home *** CD-ROM | disk | FTP | other *** search
- Xref: sparky comp.os.os2.programmer:4667 comp.os.os2.misc:28983
- Newsgroups: comp.os.os2.programmer,comp.os.os2.misc
- Path: sparky!uunet!weyrich!orville
- From: orville%weyrich@tnet.com (Orville R. Weyrich)
- Subject: Re: Does math coprocessor speed up GUI (here: OS/2 presentation mgr)?
- Message-ID: <1992Sep2.055929.24169@weyrich.UUCP>
- Sender: orville@weyrich.UUCP (Orville R. Weyrich)
- Reply-To: tnet.com!weyrich!orville
- Organization: Weyrich Computer Consulting
- References: <1992Aug31.034803.27829@unixg.ubc.ca>
- Date: Wed, 2 Sep 92 05:59:29 GMT
- Lines: 47
-
- In article <1992Aug31.034803.27829@unixg.ubc.ca> ochealth@unixg.ubc.ca (ochealth) writes:
- >
- >
- >This is pure speculation, as I haven't checked out the 386/7 or 486 instructions
- >but couldn't some PM applications seem faster because OS/2 can shove several
- >instructions at the coprocessor, and then continue with another task?
- >
- >I thought that the CPU and the FPU worked more-or-less independently of
-
- As far as I know, the O/S makes no difference -- it is entirely possible
- to have the 80x86 chip do some non-dependent ** calculations while the
- 80x87 chip is doing it's thing. There is an instruction that is the
- equivalent of a "join" -- it forces the 80x86 to wait for the 80x87 to
- complete its work before the program procedes further. But if there is
- something else that the 80x86 can do in the meantime [typical example:
- index or pointer arithmetic], the the only limitation is that of the
- programmer's or compiler's intelligence.
-
- ** A non-dependent calculation is one that can be done using input
- data that is not being produced by the co-processor.
-
- It is poor programming practice to leave out the "wait" instruction --
- even if the timing works out alright now, a future version of the
- 80x87 or 80x86 [or combined equivalent] might have different cycle times
- for the two parallel execution threads and the program might fail on those
- processors.
-
- orville
-
- -------------------------------------- **********************************
- Orville R. Weyrich, Jr. Weyrich Computer Consulting
- Certified Data Processor POB 5782, Scottsdale, AZ 85261
- Certified Systems Professional Voice: (602) 391-0821
- Certified Computer Programmer Internet: orville%weyrich@tnet.com
- -------------------------------------- **********************************
- >each other. Under DOS anyway, I thought the CPU had to wait for the FPU
- >to finish to get results back.
- >
- >jpm
- >
- >--
- >______________________________________________________________________________
- >jpm: ochealth@unixg.ubc.ca
- > Happily using OS/2 2.0 because MS Windows isNT ___
- > Insert VapourFeature ^^^
-
-
-