Yeah, new skin. Same layout as the old skin but not red, and not fixed-width font. I'm not sure I like this better, but it does have some nice improvements.

So, a new semester is upon me. With work and all, I'm taking a lighter courseload -- just 14 hours of classes:

• CS 440: Software Engineering I
• CS 473: Compiler Design
• ECON 121: Macroeconomics
• EAES 101: Intro to Earth Science

CS 440 is more about diagrams and documentation than programming, which might suck. Other than that, it's like CS 340, but in groups. Same prof even (John Bell for those who know). One big project that matures over the semester. When I took 340, I ended up with a 6000 line Java and C++ adventure game program, which subsequently went on my resume and helped me get my current …

read the whole post

System: Dual Pentium III (Coppermine) @ 1000 MHz / 133 MHz FSB, 2 GiB ECC SDRAM. Linux 2.6.32-ARCH.
Disks: four 10,000 RPM UltraSCSI 320 disks in RAID5 (hardware accelerated). Max read speed: ~50 MB/s. Max write speed: ~35 MB/s
Setup: filesystems created on a 10 GiB (10737418240 bytes) zero-filled file, mounted via loop device.

Test 1: unpack an uncompressed tar archive of the Chromium browser source code at revision 35890 (2.6 GiB, 222964 files) to the test filesystem.
Test 2: delete all files created in test 1.

(Note: CPU usage is out of 200%, and the loop device and tar command make up a large part of it)

JFS Test 1: 3:25.85, 32% CPU
JFS Test 2: 0:22.38, 64% CPU

XFS Test 1: 4:22.26, 43% CPU
XFS Test 2: 1:16.31, 89% CPU

Ext4 Test 1: 3:28.61, 28% CPU
Ext4 Test 2: 0:24.12, 72% CPU

ReiserFS Test 1: 4:28.20, 39% CPU
ReiserFS Test 2: 0:33.93, 91% CPU

Test procedure: % dd if=/dev/zero of=test_disk bs=1G count=10
% sudo losetup /dev/loop0 test_disk
% …

read the whole post

Aha. Figured out why the AthlonXP beats the pants off the Core Duo.

Too many processes.

The Core Duo seems to be much slower at creating new processes, which my test method did a lot of.

Also, with the fact that so many short-lived processes are flying around, and therefore being shuttled back and forth between the two cores, these cores' caches probably aren't being used very well, causing even more slowdown.

In the end, I rewrote my processing shell scripts as one program (a PHP script :P), and the processing time on the Core Duo went from 20 minutes down to a few seconds. Clearly, the new-process overhead is significant. The other machines saw similar improvements, to the point where I can't really tell which machine is fastest, because the script runs so fast.

read the whole post

My latest Computer Science assignment involved a large amount of number crunching and text processing, and seeing as how I have a large assortment of computers around my apartment, I figured I'd try them all out and see how they stacked up. The results proved quite surprising.

The data I needed to process was a large text file (10680 lines), where each line had a source IP address, a destination IP address, and various network statistics. I wrote two scripts to process this: the first took each IP address, looked it up in a file that mapped IPs to latitude-longitude pairs (816 lines long), and printed lines with the address substituted for the location. The second script took these two locations and calculated the distance on the globe between them.

Script #1, for each input line, grep'd the ip-location map twice (once for source, once for dest), and printed the new line with the IPs substituted for locations.

Script #2, for each line, ran a perl script that used the …

read the whole post

We're currently experiencing some problems because our DNS provider is having sporadic outages due to a DDoS targeted at them. If the problems continue I may be switching codewise.org to a different DNS provider. Please be patient with the problems.

read the whole post