Powered by Blogger

Ethel the Blog
Observations (and occasional brash opining) on science, computers, books, music and other shiny things that catch my mind's eye. There's a home page with ostensibly more permanent stuff. This is intended to be more functional than decorative. I neither intend nor want to surf on the bleeding edge, keep it real, redefine journalism or attract nyphomaniacal groupies (well, maybe a wee bit of the latter). The occasional cheap laugh, raised eyebrow or provocation of interest are all I'll plead guilty to in the matter of intent. Bene qui latuit bene vixit.

The usual copyright stuff applies, but I probably won't get enraged until I find a clone site with absolutely no attribution (which, by the way, has happened twice with some of my other stuff). Finally, if anyone's offended by anything on this site then please do notify me immediately. I like to keep track of those times when I get something right.

Google!


How to blog?

METABLOGGING

Blog Madness
Blog Portal
linkwatcher
Monitor

BLOGS (YMMV)

abuddhas memes
alamut
apathy
arms and the man
baghdad burning
bifurcated rivets
big left outside
boing boing
booknotes
bovine inversus
bradlands
bushwacker
camworld
cheek
chess log
cogent provocateur
cool tools
counterspin
crooked timber
delong
digby
drat fink
drmike
d-squared
dumbmonkey
electrolite
eschaton
estimated prophet
ezrael
fat planet
flutterby!
follow me here
geegaw
genehack
ghost
glare
gmtplus9
hack the planet
harmful
hauser report
hell for halliburton
honeyguide
hotsy totsy club
juan cole
kestrel's nest
k marx the spot
kuro5hin
lake effect
lambda
large hearted boy
leftbanker
looka
looking glass
macleod
maxspeak
medley
memepool
metagrrrl
mike's
monkeyfist
more like this
mouse farts
my dog
norbizness
off the kuff
orcinus
pandagon
pedantry
peterme
philosoraptor
pith and vinegar
plastic
portage
q
quark soup
quiggin
randomwalks
rip post
rittenhouse
see the forest
shadow o' hegemon
sideshow
simcoe
south knox bubba
slacktivist
smudge
submerging markets
sylloge
synthetic zero
talking points
tbogg
twernt
unknownnews
vacuum
vanitysite
virulent memes
whiskey bar
windowseat tv
wood s lot

TECH

Librenix
use perl
rootprompt
slashdot
freshmeat
Ars Technica
32BitsOnline
UGeek
AnandTech
Linux Today
Tom's Hardware
DevShed


"When they say, 'Gee it's an information explosion!', no, it's not an explosion, it's a disgorgement of the bowels is what it is. Every idiotic thing that anybody could possibly write or say or think can get into the body politic now, where before things would have to have some merit to go through the publishing routine, now, ANYTHING." - Harlan Ellison



JOLLY OLD PALS
Old pals Rumsy and Saddam


Other stuff of mild interest to some:
unusual literature
scientific software blog
physical oceanography glossary
computer-related tutorials and texts

Friday, June 14, 2002

TRADE LIBERALIZATION MYTHOLOGY
In
The relative impact of trade liberalization on developing countries, Mark Weisbrot and Dean Baker take issue with the conventional "wisdom" that NAFTA, TRIPS, etc. are going to do wonderful things for the economies of developing countries. Their points:
  • The removal of all of the rich countries' barriers to the merchandise exports of developing countries-including agriculture, textiles, and other manufactured goods-would result in very little additional income for the exporting countries. According to the World Bank's estimates, when such changes were fully implemented by 2015, they would add 0.6 percent to the GDP of low and middle-income countries. This means that a country in Sub-Saharan Africa that would, under present trade arrangements have a per capita income of $500 per year in 2015, would instead have a per capita income of $503.
  • Some of the most widely used economic models show that many developing countries will actually lose from trade liberalization in important sectors, such as agriculture and textiles. There are three reasons for this outcome. First, some countries will be hurt by the elimination of quotas that now allow them to sell a fixed amount of exports at a price that exceeds the competitive market price. Second, trade liberalization changes the relative prices of various goods, and some countries will find that their export prices fall relative to the price of imports (the "terms-of-trade" effect). Third, some developing countries currently benefit from access to cheap, subsidized agricultural exports from the rich countries.
  • Developing countries incur substantial problems from reducing their trade barriers. In many developing countries, tariff revenue accounts for 10-20 percent of government revenue, and in some cases considerably more. If tariffs are reduced or eliminated, these countries will have to impose large increases in other taxes in order to keep their budgets in line. The distortionary effect of these tax increases, as well as the costs and problems associated with collecting taxes from other sources, are generally ignored in economic models that project gains from eliminating trade barriers.
  • The removal of trade barriers is also likely to lead to large disruptions in agriculture. In most developing countries, a large portion of the population is still tied to the agricultural sector. If barriers to agricultural imports are removed too quickly, it can lead to large-scale displacement of the rural population. Standard economic models implicitly assume that these people are re-employed in other sectors of the economy, but rapid import liberalization can lead to substantial unemployment and underemployment, as well as dangerous levels of social and economic instability.
  • Recent trade agreements, such as the TRIPS provisions in the WTO, have sought to impose U.S.-style patent and copyright protections in developing countries. This will lead to the transfer of billions of dollars from developing to high-income countries in the form of royalties and licensing fees. In addition, the efficiency loss resulting from higher prices of patented and copyrighted items is likely to be even larger. The World Bank's estimates indicate that the cost of TRIPS to developing countries is likely to be comparable to any gains they might receive from trade liberalization.
  • As a result of increasing instability in world financial markets, developing countries have felt the need to vastly increase their holdings of foreign exchange reserves. These reserves are held in the form of short-term deposits that pay little or no real interest. By contrast, this money could otherwise be invested in building up the infrastructure or the physical and human capital in a developing country. The opportunity costs of these increased reserve holdings are also of the same magnitude as the World Bank's projections for the benefits of trade liberalization.
On the other hand, doctrine says it has to work and it will be most beneficial to the obvious suspects, whose booty will thereafter be gently trickled down upon the developing countries as is indicated by the holy scriptures.
posted by Steven Baum 6/14/2002 10:32:08 AM | link

