Android on Intel Platform

Software and Services Group

Android* on Intel platforms

Daniel Holmlund

Gayathri Murali


Agenda

Introduction

Porting and Packaging for IA

Android Testing

Intel tools


Introduction


Committed to providing the best experience on IA

Intel is the number one contributor to Linux kernel.

Intel is working on Android 64 bit operating system.

Intel recognized as the #1 external contributor2 to Jelly Bean. In line to be #1 again for next Android release

Chrome browser ported to Android and optimized to run best on IA

Intel has contributed 750+ patches to Webkit (Drives the browser for Android and iOS) in the last two months and is the 2nd largest Contributor to Chromium behind Google

Creating cross-platform tools and development environments based on new programming models

3X Faster1

Android ICS Browser

GPU accelerated

FPS: 20 60

6X Faster1

Jscript Performance

X86 codegen

Industry Leader

Best-in-class1

Optimized Android 2.3

Web Runtime

CES 2012 Review

>20% Faster1

Optimized SKIA

Rendering Library

Android 2.3


Optimizing Android for Intel

Atom Processor-Based Devices

A

pp

s A

pp

lica

tio

n

Fram

ew

ork

Home Contacts Phone Browser

Activity Manager Windows* Manager

Telephony Manager

Resource Manager

Content Providers View System

Location Manager Notification Manager Package Manager

Surface Manager

OpenGL* ES

SGL

Media Framework

FreeType

SSL

SQLite

WebKit

libc

Core Libraries

Dalvik Virtual Machine

Display Driver Camera Driver Flash Memory Driver Binder (IPC) Driver

Keypad Driver WiFi Driver Audio Drivers Power Management

Libraries

Linux* Kernel

Android* Runtime

Ap

ps

Use

r Ex

per

ien

ce

Mid

dle

war

e

Op

erat

ing

Syst

em

Apply our extensive experience optimizing Java* to the Dalvik* VM

Drivers validated & optimized for power & memory footprint

SKIA and openGL optimizations

Enhanced debugging and logging

IA assembly optimizations

We optimize web technologies such as HTML 5, WebKit and JavaScript

GPU & Video support for canvas operations

Extensive middleware development in imaging, media and DRM deliver compelling media experiences

Memory Optimizations, AVI, DivX*, and ASF container types,

WMV /VC-1 decoder. Live Streaming optimizations, HDMI

and WiDI Extended Video Modes, Video Playback DRM

Based on third party validation and sampling of Android apps using specific computer systems and/or components and reflect the approximate performance of Intel products as measured by those tests. Any difference in system hardware or software design or configuration may affect actual performance and/or results.


Smartphones with Intel Inside - 2013

Intel Yolo

Acer Liquid C1

ASUS* Fonepad - 7

Z2420 Z2580 Z2560

Lenovo* K900 5.5

ASUS Fonepad Note FHD - 6

ZTE Grand X2 4.5 ZTE* Geek 5


Tablets with Intel Inside - 2013

Samsung* Galaxy Tab 3 10.1

(Z2560)

ASUS* MeMO Pad FHD 10

(Z2560)

ASUS* Fonepad 7

(Z2420)

ASUS* Transformer Book Trio

Atom Z2580 + Core i7 4th Generation

Dell* Venue 7


Porting and Packaging for IA


NDK application

Good news! Most NDK applications run as it as on x86 platforms

Port your applications to improve performance and reliability

NDK applications categorized to 3 types

Consists of C/C++ code only that is not related to hardware Recompile the native code to run the application on the Intel Atom platform successfully.

Open the NDK project and search for Android.mk file and add APP_ABI := armeabi armeabi-v7a x86 in Android.mk and re-build the native code with ndk-build.

Uses a third-party dynamic linked library the shared library must be recompiled into x86 version for the Intel Atom platform.

Includes assembly code that is highly related to non-IA platforms code must be rewritten with IA assembly or C/C++.

Use the header file here to convert all NEON instructions into SSE instructions.


NDK Application Demo


Porting Dalvik applications

Dalvik applications run without any changes on all Intel powered devices

For a high resolution device, change the default memory allocation to meet application

requirements


PSI

TS

PIDs

Packaging APKs for multiple CPU

architectures

Two options:

One package for all (fat binary)

Embed native libraries for each architecture in one APK

Easiest and preferred way to go.

