Forums before death by AOL, social media and spammers... "We can't have nice things"
|    comp.lang.asm.x86    |    Ahh, the lost art of x86 assembly    |    4,675 messages    |
[   << oldest   |   < older   |   list   |   newer >   |   newest >>   ]
|    Message 3,809 of 4,675    |
|    Terje Mathisen to Rick    |
|    Re: Fast conversion to a boolean of 0 or    |
|    08 Mar 19 23:35:28    |
      From: terje.mathisen@nospicedham.tmsw.no              Rick wrote:       > Ranking:              Perfect prediction:       > #1 0, -1 1267       > #2 -1, 0 1285       > #3 Empty 1293       Very good branch prediction       > #4 JZ 1321       These are all effectively identical, they have the exact same critical       latency path length, just using different instructions.              > #5 Terje ADC 1546       > #6 James SBB + NEG 1554       > #7 SBB 1555       > #8 SETNZ 1554              CMOV is always quite slow, I have (almost) never found it a win.       > #9 CMOV 2077       It normally does not use value prediction, instead it will use a       three-operand instruction where the old value and the new value are       inputs along with the predicate: Minimum latency is typically 2 cycles,       all on the critical path.              Terje              --       - |
[   << oldest   |   < older   |   list   |   newer >   |   newest >>   ]
(c) 1994, bbs@darkrealms.ca