Category Archives: etc

Liveblogging and the iPad

While I wait for a few loose ends to tie before I post my first entry in the series of VisWeek 2010 papers, let me ask you all this question: have you tried liveblogging with your iPad? I’ve done that in the past with my laptop, and I am confident I can type fast enough to keep track of most of what is going on. However, I know my typing on the iPad is terrible. I was really not planning on taking my laptop with me this year, since I can do pretty much everything else on my iPad, but I think this might be a real problem.

Have you tried something like this with your iPad in another conference? Did you train yourself with some typing app before going? I know at least a handful of you have iPads, so I’d love to hear any ideas. 🙂

Your graphics puzzle for the day

Just the other day, I wrote a trivial geometry shader to tesselate cubic 2D Bezier curves on the fly, so we get nice curved edges for free. Remembering that the geometry shader sits between the vertex shader and the fragment shader, and its purpose is to create new geometry on the fly, the code looked like this (my apologies to the direct threedeese readers – I only speak OpenGL):

vec4 pp[4] = ... // control points
for (int i=0; i<=10; ++i) {
  float u = i / 10.0f // interpolation control
  float omu = 1.0f-u;
  vec4 spline_p = pp[0] * omu * omu * omu +
                  pp[1] * omu * omu   * u +
                  pp[2] * omu   * u   * u +
                  pp[3] * u   * u   * u;

This worked nicely, generated really good looking splines, and was fast. Then I wanted to backport this to the CPU in case the user doesn’t have geometry shaders. We could be using OpenGL evaluators, but it seemed simple enough, right? Like this:

float pp[4][2] = // control points
for (int i=0; i<=10; ++i) {
  float u = i / 10.0f // interpolation control
  float omu = 1.0f-u;
  float spline_p[2] = {
    pp[0][0] * omu * omu * omu +
    pp[1][0] * omu * omu * u +
    pp[2][0] * omu * u   * u +
    pp[3][0] * u   * u   * u,
    pp[0][1] * omu * omu * omu +
    pp[1][1] * omu * omu * u +
    pp[2][1] * omu * u   * u +
    pp[3][1] * u   * u   * u}

No matter what I tried, this wouldn’t work – the lines come out garbled. The endpoints are in the right places, but the interior vertices tend to bend towards a common point. The bug is fairly obvious: the second and third cubic Bernstein polynomials are wrong (I was missing a factor of 3 there).

Now for the puzzle: why did the lines look right on the geometry shader?

2-adic integers are awesome

Over at A Neighborhood of Infinity, Dan Piponi has a post about a little bit of magic that happens when you ask gcc to do a division by 7.

This would have nothing to do with visualization aside from the realization I had in the shower, just the other day, that I believe the right way to pick good numbers for an axis scale is to think about the problem in the p-adic sense. “Small”, in this case, means pretty much exactly “close to a round number”. The usual descriptions of the algorithms (at least as I remember them!) for coming up with a good set of axis markings are pretty nasty; maybe p-adic numbers can help.

Fun with my iPad

I’ve had it for a week now. There are many iPad reviews, but this one is mine.

Screen, web browsing, reading

The contrast on the screen is much better than I expected it to be, and if you’re indoors, pictures look better on the iPad than they do on decent 8″x10″ prints. Reading text is also fine, and I have mostly been using it for that. Browsing the web is perfectly fine, and there’s something strangely pleasant about scrolling down a webpage by “grabbing” the page itself. It’s hard to describe it, because the same type of interaction was there with the iPhone, but the iPad makes it feel qualitatively different. I suspect it’s a combination of responsiveness and physical display size. Apple marketing catchphrases are at times fairly idiotic, but interacting with the iPad is really a different experience compared to the iPhone. The one letdown about reading on the iPad is that the PDF rendering does not use LCD-aware hinting a la ClearType (and whatever OS X has). This means your fonts will be slightly blurrier than you might be used to, although the increased pixel density (~150DPI vs ~100DPI for large LCD screens nowadays) means you might not notice it.

Watching movies on iPad works fine (I watched Fargo for the first time on it), and the Netflix app is incredibly cool (all the streaming videos are available). However, if you’re annoyed by letterboxing or cropping, remember that the screen is 4:3. The NPR, BBC and New York Times apps have become my main news source, which is nice, because they actually employ people interested in journalism. We now only tune in for Jeopardy (that’s right, I’m a nerd): together with Netflix, it means I’ll be canceling my cable TV subscription.


