lol. Boot animation was always OK for me....
Yeah, using a stock clock would be a good idea. But I'm on 1.4 ghz anyway because that's what I've always used and always been stable on.
EDIT: I think setting FPSCap to 60 was the reason for Antutu freezing on 2d test.
EDIT2: You are using LuPuS GB kernel right? Because Turbo Kernel has backported KGSL drivers, so that's probably why we not only have different performance scores but may have different results with these configs.
---------- Post added at 01:24 PM ---------- Previous post was at 01:13 PM ----------
DUDE! New Gingerbread record!
/data/local/tmp/adreno_config.txt:
Code:
facenessCulling=off
forceAutoTextureCompression=1
triJuice=3
2D.HwBlt=1
CosmicDan said:
I can't get past the 2D/sprite test in Antutu (the little Androids) it freezes at the end lol. Boot animation was always OK for me....
Yeah, using a stock clock would be a good idea. But I'm on 1.4 ghz anyway because that's what I've always used and always been stable on.
EDIT: I think setting FPSCap to 60 was the reason for Antutu freezing on 2d test.
EDIT2: You are using LuPuS GB kernel right? Because Turbo Kernel has backported KGSL drivers, so that's probably why we not only have different performance scores but may have different results with these configs.
---------- Post added at 01:24 PM ---------- Previous post was at 01:13 PM ----------
DUDE! New Gingerbread record!
/data/local/tmp/adreno_config.txt:
Code:
facenessCulling=off
forceAutoTextureCompression=1
triJuice=3
2D.HwBlt=1
Click to expand...
Click to collapse
OMFG! LEGEND!
Gonna test on mine and report back ASAP. Gimme 10 - 20 mins, depending on when I can get a free sec @ work. Lol. xD
Edit: Unable to replicate your score at the moment. Getting stock scores, most likely something to do with those pesky init.d scripts. Deleting now and will report back shortly
In the next turbo kernel release I'll make the kernel do a symlink from /data/local/tmp/adreno_config.txt to /system/etc/adreno_config.txt (it will be linked before init starts so will apply straight away), that way we can include modified config with ROM's.
You could just make an init.d script do the same thing, but then the ROM will need to be rebooted again (because adreno driver is already loaded).
CosmicDan said:
In the next turbo kernel release I'll make the kernel do a symlink from /data/local/tmp/adreno_config.txt to /system/etc/adreno_config.txt (it will be linked before init starts so will apply straight away), that way we can include modified config with ROM's.
You could just make an init.d script do the same thing, but then the ROM will need to be rebooted again (because adreno driver is already loaded).
Click to expand...
Click to collapse
I think I'll include an init.d script with my ROM, so that in the event someone isn't using your Kernel it will still work as intended.
...At least, after a reboot.
Doing Antutu now btw, ITS CRAZY FAST OMG. About to give you results. UNO MOMENTO!
OMFG!
This actually brought a tear to my eye... Lmfao... XD
Spizzy01 said:
I think I'll include an init.d script with my ROM, so that in the event someone isn't using your Kernel it will still work as intended.
...At least, after a reboot.
Doing Antutu now btw, ITS CRAZY FAST OMG. About to give you results. UNO MOMENTO!
Click to expand...
Click to collapse
Yeah I realized that too, already done it for Turbo UI Classic (which is uploading now). This should work:
/system/etc/init.d/93adreno:
Code:
#!/system/bin/sh
#
if [ ! -h /data/local/tmp/adreno_config.txt ] then
ln -s /system/etc/adreno_config.txt /data/local/tmp/adreno_config.txt
fi
EDIT:Woohoo! Play broke the 7000 mark
Now I wonder how Jellybean on Turbo Kernel performs.... maybe closer to 8000 lol! And the visual quality of the orc fight 3D test looked OK for you yeah?
CosmicDan said:
Yeah I realized that too, already done it for Turbo UI Classic (which is uploading now). This should work:
/system/etc/init.d/93adreno:
Code:
#!/system/bin/sh
#
if [ ! -h /data/local/tmp/adreno_config.txt ] then
ln -s /system/etc/adreno_config.txt /data/local/tmp/adreno_config.txt
fi
Will be good to know if the results on LuPuS GB kernel are worse, better or the same with this config.
