I now work with lisp most of the time
30 Mar 2006
More silliness. I'm running Debian linux on my one SparStation 20 which is turned on. I was moving things around yesterday and decided to, once again, try a Hypersparc with Debian. Before I had decided that it didn't work but I thought that one more try might be worth it. Well, a single 180mhz HyperSparc does work. It booted and ran, though, even though it benchmarks faster, it seems slower. Maybe it's memory access? It wasn't perfect though, top, for example, didnt' run. So, since that worked I downloaded and installed the smp kernel and tried that with 2 Hypersparc 180s. No go. Then, since the worst I could do is void my guarantee I decided to stick both the new 85mhz SuperSparcII and the 60mhz SuperSparc in and see what happens. I expected it not to boot. Happily it booted and seems to run well. Two cpus is a nice speedup for the interactive use of this system since X basically wants one and then the other is free to run your program.
29 Mar 2006
Related to the updates below on 21 Mar, there seems to be one final change which is necessary. Link build/gmp/mpn/udiv_fp.asm to src/gmp/mpn/sparc32/udiv_nfp.asm rather than udiv_fp.asm. I'm not 100% sure why this is necessary at the moment.
On my SparcStation 20 I recently installed the VSIMM so that it has a cg14. I couldn't get this to work with Debian Sarge's XFree86 until I told it to be in 32 bit mode in the XFree86 config file. Much head scratching occured trying to figure out why 24 bit mode didn't work, since it did work on NetBSD. Still, 32 bits does and it looks nice. Not fast, but nice. Note that 24 bits is the logical choice since the cg14 is sold as a 24 bit frame buffer.
21 Mar 2006
I was trying to get one of my SparcStations to act as a serial terminal for another one using the cu command. The magic serial line is /dev/ttyS0. I kept getting Permission denied as root. It turns out that this is bug 264626. A chown to uucp fixed it for now
Finally got ecl to build easily on my SuperSparc. You need to comment out the following bit of code before you run configure in src/gmp/configure.
#sparcv8*-*-* | supersparc*-*-* | microsparc*-*-*) # gcc_cflags_cpu="-mcpu=v8 -mv8" # acc_cflags="-O2 -cg92" # path="sparc32/v8 sparc32" # extra_functions="udiv" #case $host in # *-*-solaris2.*) cc_cflags="-xtarget=native -xarch=v8 -xO4" ;; #esac #case $host in # supersparc*-*-*) path="sparc32/v8/supersparc sparc32/v8 sparc32" ;; #esac #;;
and, after you configure, you need to edit build/gmp/config.h and change it:
#define HAVE_HOST_CPU_sparcv8 1 /* #define HAVE_HOST_CPU_supersparc 1 */
Once you do that everything builds fine. I'm still testing the running
16 Mar 2006
Debian/Sparc. No, I tried, but sarge does not work on a hypersparc
Well, it wasn't too hard to port things to MySQL 5.1.7 Beta. First, things got simplier. We only have to modify the configure.in file in the top level directory. We had to change ha_eclstorage.[h|cc] to match some changes made internally in MySQL. The rest of the changes were just so that MySQL would recognize a new storage engine.
The one catch is that after you untar things you MUST
- touch -t 200602272225 configure.in
- touch -t 200602272227 configure
otherwise automake tries to run. I need to read more on automake but that can wait.
The configure line becomes ./configure --prefix=/usr/local/mysql --enable-assembler --with-mysqld-ldflags=-all-static --prefix=/usr/local/mysql --with-named-curses-libs=/usr/lib/libncursesw.so.5 --with-blackhole-storage-engine --with-eclstorage-storage-engine
These changes live in eclstorage-20060315.tar.gz
13 Mar 2006
Long delays with eclstorage because we got the new carry around system, a Mac Powerbook G3. This is an old world system, with, sadly, a defunct PRAM battery. Ergo no luck installing NetBSD. Since the Linux world boots from MacOS 9 in an old world system, we did a small install of MacOS 9 and installed Debian. Seems to work well. In fact, happily or sadly, depending on your POV, it seems faster on this 266mhz G3 than NetBSD current on my 300mhz G3 new world system.
The one key bit of info to make the display on the Powerbook G3 work is the line
A mclk of 63 also worked, but on my display it was ugly
The second half of the delay is that FinSwarms doesn't look like a viable project anymore. Because of that, the cluster of Sparc 20s and Ultras I've built up at home doesn't need to be kept turned on. Electricty is expensive. That means that I installed Debian as well on on of the Sparc 20s that I subbed a 60mhz SuperSparc in place of the dual 135mhz Hypersparc processors that were there before. It did slow down, but not as much as one might think. Apparently, unlike NetBSD, Linux 2.4 and Hypersparcs are a bad combo. Anyway that pretty much works so far, and, it's a lot quieter. It also runs Firefox as well. Now we have to get ECL and MySQL to both work and then we'll be good to go. ECL didn't work this morning but this looks to be something in the GMP library.
07 Mar 2006
Real-Time remote Backups
Perspectives from someone who worked at Amazon
For those times when someone insists on sending me a Word document
Last modified: Tue Mar 7 14:45:53 MET 2006