From ehg Mon Feb 5 23:23 EST 1990 /netlib/toms/big652 contains a version of hompack that may be newer than /netlib/hompack. I'd appreciate it if you take a look sometime and see what you think. From wmc Fri Mar 9 13:26 EST 1990 I checked with Layne Watson. He says that netlib/hompack is current. From vtcs1.cs.vt.edu!ltw%cayuga.decnet Fri Mar 9 12:20:23 1990 Received: by pyxis; Fri Mar 9 12:20:23 1990 Received: by inet.att.com; Fri Mar 9 12:20:17 1990 Date: 9 Mar 90 12:09:00 EST From: "CAYUGA::ltw" Subject: Re: hompack in netlib To: "wmc" Bill, I sure hope they are the same (netlib HOMPACK and TOMS 652). Netlib had it first, and there may be cosmetic changes in the TOMS version. If you discover any actual code differences, please let me know. If you don't object, I would like HOMPACK to remain listed in Netlib as it currently is, and not buried in the CA where few people will notice it. Thanks. Layne From wmc Mon Feb 4 15:49 EST 1991 >From cayuga.cs.vt.edu!ltw Mon Feb 4 15:48:02 0500 1991 Received: by pyxis; Mon Feb 4 15:49 EST 1991 Received: by inet.att.com; Mon Feb 4 15:49 EST 1991 Received: by cayuga.cs.vt.edu (5.57/Ultrix3.0-C) id AA17314; Mon, 4 Feb 91 15:48:02 -0500 Date: Mon, 4 Feb 91 15:48:02 -0500 From: ltw@cayuga.cs.vt.edu (Layne Watson) Message-Id: <9102042048.AA17314@cayuga.cs.vt.edu> To: wmc@research.att.com Subject: Re: netlib's version of hompack Bill, MAIN[FPS] and INNHP.DAT are the main programs and data file for testing HOMPACK after an installation, and should DEFINITELY be "active"! The INDEX file is correct and up to date. We are actively working on a new version of HOMPACK (mostly for sparse problems), but at the moment you have the latest "public" version. Thanks for checking with me. Layne From cayuga.cs.vt.edu!ltw Mon Jun 3 17:53:27 0400 1991 Received: by pyxis; Mon Jun 3 17:54 EDT 1991 Received: by inet.att.com; Mon Jun 3 17:54 EDT 1991 Received: by cayuga.cs.vt.edu (5.57/Ultrix3.0-C) id AA09608; Mon, 3 Jun 91 17:53:27 -0400 Date: Mon, 3 Jun 91 17:53:27 -0400 From: ltw@cayuga.cs.vt.edu (Layne Watson) Message-Id: <9106032153.AA09608@cayuga.cs.vt.edu> To: ehg@research.att.com Subject: HOMPACK code in netlib Eric, Someone at Iowa State tells me that the HOMPACK routines from netlib@ornl contain calls to DTIME, which is undefined. My own version, as well as the version I sent to ACM TOMS, contains no such calls. The code accessed as HOMPACK and ACM TOMS algorithm 652 should be identical (or nearly so), and neither should reference DTIME. Over the years, there have been several versions of HOMPACK in netlib, and I'm wondering if somehow a VERY OLD version of something has crept into the "official" version. Thanks. Layne ?R !/bin/mail cayuga.cs.vt.edu!ltw My copy here has $pyxis g -i dtime *.f mainf.f: REAL DTIME,TIME,TARR(2) mainf.f: TIME=DTIME(TARR) mainf.f: TIME=DTIME(TARR) mains.f: REAL SECNDS,DTIME mains.f: DTIME=TIME/100.0 mains.f: WRITE (6,50) Y(NP1),IFLAG,NFE,ARCLEN,DTIME,(Y(J),J=1,N) Most of the hompack files are dated Sep 1986, with a few fixp* and step* files dated Nov 1987. In February, you mentioned that you were working on a sparse version, but that our copy seemed to be current otherwise. ! cd From cayuga.cs.vt.edu!ltw Mon Jun 3 18:54:29 0400 1991 Received: by pyxis; Mon Jun 3 18:55 EDT 1991 Received: by inet.att.com; Mon Jun 3 18:55 EDT 1991 Received: by cayuga.cs.vt.edu (5.57/Ultrix3.0-C) id