Thursday, November 22, 2007

Guess what OS Don Knuth runs?

Check out

In the section "Wanted: A Name For High-Tech Grief" Knuth states and I quote

"I myself have often cried out for help to colleagues who have generously made house calls, in order to unwedge my highly customized Linux system"

We've known it for a while, looking at the customized fvwm2 configuration on Knuth's website, but it's good to see it in writing.

Some of you must be thinking, he has problems with Linux? We'll look at the keywords -- highly customized. Can't do that with other operating systems without hitting yourself in the head a few times or hitting your head against something. You don't hit yourself because your stupid, you do, because you brought the product :-)

Hmmm.... Now, if we can get everyone who has directly or in-directly been benefited from Knuth's work to run Linux, that would be a wonderful starting place. Of-course regular users are more than welcome to adopt and customize Linux.

Linux is a Trademark of Linus Torvalds

Tuesday, November 06, 2007

FOSS.IN delegate registration is open

Register at

The fee structure has changed this year. Do read the details before registering. This year we have project days and the main conference

Saturday, November 03, 2007

Talk at VTU

I gave a talk on Introduction to Operating Systems via EDUSAT. The lecture was broadcast to several universities via satellite. I had no live audience, but it was fun to speak into a camera. The video has been stored/archived for students to see later.
Here's a link to the slides

New Code in the Linux Kernel

I got some new code included into the mainline linux 2.6.24-rc1 kernel

cgroupstats - Is an infrastructure to allow sending control group statistics to user space using taskstats/genetlink
cpu_acct - CPU accounting subsystem for control groups

Worked on some other miscellaneous stuff as well. Overall, 2.6.24 was an interesting merge cycle, lots of new stuff went it. Check it out! Please do report regressions

A Quick Estimate

I was trying to estimate how long it would take for a timer, measured in nano-seconds using 64 bits would take to roll over. Here's my quick estimate

It would take 2^64/10^9 seconds = 2^64/2^9*5^9

Approximating on all calculations

= 2^55/5^9

5^3 = 125 ~= 2^7
5^9 = 2^21

= 2^55/2^21

= 2^34 seconds

2^34 seconds = 2^34/86500 days ~= 2^34/80*2^10 = 2^24/2^6 days = 2^18 days

1 year = 365 days, 3 years ~= 2^10 days

That leaves us with 2^8 years * 3 years

= 256*3 years before the timer overflows

With a machine, usually one would expect at-least one reboot in 800 years, if so, we would be fine for a long time to come.

See I need no calculator :-)

Dynamic programming for the binomial coefficient

More fun things, this time with some visualisation of what happens when memoisation is used and what happens when we don't. I don'...