johnl@ima.ISC.COM (12/14/87)
Can someone explain the technique of using "scoreboards" in compiler development? Any references? -- Send compilers articles to ima!compilers or, in a pinch, to Levine@YALE.EDU Plausible paths are { ihnp4 | decvax | cbosgd | harvard | yale | bbn}!ima Please send responses to the originator of the message -- I cannot forward mail accidentally sent back to compilers. Meta-mail to ima!compilers-request
meissner@mcnc.org (12/18/87)
In article <797@ima.ISC.COM>, ...!harvard!talcott!motcoh!mark writes: > Can someone explain the technique of using "scoreboards" in > compiler development? Any references? I don't have any hard references, but when I ran into the CDC 6600 hardware it was described as having a hardware scoreboard. With regard to compilers, it comes up with some of the RISC designs, where the compiler has to do instruction scheduling and branch delays. I would imagine that is where the term came from. A scoreboard keeps track of the various functional units of the machine, and will not allow access to a unit, until it finishes. For example, if divide takes 8 clock units, a scoreboard would indicate that the any reference within the 8 clock units to either the input and result registers would pend instructions. Even if the hardware provides such a scoreboard, on such machines it is usually an advantage to have the compiler know about these delays, so that it can try to do more stuff in parallel. -- Michael Meissner, Data General. Uucp: ...!mcnc!rti!xyzzy!meissner Arpa/Csnet: meissner@dg-rtp.DG.COM -- Send compilers articles to ima!compilers or, in a pinch, to Levine@YALE.EDU Plausible paths are { ihnp4 | decvax | cbosgd | harvard | yale | bbn}!ima Please send responses to the originator of the message -- I cannot forward mail accidentally sent back to compilers. Meta-mail to ima!compilers-request
wallach@seismo.CSS.GOV (12/24/87)
/* Written 9:00 am Dec 18, 1987 by johnl@ima.UUCP in convex:comp.compilers */ In article <797@ima.ISC.COM>, ...!harvard!talcott!motcoh!mark writes: > Can someone explain the technique of using "scoreboards" in > compiler development? Any references? I don't have any hard references, but when I ran into the CDC 6600 hardware it was described as having a hardware scoreboard. With regard to compilers, it comes up with some of the RISC designs, where the compiler has to do instruction scheduling and branch delays. I would imagine that is where the term came from. A scoreboard keeps track of the various functional units of the machine, and will not allow access to a unit, until it finishes. For example, if divide takes 8 clock units, a scoreboard would indicate that the any reference within the 8 clock units to either the input and result registers would pend instructions. Even if the hardware provides such a scoreboard, on such machines it is usually an advantage to have the compiler know about these delays, so that it can try to do more stuff in parallel. -- Michael Meissner, Data General. Uucp: ...!mcnc!rti!xyzzy!meissner Arpa/Csnet: meissner@dg-rtp.DG.COM -- Send compilers articles to ima!compilers or, in a pinch, to Levine@YALE.EDU Plausible paths are { ihnp4 | decvax | cbosgd | harvard | yale | bbn}!ima Please send responses to the originator of the message -- I cannot forward mail accidentally sent back to compilers. Meta-mail to ima!compilers-request /* End of text from convex:comp.compilers */ -- Send compilers articles to ima!compilers or, in a pinch, to Levine@YALE.EDU Plausible paths are { ihnp4 | decvax | cbosgd | harvard | yale | bbn}!ima Please send responses to the originator of the message -- I cannot forward mail accidentally sent back to compilers. Meta-mail to ima!compilers-request
johnl@ima.UUCP (12/25/87)
In article <808@ima.ISC.COM> xyzzy!meissner@mcnc.org writes: |In article <797@ima.ISC.COM>, ...!harvard!talcott!motcoh!mark writes: |> Can someone explain the technique of using "scoreboards" in |> compiler development? Any references? | | I don't have any hard references, but when I ran into the CDC 6600 |hardware it was described as having a hardware scoreboard. It could also be a misnomer for a "register history", which is useful on machines without any visible parallellization, to keep track of what register currently holds what variable or value. For an example, see: Jim Welsh and Michael McKag, "Structures Systems Programming", Englewood Cliffs, N.J. (Prentice-Hall), 1980. -- Send compilers articles to ima!compilers or, in a pinch, to Levine@YALE.EDU Plausible paths are { ihnp4 | decvax | cbosgd | harvard | yale | bbn}!ima Please send responses to the originator of the message -- I cannot forward mail accidentally sent back to compilers. Meta-mail to ima!compilers-request