AA10016; Mon, 3 Jun 91 18:54:29 -0400 Date: Mon, 3 Jun 91 18:54:29 -0400 From: ltw@cayuga.cs.vt.edu (Layne Watson) Message-Id: <9106032254.AA10016@cayuga.cs.vt.edu> To: ehg@research.att.com Subject: assumptions My mistake ("assuming doesn't feed the bulldog..."). I've never actually retrieved the whole thing from netlib and compared it to the ACM version, I just assumed that when TOMS algorithm 652 came out, it would also be distribued as "HOMPACK". Bad assumption. Apparently those main programs (dated 1986) were very early versions, designed just to put SOMETHING there so people would have an example. The ACM stuff is much more polished. My correspondence in February concerned whether I wanted HOMPACK to continue to be distributed both as hOMPACK and alg 652, to which I replied yes, there isn't any difference. Neither I nor anyone at AT&T bothered to check that statement. So, is it possible to have the ACM code distributed as HOMPACK also? The Iowa state guy also noted that he couldn't get the data file INNHP.DAT needed to test MAINP.f . So be sure that INNHP.DAT is distributed with MAINP.f. I'm sorry--I'm wipin' the egg off my face.... Layne !/bin/mail cayuga.cs.vt.edu!ltw Yes, and I was figuring Bob Melville would point out anything awry, or else Bill Coughran would notice. Just fell between the cracks, I'm afraid. I'll take 652 and split it up into the hompack directory. Netlib stuff is pretty backed up at the moment, but this sounds like an easy one so I'll try to push it high on the list. ! From cayuga.cs.vt.edu!ltw Wed Jun 5 08:21:52 0400 1991 Received: by pyxis; Wed Jun 5 08:23 EDT 1991 Received: by inet.att.com; Wed Jun 5 08:23 EDT 1991 Received: by cayuga.cs.vt.edu (5.57/Ultrix3.0-C) id AA11724; Wed, 5 Jun 91 08:21:52 -0400 Date: Wed, 5 Jun 91 08:21:52 -0400 From: ltw@cayuga.cs.vt.edu (Layne Watson) Message-Id: <9106051221.AA11724@cayuga.cs.vt.edu> To: ehg@research.att.com Subject: HOMPACK in netlib Eric, I did some checking, and this is what happened (I think). In 1986 I sent an early version of HOMPACK to Jack Dongarra, which he installed in netlib. There were no test programs with it, which Jack said would be a good idea to include. So I put something together quickly, not fully documented, just to give a user an idea of what a main program might look like. Later, after the ACM TOMS version was finalized, I sent Jack the new versions of the HOMPACK subroutines, but not the "test programs". A lot of people have gotten HOMPACK from netlib and no one has complained until now, because it is obvious how to "fix up" the main test programs to work with the HOMPACK subroutines. The guy from Iowa State who recently complained didn't recognize DTIME as the UNIX timing routine, and also didn't realize that netlib was returning multiple routines in the same mail message. Bottom line: the situation is not near as bad as I feared. It looks like all you need are the new test programs (3 of them) and one data file. How about if I just send you those 4 files? Also, you need to be sure that the data file is mailed out with the appropriate main program. I noticed that the phone number is wrong in the index file for hompack; can I send you an updated index file also? Thanks. Layne !/bin/mail cayuga.cs.vt.edu!ltw Subject: HOMPACK in netlib Fine, send me just the files you think I need. I'll fsplit the TOMS version and compare anyway. Eric !D .