Improve performance tweak

[dishe]

I've got a hunch regarding why the stock launcher is so slow versus other launchers, and I'd love for someone with a Marquee to try this out.

Before doing anything else (for the sake of a controlled experiment) download Quadrant, neocore, or any benchmarking app that display frames per second, and run it first, taking note of the displayed fps.We're going to see if that changes later.

Then (assuming you are rooted), remove or move libGLES_android.so from the system/lib/egl directory. I would recommend moving instead of deleting, just in case you want to put it back (as some apps may rely on it to work).

Reboot and run the benchmark again. Is the FPS higher now? How about the stock launcher? If so, its a standard issue that has to do with Android and the PowerVR GPU. Apparently, the Optimus Black runs its own launcher which knows how to utilize the GPU properly, whereas the stock Android launcher on the Marquee relies on this driver (which is conflicting with GPU acceleration). Removing the driver forces it to use the GPU (in most cases- in other cases it may cause a force close). That would really explain a lot of the initial experience complaints I was talking about!

Report back if you can!

 

[nosajsam]

Moving the libGLES_android.so file did produce different end results w/ running Quadrant Standard

1st run:
1st Graphic: 13 FPS
2nd Graphic (Hallway) : 14-20 FPS
3rd Graphic (Moons): 8-9 FPS
4th Graphic (Blue/Red): 12-14 FPS
Quadrant Score: 851

2nd Run
1st Graphic: 13 FPS
2nd Graphic (Hallway) : 20 FPS
3rd Graphic (Moons): 40 FPS
4th Graphic (Blue/Red): 20 FPS
Quadrant Score: 1207

I appreciate the effort & your suggestion here. Although my results showed improved numbers, I didn't really notice "improved performance". I am not looking to fight w/ you but if I could make a suggestion to change this threads title as it could be perceived as misleading (I came here thinking there was a tweak here to improve performance). Yes there is an increased Quadrant score but there was ZERO noticeable difference w/ the stock launcher lag (unfortunately, the lag is still there).

My opinion is the stock launcher lag may be caused from a poor value set internally for its scrolling speed & opening/closing speed...I feel this way because once I tried other Launchers, the phone seemed to fly...

Personally, I am not really looking for the stock launcher to work faster as even if it was buttery smooth I would prefer to use a Launcher that I can customize a bit more (for example, I am not a fan of having a big "D" on the homescreen that I cannot remove )

For anyone considering this phone, If you have the phone & feel its "laggy" or want to be able to change your phones appearance, try out a different Launcher app from the market. There are quite a few of them, my suggestion is to try them out & find the one that works best for you. I can tell you this, for me installing Launcher Pro has definitely improved my overall feel of the phone. If the phone came stock w/ Launcher Pro installed there would be NO ONE talking about a "laggy" Marquee

 

[dishe]

::scratches head:: you don't consider the openGL test going from 8-9FPS to a very respectable 40FPS a performance tweak?

Its a shame it didn't have any effect on the stock launcher, I was really hoping they would be related. But this is still something- as you pointed out, the stock launcher is sort of irrelevant at this point. But this tweak means certain heavy GPU related 3D games that may have had poor performance previously will run as smoothly as other modern handsets (as they should have all along).

Have you noticed any app or widget compatibility broken with this?

 

[nosajsam]

::scratches head:: you don't consider the openGL test going from 8-9FPS to a very respectable 40FPS a performance tweak?

Its a shame it didn't have any effect on the stock launcher, I was really hoping they would be related. But this is still something- as you pointed out, the stock launcher is sort of irrelevant at this point. But this tweak means certain heavy GPU related 3D games that may have had poor performance previously will run as smoothly as other modern handsets (as they should have all along).

Have you noticed any app or widget compatibility broken with this?

There is an obvious difference in numbers between 9 FPS & 40 FPS but because of how I use the phone, this change was unnoticeable. Let me clarify so to not be misunderstood, I didn't notice any change in the way MY phone was running with this change.

I use my phone for well....a phone.

