Presented at the OnAndroidConf, October 22 2013, http://www.onandroidconf.com/sessions.html Abstract: Optimizing the Android build environment to perform at world-class level is a big challenge for many Android device and chipset makers today. Churning through thousands of platform builds per week requires laser-focus on high-performance infrastructure and tooling. If you’re looking at improving your overall engineering and developer productivity, the software build use case is an obvious area to prioritize. This technical talk will focus on the following aspects of the Android platform build: Common Android platform build challenges and opportunities with real-life production references The various Android build use cases and their needs – full integration and release builds, developer incremental builds Evolution of the Android build and codebase with trends and statistics Detailed technical analysis of the Android platform build, highlighting opportunities for improvements Proposed solutions and technical tricks to optimize an Android software build environment

1. Accelerating the Android Platform Build
David Rosen, Product Manager
drosen@electric-cloud.com
@adr0sen
OnAndroidConf 2013

2. What’s this Android buzz all about?
2

3. Qualcomm Reference Design Program
<60 days TTM, and pushing for faster
Source: http://www.qualcomm.com/connect/analyst-relations/events/americas-analyst-summit-2012

4. Android Development Landscape
Real-world challenges at large APAC mobile device maker

5. Electric Cloud Android/Embedded Survey (Q1 2013)
Users of Android by Vertical
• Survey
• Focused on Android
development & delivery
Military/aero
space
3%
Other
25%
• Global embedded contacts in
Electric Cloud marketing lists
• 350+ respondents
• 175+ Android
Automotive
6%
Mobile
38%
Medical
4%
Industrial
control and
automation
6%
Consumer
electronics
18%

6. Android Delivery Process Today
Technical Challenges
Negative Business Impacts
• Slow Android Builds
• Slow, Complex CTS
Testing
• Complex B-T-R Matrix
• Custom Testing –
Volume, Complexity
• Slow Time to Market
• Poor Quality of products
• Poor productivity of
teams
Avg. release build time
More than 4
hours, 15.3
%
1-4
hours, 27.0
%
Avg. time to execute CTS
Less than 5
minutes, 7.8
%
5-20
minutes, 21.
0%
20 minutes 1
hour, 28.8%
4-8
hours, 15.0
%
More than 8
hours, 5.0%
Less than 1
hour, 22.5%
1-4
hours, 57.5
%

7. Accelerating the
Android build

8. Why look at Android builds?
• Android is everywhere
• 68% of global mobile phone market share1
• Explosion of non-phone applications
•
•
•
•
Automotive (Saab, Renault)
Entertainment (Ouya, Nvidia)
Exercise equipment (NordicTrack)
Rice cookers! (Panasonic)
• Android is everywhere
• … and that means thousands of developers building Android
• What if we can make those builds faster?
• How would your development process change?
• How much more could you achieve?
1
Source: Gartner, February 2013

9. What is (and is not) covered in this part of the talk?
• Build acceleration techniques fall into two broad categories:
• Hardware improvements like faster CPU’s and better disks
• Software improvements like smarter build tools and faster compilers
• In general, hardware and software improvements are complimentary
• If you want the fastest possible builds, leverage both!
• This talk is about software techniques for build acceleration
• Given a fixed hardware platform, how fast can we build Android?
• Or, what do you do when you reach the limits of your hardware?

10. AOSP LOC Evolution
Jan 2009 – Jan 20131
18,000,000
16,000,000
14,000,000
12,000,000
10,000,000
Linux Kernel
8,000,000
Android
Mozilla Firefox
6,000,000
4,000,000
2,000,000
0
1
Source: http://www.ohloh.net, http://en.wikipedia.org/wiki/Android_version_history (February 2013)
Chromium
Webkit

11. AOSP Build Evolution
Release Date
LOC1
LOC
Growth %
GNU Make
Build Time2
Build Time
Growth %
2.2
May 2010
8,837,858
-
28m55s
-
2.3.4
April 2011
11,492,324
30%
33m10s
15%
4.0.1
October 2011
12,827,330
12%
1h13m54s
123%
4.1.1
July 2012
15,028,331
17%
1h28m11s
19%
4.2.2
February 2013
15,266,8033
2%
1h32m56s
2%
01:40
+19%
+2%
18
16
01:26
+123%
+30%
01:12
+17%
+2%
14
12
+12%
00:57
10
00:43
8
6
00:28
+15%
4
00:14
2
00:00
0
2.2
1
Millions
Android Version
2.3.4
4.0.1
4.1.1
4.2.2
http://www.ohloh.net, 2 Builds performed on 8-core 16GB RAM server, 3 LOC data as of December 2012
Build Time
LOC

12. Common wisdom about Android builds
• I can just…
• … add more cores
• … use distcc
• … use ccache
• “The Android build is as optimized as it can be!”

