code cat

출처: http://developer.sonymobile.com/wp/2012/03/30/learn-about-the-technical-differences-between-gingerbread-and-ics/

Learn about the technical differences between Gingerbread and ICS [Updated]

Ever thought about how Gingerbread (GB) and Ice Cream Sandwich (ICS) platforms differ on a technical level? In this blog post, we’ll describe some of the technical differences between GB and ICS, and what the differences in the user experience might be. This way you can decide if ICS is right for you, or if you prefer to stay on Gingerbread. Maybe you will prefer the new UI in ICS, or do you give a higher priority to the extreme stability of the Gingerbread platform? Read more after the jump!


Now as you might have seen, we’ve continuously kept you updated on our work with the ICS upgrade, and we started by telling you about what we do to get the latest software release from Google working on our Xperia™ smartphones in the article Ice Cream Sandwich – from source code release to software upgrade. Then we released ICS alpha and ICS beta versions of the coming software upgrade.

However, although ICS is new and compelling in many ways, we would like all of our users to make an informed decision when selecting what Android™ software to use. We are actually proud to say that our Gingerbread software is very stable and has great performance, so it’s not a bad idea to stay on this release. Ice Cream Sandwich is more intensive, for example in terms of resource usage. As smartphones become more capable, our own applications, as well as the Google Mobile Services (GMS) applications, are becoming more advanced, which means that they require more CPU power, run more network activities and use more RAM. On the other hand, ICS brings a refined UI and some nice new features as described below.


Comparison of the look and feel in Gingerbread (left) and Ice Cream Sandwich (right).

New features in ICS
From a UI perspective, ICS is based on a new look and feel, the Holo theme. In order to accommodate the new look of Android, we decided to do an extensive touch up of our own assets, since the graphical assets of the Holo theme cannot be changed in any way as stated in the Android Compatibility Definition Document (CDD). New looks have been added in the platform layer as well as in the application layer. All in all, well over a thousand icons have been modified. In addition, we have deployed new wallpapers and application backgrounds, which harmonise more with the flatter graphical structures of ICS.

In ICS, the activity manager has a completely new UI, where all running apps are shown as thumbnails in a list. To close an activity, you can simply swipe it out of the list. ICS also introduces a face recognition app as a way to unlock the phone, called Face Unlock. Face Unlock uses the front-facing camera and advanced object recognition algorithms. It is included in our ICS upgrade for all phones that have a front-facing camera.

The contact list will show more information about the contacts, including updates from social networks. In the calendar, colour coding has been added and it is now possible to zoom. There is also support for a new type of voicemail that is more visual, offering transcriptions of voice messages.

When it comes to ICS, it’s a major upgrade of Android™, and there are a lot of things that have changed compared to the Gingerbread release. Some of these changes affect the performance and stability of the system, for example by using more CPU power and RAM. ICS was developed with Galaxy Nexus in mind, which is based on a TI platform with dual-core processor and 1GB RAM. We are now adapting ICS to run on our 2011 Xperia™ smartphones, which are all built on a Qualcomm platform with single core and 512 MB RAM. This means that in some cases, the resource usage in ICS is heavier on the system compared to Gingerbread. The following sections identify some key areas where there is a difference between ICS and Gingerbread.

Increased RAM usage
In general, it can be said that the RAM is the working memory in the phone, used by running processes in contrast to the flash memory, which is mainly used to store things. As you might understand, this is a simplified explanation and might not be entirely true in all cases. However, it can serve as a help to understand the difference between the RAM and the flash memory of the phone. To see how much RAM is currently used, go to Applications in the Settings app of your Xperia™ phone.

Now, let’s look at how the RAM is used. Out of our 512MB RAM, about a third is used for functions that require a dedicated memory slot to operate fast enough. For example, this is the case for certain multimedia functions. The remaining space, which is at least 340MB, is reserved for the Linux user space, as required in the Android Compatibility Definition Document (CDD). Within the Linux user space, functions like the activity manager and Home screen app are running.

