Joined
·
92 Posts
This post is being edited and will be back up with a brand new OP shortly
View attachment 571
Attachments
-
1.8 KB Views: 580
have you even tried the build.prop? im guessing you havent, it dramatically improves the phones, speed, smoothness, and battery lifeAsX said:I thought Google enabled JIT on all 2.2 versions. Maybe some manufacturers chose to overwrite this but on our phone it is definitely enabled by default.
The bigger heapsize, the less often garbage collector has to run, which is definitely a good thing as the garbage collection is very expensive thing in terms of resources. There is a tradeoff between the total memory amount and the size of the heap, but on our phones 48MB seems right. It certainly should be better than 32MB. 32 was probably a tweak for some other phones with 256MB of RAM that actually increased it from even lesser default number. On our phone decreasing it doesn't make much sense, you'd only lower performance and decrease battery life in the end.
There is another tweak that stands out as a little bit excessive. It's windowsmgr.max_events_per_sec set to 200. This value tells how often Windows Managers can send events to a window. The default value is 55, which is about every 18ms. 200 changes it to 5ms. Looks too extreme, there is no need to put that much strain on the manager and consequently the CPU. It may increase responsiveness slightly, but at the cost of utilizing more CPU and battery. I'd leave this at default or at lease set it to something way more conservative, like 16-14ms (values between 60 and 70.)
That's exactly my point. I try to avoid putting anything on my phone without understanding what it does.dsexton702 said:seriously though i do appreciate your opinionit is always good to know EXACTLY what your about to put into your phone
yea i believe bali does set it to 48 as an automaticAsX said:That's exactly my point. I try to avoid putting anything on my phone without understanding what it does.
It's hard to test our phone by just running it, it's so fast already that any perceptive increase or decrease in speed can be attributed to a placebo effect.The battery life is hard to test either because you'd need to run the phone exactly the same way for few days to compare the results, which is course impossible. There are objective tests that can be created, but they really have to be written to test certain aspects of the phone. For example to test the Dalvik heapsize performance one would need to write a program that allocates and deallocates lots of memory, time it for a while, and compare results.
Btw, I there is still a question was the 48MB of heap size the Samsung/Android default or was it tuned by the Bali kernel. If it was Bali, maybe there is a reason to lower it to 32B after all.
I that case you'd be competing with mighty drhonk himselfdsexton702 said:yea i believe bali does set it to 48 as an automatic
Interesting, would you mind if I used the tweak block as a starting base to find similar tweaks for the Nexus S? Proper credit and everything of course, but it's up to you. Particularly interesting because being a T-Mobile device too it could be very similar except the 4G parts (well HSPA)."dsexton702 said:Hey guys, ive been tweaking the build.prop for the sgs4g for a day or two now and i think i finally hit the sweet spot. Enjoy
To install simply put on your sdcard and flash in cwm
for GB users here's one for you guys(bottom of OP)
UPDATE: improved build.prop 7/17/11
http://www.mediafire.com/file/x8k9igfa84aox4m/Performance.zip