Return to BSD News archive
Path: sserve!manuel!munnari.oz.au!spool.mu.edu!olivea!sgigate!odin!sgi!rigden.wpd.sgi.com!rpw3 From: rpw3@rigden.wpd.sgi.com (Rob Warnock) Newsgroups: comp.unix.bsd Subject: Re: /386bsd: reject 65340 Message-ID: <q5cfo20@sgi.sgi.com> Date: 22 Sep 92 04:13:36 GMT Sender: rpw3@rigden.wpd.sgi.com Organization: Silicon Graphics, Inc. Mountain View, CA Lines: 34 barry@ictv.com (Barry Lustig) writes: +--------------- | rpw3@rigden.wpd.sgi.com (Rob Warnock) writes: | I was doing fine with my WD8013 "Combo" until I applied Barry Lustig's | patch to turn on 16-bit transfers. Then the "reject"s started: | Sep 19 04:33:01 jolly /386bsd: reject 65531reject 65531reject 65531 | I had to back out the change. (Barry? *Help!*) | | Sorry about the goof folks. laar.laar_byte and wem.ms_byte should | both be initialized to 0 before use... | wem.ms_byte = 0; | laar.laar_byte = 0; +--------------- Thanks! That did the trick. So, now I have Barry's 16-bit transfers, Bakul Shah's 80x86-specific in_cksum.c, and I changed the we0 config to use "iosiz 16384" (it's an Elite 16), so now I get 573 KB/s of "ttcp" receiving from a fast host (an SGI box) and 455 KB/s transmitting. Not great (many RISC workstations can get over 1000 KB/s), but not bad for an interface which bcopy's instead of DMA'ing. The CPU usage when transmitting (80-90%) is *much* higher than when receiving (<30%), so I suspect there's room for tuning the send speed... ;-} -Rob ----- Rob Warnock, MS-9U/510 rpw3@sgi.com Silicon Graphics, Inc. (415)390-1673 2011 N. Shoreline Blvd. Mountain View, CA 94043