tas@tasman.cc.utas.edu.au (Tasman Derk Van Ommen) (12/10/90)
I guess this is not purely a hardware question, but readers of this group are probably best in a position to answer it... I recently purchased a 386sx machine and since it came with the machine I installed Quarterdeck's QEMM memory manager. I then ran some benchmarks - and found that my performance on floating point operations (I don't have a coprocessor) was down by 25% or thereabouts. Does anyone have an explanation? Tas van Ommen Physics Department email: tas@physvax.phys.utas.edu.au University of Tasmania
chuck@umbc5.umbc.edu (Chuck Rickard; ACS (UGRAD)) (12/10/90)
In article <tas.660781039@tasman> tas@tasman.cc.utas.edu.au (Tasman Derk Van Ommen) writes: >I recently purchased a 386sx machine and since it came with the machine >I installed Quarterdeck's QEMM memory manager. I then ran some benchmarks - >and found that my performance on floating point operations (I don't have >a coprocessor) was down by 25% or thereabouts. Does anyone have an >explanation? I'm having similar problems with HIMEM.SYS that comes with MS Windows 3.0. Without it installed, the Norton 4.5 System Info (SI) rating is 17.1, but with it installed, the SI drops to 14.9. This seems to be mighty big CPU hog if I ever saw one. I tried QEMM and got no slowdown, but I had a compatibility problem with WP 5.1 (It crashed when attempting a spell check). Anyone else have any clues? Oh, I might want to add that I do have a math coprocessor installed (387sx). Chuck Rickard (chuck@umbc5.umbc.edu)
reisert@ricks.enet.dec.com (Jim Reisert) (12/11/90)
In article <tas.660781039@tasman>, tas@tasman.cc.utas.edu.au (Tasman Derk Van Ommen) writes: > >I recently purchased a 386sx machine and since it came with the machine >I installed Quarterdeck's QEMM memory manager. I then ran some benchmarks - >and found that my performance on floating point operations (I don't have >a coprocessor) was down by 25% or thereabouts. Does anyone have an >explanation? QEMM puts the processor in protected mode. In this mode, interrupts slow down. FP coprocessor uses interrupts to signal when results are ready. Hence the slowdown. - Jim =-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-= "The opinions expressed here in no way represent the views of Digital Equipment Corporation." James J. Reisert Internet: reisert@ricks.enet.dec.com Digital Equipment Corp. UUCP: ...decwrl!ricks.enet!reisert 77 Reed Road Hudson, MA 01749-2895
reisert@ricks.enet.dec.com (Jim Reisert) (12/11/90)
In article <3599@ryn.mro4.dec.com>, I wrote, " QEMM puts the processor in protected mode. In this mode, interrupts slow down. FP coprocessor uses interrupts to signal when results are ready. Hence the slowdown." Of course, I failed to read the original posting that said he didn't have a coprocessor installed. Never mind. Though I would like to hear the real answer. - Jim >=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-= > >"The opinions expressed here in no way represent the views of Digital > Equipment Corporation." > >James J. Reisert Internet: reisert@ricks.enet.dec.com >Digital Equipment Corp. UUCP: ...decwrl!ricks.enet!reisert >77 Reed Road >Hudson, MA 01749-2895 =-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-= "The opinions expressed here in no way represent the views of Digital Equipment Corporation." James J. Reisert Internet: reisert@ricks.enet.dec.com Digital Equipment Corp. UUCP: ...decwrl!ricks.enet!reisert 77 Reed Road Hudson, MA 01749-2895
johnl@iecc.cambridge.ma.us (John R. Levine) (12/11/90)
In article <3599@ryn.mro4.dec.com> reisert@ricks.enet.dec.com (Jim Reisert) writes: >QEMM puts the processor in protected mode. In this mode, interrupts slow >down. Quite true. >FP coprocessor uses interrupts to signal when results are ready. >Hence the slowdown. Not true at all, there's a dedicated READY line. But you're close. The original question was about a machine with no 387, and the usual way to emulate a 387 is to trap all of the floating point instructions and simulate them. Since the traps are slower, the emulation is slower. -- John R. Levine, IECC, POB 349, Cambridge MA 02238, +1 617 864 9650 johnl@iecc.cambridge.ma.us, {ima|spdcc|world}!iecc!johnl "Typically supercomputers use a single microprocessor." -Boston Globe