Click to expand...
Click to collapse
My test was done on LuPuS GB.
Sorry - I moved back from your Kernel last night. =x
Thank's for the script. I'll add it to v0.5 Aurora now. xD
Edit: Agreed. JellyBean should have crazy scores... Right, I'm gonna test a few of the other configs. From that list you've done, can you 'tick' off which you have already checked please? Just so I can continue where you left off at.
Orc fight looked perfectly fine. If it were a HD movie I'd say it had a low bit-rate, but it's not... So I have no idea what to call it, but it does look ever so slightly more grainy. But this is barely noticeable at all.
I looked in /data/local/tmp/ and there was no adreno_config.txt I'm using joka wild any ideas did I have to do something before hand using LuPuS v6 480p I'm very interested because I use this a a gaming device as I have a nexus 4
Spizzy01 said:
My test was done on LuPuS GB.
Sorry - I moved back from your Kernel last night. =x
Thank's for the script. I'll add it to v0.5 Aurora now. xD
Edit: Agreed. JellyBean should have crazy scores... Right, I'm gonna test a few of the other configs. From that list you've done, can you 'tick' off which you have already checked please? Just so I can continue where you left off at.
Click to expand...
Click to collapse
I edited my last post since you uploaded results
That's OK, I'll still win the record by being first to test Turbo UI (JB) score lolz
EDIT: OK, I'll edit that list.
extremetempz said:
I looked in /data/local/tmp/ and there was no adreno_config.txt I'm using joka wild any ideas did I have to do something before hand using LuPuS v6 480p I'm very interested because I use this a a gaming device as I have a nexus 4
Click to expand...
Click to collapse
You need to move the file there yourself.
See attached.
Move the file to /Data/Local/tmp and reboot. Ensure you have no init.d scripts that clear cache or tmp though, as it will not work.
OK here's what my data is.
Untested:
clockGating - Turning this on might save power consumption. Will need to test if it has a performance hit.
shadowGmemInAppBuffers - disabling might improve performance but break some things
preserveZStencilOnSwap - enabling might improve performance at the increased risk of visual artifacts
fullSurfaceDynamicUpdatePath - enabling might improve performance at the increased risk of visual artifacts
oxiliForceVsSingleThread - enabling might improve or reduce performance
oxiliForceFsSingleThread - enabling might improve or reduce performance
oxiliForceSingleSp - enabling might improve or reduce performance
Dangerous:
facenessCulling - Turning on causes crash on boot
FPSCap - setting to anything other than 0 (even to 60 or 100) causes unstable 2D rendering
Performance Boosts:
forceAutoTextureCompression - Turning this on seems to help a LOT with 3D performance. Might increase load times by a a tiny amount.
triJuice - setting it to the max value of 3 increases particle/shader/lighting performance a LOT with a minor loss in quality
forceMSAAMode - enabling this will force no anti-aliasing as long as MSAAMode is left at 0. Could increase performance on some things but make them look very chunky
2D.HwBlt - Enabling this should enhance GPU hardware acceleration in gingerbread. It says default is enabled but I've set it to 1 anyway.
Seemed to hurt performance: (I did not test these much, could do with more tests one-by-one)
2D.eglSwapMode - changing to interrupt mode might.... do something. LETS TRY IT!
leiaEnableLrzExpansion - enabling might.... do something.
leiaEnableFastLrzUnresolves - enabling might improve performance
oxiliDisableChunkedUpdates - enabling might improve performance
oxiliForceShaderDirectUpdates - enabling might improve performance
oxiliForceConstantDirectUpdates - enabling might improve or reduce performance
oxiliForceShaderSingleContext - enabling might improve performance at cost of quality
oxiliForceSuperthreadMode - enabling might improve performance or reduce it. Probably conflicts with above one.
suppressTimestampInterrupts - enabling might improve or reduce performance
I made a quick Update ZIP to flash the mod/script and updated the first few posts.
You have been fully credited of course in the updater_script
Gonna look into other settings now.
Edit: I'm not sure I like the new Mediafire layout :|