What are the effects of exceptions on performance in Java?

Question

Question: Is exception handling in Java actually slow?

Conventional wisdom, as well as a lot of Google results, says that exceptional logic shouldn't be used for normal program flow in Java. Two reasons are usually given,

1. it is really slow - even an order of magnitude slower than regular code (the reasons given vary),

and

2. it is messy because people expect only errors to be handled in exceptional code.

This question is about #1.

As an example, this page describes Java exception handling as "very slow" and relates the slowness to the creation of the exception message string - "this string is then used in creating the exception object that is thrown. This is not fast." The article Effective Exception Handling in Java says that "the reason for this is due to the object creation aspect of exception handling, which thereby makes throwing exceptions inherently slow". Another reason out there is that the stack trace generation is what slows it down.

My testing (using Java 1.6.0_07, Java HotSpot 10.0, on 32 bit Linux), indicates that exception handling is no slower than regular code. I tried running a method in a loop that executes some code. At the end of the method, I use a boolean to indicate whether to return or throw. This way the actual processing is the same. I tried running the methods in different orders and averaging my test times, thinking it may have been the JVM warming up. In all my tests, the throw was at least as fast as the return, if not faster (up to 3.1% faster). I am completely open to the possibility that my tests were wrong, but I haven't seen anything out there in the way of the code sample, test comparisons, or results in the last year or two that show exception handling in Java to actually be slow.

What leads me down this path was an API I needed to use that threw exceptions as part of normal control logic. I wanted to correct them in their usage, but now I may not be able to. Will I instead have to praise them on their forward thinking?

In the paper Efficient Java exception handling in just-in-time compilation, the authors suggest that the presence of exception handlers alone, even if no exceptions are thrown, is enough to prevent the JIT compiler from optimizing the code properly, thus slowing it down. I haven't tested this theory yet.

Solution 1

It depends how exceptions are implemented. The simplest way is using setjmp and longjmp. That means all registers of the CPU are written to the stack (which already takes some time) and possibly some other data needs to be created... all this already happens in the try statement. The throw statement needs to unwind the stack and restore the values of all registers (and possible other values in the VM). So try and throw are equally slow, and that is pretty slow, however if no exception is thrown, exiting the try block takes no time whatsoever in most cases (as everything is put on the stack which cleans up automatically if the method exists).

Sun and others recognized, that this is possibly suboptimal and of course VMs get faster and faster over the time. There is another way to implement exceptions, which makes try itself lightning fast (actually nothing happens for try at all in general - everything that needs to happen is already done when the class is loaded by the VM) and it makes throw not quite as slow. I don't know which JVM uses this new, better technique...

...but are you writing in Java so your code later on only runs on one JVM on one specific system? Since if it may ever run on any other platform or any other JVM version (possibly of any other vendor), who says they also use the fast implementation? The fast one is more complicated than the slow one and not easily possible on all systems. You want to stay portable? Then don't rely on exceptions being fast.

It also makes a big difference what you do within a try block. If you open a try block and never call any method from within this try block, the try block will be ultra fast, as the JIT can then actually treat a throw like a simple goto. It neither needs to save stack-state nor does it need to unwind the stack if an exception is thrown (it only needs to jump to the catch handlers). However, this is not what you usually do. Usually you open a try block and then call a method that might throw an exception, right? And even if you just use the try block within your method, what kind of method will this be, that does not call any other method? Will it just calculate a number? Then what for do you need exceptions? There are much more elegant ways to regulate program flow. For pretty much anything else but simple math, you will have to call an external method and this already destroys the advantage of a local try block.

See the following test code:

public class Test {
    int value;


    public int getValue() {
        return value;
    }

    public void reset() {
        value = 0;
    }

    // Calculates without exception
    public void method1(int i) {
        value = ((value + i) / i) << 1;
        // Will never be true
        if ((i & 0xFFFFFFF) == 1000000000) {
            System.out.println("You'll never see this!");
        }
    }

    // Could in theory throw one, but never will
    public void method2(int i) throws Exception {
        value = ((value + i) / i) << 1;
        // Will never be true
        if ((i & 0xFFFFFFF) == 1000000000) {
            throw new Exception();
        }
    }

