Just...WOW...massive speed increase using ART cache

[Johnly]

I read at XDA that titanium doesnt work with ART turned on.

Does not surprise me.

Sent from my Nexus 5 using Tapatalk

 

[scorpiodsu]

I will try this. The only thing I maybe concerned about is the extra storage spacethat could be taken up. For small apps, no big deal but for some of the larger could make a difference on a device with less than 16GB usable space. But hey, for better performance, could be worth it.

 

[Shilohcane]

I assume it will let you convert back too. I am still seeing the Dalvik setting in developer options.

I understood Dalvik to ART since I have worked on both Emulators and Compilers and know why ART will be faster. ART is still in Beta and I want to know about all the possible side effects first. With the ART speed advantages that brings great power it must have some issues since ART didn't become the Kit Kat default.

 

[scorpiodsu]

I understood Dalvik to ART since I have worked on both Emulators and Compilers and know why ART will be faster. ART is still in Beta and I want to know about all the possible side effects first. With the ART speed advantages that brings great power it must have some issues since ART didn't become the Kit Kat default.

Google tends to run stuff as better for a loooooooong time. Navigation is technically still in beta. Also on the other hand, Google sometimes removes the beta tag when something is still a beta product..... google tv as an example lol. Very hard to interpret beta from Google.

 

[Shilohcane]

This is a interesting Article.

Android 4.4 KitKat offers ART as alternative runtime to Dalvik

Google and Oracle have been tied up in a fight for a while over how Java works in the Android operating system. In 2012, Oracle specifically targeted the Dalvik virtual machine that runs on Android devices in its legal assault against Google, though Google eventually prevailed in that case. Google has been working toward replacing Dalvik in secret and users running Android 4.4 will be able to use Google’s new alternative.

 

[Jdane07]

I read an article where it warned against using it on AOSP ROMs. Is everyone rooted? Or running stock software? I'll probably wait a while before I try it, my N5 is a lightning bolt as it is.

Posted via Android Central App

 

[JeffDenver]

I understood Dalvik to ART since I have worked on both Emulators and Compilers and know why ART will be faster. ART is still in Beta and I want to know about all the possible side effects first. With the ART speed advantages that brings great power it must have some issues since ART didn't become the Kit Kat default.

If I was to guess, I'd say it is beta because they need to solve the compatibility issues with legacy devices. The Titanium backup example is a good one, and something I myself would never have thought about until it was too late.

 

[JeffDenver]

I read an article where it warned against using it on AOSP ROMs. Is everyone rooted? Or running stock software? I'll probably wait a while before I try it, my N5 is a lightning bolt as it is.

I am running stock. I have no plans to root yet.

 

[greydarrah]

It would be interesting to see Quadrant scores before and after enabling ART. I know a lot of trash gets thrown at benchmarks, but I think they have relevance when comparing a change you make on your own phone as opposed to using them to compare your phone to others.

 

[javier180]

i am rooted and it works great on my phone

 

[javier180]

what im also curious if this could improve battery performance.

 

[zkSharks]

what im also curious if this could improve battery performance.

On mobile devices, computational efficiency typically has a strong correlation with power draw. The degree of improvement we'll see from ART, however, I can't say. I'll be using it this evening, and quite a few people on Google+ (and the android subreddit) are talking about it.

ART has issues with Titatnium Backup as well as WhatsApp.

 

[JeffDenver]

It would be interesting to see Quadrant scores before and after enabling ART. I know a lot of trash gets thrown at benchmarks, but I think they have relevance when comparing a change you make on your own phone as opposed to using them to compare your phone to others.

My benchmark scores have not really changed. I ran Antutu and 3D mark.

 

[Shilohcane]

If I was to guess, I'd say it is beta because they need to solve the compatibility issues with legacy devices. The Titanium backup example is a good one, and something I myself would never have thought about until it was too late.

I would think they would have do what we do backing up data bases on large storage subsystem. Different storage manufactures reduces the physical size of data using propitiatory algorithm to shrink the total size. Reducing the data size on the disk is called dehydrating and when you go to back up the data you always rehydrate to the original data that is portable to any environment.

 

[Shilohcane]

My benchmark scores have not really changed. I ran Antutu and 3D mark.

That is because you aren't testing the ART improvement that way. You only reduced the CPU overhead one time of loading that one benchmark Application by eliminating the conversion from bytecode to snapdragon 800 native executable instructions. With ART it is the process of starting up the Application over and over that reduces the need to convert from bytecode to native snapdragon 800 instructions.

With Dalvik using JIT if you start up that benchmark a 1000 times you would have to convert from bytecode to native snapdragon 800 instructions every time the App loaded. With ART you convert the bytecode to native snapdragon 800 native executable instructions when you installed the benchmark application on your N5. Now with ART every time you start the benchmark it never had to do that conversion again. With ART you eliminated the short conversions loading the benchmark but what you timed was the benchmark running.

 

[Adranalyne]

My benchmark scores have not really changed. I ran Antutu and 3D mark.

Try Quadrant with and without. I think you'll see the difference there.

 

[JeffDenver]

Try Quadrant with and without. I think you'll see the difference there.

12,392. Did not run before converting to ART...don't want to spend the time to go back to see what it'd be like before. Thats a good score though (the same as the HTC One).

 

[JeffDenver]

That is because you aren't testing the ART improvement that way. You only reduced the CPU overhead one time of loading that one benchmark Application by eliminating the conversion from bytecode to snapdragon 800 native executable instructions. With ART it is the process of starting up the Application over and over that reduces the need to convert from bytecode to native snapdragon 800 instructions.

With Dalvik using JIT if you start up that benchmark a 1000 times you would have to convert from bytecode to native snapdragon 800 instructions every time the App loaded. With ART you convert the bytecode to native snapdragon 800 native executable instructions when you installed the benchmark application on your N5. Now with ART every time you start the benchmark it never had to do that conversion again. With ART you eliminated the short conversions loading the benchmark but what you timed was the benchmark running.

I understand what you're saying, but how would that explain the IN-APP improvements I have seen in the gallery and Maps and Google Earth? It is having some kind of effect. They are definitely faster with ART. And not just start-up.

 

[chrisgaza]

where do i get the art cache

 

[dammitcubs]

LOL...startup...is that how we measure performance now? What is this?....2008?