RFE: option to have AUTORESET reset flops to x rather than 0

[jwise]

I've been trying to get a little better at leaving invalid state machine parameters initialized to x rather than 0 in my resets (knowing that my synthesis tool will squash them to 0 eventually), so that I can get earlier feedback in simulation if something takes a path I do not expect it to. It would be nice to have an option to have AUTORESET set flops to x as a default, and then have to explicitly set flops to either asserted or deasserted if I care.

[wsnyder]

That should be easy to add (e.g. AUTORESETX?), but I feel the need to caution this can be a dangerous methodology.

Synthesis is allowed to pick any value it wants that optimizes well (not just 0), and simulation might not catch it given that in IEEE "if ('x)" evaluates false. So, assigning Xs is a known route to synthesis-simulation mismatch bugs.

To counteract that, with such a methodology the user needs to be certain to be using formal or x-pessimism simulator (or verilator -x-assign unique).