You do the same with fp. First floating point instruction traps and the process is then flagged as needing fp. This is then used in the context switch to also i

You do the same with fp. First floating point instruction traps and the process is then flagged as needing fp. This is then used in the context switch to also include fp instructions. The advantage is you avoid storing fp registers unless you are going to use them. That flag could easily determine what you can run where.