Multiple APKs

One APK per architecture

If you have good reasons to do so (ie your fat binary APK would be larger than 50MB).


The application will be filtered on ABI support during installation (after download).

Use lib/armeabi libraries

Use lib/armeabi-v7a libraries

Use lib/x86 libraries

Fat Binaries


Version code convention (proposal)

APKs that targets different platforms must have different version codes.

The store will choose the compatible APK with the higher version code number.

Using this convention, the chosen APK will be the one that run best on the device.


Android Testing


Options available to test

Emulators

Part of the SDK or stand alone emulators that aids in testing application in your development environment.

An Emulator mimics the hardware and software features of a typical mobile device

Cloud based testing

Actual devices are available on cloud for testing

Developers can create test scripts and run the tests on cloud on actual devices

Real device testing

Developers have access to the actual device to test the application functionality


Emulators - Create and starting

AVD Launch the Android Virtual

Device Manager: In Eclipse,

click Android Virtual Device

Manager from the toolbar.

In the Android Virtual Device Manager panel,

click New.

Fill in the details

Make sure you select an API that is supported by an

x86 system image,

CPU/ABI is set to x86, and

you have enabled GPU

(OpenGL ES*) emulation.

You can launch the x86 AVD by clicking on Start,

then Launch.


Cloud based testing

A device laboratory of multiple real devices are available for the developers to test

A test is carried out on the cloud based on test scripts either created by the developers or built in test scripts provided the service

provider

There may be a cost associated in using the devices

Developers may not be able to see the actual test in progress, but gets a detailed report out with screenshots

Ability to adopt to changing condition

Scale up and down


Real device testing on Cloud - AppTHwack

AppTHwack provides real device testing services on Android, iOS and other mobile web apps.

AppTHwack maintains a device lab with almost all devices available in market which are available for a period of testing. Users can

either use the built in test created by AppTHwack to test their apps

or create a custom test script.

Check out the pricing details here

Intel Software sponsors free testing of devices running on Intel platforms.

Check out the device lab here


Intel Hardware Accelerated

Execution Manager (HAXM)


Pre-built Atom Images Android SDK manager has x86 emulation

images built-in

To emulate an Atom-based Android phone, install the Intel Atom x86 System Image available in the Android SDK Manager

Much Faster Emulation Intel Hardware Accelerated Execution Manager

(HAXM) for Mac and Windows uses Intel

Virtualization Technology (VT) to accelerate

Android emulator by 2-4x

Intel VT is already supported in Linux* (KVM) for x86.

See www.intel.com/software/android for demo videos & installation tips

Intel x86 Emulator

Accelerator

Faster Android Emulation using an IA-based host

PC

Intel x86 Atom System Image


Intel HAXM - DEMO


Intel Graphics Performance

Analyzers


25

Intel Graphics Performance Analyzers System Analyzer for Android

Real-time system-level performance analysis for Intel-based Android devices

Immediate experiments and overwrites enable developers to isolate CPU and GPU performance problems

Metrics for CPU, GPU, API, memory, network, power, etc.

Copyright 2012, Intel Corporation. All rights reserved. *Other brands and names are the property of their respective owners.

For on-device app testing: Intel GPA System Analyzer Real-time Remote Performance and Power Profiling

Intel GPA System

Analyzer running on

a PC Android

Smart

Phone USB adb Connection

Wi-Fi Connection

Data

Real-time charts: CPU metrics Graphics API level metrics Processor graphics hardware metrics Power metrics

Download Intel GPA for FREE at intel.com/software/GPA


Intel GPA - DEMO


Intel HTML5 Development Environment

(XDK)

more on:

html5dev-software.intel.com

Create, Debug and Build HTML5 apps for multiple platforms from a single code base.


Intel Threading Building Blocks

(TBB)


Key Features of TBB

You can specify tasks instead of manipulating threads

TBB maps your logical tasks onto threads with full support for nested parallelism

Targets threading for scalable performance

Uses proven efficient parallel patterns

Uses work-stealing to support the load balance of unknown execution time for tasks.

