What makes JNI calls slow?

I know that 'crossing boundaries' when making a JNI call in Java is slow.

However I want to know what is it that makes it slow? What does the underlying jvm implementation do when making a JNI call that makes it so slow?

First, it's worth noting that by "slow," we're talking about something that can take tens of nanoseconds. For trivial native methods, in 2010 I measured calls at an average 40 ns on my Windows desktop, and 11 ns on my Mac desktop. Unless you're making many calls, you're not going to notice.

That said, calling a native method can be slower than making a normal Java method call. Causes include:

• Native methods will not be inlined by the JVM. Nor will they be just-in-time compiled for this specific machine -- they're already compiled.
• A Java array may be copied for access in native code, and later copied back. The cost can be linear in the size of the array. I measured JNI copying of a 100,000 array to average about 75 microseconds on my Windows desktop, and 82 microseconds on Mac. Fortunately, direct access may be obtained via GetPrimitiveArrayCritical or NewDirectByteBuffer.
• If the method is passed an object, or needs to make a callback, then the native method will likely be making its own calls to the JVM. Accessing Java fields, methods and types from the native code requires something similar to reflection. Signatures are specified in strings and queried from the JVM. This is both slow and error-prone.
• Java Strings are objects, have length and are encoded. Accessing or creating a string may require an O(n) copy.

Some additional discussion, possibly dated, can be found in "Java(tm) Platform Performance: Strategies and Tactics", 2000, by Steve Wilson and Jeff Kesselman, in section "9.2: Examining JNI costs". It's about a third of the way down this page, provided in the comment by @Philip below.

The 2009 IBM developerWorks paper "Best practices for using the Java Native Interface" provides some suggestions on avoiding performance pitfalls with JNI.

It is worth mentioning that not all Java methods marked with native are "slow". Some of them are intrinsics that makes them extremely fast. To check which ones are intrinsic and which ones are not, you can look for do_intrinsic at vmSymbols.hpp.

Basically the JVM interpretively constructs the C parameters for each JNI call and the code is not optimized.

There are many more details outlined in this paper

If you are interested in benchmarking JNI vs native code this project has code for running benchmarks.

When talking about JNI, there are two directions: java calling C++, and C++ calling java. Java calling C++ (or C) via the "native" keyword is very fast, around 50 clock cycles. However, C++ calling Java is somewhat slow. We do a great deal of Java/C++ integration, and my rule of thumb is 1000 clock cycles per call, so you can get around 2M calls/second. I cannot answer your actual question of "why is it slow", but I'll hazard a guess that a lot of work has to be done to transfer arguments from the native C++ stack using varargs, onto the Java stack, validate whatever conformance is needed, and vice-versa on the return value.

However, also remember that once you make a call into a Java method from C++, if that method returns a complex data structure, you'll need to make JNI calls for all accesses into the result, as well. The same applies for converting complex C++ structure to Java. We've found in practice for example that it is much faster to serialize a C++ std::map to JSON, hand the string across JNI, and have Java deserialize it into a Map, assuming you want the entire map converted to Java.