home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!cs.utexas.edu!sun-barr!ames!purdue!mentor.cc.purdue.edu!pop.stat.purdue.edu!hrubin
- From: hrubin@pop.stat.purdue.edu (Herman Rubin)
- Newsgroups: comp.arch
- Subject: Re: CISC Microcode (was Re: RISC Mainframe)
- Message-ID: <54987@mentor.cc.purdue.edu>
- Date: 24 Jul 92 13:14:48 GMT
- References: <Brsx7o.G69@zoo.toronto.edu> <1992Jul22.163956.57436@cc.usu.edu> <Brus8r.2K7@zoo.toronto.edu>
- Sender: news@mentor.cc.purdue.edu
- Organization: Purdue University Statistics Department
- Lines: 31
-
- In article <Brus8r.2K7@zoo.toronto.edu> henry@zoo.toronto.edu (Henry Spencer) writes:
- >In article <1992Jul22.163956.57436@cc.usu.edu> ivie@cc.usu.edu (CP/M lives!) writes:
- |>> Item 3 can still be
- |>> an issue, particularly on old architectures with clunky interrupt handling,
-
- |>What do you mean by "clunky interrupt handling"?
-
- >Many, many CPU cycles needed to get into and out of an interrupt, so it's
- >impossible to take very many of them or respond quickly to one. You have
- >to figure in both hardware and software interrupt overhead; the relevant
- >metric is elapsed time between stopping mainline execution and starting
- >the interrupt handler *for that specific interrupt*. Time spent figuring
- >out deciding who gets to handle this one, or saving and restoring registers
- >so he can be invoked safely, is part of the overhead regardless of whether
- >it's in software or hardware.
-
- >Pretty much the inescapable minimum is two pipeline breaks plus a cycle for
- >the "return from interrupt" instruction. There are CPUs that approach this
- >for simple cases where the interrupt can be dealt with quickly (they do
- >typically incur somewhat more overhead if the interrupt handler wants to
- >run for a while and use substantial resources).
-
- This "inescapable minimum" is at least partly due to the failure to realize
- that planned interrupts are becoming more necessary. The initial overhead
- of an interrupt should be no greater than that for a subroutine call, and
- possibly less. FULL context switching needs speeding up.
- --
- Herman Rubin, Dept. of Statistics, Purdue Univ., West Lafayette IN47907-1399
- Phone: (317)494-6054
- hrubin@pop.stat.purdue.edu (Internet, bitnet)
- {purdue,pur-ee}!pop.stat!hrubin(UUCP)
-