Charles Mills wrote: > Do I recall correctly hearing that BCTR Rn,0 is no longer the favored way of > decrementing a register, perhaps because the cache logic sees it as a > potential branch, and that AHI Rn,-1 should be substituted?
Seems unlikely to me. I'd think even the worst branch prediction algorithm is going to predict a BCTR Rn,0 as "not taken" (what're you going to do - EXecute it and change the second register to non-zero?) AHI is a larger instruction, and it has potential program checks, which BCTR hasn't. I'd guess that makes it harder to set up for execution. > Similarly, that AHI is preferable to LA for adding a small increment to a > register, perhaps because LA may invoke or "reserve" the address translation > logic? Again, seems unlikely. LA can't program check, and although it sounds like an addressing instruction, doesn't deal with things like access registers or translation. But I'm not a z hardware or millicode designer. Tony H. ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