Another interesting thing is that many apps use slightly more RAM in ICS. For example, the web browser is quite intensive, and our measurements indicate that it uses 20-30MB more in ICS compared to Gingerbread. All in all, there are a lot of changes that together result in greater RAM requirement.

Illustration of the RAM usage.

When running low on RAM, typically with less than approximately 40MB left, the activity manager will start to close processes according to priority. At first, idle background activities are killed. The last thing to be closed down is the foreground activity. We have described this briefly in the table below. For more information, check out Android developers. (Please note that all figures mentioned about RAM usage are approximations and will differ depending on phone model and use case.)


Table showing different types of processes. When running out of RAM, the activity manager starts shutting down processes from the bottom and up, so that the last things to close are foreground and persistent activities.

Processes that are closed will obviously have to be restarted when the user enters the app again, which takes time and slows the system down. For example, when running a heavy game that uses all available RAM, the activity manager will be forced to kill all processes running in the background. This might include vital functions like the dialler and even the Home screen application. When you exit your game, there is a risk that the phone is perceived as slow, since the Home screen app will have to be restarted, just like every other activity you access afterwards.

Slower interaction with the SQL database
Another change in ICS compared to Gingerbread is that Google has moved a lot of the SQL handling from the native to the Java layer. In our internal studies, we have seen that read and write operations to the SQL database takes longer time, which slows down the apps. Many applications perform a lot of SQL operations when started, which greatly impacts the start-up time.

According to good practice, database operations or http requests should not be performed in the main thread. However, we know that there are quite a few applications that perform these kinds of operations directly in the main thread, which might cause them to hold up other operations. Also, when reading feedback on ICS software out on the market now, we’ve seen comments about people having problems with some applications and games.

If an operation takes too long, there is a risk of getting an Application Not Responding (ANR) as a result. An ANR occurs when an application doesn’t answer an intent, or responds to an input event, within a certain time limit. In case of intent, the time out is set to five seconds. For the input event, such as screen touch or button click, it’s ten seconds.

This can result in a user experience that is perceived as slower and less stable, due to longer response times and increased ANRs.

Introducing full hardware acceleration
Yet another change in ICS, is that the graphics hardware acceleration is on by default for all apps from API level 14. For apps at lower API levels, it can be turned on in the manifest with the attribute android:hardwareAccelerated=”true”. Hardware acceleration means that the GPU is used to render graphics, which enables a smooth user interface. However, it also results in at need to load additional graphic libraries for certain apps, which makes them use even more RAM.

When we performed internal tests on our applications, we saw that the Settings app consumed 1-2MB more RAM, and actually took longer time to start with HW acceleration, compared to without. Once the app is running, the UI is HW accelerated, but unless the app performs advanced graphics, the user will not see the difference.

Another effect of the hardware acceleration is that it can make the battery drain faster in some cases. An example of this is video playback, where the hardware acceleration requires every video frame to be run through the GPU, thus making the system use more power than it would have without HW acceleration.

As a developer, you should therefore evaluate if HW acceleration is required or not, as it comes with a cost in terms of RAM usage, start-up time and possibly even battery duration which can have negative effects on the user experience. You can read more about hardware acceleration in Ice Cream Sandwich on the Android Developers blog.

So, what will be your platform of choice? We hope this article clarifies some of the aspects to consider when making the decision. As always, we are eager to hear your opinion, so drop us a comment below and let us know! For more details on timing and practicalities on the ICS upgrade, check out this latest post on the Sony Xperia™ Product Blog.

Updated – comment from the Developer World team:

We we would like to clarify that above mentioned “challenges” have already been addressed by our SW engineering teams. For instance, we have not only optimised the RAM management by making the RAM usage for internal apps as good as possible, but we will also introduce a Performance assistant at start up when running ICS. In this Performance assistant, you can enable and disable certain services that you might not want to run on your phone, in order to optimise the performance of your phone.

