Sun-Spots-Request@RICE.EDU (William LeFebvre) (02/16/88)
SUN-SPOTS DIGEST Monday, 15 February 1988 Volume 6 : Issue 15 Today's Topics: Re: Sun-4 bcopy warning Re: Adding a node Re: Sun vs. DELNI Re: more spurious level 3 interrupts! Re: Lisp for Sun-4? (2) Make bug in SunPro make Bad blocks "fig" program problems Handling Events in SunCore? Diskless CAD Stations? Poor UUCP performance on Sun 3's and 4's? Press Release: Sun Common Lisp for Sun-4 workstations Send contributions to: sun-spots@rice.edu Send subscription add/delete requests to: sun-spots-request@rice.edu Bitnet readers can subscribe directly with the CMS command: TELL LISTSERV AT RICE SUBSCRIBE SUNSPOTS My Full Name Recent backissues are stored on "titan.rice.edu". For volume X, issue Y, "get sun-spots/vXnY". They are also accessible through the archive server: mail the word "help" to "archive-server@rice.edu". ---------------------------------------------------------------------- Date: Tue, 2 Feb 88 15:56:23 EST From: Root Boy Jim <rbj@icst-cmr.arpa> Subject: Re: Sun-4 bcopy warning > The bcopy routine on the Sun-4's apparently has not been optimized for the > machine at all....An unrolled loop, moving a word at at time (except > for the last few bytes) does much better. Hey, remember this trick? bcopy(from,to,count) register int count; register char *from, *to; { switch (count & 7) do { case 0: *to++ = *from++; case 7: *to++ = *from++; case 6: *to++ = *from++; case 5: *to++ = *from++; case 4: *to++ = *from++; case 3: *to++ = *from++; case 2: *to++ = *from++; case 1: *to++ = *from++; } while ((count -= 8) > 0); } I may have misquoted it, but at least it compiles. You get a warning about loop not entered at top. (Root Boy) Jim Cottrell <rbj@icst-cmr.arpa> National Bureau of Standards Flamer's Hotline: (301) 975-5688 [[ It doesn't do the appropriate thing when "count == 0". :-) --wnl ]] ------------------------------ Date: 2 Feb 88 20:48:44 GMT From: esosun!pete@seismo.css.gov (Pete Ware) Subject: Re: Adding a node Barry Shein described how to add an extra client without repartitioning a disk (assuming you have enough space!). Add in his description of how to decide where to put the new client and then follow these procedures: (BE CAREFUL! MISTAKES CAN BE FATAL) To install a new client. Assumptions: 1. Sun 3.0 to 3.4. 2. Server is the yellow pages master. 3. Server is running a Fuji eagle disk. 4. Have at least one running client. Steps: All of the files changed on the server, unless otherwise indicated (if server is not YP master, a couple should be edited on the YP master). [[ 0. Back up the ENTIRE disk! --wnl ]] 1. Add swap and root partitions to /etc/nd.local. A 3/50 needs about 8M (18 cyl. or 16560 sectors) for the root partition and 16M (37 cyl. or 34040 sectors) for the swap partition. See the nd manual page for details on the format. 2. Add internet address and name to /etc/hosts (YP file) 3. Add host name to /etc/hosts.equiv (be sure this is not a security problem before doing so!). 4. Power up machine to get the ethernet address (looks like 8.0.20.1.ef.54). 5. Add ethernet address and host name to /etc/ethers (YP file). 6. Update yellow pages (if appropriate): # cd /etc/yp; make 7. Tell the server about the new nd information: # /etc/nd - < /etc/nd.local 8. Make sure /dev/ndl? exists (where the "?" is the nd client number. # cd /dev # MAKEDEV ndl? 9. Create the actual root file system. Since this is a "soft" partition, newfs will not work on it. The following creates an 8 Mbyte partition on a Fufi eagle disk: # mkfs /dev/ndl? 16560 46 20 8192 1024 For further information, see "Sun Network Services," pg. 14 and the mkfs manual page. IF YOU HAVE SCREWED UP /etc/nd.local, THIS STEP WILL CAUSE MUCH TROUBLE! 10. Mount the root partition on the server: # mount /dev/rndl? /host-name 11. Copy an existing clients root partion over (this is also on the distribution tapes, in which case you must update some further files): # dump 0f - /dev/rndl0 | (cd /host-name; restore xf -) There will probably be an error message about an existing lost+found and possibly some "impossible file types", ignore these. Should take about 2-3 minutes. 