    // This one will regularly throw one
    public void method3(int i) throws Exception {
        value = ((value + i) / i) << 1;
        // i & 1 is equally fast to calculate as i & 0xFFFFFFF; it is both
        // an AND operation between two integers. The size of the number plays
        // no role. AND on 32 BIT always ANDs all 32 bits
        if ((i & 0x1) == 1) {
            throw new Exception();
        }
    }

    public static void main(String[] args) {
        int i;
        long l;
        Test t = new Test();

        l = System.currentTimeMillis();
        t.reset();
        for (i = 1; i < 100000000; i++) {
            t.method1(i);
        }
        l = System.currentTimeMillis() - l;
        System.out.println(
            "method1 took " + l + " ms, result was " + t.getValue()
        );

        l = System.currentTimeMillis();
        t.reset();
        for (i = 1; i < 100000000; i++) {
            try {
                t.method2(i);
            } catch (Exception e) {
                System.out.println("You'll never see this!");
            }
        }
        l = System.currentTimeMillis() - l;
        System.out.println(
            "method2 took " + l + " ms, result was " + t.getValue()
        );

        l = System.currentTimeMillis();
        t.reset();
        for (i = 1; i < 100000000; i++) {
            try {
                t.method3(i);
            } catch (Exception e) {
                // Do nothing here, as we will get here
            }
        }
        l = System.currentTimeMillis() - l;
        System.out.println(
            "method3 took " + l + " ms, result was " + t.getValue()
        );
    }
}

Result:

method1 took 972 ms, result was 2
method2 took 1003 ms, result was 2
method3 took 66716 ms, result was 2

The slowdown from the try block is too small to rule out confounding factors such as background processes. But the catch block killed everything and made it 66 times slower!

As I said, the result will not be that bad if you put try/catch and throw all within the same method (method3), but this is a special JIT optimization I would not rely upon. And even when using this optimization, the throw is still pretty slow. So I don't know what you are trying to do here, but there is definitely a better way of doing it than using try/catch/throw.

Solution 2

FYI, I extended the experiment that Mecki did:

method1 took 1733 ms, result was 2
method2 took 1248 ms, result was 2
method3 took 83997 ms, result was 2
method4 took 1692 ms, result was 2
method5 took 60946 ms, result was 2
method6 took 25746 ms, result was 2

The first 3 are the same as Mecki's (my laptop is obviously slower).

method4 is identical to method3 except that it creates a new Integer(1) rather than doing throw new Exception().

method5 is like method3 except that it creates the new Exception() without throwing it.

method6 is like method3 except that it throws a pre-created exception (an instance variable) rather than creating a new one.

In Java much of the expense of throwing an exception is the time spent gathering the stack trace, which occurs when the exception object is created. The actual cost of throwing the exception, while large, is considerably less than the cost of creating the exception.

Solution 3

Aleksey Shipilëv did a very thorough analysis in which he benchmarks Java exceptions under various combinations of conditions:

• Newly created exceptions vs pre-created exceptions
• Stack trace enabled vs disabled
• Stack trace requested vs never requested
• Caught at the top level vs rethrown at every level vs chained/wrapped at every level
• Various levels of Java call stack depth
• No inlining optimizations vs extreme inlining vs default settings
• User-defined fields read vs not read

He also compares them to the performance of checking an error code at various levels of error frequency.

The conclusions (quoted verbatim from his post) were:

1. Truly exceptional exceptions are beautifully performant. If you use them as designed, and only communicate the truly exceptional cases among the overwhelmingly large number of non-exceptional cases handled by regular code, then using exceptions is the performance win.

2. The performance costs of exceptions have two major components: stack trace construction when Exception is instantiated and stack unwinding during Exception throw.

3. Stack trace construction costs are proportional to stack depth at the moment of exception instantiation. That is already bad because who on Earth knows the stack depth at which this throwing method would be called? Even if you turn off the stack trace generation and/or cache the exceptions, you can only get rid of this part of the performance cost.

4. Stack unwinding costs depend on how lucky we are with bringing the exception handler closer in the compiled code. Carefully structuring the code to avoid deep exception handlers lookup is probably helping us get luckier.

5. Should we eliminate both effects, the performance cost of exceptions is that of the local branch. No matter how beautiful it sounds, that does not mean you should use Exceptions as the usual control flow, because in that case you are at the mercy of optimizing compiler! You should only use them in truly exceptional cases, where the exception frequency amortizes the possible unlucky cost of raising the actual exception.

6. The optimistic rule-of-thumb seems to be 10^-4 frequency for exceptions is exceptional enough. That, of course, depends on the heavy-weights of the exceptions themselves, the exact actions taken in exception handlers, etc.

The upshot is that when an exception isn't thrown, you don't pay a cost, so when the exceptional condition is sufficiently rare exception handling is faster than using an if every time. The full post is very much worth a read.

Solution 4

My answer, unfortunately, is just too long to post here. So let me summarize here and refer you to http://www.fuwjax.com/how-slow-are-java-exceptions/ for the gritty details.

