An Android application will run on a mobile device with limited computing power and storage, and constrained battery life. Because of this, it should be efficient. Battery life is one reason you might want to optimize your app even if it already seems to run "fast enough". Battery life is important to users, and Android's battery usage breakdown means users will know if your app is responsible draining their battery.
This document covers these topics:
Note that although this document primarily covers micro-optimizations, these will almost never make or break your software. Choosing the right algorithms and data structures should always be your priority, but is outside the scope of this document.
There are two basic rules for writing efficient code:
As you get started thinking about how to design your application, and as
you write it, consider
the cautionary points about optimization that Josh Bloch makes in his book
Effective Java. Here's "Item 47: Optimize Judiciously", excerpted from
the latest edition of the book with permission. Although Josh didn't have
Android application development in mind when writing this section — for
example, the java.awt.Component class
referenced is not available in Android, and Android uses the
Dalvik VM, rather than a standard JVM — his points are still valid.
There are three aphorisms concerning optimization that everyone should know. They are perhaps beginning to suffer from overexposure, but in case you aren't yet familiar with them, here they are:
More computing sins are committed in the name of efficiency (without necessarily achieving it) than for any other single reason—including blind stupidity.
—William A. Wulf 1
We should forget about small efficiencies, say about 97% of the time: premature optimization is the root of all evil.
—Donald E. Knuth 2
We follow two rules in the matter of optimization:
- Rule 1. Don't do it.
- Rule 2 (for experts only). Don't do it yet — that is, not until you have a perfectly clear and unoptimized solution.
—M. A. Jackson 3
All of these aphorisms predate the Java programming language by two decades. They tell a deep truth about optimization: it is easy to do more harm than good, especially if you optimize prematurely. In the process, you may produce software that is neither fast nor correct and cannot easily be fixed.
Don't sacrifice sound architectural principles for performance. Strive to write good programs rather than fast ones. If a good program is not fast enough, its architecture will allow it to be optimized. Good programs embody the principle of information hiding: where possible, they localize design decisions within individual modules, so individual decisions can be changed without affecting the remainder of the system (Item 13).
This does not mean that you can ignore performance concerns until your program is complete. Implementation problems can be fixed by later optimization, but pervasive architectural flaws that limit performance can be impossible to fix without rewriting the system. Changing a fundamental facet of your design after the fact can result in an ill-structured system that is difficult to maintain and evolve. Therefore you must think about performance during the design process.
Strive to avoid design decisions that limit performance. The components of a design that are most difficult to change after the fact are those specifying interactions between modules and with the outside world. Chief among these design components are APIs, wire-level protocols, and persistent data formats. Not only are these design components difficult or impossible to change after the fact, but all of them can place significant limitations on the performance that a system can ever achieve.
Consider the performance consequences of your API design decisions. Making a public type mutable may require a lot of needless defensive copying (Item 39). Similarly, using inheritance in a public class where composition would have been appropriate ties the class forever to its superclass, which can place artificial limits on the performance of the subclass (Item 16). As a final example, using an implementation type rather than an interface in an API ties you to a specific implementation, even though faster implementations may be written in the future (Item 52).
The effects of API design on performance are very real. Consider the
getSizemethod in thejava.awt.Componentclass. The decision that this performance-critical method was to return aDimensioninstance, coupled with the decision thatDimensioninstances are mutable, forces any implementation of this method to allocate a newDimensioninstance on every invocation. Even though allocating small objects is inexpensive on a modern VM, allocating millions of objects needlessly can do real harm to performance.In this case, several alternatives existed. Ideally,
Dimensionshould have been immutable (Item 15); alternatively, thegetSizemethod could have been replaced by two methods returning the individual primitive components of aDimensionobject. In fact, two such methods were added to the Component API in the 1.2 release for performance reasons. Preexisting client code, however, still uses thegetSizemethod and still suffers the performance consequences of the original API design decisions.Luckily, it is generally the case that good API design is consistent with good performance. It is a very bad idea to warp an API to achieve good performance. The performance issue that caused you to warp the API may go away in a future release of the platform or other underlying software, but the warped API and the support headaches that come with it will be with you for life.
Once you've carefully designed your program and produced a clear, concise, and well-structured implementation, then it may be time to consider optimization, assuming you're not already satisfied with the performance of the program.
