Oh this takes me back to a horrible time in my life when I was tracking down a terrible intermittent crash on a LEON (SPARC) processor. I was down at the assemb

Oh this takes me back to a horrible time in my life when I was tracking down a terrible intermittent crash on a LEON (SPARC) processor. I was down at the assembly level, stepping over instructions. It was a bit mind bending to always have an instruction execute AFTER the branch (you wouldn't know if the branch was taken or not until AFTER the delay slot instruction). Often the delay instruction was a NOP, showing the difficulty in filling the delay slots as you mentioned. If you were really lucky, your branch would also result in a register window overflow. By the time handling that was done, you lost your mental context of what was going on. shudder