[sitkack]

My feeling is that has more to do with memory traffic and the cache hierarchy than the difference between register and stack VMs. Stack VMs are going to generate more read-modify-write cycles than a register VM. Of course experiments are in order (or out of order, ha!).

[questerzen]

I imagine you are right. But there should be plenty of scope for optimising this in the VM itself. Most stack operations are immediately preceded by a push, so hold the top of the stack in a register and the last push in another and you could avoid a high proportion of memory calls in the most common cases. Does anyone know if JVM implementations do this?

[questerzen]

From the Wikipedia page on the JVM: "[code verification]...allowing the JIT compiler to transform stack accesses into fixed register accesses. Because of this, that the JVM is a stack architecture does not imply a speed penalty for emulation on register-based architectures when using a JIT compiler."