susi22

You can get another 2x speedup if you choose the pivot better. Right now it seems you're doing median of 3.

When you're doing Order Statistics (ie Selection/Quickselect) with 1M elements you should be very very close to the theoretical optimum (1.5N) for complexity. See my comment here:

https://news.ycombinator.com/item?id=6629117

If you're interested in choosing the best strategy.

HTH

This is a great list, but IMO lacks the most powerful (but unfortunately unpopular) one:

https://github.com/dbro/csvquote

Apply it first, then do the normal processing with GNU coreutils and you'll cover most use cases.

"The Secret To 10 Million Concurrent Connections -The Kernel Is The Problem, Not The Solution":

http://highscalability.com/blog/2013/5/13/the-secret-to-10-m...

https://news.ycombinator.com/item?id=5711232

Could you give some examples/sources to "Multi-Tenant cloud services are problematic and insecure"?

Q: Do people have root on the containers?

aatish, since you seem interested, I'll throw another fun-fact at you that I found very interesting even after years of working with Fourier transforms:

You can view the Fourier transform as a fitting problem. Yes, you fit the data to a function. Ie you take the data points and fit it to a sum of exponential functions. There is actually a much more general approach called "Prony method" that extends the concept and adds a dampening factor into the function to fit:

http://www.engr.uconn.edu/~sas03013/docs/PronyAnalysis.pdf

You can take it further and use matrix pencil methods and eventually you'll see connections to ESPRIT algorithm and even least squares algorithm. It's really interesting how they're all actually connected.

Cheers

I work in this field (though I'm none of the authors). If anybody has any questions about the sFFT, let me know.

According to Intel's roadmap they'll arrive in the next 2 years:

http://media.bestofmicro.com/M/F/333879/original/Intel-Roadm...

FYI this algorithm is called the "Water filling algorithm" and is used extensively in Communications to optimize the allocation power for channels.

You can get a solution with simple Lagrangian method (which I believe is the linear solution).

http://www.eecs.berkeley.edu/~dtse/Chapters_PDF/Fundamentals...

(pages 183 - 185)

Note that you shouldn't use the algorithm in practice (unless you need certain guarantees) since the average performance isn't that great. This is just like Mergesort vs Quicksort where Mergesort guarantees O(NlogN) yet Quicksort is usually much faster. The constant for the number of comparisons is quite high. Compare this to the average complexity for Quickselect with these pivot strategies:

- Random pivot (ie Median of 1) has 3.39N comparisons

- Median of 3 has 2.75N

If you want the fastest possible algorithm you should follow

http://dx.doi.org/10.1109/TSP.2012.2197394

which proposes to choose the median of N^{2/3}1/(4pi)^{1/3} elements as the pivot and then also uses a different strategy to choose the second pivot (close to the median yet "safe").

Another fast algorithm is Floyd & Rivest's SELECT (yes the same crypto Rivest) which is also asymptotically optimal and implementations exist. (Paper title: "Expected time bounds for selection")

Google translate + own improvments:

German Chancellor Angela Merkel may have been the target of U.S. intelligence agencies over year. Serious indications have led Merkel to complain directly to U.S. President Barack Obama on Wednesday.

This evidence suggests that U.S. intelligence agencies have declared Merkel's cell phone as a target. In the telephone conversation with U.S. President Barack Obama, Angela Merkel asked for a comprehensive clarification of the allegations.

The Chancellor has made ​​clear "that if the information should materialize such practices clearly disapproved of and sees as completely unacceptable ," said her spokesman Steffen Seibert . "This would be a serious breach of trust. Such practices should be stopped immediately."

The cause of this unusually sharp reaction of the federal government is a recent SPIEGEL investigation. Following a review by the Federal Intelligence Service and the Federal Office for Security in Information Technology, the government found the allegations for plausible enough to confront the U.S. government so that.

German Chancellor Angela Merkel requested in the call with President Obama, the expectation "that the U.S. authorities will clarify about the possible overall scope of such monitoring practices against Germany and thus answer questions, which had been made several months ago," Seibert said.

"As a close ally of the United States of America, the Federal Government expects for the future a clear contractual basis on the activities of the services and their cooperation . "

A spokeswoman for the National Security Council of the United States declared to the reproaches told SPIEGEL : "The president has assured the Chancellor that the United States does not monitor their communications and not also be monitored."

The spokeswoman did not say to SPIEGEL, whether this also applies to the past. This point was noted on Wednesday evening in Berlin government circles.

MS Visual C initializes it to 0xCCCC so it's easy to spot. It helped me find a few errors.

If you need the most packages and the quickest then you should use ARCH linux. Otherwise use your package manager (like others have pointed out)

Seems like a classical low pass filter which can be implemented quite efficiently using an FFT. Or am I missing something?

In a corporate environment I'd expect crucial data to be on the network drive and snapshotted every few hours. We run ZFS on our network and all the secretaries have to do their doc/excel work on the drive. Nowadays that everybody has a Gigabit Ethernet connection read/writes are extremely quick.

Use ZFS and make read only snapshots that are only accessible to the sysadmins. You'll solve many problems that way. We do snapshots at 6am,noon and 6pm and then keep the 6pm one for 7, 14 and 30 days.

Will it be waterproof? ipx7? What about vibrations? The housing for a Di2 battery case are enormous compared to the battery it contains.

