On Mon, Jan 15, 2018 at 11:54:41AM -0600, Bill Schmidt wrote: > > I think we settled on calling the option -mmispredict-indirect-jumps; > > please let me know if you still agree with that. Or have thought of a > > better name :-) > > Looks like we are now looking at -m[no-]speculate-indirect-jumps with > default to true. LLVM folks are in agreement too.
A fine name! > >> For the indirect jump, I don't see a way around it due to the > >> expected form of indirect jumps in cfganal.c. > > > > I'm not sure what you are getting at here, could you explain a bit? > > An indirect jump is expected to be of the form (set (pc) (...other stuff)); > otherwise it might get missed and blocks that are only reachable from > an indirect jump can get deleted. I found this out when I tried to do > something involving a parallel that was not very bright; that doesn't > actually prevent anything if you are doing things right. And the > clobber solution you suggest should be just fine. (rtlanal.c:computed_jump_p). Yeah, PARALLELs work just fine. Segher
