Path: utzoo!utgpu!jarvis.csri.toronto.edu!rutgers!gatech!purdue!mentor.cc.purdue.edu!l.cc.purdue.edu!cik From: cik@l.cc.purdue.edu (Herman Rubin) Newsgroups: comp.arch Subject: Re: Register Scoreboarding Summary: Scoreboarding can be important even without load bottlenecks Message-ID: <1310@l.cc.purdue.edu> Date: 18 May 89 10:50:42 GMT References: <24821@lll-winken.LLNL.GOV> <25484@ames.arc.nasa.gov> Organization: Purdue University Statistics Department Lines: 23 In article <25484@ames.arc.nasa.gov>, lamaster@ames.arc.nasa.gov (Hugh LaMaster) writes: > > In article <104999@sun.Eng.Sun.COM> khb@sun.UUCP (Keith Bierman - SPD Languages Marketing -- MTS) writes: > ......................... > The main point is that if you want to have multiple outstanding loads, > indeterminate memory delays due to bank busy conditions, multiple > outstanding instructions (all or almost all the ALU operations on these > machines are fully segmented- typically at one per clock cycle, though > not always with divide etc.), and, fairly long memory latencies > (e.g. 5-13 clock periods) scoreboarding is a nice way to handle it. Scoreboarding and related items can be important even if load time is not the bottleneck. Much mathematical code can use this, as operations need the result of previous operations. This is especially the case if operations do not take constant amounts of time, or if branches occur. It is sometimes even possible to arrange operations to take into account unusual timings, letting scoreboarding keep things efficient. -- Herman Rubin, Dept. of Statistics, Purdue Univ., West Lafayette IN47907 Phone: (317)494-6054 hrubin@l.cc.purdue.edu (Internet, bitnet, UUCP)