12. Edit /host-name/etc/rc.boot to change the hostname to host-name. 13. Add symbolic link to /tftpboot. The name of the link is the hexadecimal version of the internet address with each field in the address taking up two digits of the hex number (leading 0's). 14. Unmount the client's partition. 15. Boot the client. For further information see: boot(8) nd(5) mkfs(8) mount(8) "Sun Network Services" esosun!pete@seismo.CSS.GOV (Pete Ware) (619) 458-2520 {seismo,sdcsvax}!esosun!pete [[ I have not been reading these descriptions too closely. But I want to warn everyone who is attempting these procedures: if you accidentally overlap two client's partitions, you will lose BOTH partitions (more if the overlap is large enough). You may only be making changes to one client's description, but it could easily effect any other client on the server if you make a mistake. Don't be content with backing up just one client's partition.....do the entire disk! A mistake like that could be costly! William's first rule of system management: backup, check, double check, and triple check! --wnl ]] ------------------------------ Date: Tue, 02 Feb 88 18:37:22 +0000 From: Jim Reid <mcvax!cs.strath.ac.uk!jim@uunet.uu.net> Subject: Re: Sun vs. DELNI We have a couple of Sun-2's running SunOS2.0 that work just fine with a DELNI. One (a 2/120) has a Multibus ethernet interface (3 Com?) and the other (a 2/50?) has a Sun interface on the CPU/memory/everything board. The installation engineer had to change a jumper on the 2/50 so that it would work with the DELNI. According to an old 2/160 installation manual, jumper J704 on the 2050 CPU board has to be "taken out for Level 2 transceivers such as the DELNI or a TCL 2010I". The boards are shipped with this jumper IN and so by default, the board is set up for "Level 1 transceivers such as 3Com 3C100 and TCL2010E". Jim ------------------------------ Date: Wed, 3 Feb 88 08:36:24 CST From: fritz@astro.as.utexas.edu (Fritz Benedict) Subject: Re: more spurious level 3 interrupts! Jonathan Eisenhamer, UCLA Astronomy writes: > What does the following message mean when it appears on the console? > spurious level 3 interrupt ... Our Sun-3/160C crashes when we see these, again always late at night, during times of low activity. We destalled a SKY Warrior AP and the frequency of crash has declined a lot. We haven't yet figured out why the AP generates spurious level 3 interrupts. Fritz Benedict (512)471-3448 uucp: {allegra,ihnp4}!{noao,ut-sally}!utastro!fritz internet: fritz@astro.AS.UTEXAS.EDU snail: Astronomy, U of Texas, Austin, TX 78712 ------------------------------ Date: 3 Feb 1988 08:51-EST From: Ralph.Hyre@ius3.ius.cs.cmu.edu Subject: Re: Lisp for Sun-4? (1) The hearsay around here is that 'Lucid Common Lisp on a Sun-4 is only twice as fast as on the Sun-3', so this implies that there is a solution that may be available directly from Lucid. It seems like Kyoto Common Lisp might work, since it compiles to C-code rather than directly to machine code. Hopefully any Sun-4 Lisp will make good use of the SPARC capabilities. (described in the edited netnews articles below) __________ > From: garner@Sun.COM (Robert Garner) > Subject: Re: Why is SPARC so slow? [...] > Register windows have several advantages over a fixed set of registers, > besides reducing the number of loads and stores by about 30%: They work > well in LISP (incremental compilation) and object-oriented environments > (type-specific procedure linking) where IRA is impractical. They can also > be used in specialized controller applications that require extremely fast > context switching: a pair of windows (32 registers) can be allowed per > context. > > Robert Garner > Sun Microsystems > > DISCLAIMER: I speak for myself only and do not represent the views > of Sun Microsystems, or any other company. __________ > From: hedrick@athos.rutgers.edu (Charles Hedrick) > Subject: hardware tagging in SPARC and DEC-20 > > Some comments on the use of tags in the SPARC and DEC-20. We should not > give the impression that either of these machines have hardware tagging in > quite the generality of the Lisp Machine. It is true that both of these > machines were designed specifically with Lisp in mind. [...] > > Clearly the SPARC designers were thinking of Lisp. There are special > opcodes for tagged add and subtract. The tag is the low-order two bits. > Integer arithmetic is done. [...] > > (Note however that I have no plans to do a SPARC Lisp. Several quite > competent groups are doing it already.) Obviously a tag value of zero is > used for 30-bit integers. I think I would use tag values of 1 and 3 for > 31-bit reals and tag value 2 for numbers that need to be pointers (bignums > and larger reals). > > In summary it seems to me that the SPARC design would get roughly the same > results as the DEC-20, namely a slowdown of about a factor of 2 in > arithmetic instructions. Since not all of your program is arithmetic > instructions, the actual results should be better than that, like maybe > 1.5. [..] > > In addition, tag support on SPARC could be used to help in implementing > type security. [...] ------------------------------ Date: Wed, 3 Feb 88 00:40:25 CST From: steve@ncsa.uiuc.edu (Steve Christensen) Subject: Re: Lisp for Sun-4? (2) In volume 6n10 you ask about Lisp for the SUN-4. There are several options: The SUN SCLisp Release 3.0 is now in beta test and the last I was told had an April 1988 release date. I do not have the cost yet. In the past the lisps from SUN have run around $3,000. The Symbolic Programming Environment (SPE) is also in beta test. The IBUKI people have announced a SUN-4 version of their brand of Kyoto Common Lisp. I have also been told that a version of Lucid CL will be ready for the 4 if it isn't already. I have seen the benchmarks for the Franz/SUN version (admitted a pre beta version) and the 4/260 seems to run the Gabriel benchmarks about 2-3 times faster than a 3/260, 3-4 times faster than a 3/60 and significantly faster than a 3/160. I have been told that a lot more optimization is possible using the new facilities provided by the SPARC chip and that even better performance is expected. Naturally, the more memory you add the better off you are. When I get my copy of the SUN Lisp, I will do an independent running of various benchmarks including a compile of a large code like Macsyma (Vaxima) to see how that times on a 3/160 versus a 4/260 and send the information to Sun-Spots if anyone is interested. I would like to hear from other Sun-4 Lisp users also. Steve Christensen NCSA, Illinois steve@spock.ncsa.uiuc.edu [[ Steve later sent the following addendum: ]] In my previous note about SUN-4 Lisp I mentioned a Release 3.0 version. SUN has now announced a version 2.1 that is based on the Lucid lisp. I suspect there is still a 3.0 version coming, but the details of that are now fuzzy to me. Steve Christensen NCSA, Illinois [[ Indeed they have! I just got the press release announcing Sun Common Lisp 2.1. I have included it at the end of this digest for all those who are interested. --wnl ]] ------------------------------ Date: 2 February 1988 1307-PST (Tuesday) From: @nprdc.arpa:lim@hull (Bill Lim) Subject: Make bug in SunPro make Has anybody run across this bug using Sun's SunPro make: If the line .KEEP_STATE: is present in the makefile, then make is supposed to create a file called .make.state. One of the things this file is supposed to contain are all of the include files each target depends on. However, it doesn't. Only some of the include files are listed. And, it only does this if one is building libraries. It works fine if you aren't building libraries. As a consequence, make compiles everything twice! Even if the library modules are up to date. Here is an example: SAMPLE CODE: #include <suntool/sunview.h> /* this