I did some minor simplification and wire management. The biggest change is I removed the clock from the registers making them asynchronous. Their load signal is robust enough for the task and this means less wires and logic. I left the clock on the RAM, but it could probably be easily removed as well for more efficiency.
I moved the number display to the program memory as it more accurately displays the line of code being processed. As the double-dabble and 7-segment drivers in the display are relatively complex, you can speed up the simulation by removing that component. I left it in for ease of de-bugging code.
The previous code and compiler works for this version, though I did update the micro-instructions B ROM to have the high and low outputs mirrored so either could be hooked up.