Hi, Can I - for a special mechanism that is defined by some special semantics - add a parallel side effect setting a register (e.g. from unspec) to a >>jump_insn? When doing that, it looks like CSE doesn't understand this. Is there a way (during expand by generating proper RTL) to tell each optimization phase that a side effect of a certain jump_insn is writing to a certain register? Thanks, Regards, Hendrik Greving