We have also worked with quite a few partners in regards to architecture optimisations for SQL handling. In addition, we have also optimised the hardware usage. And as a result of this article, a number of app developers have notified us that they are evaluating if HW optimisation will be needed or not for their apps.

The aim of this article was to share our knowledge regarding the different characteristics for ICS and Gingerbread in an open way, as we strive to have an open communication with the developer community. All in all, we would like to point out that it’s our clear aim to deliver an as good ICS update as ever possible. As you might have seen on the Sony Xperia Product Blog, we’re not far from releasing it now. Thanks for all the feedback!

Android 소스를 다운 받고 환경 구축을 할 때, jdk를 깔아야 한다.

Android 웹사이트에는 다음과 같이 하라고 나와 있다.

위와 같은 방법은 우분투 10.x 에서는 잘 되나, 11.x(11.04 특히)에서는 되지 않는다.

그럴 땐 다음과 같이 하면  jdk6을 깔 수 있다.

sudo add-apt-repository ppa:ferramroberto/java
sudo apt-get update
sudo apt-get install sun-java6-jdk sun-java6-plugin

정적 라이브러리

컴파일 시 링커로 라이브러리의 오브젝트 코드를 만들고자 하는 타겟 바이너리에 추가된다.

동적 라이브러리

컴파일 시, 라이브러리가 사용된 곳에 공유 라이브러리를 쓴다고 해놓고, 런타임시 동적링크로 링크한다.

정적 라이브러리 만들기

언제나와 같이 예제로 살펴보자.
우선 foo1.c라는 파일을 하나 만들어 다음과 같이 넣어보자.

#include <stdio.h> void foo1() { printf("Hello World\n"); }

자 이제는 foo2.c를 만들어 다음과 같이 만들자.

#include <stdio.h> void foo2() { printf("It's foo2 greeting!\n"); }

마지막으로 이 둘을 interface해 줄 역활의 foolib.h을 만들자.

void foo_1(void); void foo_2(void);

코드 입력을 다 끝냈으면 이제 foo1.c와 foo2.c를 컴파일 하자.

그러면 foo1.o 와 foo2.o라는 오브젝트 파일이 만들어지는데, 이 둘을 다음과 같이 묶어주자.

그러면 libfoo.a라는 파일이 생성된다.  이 파일이 이제 우리가 쓸 수 있는 정적라이브러리 파일이다. 제대로 동작하나 보기 위해, example.c를 만들자.

#include <stdio.h> #include "foolib.h" int main() { foo1(); foo2(); return 0; }

컴파일은 다음과 같이 한다.

-L 은 gcc에게 라이브러리를 찾을 디렉터리를 지정한다.

-l은 링크할 라이브러리를 지정한다. l뒤에 붙은 이름은 라이브러리 이름에서 lib을 뗀(libfoo의 경우 foo만 남는다) 것이며, -l 옵션은 컴파일 할 대상의 소스가 지정되는 뒤에 두면 된다.

동적 라이브러리 만들기

동적 라이브러리를 살펴 볼때도 위에서 나열한 소스를 쓰자. (재활용은 위대하다.)

-fPIC은 컴파일러에게 위치에 상관없이 수행할 수 있는 코드로 컴파일 하라고 하는 것이다. 이제 이렇게 해서 나온 오브젝트 파일을 정적 라이브러리와 합쳐(?)보자.

다음과 같이 쳐보자.

-shared는 동적 라이브러리를 우선 시해서 링크 하라고 지시한다.

-soname은 정적 라이브러리의 soname을 정한다.

자 이제 gcc 링크와 동적 링크를 위한 심볼릭 링크만 생성해 주자.

libfoo.so는 컴파일 시에 실행 파일들을 링크 할 때 링커가 찾을 수 있게 만든 것이며, libfoo.so.0은 실행 시에 동적 링커가 soname으로 찾을 수 있게 soname을 파일명으로 하는 libfoo.so.0.0.0파일의 심볼릭 링크를 만든 것이다. 