Thursday, June 13, 2002

AN ECONOMIST ENCOUNTERS REALITY
The strongest argument for the "dismal science" not being a science is the proclivity of the practitioners tbereof (especially those of a right-wing bent, although that seems to be most of them these days) to choose - when data contradicts their favorite theories - to believe the latter over the former. Perhaps the most notorious recent example of this involved the predictions in the early 1990s of impending economic disaster should the proles be granted an increase in the minimum wage. Even well after the economy continued to grow after the minimum wage increase, the usual boneheaded suspects continued to ignore reality and predict doom. To be blunt, such behavior is more typical of religion than science, and should be treated as such. That such nonsense is the rule rather than the exception makes it most refreshing when one of the shamans encounters data belying one of his precious doctrines and *GASP* accepts the reality of the data.

Such a refreshing example has been provided by Stan Liebowitz (via Slashdot), usually one of the house hacks at the Cato Institue who can be counted on to churn out report after report affirming the house ideology. It seems that Stan, who wrote a report back in May assuring his paymasters in the recording industry that peer-to-peer file sharing would eventually leave the industry in ruins (and the poor executives unable to afford a coca leaf much less a mountain of snow), is losing his religion. Here's Stan on why he's no longer certain that what the record industry paid him to say will happen will indeed happen.

Actually, I'm not sure. It took six months to get [the Cato piece] out. Now I'm stepping back a bit. In the Cato piece, what I said was that [file sharing] seems like it should be causing a lot of harm. But we're not seeing it. The explanation I gave was that maybe there weren't enough people who owned CD writers during that period. In order for downloading to really have an impact on CD sales, it needs to be a substitute for CDs. If file sharing is not a good substitute, then you can download all you want and it may be a new form of listening but it may not hurt CD sales.

The problem is that the number of downloads appears to be larger than the total number of CDs purchased. Worldwide annual downloads, according to estimates from places like Webnoize, would indicate that the number of downloads -- if you assume there are 10 songs on a CD -- is something like five times the total number of CDs sold in the U.S. in a year, and one-and-a-half times the worldwide sales. That's so large that you have to say: Look, if downloads are substitutes [for CDs] in any significant way, we should see really big declines -- unless there's something else going on.

The reason I gave in the paper is that maybe people aren't shifting their music [from MP3s to CDs]. But I've also seen some recent numbers on households that have CD writers, and it's something approaching 30 percent. We should see an impact. There's a 5 percent decline in CD sales this year, but that's what you might expect in a recession. So we're still not seeing much. And what I'm beginning to suggest now is that perhaps people aren't going to replace the purchase of CDs with these MP3s.