Open source and licensed versions available on :

Linux, Windows, Mac OS X* and Android*

30


Intel TBB Integration

#for including tbb in your project:

include $(CLEAR_VARS)

LOCAL_MODULE := tbb

LOCAL_SRC_FILES := $(TBB_PATH)/lib/android/libtbb.so

LOCAL_EXPORT_C_INCLUDES := $(TBB_PATH)/include

include $(PREBUILT_SHARED_LIBRARY)

#for calling tbb from your lib:

LOCAL_CPP_FEATURES := rtti exceptions

LOCAL_SHARED_LIBRARIES += tbb

LOCAL_CFLAGS += -DTBB_USE_GCC_BUILTINS -std=c++11

Android.mk

APP_ABI := x86

APP_STL := gnustl_shared Application.mk

System.loadLibrary("gnustl_shared");

System.loadLibrary("tbb");

System.loadLibrary("YourLib"); Java


Intel TBB - Download

GPLv2 with runtime exception available from threadingbuildingblocks.org/download :

The commercial version with support is here: software.intel.com/en-us/intel-tbb

Copyright 2012, Intel Corporation. All rights reserved. *Other brands and names are the property of their respective owners.

Where to Get More Info

Intel Developer Zone (IDZ) Real developers sharing knowledge and

offering help Dedicated communities and forums

focused on your interests Worldwide reach News and insights on cutting edge

technology

Intel Android* Developer Website Great content you wont find anywhere else Technical articles, tools, and How-To

guides Native app porting tips & case studies Info on x86 emulator and Intel Hardware

Accelerated Execution Manager Active forums and blogs written by Intel and

community experts

software.intel.com/android


Backup


Intel C++ Compiler for Android


Intel C++ Compiler Optimized System and Application Performance

36

High Performance

Boosts performance with a single recompile

Binary and source compatible with GNU* GCC compiler

Intel Cilk Plus - task and data parallelism

Standards support

Optimization

Intel Atom - optimized common libc/libm functions

Streaming SIMD Extensions 3 (SSE3) instruction support

Vectorization support

Interprocedural optimization

Profile guided optimization

High performance. GNU compatibility. Standard support.


Most Android* Apps Will Just Work on Atom

Android Dalvik and HTML5 apps These should just work we do extensive testing of pre-production systems

Android native (NDK) apps Most will run w/o any recompilation but A simple recompile using the Android NDK yields the best performance If the developer uses ARM-specific features, add equivalent Intel SSE

features

Top Android Market apps Intel validates and tests on Intel Atom processor platforms (Mix of Dalvik and native (NDK) apps)

Android Runtime

Core Libraries

Dalvik Virtual

Machine


Optimization

Use Inline instead of Frequently Used Short

Use Float instead of Double

The Intel Atom platform has FPU enabled. In most cases, using Float instead of Double speeds up the data computing process and saves memory bandwidth in

Intel Atom processor-based devices.

Multi-thread coding

Multi-thread coding allows you to use the hyper-threading function of the Intel Atom processor to increase throughput and improve overall performance

High Performance compiler flags

Adding specific flags according to your target platform is highly recommended. You can add the following recommended flags during

compilation on the Intel Atom platform:

-march=atom

-msse4

-mavx

-maes


Intel VTune Amplifier for Android


Intel VTune Amplifier XE 2013 Performance & Power Profiler

Where is my application

Spending Time? Wasting Time? Consuming Power?

Focus tuning on functions taking time

See call stacks Drill to source and asm code

See cache misses on your source

See functions sorted by # of cache misses

See Frequency and Sleep states

See Wakeup reasons

Advanced profiling for Android Mobile Devices


Intel VTune Amplifier XE 2013 CPU Performance Analysis

Identify Hotspots and inefficient code

Cache Misses, Branch Mispredictions

Drill down from Java & C/C++ to assembler code


CPU Power Analysis

Use case: Collecting data from sensor

Continuously collecting accelerometer data

Use power analysis to find out impact of sampling frequency

2/11/2014

SENSOR_DELAY_FASTEST SENSOR_DELAY_NORMAL

Lowest Power consumption

High Power consumption

Android on Intel Platform

Documents

services group android

ia intel

ia software

kernel android

services group porting

services group tablets

services group smartphones

class1 optimized android