bukys@rochester.UUCP (Liudvikas Bukys) (06/25/84)
It does not bode well that the DMFDMA option causes two syntax errors (albeit minor ones) in vaxuba/dmf.c. Does anyone have any experience with using DMA on the DMF32? I'd like to know whether I should even consider buying (an emulation of) one at this point. Liudvikas Bukys rochester!bukys (uucp) via allegra, decvax, seismo bukys@rochester (arpa)
sdyer@bbncca.ARPA (Steve Dyer) (06/29/84)
Can anyone comment on the advantages of purchasing an ersatz DMF32 (e.g., Able VMZ32) instead of the corresponding DH/DM clone? As some of you have heard, I am trying to configure several VAX systems, and the advantages of one over the other of these terminal multiplexers is beyond me. Also, comments on the DZ11/KMC combination are welcome, too. For background, 20-25 active terminal lines at 9600 baud in a software development enviroment. Ah, for the good old days when only a DH-11 was reasonable... -- /Steve Dyer {decvax,linus,ima}!bbncca!sdyer sdyer@bbncca.ARPA
chris@mddc.UUCP (Chris Maloney) (06/29/84)
Reply to Bill Shannon: I recently did a lot of work on your DMF32 driver so that it would decide on SILO or DMA mode dynamically. At least on the CS11/F1 from EMULEX (a dmf32 emulator) this cut the CPU time by about 10%. My driver basically works like the DEC driver deciding on DMA when there are a lot of chars. and using SILO for a small number of chars. Chris Maloney Management Decisions Development Corp. 7209 Dixie Highway Fairfield, Ohio 45014 (513)874-6464 ...{ucbvax,decvax,inhp4,mhuxi}!cbosgd!mddc!chris (uucp) cbosgd!mddc!chris@BERKELEY (arpa)
johnsson@decwrl.UUCP (Richard Johnsson) (07/04/84)
I can't comment on performance, but I can report that the DMF-32 driver in 4.2 BSD works. It does not support DMA and I believe it does not support the parallel ports.
shannon@sun.uucp (Bill Shannon) (07/10/84)
When I wrote the original DMF driver at DEC, I did some crude measurements with and without DMA on a VAX-11/730. At that time, with a prototype DMF-32, my measurements showed that it was more efficient* to not use DMA. That's why the DMA code is commented out in the distributed driver. (The code used to work under 4.1 but was probably not modified for 4.2.) Has anyone done any measurements recently with real and look-alike DMF's and other cpu types? Bill Shannon Sun Microsystems, Inc. * more efficient == greater number of aggregate characters per second with less cpu overhead.
pete@lvbull.UUCP (Pete Delaney - Rocky Mountain UNIX Consultants) (07/21/84)
I use tracing in the network stuff, you might find that usefull. Most kernel hackers I've shown it to like it.