The rest of the article is an equally fun read, although I don't think Stan the Apostate's going to be getting a new Mercedes real soon.
posted by Steven Baum 6/13/2002 03:58:50 PM | link

DIRTY BOMBS
"Dirty bombs" must be to John Ashcroft what "dirty girls" are to others, although one suspects that Ashcroft's loud trumpeting of the supposed dire and imminent threat posed by someone arrested over a month ago is more a matter of the Cabal attempting to spin the news away from investigations of their pre-9/11 screw-ups than it is of reality. The only real news related to this non-event is a quote from Paul Wolfowitz, usually among the loudest of saber-rattlers in the Cabal.
"I don't think there was actually a plot beyond some fairly loose talk and (Al Muhajir's) coming in here obviously to plan further deeds."
It's just not like Paul to let reality intrude on his public statements.
posted by Steven Baum 6/13/2002 03:51:27 PM |
link

Wednesday, June 12, 2002

HYPERTHREADING
Looking for the best floating point performance per dollar, we decided to try out a dual processor Intel set-up to see what we could get. The details:
After getting the whole thing for just under $2800 by buying the individual parts (i.e. those listed plus a case, a SCSI hard drive, a floppy drive, a CD drive, a mouse and a keyboard), I assembled everything and attempted to install Red Hat 7.2. This didn't work, and after a bit of a search I attempted the fix recommended by Tyan. After farting around with that for a while, I decided to download and burn the three ISO images comprising the recently released Red Hat 7.3, which contained the newer kernel recommended by Tyan. This worked like a charm.

After booting up this screamer for the first time and running top to see if both processors were running, I was quite surprised to see four processors up and running. Just to be sure, I opened up the case to see if any extra processors had been hidden somewhere. Nope. After another search on the web, I found an answer: hyper-threading, an Intel technology that enables individual processors to have more than one hyperthreaded "core". The current Linux kernel supports two "siblings" per CPU, although apparently more are allowed by the specification. More can be gleaned from a hyperthreading thread at the Linux Kernel Mailinglist archive, including a statement by Alan Cox that hyperthreading performance improvement is typically 10-30%. This has thus far proved correct in my benchmarking tests.

Speaking of which, I'm using the ROMS (Regional Ocean Modeling System) code, written in Fortran, for benchmarking. Why? Because it's the code I'm presently using for research. It compiled and ran using the Intel Fortran Compiler with just a minor bit of tweaking, and the performance I've seen so far is most impressive. A single P4 2.2 GHz processor is almost exactly four times faster than a single 800 MHz P3 processor (with the same compilation options), indicating that the Xeon is doing quite a bit better than just a linear performance increase due to increased processor speed. Then is started tweaking the compiler options, with the following results:

  • the vectorization switch (-xW) increased performance typically by around 25%
  • the -O2 optimization switch typically produces better results than the -O3 switch, probably due to the complex additional -O3 optimization attempts conflicting with other optimization switches
  • the -openmp SMP parallelization switch (which enables the OpenMP statements buried in the code to be compiled) showed little effect with 2 processors specified, but showed about a 10-25% increase with 4 specified, so apparently the hyperthreading is working better than the OpenMP
Another interesting tidbit about hyperthreading is that Linux was the first OS implementation thereof, which apparently came about when Intel submitted a kernel patch. Microsoft and Intel are having problems getting together on this one, purportedly because of a battle over license fees, i.e. whether each hyperthreaded processor "sibling" is considered a separate processor when it comes to charging fees.

Maybe I'll run some of the standard benchmarks to give a better feel as to this box's relative performance, although it's already most impressive running the ROMS code as compared to other machines on which we've been running it.
posted by Steven Baum 6/12/2002 10:22:31 AM | link


Comments?
Archive

LISTS

Books
Software

uPORTALS

cider
crime lit
drive-in
fake lit
hurricanes
os
scripting
sherlock
texas music
top 100
weirdsounds
wodehouse

LEISURE

abebooks
alibris
amazon
bibliofind
bookfinder
hamilton
powells

all music guide
best used cds
cd bargains
second spin
raven's links

ampol
arts & letters
atlantic
art history
attrition
bibliomania
bitch
bizarre
bizarro
bloom country
bob 'n' ed
bob the angry flower
callahan
chile pepper
classical music
cnnsi
crackbaby
cult films
culture jamming
discover
disinformation
dismal scientist
electric sheep
espn
exquisite corpse
feed
fine cooking
fishbowl
fluble
fried society
fry and laurie
hotel fred
hotendotey
hypocrisy network
jerkcity
last cereal
leisure town
logos
london times
mappa mundi
miscmedia
mp3lit
mr. chuck show
mr. serpent
national geographic
new scientist
no depression
not bored
obscure store
onion
on-line books
parking lot is full
pearly gates
phrase and fable
probe
red meat
rough guides
salon
Simpleton
sluggy freelance
spacemoose
spike
straight dope
strenua inertia
suck
superosity
tawdry town
too much coffee man
toon inn
verbivore
vidal index
yes minister
you damn kid





Powered by Blogger