includes many files */ #include <stdio.h> main () { return; } SAMPLE MAKEFILE: FLAGS = -g -m68010 TARGET_ARCH = -mc68010 LIB = libtest.a INCLUDE = OBJS = $(LIB)(test.o) .KEEP_STATE: .PRECIOUS: $(LIB) makelib: $(MAKE) $(LIB) ranlib $(LIB) $(LIB) : $(OBJS) SAMPLE OUTPUT: % make make libtest.a cc -mc68010 -c -o test.o test.c ar rv libtest.a test.o r - test.o rm -f test.o cc -mc68010 -c -o test.o test.c ar rv libtest.a test.o r - test.o rm -f test.o `libtest.a' is up to date. ranlib libtest.a SAMPLE .make.state FILE: .MAKE_VERSION: VERSION-1.0 .BUILT_LAST_MAKE_RUN: libtest.a(test.o): /usr/include/suntool/sunview.h /usr/include/signal.h /usr/include/pixrect/pixrect_hs.h /usr/include/sys/types.h /usr/include/sys/sysmacros.h /usr/include/pixrect/pixrect.h /usr/include/pixrect/pr_dblbuf.h /usr/include/pixrect/pr_line.h /usr/include/pixrect/pr_planegroups.h /usr/include/pixrect/pr_util.h /usr/include/pixrect/traprop.h /usr/include/pixrect/bw1var.h /usr/include/pixrect/cg1var.h /usr/include/pixrect/bw2var.h /usr/include/pixrect/cg2var.h libtest.a(test.o): .BUILT_LAST_MAKE_RUN: makelib:.RECURSIVE . libtest.a Makefile .BUILT_LAST_MAKE_RUN: makelib: make libtest.a ranlib libtest.a As you can see in the sample .make.state file, stdio.h is not listed, as well as half of the suntools files. (To see all of the include files run cpp -M.) I have talked to sun about this, but they can not recreate it. Has anybody else seen this? Or, am I doing something wrong? I am running SunOS 3.4, on a Sun-3/140 with 8-megs. Bill Lim NPRDC, San Diego. (619) 553-7684 lim@nprdc.arpa ihnp4 \ akgua \ decvax >---- !sdcsvax!sdics!nprdc!lim dcdwest / ucbvax / ------------------------------ Date: Tue, 2 Feb 88 10:14:25 +0100 From: mcvax!solaris!wyle@uunet.uu.net (Mitchell Wyle) Subject: Bad blocks I use diag. To avoid the '+' plus command in diag, I use: rsh orion cat /private/usr/adm/m* | grep blk | awk ' { abs1=$12; abs=substr(abs1,2,length(abs1)); cyl=abs/920; blk=abs%920; head=blk/46; sect=blk%46; printf("abs #%d cyl #%d blk #%d head #%d sect #%d\n", abs, cyl, blk, head, sect); }' to find out the correct disk info. It takes 1/2 hour to slip or map the disk. Don't be afraid. -Mitch -Mitchell F. Wyle wyle@ethz.uucp Institut fuer Informatik wyle%ifi.ethz.ch@relay.cs.net ETH Zentrum 8092 Zuerich, Switzerland +41 1 256-5237 I'm prepared for all emergencies but totally unprepared for everyday life. ------------------------------ Date: Wed, 3 Feb 88 10:02:51 EST From: Bob Chatham <bchatham@bfly-vax.bbn.com> Subject: "fig" program problems I retrieved the "fig" program described in v6n7 from sally.utexas.edu and have been having problems un-tarring it. I get the message "tar: directory checksum error (0 != -1130)". I can't retrieve the file again since sally has been unreachable for a few days. Can you point me to a new source for the program, or do you know what might be causing the tar error? Thanks, Bob Chatham bchatham@bfly-vax.bbn.com [[ Did you retrieve the file with transfer TYPE set to IMAGE? Tar files contain binary data, and should be transferred as such. After connecting with FTP, use the command "type image" before the transfer. I hope to have a copy of "fig" in the sun-spots archives real soon now... --wnl ]] ------------------------------ Date: 2 Feb 88 15:23:37 GMT From: has@comp.lancs.ac.uk (Howard Siwek) Subject: Handling Events in SunCore? Has anyone ever attempted to catch window events, like window entry and exit, from within SunCore? I would like to know which window the mouse is positioned over at any time. Please read 'window' as 'view surface' within SunCore. Any suggestions would be appreciated. Thanks in advance Howard. Please reply to: UUCP: ..seismo!mcvax!ukc!dcl-cs!has DARPA: has@uk.ac.lancs.comp@ucl-cs JANET: has@uk.ac.lancs.comp ------------------------------ Date: 2 Feb 88 22:24:27 GMT From: cnt.mn.org!rod@rutgers.UUCP (rod merry) Subject: Diskless CAD Stations? We are looking for people with experience running CAD software packages (like