However, a few days ago, the good folks at distributed.net released the first public beta (for x86-64 Linux) of a GPU client for NVIDIA cards using the CUDA platform. The client currently only supports the RC5-72 project and not OGR, which might be disappointing for some, as the RC5-72 project is of dubious scientific or mathematical value, though there is a prize involved.

In any case, I have been testing the client over the past few days on my NVIDIA 512MB GeForce 8800 GTS and here are my observations:

The thing is ridiculously fast. My Intel Core 2 Quad Q6600 overclocked to 2.97Ghz is already quite fast, but the graphics card completely smokes it. Using all four cores, the Q6600 will average about 800 RC5-72 blocks per day. If utilized all day on my aforementioned graphics card, the CUDA-enabled client would complete approximately 6200 blocks. That’s almost eight times as fast. Granted, these GPUs can only be used in some problem domains, but they clearly have an advantage in those areas where they are well-suited. Nonetheless, there remain a few issues that will prevent me from using the client to its full capacity.

First, the client pegs one of the CPU cores at 100%, meaning you’re trading one CPU for the performance on the GPU. All things considered, this is probably a fair trade, though if you like to run several different projects, it’s more limiting. (RC5-72 probably isn’t interesting enough to me to run on my CPUs at the moment, though that can change at any moment.) At GPUGRID, they seem to have been able to reduce the CPU load caused by the client to a few percent on Linux, so this problem may yet be solvable to some degree.

Perhaps more importantly, however, desktop performance becomes quite sluggish while the client is running, to the point of being unusable. Perhaps I bring this upon myself by using maximized applications on a 1920×1200 screen, but it’s an issue nonetheless. A separate test client (using 64 threads instead of 128) made available by one of the authors was slightly better in this regard, but probably not enough for me to tolerate. However, this problem in no way prevents me from running the client while I sleep or am otherwise not at the computer, which can add up to a lot of hours a week. (The past couple of days, I’m averaging about 4000 RC5-72 blocks per day.)

So, while the client is somewhat rough about the edges, the performance enhancements it brings make the whole distributed computing arena about to become much more interesting. My only hope is that other suitable projects begin to adopt the technology. I know not all of them will be able to take advantage at this stage (especially those requiring double-precision floating point math), but it’s still something to look forward to in the future. It’s an exciting time for the technology enthusiasts of the world.

So, when my trusty X5L recently suffered an unfortunate accident (in which the power button managed to break itself, perhaps from excessive use), I was tasked with finding a new player. After doing some quick research, I determined that support for the free formats was almost exclusively the domain of Cowon and iRiver. Being satisfied with my previous Cowon purchase, I went down that road again. The larger players were all out of my price range, but I eventually settled on the 8GB model of the Cowon D2, which has the following major features:

• Supports MP3, WMA, FLAC, WAV, APE, OGG Vorbis
• USB 2.0 mass storage device for high compatibility
• Comes in sizes ranging from 2GB to 16GB.

The Cowon D2 has an intuitive interface, and I was able to configure it the way I like it in almost no time at all. It operates on something of a dynamic playlist, where you can choose to limit its playlist to either one track, one folder, or everything. It can then play that playlist in order or shuffle it. It may support custom playlists, but I don’t use such a feature, so I don’t know.

Sound-wise, the quality is great. Music is clear, with no audible distortion. The equalizer performs well at enhancing the particular frequency ranges I like to listen to. In addition, it has the built-in ability to control playback speed, for listening to podcasts at a slightly faster clip (though I preprocess my podcasts, as the D2 doesn’t support my preferred 160% speed—it maxes out at 150%).

That said, the player isn’t perfect. Initially, I was getting atrocious transfer speeds, though I no longer have this problem, so it may have been a transient issue with my setup. It has no line out, but the headphone jack at volume level 40-45 seems to produce good results. Anything louder tends to cause clipping with my audio. I find podcasts do better at a slightly higher volume level than most music. Finally, the ability to queue a track would be nice, but is missing from the software.

The D2 also includes a built-in scientific calculator, can operate as a voice recorder, can view pictures and videos, and even purports to be able to play Flash media (though I’ve had limited success with that option).

In conclusion, the Cowon D2 is a nice player with enough space for most needs. The battery life is fairly long, and I find I need to charge it no more often than once a week (using it for at least two hours a day on average). It especially excels at podcast listening, though it’s no slouch when it comes to your favorite music. The ability to save presets makes switching between car mode and headphone mode quick and painless. If you or a family member are in the market for a great portable music player with an average amount of space, give the Cowon D2 a look.

Contrary to the experiences of most, this is one area on Linux that I’ve usually had a great experience with. My Sound Blaster Audigy card is capable of mixing several separate audio streams in hardware. No annoying blocking of the sound device. I’ve never had a need for the various software mixing solutions, such as ESD, aRts, ALSA’s dmix, and most recently, PulseAudio.