Recall that Jackson's two rules of optimization were "Don't do it," and "(for experts only). Don't do it yet." He could have added one more: measure performance before and after each attempted optimization. You may be surprised by what you find. Often, attempted optimizations have no measurable effect on performance; sometimes, they make it worse. The main reason is that it's difficult to guess where your program is spending its time. The part of the program that you think is slow may not be at fault, in which case you'd be wasting your time trying to optimize it. Common wisdom says that programs spend 80 percent of their time in 20 percent of their code.
Profiling tools can help you decide where to focus your optimization efforts. Such tools give you runtime information, such as roughly how much time each method is consuming and how many times it is invoked. In addition to focusing your tuning efforts, this can alert you to the need for algorithmic changes. If a quadratic (or worse) algorithm lurks inside your program, no amount of tuning will fix the problem. You must replace the algorithm with one that is more efficient. The more code in the system, the more important it is to use a profiler. It's like looking for a needle in a haystack: the bigger the haystack, the more useful it is to have a metal detector. The JDK comes with a simple profiler and modern IDEs provide more sophisticated profiling tools.
The need to measure the effects of attempted optimization is even greater on the Java platform than on more traditional platforms, because the Java programming language does not have a strong performance model. The relative costs of the various primitive operations are not well defined. The "semantic gap" between what the programmer writes and what the CPU executes is far greater than in traditional statically compiled languages, which makes it very difficult to reliably predict the performance consequences of any optimization. There are plenty of performance myths floating around that turn out to be half-truths or outright lies.
Not only is Java's performance model ill-defined, but it varies from JVM implementation to JVM implementation, from release to release, and from processor to processor. If you will be running your program on multiple JVM implementations or multiple hardware platforms, it is important that you measure the effects of your optimization on each. Occasionally you may be forced to make trade-offs between performance on different JVM implementations or hardware platforms.
To summarize, do not strive to write fast programs — strive to write good ones; speed will follow. Do think about performance issues while you're designing systems and especially while you're designing APIs, wire-level protocols, and persistent data formats. When you've finished building the system, measure its performance. If it's fast enough, you're done. If not, locate the source of the problems with the aid of a profiler, and go to work optimizing the relevant parts of the system. The first step is to examine your choice of algorithms: no amount of low-level optimization can make up for a poor choice of algorithm. Repeat this process as necessary, measuring the performance after every change, until you're satisfied.
—Excerpted from Josh Bloch's Effective Java, Second Ed. (Addison-Wesley, 2008).
1 Wulf, W. A Case Against the GOTO. Proceedings of the 25th ACM National Conference 2 (1972): 791–797.
2 Knuth, Donald. Structured Programming with go to Statements. Computing Surveys 6 (1974): 261–301.
3 Jackson, M. A. Principles of Program Design, Academic Press, London, 1975. ISBN: 0123790506.
One of the trickiest problems you'll face when micro-optimizing Android apps is that the "if you will be running your program on ... multiple hardware platforms" clause above is always true. And it's not even generally the case that you can say "device X is a factor F faster/slower than device Y". This is especially true if one of the devices is the emulator, or one of the devices has a JIT. If you want to know how your app performs on a given device, you need to test it on that device. Drawing conclusions from the emulator is particularly dangerous, as is attempting to compare JIT versus non-JIT performance: the performance profiles can differ wildly.
Object creation is never free. A generational GC with per-thread allocation pools for temporary objects can make allocation cheaper, but allocating memory is always more expensive than not allocating memory.
If you allocate objects in a user interface loop, you will force a periodic garbage collection, creating little "hiccups" in the user experience.
Thus, you should avoid creating object instances you don't need to. Some examples of things that can help:
A somewhat more radical idea is to slice up multidimensional arrays into parallel single one-dimension arrays:
Generally speaking, avoid creating short-term temporary objects if you can. Fewer objects created mean less-frequent garbage collection, which has a direct impact on user experience.
Previous versions of this document made various misleading claims. We address some of them here.
On devices without a JIT, it is true that invoking methods via a
variable with an exact type rather than an interface is slightly more
efficient. (So, for example, it was cheaper to invoke methods on a
Map map than a HashMap map, even though in both
cases the map was a HashMap.) It was not the case that this
was 2x slower; the actual difference was more like 6% slower. Furthermore,
the JIT makes the two effectively indistinguishable.
