Just a short note to congratulate you and the staff on Linux Journal. I've subscribed from issue 1 and have enjoyed watching the magazine grow and mature much like Linux itself. Keep up the good work.
On to the suggestion—I was wondering if you plan to run an article on the Linux SPARC port with news and views from the developers. With a number of elderly SPARCs hanging around I'm looking forward to the possibility of running Linux on them.
Progress on that front has been amazingly rapid lately, and we expect to run an article on the topic some time in the near future. While development continues, Linux/SPARC currently runs both native binaries and SunOS 4 binaries, including X-based ones. For example, the SunOS Netscape runs under Linux/SPARC. It can netboot with the recent NFS root support in Linux. Debian/SPARC and Red Hat/SPARC are both underway.
I recently purchased the book Linux Unleashed and transferred the files on the CD-ROM to a SyQuest 270 MB cartridge. Linux (Slackware version 1.2.1) is now booting and running fine on my computer.
Since 1982, I've done hundreds of Xenix and Unix installations. The only drawbacks I see to Linux are as follows:
1. The termcap file is extremely sparse, not even including the setup for a Wyse-60 terminal!
2. The printcap file is also extremely sparse, not supporting either HPGL or PostScript or Epson dot matrix printers!
3. SCO binaries cannot be run!
My question to you is: Is there a newer version of Linux I could download or purchase that rectifies each of these problems? Unfortunately, I don't have the time to experiment or develop my own version; I need one already prepared that can run SCO binaries and has a comprehensive termcap and printcap.
Thanks for your help. —Ronald W SatzSystems Engineer
The version you have purchased is very old; Slackware has gone through two major revisions since. Most Linux distributions now include the standard BSD termcap and terminfo files, which do include support for the Wyse-60. Printing support under Linux is much richer than your question suggests. There is a document called the Printing HOWTO available as part of the Linux Documentation Project. Newer distributions include the LDP documents in electronic form, and several vendors sell the documents in printed form, as well. Finally, support for SCO binaries (through the “iBCS2” package) is now a part of every major Linux distribution. While it doesn't run every SCO binary, it does run the great majority. One caution: support for SCO OpenServer 5 has been added in only development versions of the iBCS2 package, and that support won't be added to distributions for a few months.
In his “The Quintessential Linux Benchmark” Wim van Dorst writes: “The table also shows that the Pentium processor doesn't have the expected extrapolated multiplication factor. This is due to the fact that the specific busy-loop algorithm is not optimized for the parallelism of the Pentium processor.”
But won't every everybody ask, “What if it were optimized?” and “How would it be optimized?”
I have the answer: the optimization for the Pentium is to put the decrement and branch instructions apart by inserting a nop instruction between them. This allows the branch to be predicted correctly, and the BogoMips number becomes clock * 1.00 +/- 0.003 (based on two measurements on Pentium 75 and 90).
One of the reasons it is not done in the distribution code is the idea to have one code per architecture (instruction set), and another reason is quite obvious as well—the loop code is only used for timing, and the slower each loop iteration, the less the energy consumption by the processor. You don't want to fry eggs on your chip while measuring BogoMips, do you?
--Leonid A. Broukhis firstname.lastname@example.org
I love LJ and finally got a subscription a few months ago because I missed out on a month at the news stand. But, there is just one little thing...
LJ is too short. I get it and two days later I have read it cover to cover twice. Do you think you could make LJ a little bit longer so I don't read it up so quickly? *smile*
--Joel M. Lindell email@example.com