On 3/11/10 4:57 AM, Peter Lobsinger wrote:
It would be great if we could somehow get low-level, easily jittable ops and easily validatable ops at the same time. Striving for both, or at least the right balance, is something we should probably be thinking about when designing Lorito.
The idea for Lorito is that there will be two levels of ops: a) the low-level ops that are the core of the VM, and b) high-level "virtual" ops composed of those low-level ops, at about the same semantic level as our current opcodes. The low-level ops will be directly JITed (as in, we provide LLVM IR or other representations for each one) while the high-level ops will be JITed the same way as subroutines, composing the total block to JIT from the small, low-level ops.
The point to stick the security categorizations for sandboxing is on the high-level ops, rather than the low-level ops. Any work that goes into categorizing our current opcodes will be carried across into Lorito.
Allison _______________________________________________ http://lists.parrot.org/mailman/listinfo/parrot-dev