I call, text, occasionally check email/go online. I am not a high tech, 3d gamer type looking to hopefully one day use their phone to launch the next space shuttle but more of a casual, average daily phone user.......yep, basically I'm boring.

With that said, I might be the wrong person to be doing this type of test as I really don't have many apps installed. Except for extended controls, I don't have or use any widgets & I don't use my phone to play games etc....but for someone who does, yes it is an improvement.

I do have a little bit more than average tech knowledge & experience (just unfortunately nowhere near enough to really get into & dissect this phone)

Since I do have the phone I am willing to try out any apps/widgets/games etc. that you think may benefit from this to see if there is any problems/compatibility issues....let me know.

 

[nosajsam]

follow up
Out of curiosity I moved the libGLES_android.so file back, rebooted & ran quadrant standard again.
3rd run:
1st Graphic: 13 FPS
2nd Graphic (Hallway) : 14-20 FPS
3rd Graphic (Moons): 8-9 FPS
4th Graphic (Blue/Red): 12-14 FPS
Quadrant Score: 1158

As you can see although the FPS numbers have once again decreased while the overall end score isn't too far off. Maybe you can add some incite as to why this is (sometimes these end numbers don't always make sense to me).

 

[dishe]

There is an obvious difference in numbers between 9 FPS & 40 FPS but because of how I use the phone, this change was unnoticeable. Let me clarify so to not be misunderstood, I didn't notice any change in the way MY phone was running with this change.

I use my phone for well....a phone.

I call, text, occasionally check email/go online. I am not a high tech, 3d gamer type looking to hopefully one day use their phone to launch the next space shuttle but more of a casual, average daily phone user.......yep, basically I'm boring.

Wow, you sound like someone who has been using a Touch Pro 2 device for too long... Oh wait.

I'm just messing with you. But people do expect their handsets to be more than "just a phone" these days.

Anyway, back on topic- Things like augmented reality, 3D simulations and real time processing effects are likely effected by tweaks like this, not just games. Heck, having the GPU handle more of the load might actually take some stress off the CPU and save battery power, even if you don't notice more fps in benchmarks.
Who knows. Maybe keep an eye on that and see what else is effected by this change?

Since I do have the phone I am willing to try out any apps/widgets/games etc. that you think may benefit from this to see if there is any problems/compatibility issues....let me know.

Not sure off the top of my head, but while you're at it try neocore from the market. Just for giggles, I'm curious.

follow up
Out of curiosity I moved the libGLES_android.so file back, rebooted & ran quadrant standard again.
3rd run:
1st Graphic: 13 FPS
2nd Graphic (Hallway) : 14-20 FPS
3rd Graphic (Moons): 8-9 FPS
4th Graphic (Blue/Red): 12-14 FPS
Quadrant Score: 1158

As you can see although the FPS numbers have once again decreased while the overall end score isn't too far off. Maybe you can add some incite as to why this is (sometimes these end numbers don't always make sense to me).

Well, the benchmark tool tests all kinds of things- databse queries, h.264 encoding/decoding, memory read/write speeds, etc. The problem is, sometimes those numbers can vary wildly based on what the handset is doing. For example, a database query could be cached in a particularly fast memory address and give the result sooner on one test versus another. Or perhaps Android had to close an application in the background the first time in order to make room for the H.264 decoding results. Android is very complex, and this is why benchmark scores must be taken with a grain of salt. There is usually a range of values- I try to take at least 3 tests and average them out.

Meanwhile, that's why I specifically wanted you to note the fps. I'm aware that the final score numbers can change regardless of the GPU optimization, and don't particularly care about those right now. I wanted to see how fast it can output to the screen, and that should be relatively consistent.
(note that in the video I made for the transform ultra, I was impressed that the moon test displayed over 80fps. That's what got me more than the final score)

 

[nosajsam]

just a fyi, for each result above I had rebooted before running the quadrant standard (& ran it a couple of times) but the above posted results were the same

 

[tajai perry]

ahhh look my phone is a tmobile prism thats rooted and it has cm7 on it.my fps used to be around 19 fps but i got it to 60 fps using ram booster pro and ram expander with a swap file about 2.4gb. try these apps cause they helped me