Most native methods are implemented using JNI as mentioned in other answers.
However, performance critical methods such as Object.hashCode are typically implemented as intrinsics. When the byte code is compiled into machine code, the Java compiler recognises the method call and inlines appropriate code directly. This is obviously going to be much faster than going through JNI for a trivial method.
Many people will claim that Object.hashCode will return the address of the object representation in memory. In modern implementations objects actually move within memory. Instead an area of the object header is used to store the value, which may be lazily derived from the memory address at the time that the value is first requested.