13. The common wisdom is wrong.

14. Sidebar: how much faster is “interesting”?
• Some large mobile device makers run > 50K builds/week
• At 15 minutes per build that’s 12,500 hours of build time
• A reduction of just one minute would save 800 hours every week
• What about wait time?
• Faster builds = more iterations
• Faster builds = higher quality
• Faster builds = lower time to market

15. How fast is GNU make?
• Best time: ~15m
• 48-cores
• 128GB RAM

16. Problem: Visibility
• Knowing the end-to-end time is great but not actionable
• We need visibility
• Where is the time going?
• What are the bottlenecks?
• Are there opportunities for improvement?
• How do we get that visibility?
• Instrument the build… somehow! strace, echo, hack gmake, or…
• ElectricInsight
• Visualization and analysis of builds run with ElectricMake
• Let’s pump this build through emake and see how it looks!

17. Solution: ElectricInsight

18. What can we see from here?
• An ideal build would look like a tightly packed box
• Overall looks almost perfect – well done, Android team!
• But! a couple of things stand out:
• Crazy long parse time1
• Gaps at the end of the build, indicative of serializations
• Some very long serial chunks at the bottom of graph
• We’ll investigate these issues, but first: what if we just use
more cores?
1
emake parsing may be slower than gmake parsing

19. Longest serial chain
Best possible is about 15m
• Best possible runtime is about 15m

20. Projected runtime with more cores

21. What else can we learn?
• Lots of non-compile work in the build - ~25% of total time

22. Problem: long parse time
• Why do we build?
• To transform sources into deliverables (programs, docs, etc).
• Does parsing makefiles transform sources into
deliverables?
• Nope.
• Parsing makefiles is pure overhead
• But you have to tell make what to do somehow
• Unless you want to manually invoke the compiler, linker, etc.

23. Solution: parse avoidance
• What if we don’t parse (every time)?
• Makefiles don’t change very often, so why reparse every time?
• Reuse parse results from a previous build, as long as…
• Makefiles are unchanged (MD5)
• Command-line is unchanged
• Environment is unchanged
• How do we do it?
• Electric Make already has parse results in a reloadable form, just
need to add cache management
• GNU make doesn’t, but could be modified

24. Parse avoidance impact
Build time reduced to about 13m30s

25. Problem: serializations
• Gaps in visualization suggest serializations
• Q: How many dependencies are there in the
Android build?
• A: More than you think!

26. Android Dependency Graph
• Dump all makefile rules
• 100,000 files/targets
• 1,990,628 dependencies

27. Dependencies in Android
• ~19 dependencies per file: why so many?
• Consider a typical build structure:
lib.a: foo.o bar.o
foo.o: foo.c foo.h util.h
bar.o: bar.c bar.h util.h
• Some files will have many dependencies
• Most have only a few
• What is going on in Android?

28. Superfluous dependencies in Android
• Do we really need all 1.9M dependencies?
• The filesystem can tell us!
• Collect a list of files actually used to generate a target
• Compare to the list of prerequisites specified in the makefile
• Example:
foo.txt:
echo “Hello” > foo.txt
bar.txt: foo.txt
echo “World” > bar.txt
• Why not run foo.txt and bar.txt in parallel?

29. Actual dependencies in Android
• 100,000 files/targets
• 288,804 dependencies

30. Specified vs. Actual dependencies in Android
Specified
Actual

31. Solution: dependency optimization
• Impossible to manually eliminate superfluous
dependencies
• Electric Make can prune them automatically
• If you use gmake, you can achieve the same effect:
• Remove existing dependency specifications
• Generate minimal dependency specifications from emake data
• Before:
$(OBJECTS): %.o: %.cpp $(ALL_HEADERS) $(OTHER_DEPS)
$(COMPILE.CC)
• After:
$(OBJECTS): %.o: %.cpp
$(COMPILE.CC)
foo.o: foo.h util.h
bar.o: bar.h util.h generated.h

32. Dependency optimization impact
Build time reduced to about 12m30s

33. Problem: long serial jobs
• Long serial jobs prohibits further parallelization
• Dominates longest serial chain:
• In Android, Javadoc generation is a performance albatross
• Typically a long, monolithic job
• Not decomposable/parallelizable, not incremental

34. Solution: Javadoc Caching
• Javadoc generation can be cached
• Stash output after running javadoc
• Reuse output in future builds if inputs match
• Like ccache, but for javadoc
Carving 4 minutes away from job!

35. Javadoc Caching impact
Build time reduced to about 9m30s

36. Summary
• Google has done a great job of optimizing Android builds
• But there’s still room for improvement!
• By caching parse results and javadoc output while pruning
dependencies, we have shown to carve off 5m+ or ~37% from build
21:36
19:30
17:45
19:12
16:48
15:00
14:24
15:00
12:00
11:55
Gmake
09:36
Emake
09:30
07:12
04:48
02:24
00:00
24
32
48

37. Questions?
David Rosen, Product Manager
drosen@electric-cloud.com
@adr0sen
http://www.electric-cloud.com/eade