All with source code:

http://asymptote.sourceforge.net/gallery/animations/

http://www.piprime.fr/developpeur/asymptote/animation-asy_as...

http://www.marris.org/asymptote/animations/index.html

Ah Mathematica:

    $ curl -s http://4.bp.blogspot.com/-MPv_CwvvwKQ/Ulrw3TfdgyI/AAAAAAAAAEw/YsRPmU6C5xM/s1600/trefoil_rotate_white.gif |strings|grep -i created
    UCreated by Wolfram Mathematica 9.0 for Students - Personal Use Only : www.wolfram.com

Would love to see the source for them.

There is a HUGE controversy about Adblock plus. One blogger uncovered some of their business methods. They have Mafia like structures and extort businesses. Unfortunately it's only in German and I have yet to find an English site report about it.

Main article (very detailed):

http://www.mobilegeeks.de/adblock-plus-undercover-einblicke-...

They've been trying to cover up their methods and simply lied about it. Stay away from Adblock Plus.

Some shorter article:

http://www.heise.de/newsticker/meldung/Adblock-Plus-Weitere-...

Yeah but woulnd't you agree this is more fun:

http://asymptote.sourceforge.net/doc/Tutorial.html#Tutorial

One of the members of waybury.com is this dude with this repo:

https://github.com/somerandomdude/Iconic

I'd use an existing programming language (asymptote or scala) and build a framework to create a bunch of EPS/PNG/SVG icons.

No new file format....

This is what I mean with DSL:

http://asymptote.sourceforge.net/doc/Tutorial.html#Tutorial

I don't say that designers should program instead of using their tools. What I say is: Stop the craziness of 50 different projects which give you some set of icons. Each a little different but each repeating things over and over again. Let's define one framework and go from their and satisfy 95% of all projects.

Postscript is so unpopular to use as a programming language that github doesn't even have it in their languages. I'm not even sure if you're just trolling or serious?

Here:

https://github.com/FortAwesome/Font-Awesome/issues?page=1&st...

https://github.com/FortAwesome/Font-Awesome/pulls

Result:

https://github.com/FortAwesome/Font-Awesome/graphs/commit-ac...

You can do cool things with immediate updates:

http://vimeo.com/36579366

The great thing about hosting on github is that people can send pull request and you have real collaboration. That's alsmost impossible with Illustrator files. In reality many OSS is abondoned and change leaders or simply die. If you have a source that is accessible by 100% of the programmers (and with icons there is a huge crows who's interested) then your project probably won't die.

Want will the kickstarter accomplish? Sure it'll give you great quality icons, but then what? What if the project is delivered and put open sourced? Is it going to be maintained indefinitely? For 2 years? 5 years? What about in 10 years? Can we still use it?

But now we're really off topic, I'm just re-elaboration why open source (and line based source code) is better. I could also argue why documentation in Markdown is better than word for highly collaborative projects....

I meant a DSL (or at least a framework) to _create_ graphics. I don't propose a new file format. In my approach there is no such thing as a GUI tool since that usually means that you'll be editing a complex (non gittable) file format. I want to compile the source code to SVG/EPS, then use open source tools (such as fontforge) to create the font. But you don't have to, you can of course also create PNGs. I want a DSL that is optimized to describe graphics for humans (so no SVG).

And with a framwork you can easily adjust the fonts to your liking. Rounded edges in general? no problem. Rounded arrows? No problem. All icons with background of gray? No problem. Compile to 173x294 pixels? No problem...

It's much much more flexible than any Illustrator format will ever be and it'll allow real collaboration and source code versioning. I think asymptote would be great candidate and it's very close to the DSL I'd like. The good thing is that most of it is also written in asymptote itself so it's very extensible (unlike some other graphics languages)

Yes that's exactly what I meant. Though, I think I wasn't clear and many people misunderstood me. SVG is to describe graphics and not to create graphics. We need a DSL that has loops, objects, methods, overlaoding etc so we can actually create a framework that allows creating EPS/PDF/SVG etc..

Some exist, there is very simple gnuplot. There is python matplotlib but it's a general purpose programming language and no DSL specific one. I think the closest is asymptote (which is very little known). Then is is processing but it's geared toward animations and also does mostly raster graphics. There is R which is mostly for statistics. Then there is TikZ which is awful since TeX was never meant to be a programming language so it'f far from a DSL.

I want to be able to describe shapes and extend them easily. I think Scala would be wonderful and somebody fluent in it could create a DSL and still allow extreme extensibility. We need a language which has multiple layers of API. One very low which can easily do lines, bezier curves and the usual. And then some higher APIs that allow _easy_ composition and intuitive transformations.

SVG is like PDF, a format to display but nobody wants to write PDF or SVG. It's no fun.

Yeah but it's more fun to click "edit" on a file in github. Propose the changes and click "submit pull request". Instead of:

1. Download current design file

2. Edit

3. Submit pull request

4. Repo owner downloads file

5. Repo owner figures out what has changed (somehow hopefully) and hopes nothing else was broken with the edit due to different versions, setups etc.

6. Repo owners accepts pull request (while the change cannot be seen in github since it can't diff .ai files).

I can already edit EPS pretty well in inkscape and scribus but many developer still won't do it. To collaborate in graphics isn't easy.

No, even if Illustrator was open sources I would prefer to see real source code for icons. IMO Illustrator is over kill for something so (relatively) simple. Image a project on github that contains a very descriptive syntax, a hierarchy of re-usable shapes and which can be plugged together to generate custom icons. Then everybody can contribute to it with a pull request and we have true collaborative open sourced icon fonts.