CASE Technologies) on diskless workstations (like Suns). We are worried about response time in this type of environment and have not got a satisfactory answer from a number of vendors. Thanks rod Rod Merry rod@cnt.MN.ORG Computer Network Technology {quest|bungia}!cnt!rod 9440 Science Center Drive Minneapolis, MN 55428 (612)535-8111 ------------------------------ Date: Tue Feb 2 17:12:03 1988 From: psuvax1!lll-winken!ddsw1!spl1!raj@rutgers.edu (Robert Alan Johnson) Subject: Poor UUCP performance on Sun 3's and 4's? I have been trying to figure out why uucp performance is so bad on sun 3 and sun 4's. At first I was told it was due to interrupt loads generated by the old alm boards. But, even on the new ones it still seems too slow, with throughputs of <160CPS on a sun 4 "unloaded" at 2400BPS! Sun seems stumped! Bob Johnson The Software Public Library ...ihnp4!spl1!raj ...1+312=248-5777 ------------------------------ Date: Wed, 3 Feb 88 13:25:32 PST From: marleen@sun.com (Marleen Martin McDaniel) Subject: Press Release: Sun Common Lisp for Sun-4 workstations SUN COMMON LISP FOR SUN-4 WORKSTATIONS SANTA CLARA, CA -- February 3, 1988 -- Sun Microsystems today announced Sun Common Lisp for the Sun-4 family of supercomputing workstations. The new Sun-4 Common Lisp, release 2.1, is source code compatible with the Sun-2 and Sun-3 Sun Common Lisp implementations, so Sun's many third-party artificial intelligence (AI) vendors can easily move their AI applications to the new Sun-4 SPARC-based platform. "The SPARC architecture supports symbolic processing languages and is ideally suited for AI applications. The new version of this Lisp will open a large market for the Sun-4 family and bring a new level of performance to the AI industry," said Dr. Eric Schmidt, vice president and general manager of the Software Products Division at Sun. "Sun Common Lisp on the Sun-4/110 introduced today makes an extremely powerful, low-cost desktop workstation for the development and delivery of AI applications," he concluded. "We are very enthusiastic about Sun Common Lisp 2.1 running on the Sun-4 family and the resulting portability," said Chuck Williams, executive vice president of Inference Corp., which markets expert system development tools and services. "Together, the new Sun Common Lisp and the Sun-4 family of workstations will provide a very powerful platform for commercial expert system development." Sun Common Lisp 2.1, which is based on Common Lisp from Lucid, Inc., of Menlo Park, California, is enhanced with tools that optimize its performance on Sun workstations and integrate Lisp capabilities into Sun's multipurpose programming environment, based on the UNIX operating system. Common Lisp is the favorite language of AI researchers because of its powerful symbolic processing capabilities. It has served as the host for a range of AI applications in natural language, robotics, vision and expert systems. In a related announcement, the company stated that its Symbolic Programming Environment (SPE) will be available on Sun Common Lisp 2.1 for the Sun-3 and Sun-4 in June. The SPE is a complete development environment for Common Lisp with a set of tools for debugging, analyzing and developing Lisp-based applications. Sun Common Lisp for Sun-4 workstations is available in March 1988. Pricing starts at $3,500 (U.S. list). Volume discounts are available. Sun-4 workstations are based on the Scalable Processor Architecture (SPARC), a reduced instruction set computer (RISC) microprocessor architecture designed and licensed by Sun. Sales of Sun products for AI applications account for about five percent of Sun's revenues. Sun Microsystems, Inc., is a leading supplier of network-based distributed computing systems, including technical workstations, servers, and UNIX operating system and productivity software. Press Contact: Sun Microsystems, Inc., Cathleen Beall Garfield (415) 691-6536 ------------------------------ End of SUN-Spots Digest ***********************