The on-screen keyboards for the two orientations feel completely differently from each other, much in the same way that the iPhone keyboards do. The vertical keyboard on the iPad is good enough for thumbing, but the horizontal keyboard is simply too large. However, after having a couple of weeks of practice, I can now mostly touch-type on the horizontal keyboard. I suspect that in a couple of months I will be typing at a serviceable rate. However, typing on a real keyboard is much nicer, and I have bought the keyboard dock. There’s no way I could use the onscreen keyboard to write this blog post, for example.

PDF Annotation: the killer app for nerds

Even before the iPad was announced, I had been looking for a way of keeping track of all the PDFs I print and annotate. A tablet of some sort seemed like a great fit: I could carry many PDFs with me, annotate them directly on the screen, and then save the resulting files in a safe, searchable place, so (say) I never lose a paper I reviewed again (making the inevitable second round of reviews much more convenient).

Originally, I had planned to write my own iPad PDF annotation software, and eventually I might come around to doing that. However, I need to read papers right now, and iAnnotate exists right now (and it only costs $9.99 or so). iAnnotate has a quirky user interface, but when you get past that, it is a serviceable, fast PDF reader in which you can add your own notes, highlight pieces of text (works well on the LaTeX PDF outputs I tried, but some people have had some issues), scribble comments, etc. Right now, the only way to get files out of iAnnotate is via their (free as in beer) Aji Reader Service, available for Macs and Windows. My current setup for automatically saving the annotated files in the right place is a combination of Aji Reader Service, Dropbox, and some homebrew hacks with inotify. Not pretty, but functional (email me if you’re interested)


My only use case for multitask has always been Pandora, and this continues to be the case. However, most people will want to use instant messaging while they browse the web and listen to Pandora, so I can see it annoying them. Does not bug me much.


I am a big fan of the Desktop version of Keynote, Apple’s presentation software. I have not yet played around with the iPad version too much, but I have started to create a talk on it to see how far I can get. It is fairly painful to not have easy access to arbitrary pieces of LaTeX via LaTeXIt, so I’m not sure how much use this will end up getting. Getting files in and out of Keynote for iPad is also ridiculously dumb and complicated – it needs serious fixing. All iPad apps in general have this insular feel (which is part of the reason for the above dropbox/inotify hacks). If this annoys you, stay away from iPad 1.0.


So, is it ultimately worth $630? Probably not for everyone. I would be willing to pay $1000 for a tablet with integrated decent PDF annotation and management, but not everyone would. If you read a couple of papers every day, though, I cannot recommend it strongly enough. I have tried reading research papers on the Kindle DX – it’s just not as good. And have I mentioned plants vs zombies?

Ubuntu 10.04 beta 2, NVIDIA, x86_64

Are you testing Ubuntu’s latest beta on a 64-bit machine with an NVIDIA video card? Chances are you’ve hit the bug where the proprietary drivers won’t install correctly. It turns out that the nouveau kernel module steals the nvidia device, which in turns prevents from loading successfully, which makes the installer think the whole process failed (even though it didn’t, aside from actually loading the module). So you need to add nouveau and vga16fb to your kernel module blacklist configuration file, as described here.

No longer a mountain blog

Someone should write an article here every once in a while.

Three months, three submissions. I figure I should change that header to something closer to the vistas around here, so there you have it. I have been with AT&T for more than six months, after all (wow). According to the baby websites, I should have doubled my baby weight, and have discovered myself in the mirror. Supposedly I should be happy talking anytime now. Personally, I’m blaming the lack of happy talk on the Vis deadline (I hope all your submissions went well!)

If you haven’t yet done so, you should read Robert’s spot-on post on the visualization cargo cult. If we don’t define what we mean by effective visualizations, someone else will. We won’t like the result.

Arduino hacking: differential drive for piezos and the Tone Library

My Christmas present this year was a bunch of Arduino gear, so I have spent some time playing with it. I bought a piezo to see what beeping noises I could make. To get accurate tones, it turns out that you need to use the hardware timers to generate the square waves. The Tone library does exactly that, with a nice interface to boot.

To learn my way around larger pieces of arduino code, I decided to change the tone library so that it optionally uses differential drive. This is how you get twice the voltage, and twice the volume, of a regular pin-to-ground piezo setup. By connecting both ends to different pins and making sure their outputs alternate, we can get 10V edges out of the arduino’s 5V pins. While it is well short of my piezo’s 30V range, it’s loud enough to hear it across the apartment (ie sufficiently loud to annoy your SOs).

Anyway, here’s the patched Tone.cpp file. When creating a Tone object, you have the option of passing an extra parameter which is the second pin. If you ignore it, the code won’t tie up any extra pins (but it will use more memory).