> Good point... if we wanted to be completely safe we'd either have to > require a prototype to be in scope or forbid using instruction > operands directly as function call operands altogether.
I feel like I can probably help with some useful ideas, but I don't have a good enough picture of what's going on. Anyone want to do a 30-60 minute phone call to sort this out? If Gabe did a little bit of prep so he could figure out some code to walk us through (and telling us to do some prep like compiling the simulator is fine too), I think we could hammer a few things out quickly. I'm not clear on exactly what's going on here, but it seems to me that proxy classes could solve some of the type problems. For example, with partial template specialization, you can basically put a simple no-op wrapper around most types but flag any use of a non-const reference as a compiler error. Nate _______________________________________________ m5-dev mailing list [email protected] http://m5sim.org/mailman/listinfo/m5-dev
