Synchronization of data in multiple-core environment (Java-based)

Question

this is my first question ever so please be gentle on me.

What happens when two threads, say t1 and t2, running on separate CPU cores invoke a synchronized method on a shared object AT THE SAME TIME, i.e. in the same nanosecond/processor cycle? How and by what element of the architecture is the lock resolved?

It may seem like an obvious question but i could not find anything on the internet. Thanks!

Answer 1

The spec simply says the synchronized method may only be entered by one thread at a time. What architecture mechanism is used would depend on the JVM.

In the case of Oracle HotSpot (the JVM that almost everyone runs): the C++ code is GNU, and probably holds the answer for you.

But if you just need the assurance that it "just works", the specification gives that to you.

Answer 2

These days this is actually realized in hardware, so the processor has a special instruction to grant one thread out of many access. This has evolved from a pure software-based solution from last century, and the state of the art these days is what the AtomicInteger represents: a single, mutable integer, that is guaranteed to be thread-safe but never blocks.

Example: The Intel Haswell Architecture.