Yes, the observable semantics is 100% identical to running the same code in an existing JS engine. That's the genius behind Emscripten (i.e., the genius of Alon Zakai) -- he figured out that you can effectively simulate the semantics of machine integer operations by exploiting the various parts of the JS semantics that internally does ToInt32/ToUint32 on its arguments.
What asm.js is simply formalize those tricks in order to guarantee that you're only using those parts of the JS semantics, so that it's sound for an optimizing engine to directly implement them with machine arithmetic and unboxed integers. But the behavior is absolutely identical to a standard JS interpreter/JIT/etc.
What asm.js is simply formalize those tricks in order to guarantee that you're only using those parts of the JS semantics, so that it's sound for an optimizing engine to directly implement them with machine arithmetic and unboxed integers. But the behavior is absolutely identical to a standard JS interpreter/JIT/etc.