동적 링크에 대한 작업도 필요하다.  libfoo.so.0.0.0파일이 있는 디렉터리를 /etc/ld.so.conf에 추가해 줘야 하며, ldconfig로  /etc/ld.so.cache를 새로 생성한다.

마지막으로 다음과 같이 실행하여 컴파일 하자.

sscanf에 이어 이번에는 sprintf의 사용법에 대해서 알아보자 (sscanf에 대한 사용법은 여기를 참조http://codecat.tistory.com/entry/sscanf-사용법)

sprintf는 man page에 다음과 같이 선언되어 있다.

int sprintf(char *str, const char *format, ...);

sscanf 사용법에서, scanf가 standard input stream, stdin, 에서 input을 읽어오는 반면, sscanf는 str이라는 문자열(character string)에서 입력을 받아온다라고 했는데, sprintf 또한 input<->output이라는 점만 제외하면 이와 유사하다.

즉 printf가 standard output stream, stdout에 output을 출력한다면, sprintf는 str이라는 문자열에 출력을 한다고 생각하면 된다.

sscanf 사용법에 썼던 예제를 다시 한번 보면, (예제출처: 초보자를 위한 Linux & Unix C 프로그래밍)

#define MAX 100 main() { int i = 123; float j = 23.456; char buff1[MAX]; char buff2[MAX]; sprintf(buff1, "%d", i); sprintf(buff2, "%.3f", j); sscanf(buff1, "%o", &i); printf("%d\n", i); sscanf(buff2, "%e", &j); printf("%e\n", j); }

sprintf(buff1, "%d", i) 라는 부분이 있다.  이는 위에서 이미 정의된 i를 %d에 맞게 buff1에 출력(저장이라고 보는게 맞겠다)한다고 보면 된다.  (포맷에 관련해서는 sscanf 사용법을 참조하자.)

그럼 이걸 언제 써먹냐면은... 물론 여러가지 용도가 있겠지만, 대표적으로 생각나는게 여러가지 옵션을 모아서 하나의 커맨드를 만들어서 실행시킬 때 사용할 수 있겠다. 

다음과 같이 실행 커맨드와 옵션들을 긁어 모아서 system()으로 한방에 실행시킬 수 있다.

sscanf는 manpage에 다음과 같이 정의되어 있다.

#include <stdio.h>
int sscanf(const char *str, const char *format, ...);



scanf가 standard input stream, stdin, 에서 input을 읽어오는 반면, sscanf는 str이라는 문자열(character string)에서 입력을 받아온다
즉, 메모리에서(const char *str) 받은 값을 포맷 형식에 맞게 읽어오는 것이다.
간단한 예제를 보자.

 
예제출처: 초보자를 위한 Linux & Unix C 프로그래밍

위의 예제에서는 이미 정의되어 있는 값을 sprintf를 이용해 buff에 넣어두었다가, sscanf를 통해 포맷에 맞게 읽어오는 예제이다.

가끔 이 포맷으로 간편한 작업을 해낼 수도 있는데, 다음의 예제를 보면,

예제출처: Microsoft Support @http://support.microsoft.com/kb/38335/eng

sscanf를 이용해서 내부 버퍼에서 ','(컴마)로 문자를 잘라서 읽어오는 것이다.  (이렇게 하기 위해서 필요한 포맷은 %[^',']이다)  물론 일반적인 토큰화를 해야 한다면 strtok을 사용하는 걸 추천한다. 

말 나온 김에 포맷에 대해서 정리해 보자.

주의 할 점은 포맷을 잘못 사용 할 경우, 이어지는 모든 데이터가 다 틀어지게 된다는 것이다.  sscanf를 사용해서 받은 값이 이상하거나 할 때는, 포맷을 제대로 썼는지 확인하자.