The real question here is not "How slow are 'failures reported as exceptions' compared to 'code that never fails'?" as the accepted response might have you believe. Instead, the question should be "How slow are 'failures reported as exceptions' compared to failures reported other ways?" Generally, the two other ways of reporting failures are either with sentinel values or with result wrappers.

Sentinel values are an attempt to return one class in the case of success and another in the case of failure. You can think of it almost as returning an exception instead of throwing one. This requires a shared parent class with the success object and then doing an "instanceof" check and a couple casts to get the success or failure information.

It turns out that at the risk of type safety, Sentinel values are faster than exceptions, but only by a factor of roughly 2x. Now, that may seem like a lot, but that 2x only covers the cost of the implementation difference. In practice, the factor is much lower since our methods that might fail are much more interesting than a few arithmetic operators as in the sample code elsewhere in this page.

Result Wrappers, on the other hand, do not sacrifice type safety at all. They wrap the success and failure information in a single class. So instead of "instanceof" they provide an "isSuccess()" and getters for both the success and failure objects. However, result objects are roughly 2x slower than using exceptions. It turns out that creating a new wrapper object every time is much more expensive than throwing an exception sometimes.

On top of that, exceptions are the language supplied the way of indicating that a method might fail. There's no other way to tell from just the API which methods are expected to always (mostly) work and which are expected to report failure.

Exceptions are safer than sentinels, faster than result objects, and less surprising than either. I'm not suggesting that try/catch replace if/else, but exceptions are the right way to report failure, even in the business logic.

That said, I would like to point out that the two most frequent ways of substantially impacting performance I've run across are creating unnecessary objects and nested loops. If you have a choice between creating an exception or not creating an exception, don't create the exception. If you have a choice between creating an exception sometimes or creating another object all the time, then create the exception.

Solution 5

I've extends the answers given by @Mecki and @incarnate, without stacktrace filling for Java.

With Java 7+, we can use Throwable(String message, Throwable cause, boolean enableSuppression,boolean writableStackTrace). But for Java6, see my answer for this question

// This one will regularly throw one
public void method4(int i) throws NoStackTraceThrowable {
    value = ((value + i) / i) << 1;
    // i & 1 is equally fast to calculate as i & 0xFFFFFFF; it is both
    // an AND operation between two integers. The size of the number plays
    // no role. AND on 32 BIT always ANDs all 32 bits
    if ((i & 0x1) == 1) {
        throw new NoStackTraceThrowable();
    }
}

// This one will regularly throw one
public void method5(int i) throws NoStackTraceRuntimeException {
    value = ((value + i) / i) << 1;
    // i & 1 is equally fast to calculate as i & 0xFFFFFFF; it is both
    // an AND operation between two integers. The size of the number plays
    // no role. AND on 32 BIT always ANDs all 32 bits
    if ((i & 0x1) == 1) {
        throw new NoStackTraceRuntimeException();
    }
}

public static void main(String[] args) {
    int i;
    long l;
    Test t = new Test();

    l = System.currentTimeMillis();
    t.reset();
    for (i = 1; i < 100000000; i++) {
        try {
            t.method4(i);
        } catch (NoStackTraceThrowable e) {
            // Do nothing here, as we will get here
        }
    }
    l = System.currentTimeMillis() - l;
    System.out.println( "method4 took " + l + " ms, result was " + t.getValue() );


    l = System.currentTimeMillis();
    t.reset();
    for (i = 1; i < 100000000; i++) {
        try {
            t.method5(i);
        } catch (RuntimeException e) {
            // Do nothing here, as we will get here
        }
    }
    l = System.currentTimeMillis() - l;
    System.out.println( "method5 took " + l + " ms, result was " + t.getValue() );
}

Output with Java 1.6.0_45, on Core i7, 8GB RAM:

method1 took 883 ms, result was 2
method2 took 882 ms, result was 2
method3 took 32270 ms, result was 2 // throws Exception
method4 took 8114 ms, result was 2 // throws NoStackTraceThrowable
method5 took 8086 ms, result was 2 // throws NoStackTraceRuntimeException

So, still methods which returns values are faster, compared to methods throwing exceptions. IMHO, we can't design a clear API just using return types for both success & error flows. Methods which throws exceptions without stacktrace are 4-5 times faster than normal Exceptions.

Edit: NoStackTraceThrowable.java Thanks @Greg

public class NoStackTraceThrowable extends Throwable { 
    public NoStackTraceThrowable() { 
        super("my special throwable", null, false, false);
    }
}

View more solutions