On devices without a JIT, caching field accesses is about 20% faster than repeatedly accesssing the field. With a JIT, field access costs about the same as local access, so this isn't a worthwhile optimization unless you feel it makes your code easier to read. (This is true of final, static, and static final fields too.)
If you don't need to access an object's fields, make your method static. Invocations will be about 15%-20% faster. It's also good practice, because you can tell from the method signature that calling the method can't alter the object's state.
In native languages like C++ it's common practice to use getters (e.g.
i = getCount()) instead of accessing the field directly (i
= mCount). This is an excellent habit for C++, because the compiler can
usually inline the access, and if you need to restrict or debug field access
you can add the code at any time.
On Android, this is a bad idea. Virtual method calls are expensive, much more so than instance field lookups. It's reasonable to follow common object-oriented programming practices and have getters and setters in the public interface, but within a class you should always access fields directly.
Without a JIT, direct field access is about 3x faster than invoking a trivial getter. With the JIT (where direct field access is as cheap as accessing a local), direct field access is about 7x faster than invoking a trivial getter. This is true in Froyo, but will improve in the future when the JIT inlines getter methods.
Consider the following declaration at the top of a class:
static int intVal = 42; static String strVal = "Hello, world!";
The compiler generates a class initializer method, called
<clinit>, that is executed when the class is first used.
The method stores the value 42 into intVal, and extracts a
reference from the classfile string constant table for strVal.
When these values are referenced later on, they are accessed with field
lookups.
We can improve matters with the "final" keyword:
static final int intVal = 42; static final String strVal = "Hello, world!";
The class no longer requires a <clinit> method,
because the constants go into static field initializers in the dex file.
Code that refers to intVal will use
the integer value 42 directly, and accesses to strVal will
use a relatively inexpensive "string constant" instruction instead of a
field lookup. (Note that this optimization only applies to primitive types and
String constants, not arbitrary reference types. Still, it's good
practice to declare constants static final whenever possible.)
The enhanced for loop (also sometimes known as "for-each" loop) can be used for collections that implement the Iterable interface and for arrays. With collections, an iterator is allocated to make interface calls to hasNext() and next(). With an ArrayList, a hand-written counted loop is about 3x faster (with or without JIT), but for other collections the enhanced for loop syntax will be exactly equivalent to explicit iterator usage.
There are several alternatives for iterating through an array:
static class Foo {
int mSplat;
}
Foo[] mArray = ...
public void zero() {
int sum = 0;
for (int i = 0; i < mArray.length; ++i) {
sum += mArray[i].mSplat;
}
}
public void one() {
int sum = 0;
Foo[] localArray = mArray;
int len = localArray.length;
for (int i = 0; i < len; ++i) {
sum += localArray[i].mSplat;
}
}
public void two() {
int sum = 0;
for (Foo a : mArray) {
sum += a.mSplat;
}
}
zero() is slowest, because the JIT can't yet optimize away the cost of getting the array length once for every iteration through the loop.
one() is faster. It pulls everything out into local variables, avoiding the lookups. Only the array length offers a performance benefit.
two() is fastest for devices without a JIT, and indistinguishable from one() for devices with a JIT. It uses the enhanced for loop syntax introduced in version 1.5 of the Java programming language.
To summarize: use the enhanced for loop by default, but consider a hand-written counted loop for performance-critical ArrayList iteration.
(See also Effective Java item 46.)
Enums are very convenient, but unfortunately can be painful when size and speed matter. For example, this:
public enum Shrubbery { GROUND, CRAWLING, HANGING }
adds 740 bytes to your .dex file compared to the equivalent class with three public static final ints. On first use, the class initializer invokes the <init> method on objects representing each of the enumerated values. Each object gets its own static field, and the full set is stored in an array (a static field called "$VALUES"). That's a lot of code and data, just for three integers. Additionally, this:
Shrubbery shrub = Shrubbery.GROUND;
causes a static field lookup. If "GROUND" were a static final int, the compiler would treat it as a known constant and inline it.
The flip side, of course, is that with enums you get nicer APIs and some compile-time value checking. So, the usual trade-off applies: you should by all means use enums for public APIs, but try to avoid them when performance matters.
If you're using Enum.ordinal, that's usually a sign that you
should be using ints instead. As a rule of thumb, if an enum doesn't have a
constructor and doesn't define its own methods, and it's used in
performance-critical code, you should consider static final int
constants instead.
Consider the following class definition:
public class Foo {
private int mValue;
public void run() {
Inner in = new Inner();
mValue = 27;
in.stuff();
}
private void doStuff(int value) {
System.out.println("Value is " + value);
}
private class Inner {
void stuff() {
Foo.this.doStuff(Foo.this.mValue);
}
}
}
The key things to note here are that we define an inner class (Foo$Inner) that directly accesses a private method and a private instance field in the outer class. This is legal, and the code prints "Value is 27" as expected.
The problem is that the VM considers direct access to Foo's private members from Foo$Inner to be illegal because Foo and Foo$Inner are different classes, even though the Java language allows an inner class to access an outer class' private members. To bridge the gap, the compiler generates a couple of synthetic methods:
/*package*/ static int Foo.access$100(Foo foo) {
return foo.mValue;
}
/*package*/ static void Foo.access$200(Foo foo, int value) {
foo.doStuff(value);
}
The inner-class code calls these static methods whenever it needs to access the "mValue" field or invoke the "doStuff" method in the outer class. What this means is that the code above really boils down to a case where you're accessing member fields through accessor methods instead of directly. Earlier we talked about how accessors are slower than direct field accesses, so this is an example of a certain language idiom resulting in an "invisible" performance hit.
We can avoid this problem by declaring fields and methods accessed by inner classes to have package scope, rather than private scope. This runs faster and removes the overhead of the generated methods. (Unfortunately it also means the fields could be accessed directly by other classes in the same package, which runs counter to the standard practice of making all fields private. Once again, if you're designing a public API you might want to carefully consider using this optimization.)
As a rule of thumb, floating-point is about 2x slower than integer on Android devices. This is true on a FPU-less, JIT-less G1 and a Nexus One with an FPU and the JIT. (Of course, absolute speed difference between those two devices is about 10x for arithmetic operations.)
In speed terms, there's no difference between float and
double on the more modern hardware. Space-wise, double
is 2x larger. As with desktop machines, assuming space isn't an issue, you
should prefer double to float.
Also, even for integers, some chips have hardware multiply but lack hardware divide. In such cases, integer division and modulus operations are performed in software — something to think about if you're designing a hash table or doing lots of math.
In addition to all the usual reasons to prefer library code over rolling
your own, bear in mind that the system is at liberty to replace calls
to library methods with hand-coded assembler, which may be better than the
best code the JIT can produce for the equivalent Java. The typical example
here is String.indexOf and friends, which Dalvik replaces with
an inlined intrinsic. Similarly, the System.arraycopy method
is about 9x faster than a hand-coded loop on a Nexus One with the JIT.
(See also Effective Java item 47.)
Native code isn't necessarily more efficient than Java. For one thing, there's a cost associated with the Java-native transition, and the JIT can't optimize across these boundaries. If you're allocating native resources (memory on the native heap, file descriptors, or whatever), it can be significantly more difficult to arrange timely collection of these resources. You also need to compile your code for each architecture you wish to run on (rather than rely on it having a JIT). You may even have to compile multiple versions for what you consider the same architecture: native code compiled for the ARM processor in the G1 can't take full advantage of the ARM in the Nexus One, and code compiled for the ARM in the Nexus One won't run on the ARM in the G1.
Native code is primarily useful when you have an existing native codebase that you want to port to Android, not for "speeding up" parts of a Java app.
(See also Effective Java item 54.)
One last thing: always measure. Before you start optimizing, make sure you have a problem. Make sure you can accurately measure your existing performance, or you won't be able to measure the benefit of the alternatives you try.
Every claim made in this document is backed up by a benchmark. The source to these benchmarks can be found in the code.google.com "dalvik" project.
The benchmarks are built with the Caliper microbenchmarking framework for Java. Microbenchmarks are hard to get right, so Caliper goes out of its way to do the hard work for you, and even detect some cases where you're not measuring what you think you're measuring (because, say, the VM has managed to optimize all your code away). We highly recommend you use Caliper to run your own microbenchmarks.