However, since switching from my well-functioning Gentoo installation to Ubuntu, it’s been just nothing but trouble. Applications using the sound system (Pidgin, mplayer, Amarok) seem to randomly freeze and need to be restarted. After having some mixer issues with mplayer, I noticed that applications seem to be using PulseAudio by default, even if I’ve configured them to use ALSA. (The default ALSA device seems to be going through PulseAudio for reasons I can’t figure out. Either that, or through some other goofy ALSA plugins. It’s definitely not going straight to the hardware, though, and whatever it’s using has its own independent mixer.) The whole thing is awfully voodoo-like. If there’s one thing I don’t like, it’s computer voodoo.

At this point, I’ve managed to make mplayer behave by forcing it to use the hardware directly (-ao alsa:device=hw=0), but that doesn’t help the other applications. Amarok, however, at least has the advantage of also having had this problem in Gentoo, so it seems more likely to be a bug in Amarok, xine, or even the audio driver.

So, it seems that I have little need for PulseAudio at all, but since it’s depended on by the ubuntu-desktop package (which I prefer to keep installed), it’s not going anywhere. It’s issues like this that make me want to run back to Gentoo, despite shortcomings of its own.

It’s always something, isn’t it. A bug-free environment would just be too much to ask.

OGR stands for Optimal Golomb Ruler, and while I’m not sure these higher-order rulers have any practical value, they are an interesting mathematical curiosity for some of us.

In any case, I’m been somewhat interested in the statistical side of it, and have thus been working on producing something of interest, so I’ve uploaded an OGR-26 statistics page. It currently only has my data on what the distribution of stub sizes is, but as time goes on, I hope to include my own estimates of when the project will complete. It’s not incredibly useful, but it might be interesting to someone.

One barrier to working on this is good access to score data for the various sports. There are a few good resources, like Massey Ratings, that can help to facilitate this, but some of the data is outdated and doesn’t necessarily have all the information I want. For my previous site, I wrote a script that parsed results from various live sources on the web, but I think we can do better than that.

To that end, I’ve launched a new project I’ve (tentatively) named the Penultimate Reality Sports Data Project, which aims to provide free and open access to sports score data for as many sports as possible and as many years as possible. The only thing real content I’ve put so far is an older draft of the specification for the data format that I originally wrote a couple of years ago. (I temporarily shelved the project after that.) That said, I’ve come to want to relaunch it, and if anyone out there is interested in helping in any way, please get in contact with me. Otherwise, I’ll press on alone and hope somebody finds my work useful. I at the very least hope to get a new draft of the specification online by this time next week. We shall see.

For those of you who primarily want to read about programming and Linux… well, uh, sorry. Maybe next post. This is my personal blog, after all.

UPDATE: I had some time to kill, so I went ahead and revised the specification. The changes weren’t incredibly substantial, but I am now actively seeking comments. Whether or not I get any is another question altogether.

Maybe I should dig out my old 286 and play around with that. Only downside is that I can’t install Linux on it. Besides, I think it needs a new motherboard, and finding a replacement might be a bit difficult nowadays.

Long story short, a couple of years ago, I managed to install Gentoo on the thing, and it ran fine (for some definition of fine–it was difficult to compile anything, as I had set it up using another computer to compile binary packages). A couple of days ago, I decided to revive it and install Debian on it, since that would be marginally more effective. (It’s run Debian before.) Unfortunately, the Debian install CD was refusing to recognize the hard drive and CD-ROM drive.

Eventually, I discovered that I had to manually modprobe the ide-generic module. My, that’s inconvenient. In any case, Debian’s happily installing now (over the blisteringly fast 115.2k serial link, no less). There are plenty of other quirks about using this laptop, but I should probably save them for some sort of book on the horrors of buying IBM. At least I got the thing for free.

I’ve become increasingly bored with maintaining my Gentoo machines. This culminated with me reinstalling Ubuntu on my laptop a few weeks ago, At this point, I don’t even notice except that I am no longer repeatedly compiling packages every day. The biggest thing I’ve noticed is that package versions are sometimes a little outdated (which is to be expected, given Ubuntu’s six month update cycle). This led me to examine what versions of packages are in the sid repository of Debian. As it turns out, it’s quite updated, with new versions of packages appearing very quickly. It’s led me to believe it might be all right for a desktop system.

My desktop system is rapidly becoming obsolete, and it slowly feels slower and slower as time goes on. (It’s using an AMD Athlon XP 1600+ which was originally purchased in late 2002.) Keeping the copy of Gentoo I have running on it up to date is increasingly becoming a dull procedure, and given my penchant for micromanagement… In any case, I’m slowly coming around to consider installing Debian on it, though I don’t know if I want to risk breaking it. After all, it isn’t broken, and I intend on replacing it within the next year. Maybe I just play around too much…

I could always just install Linux from Scratch. That’d be great for a laugh.

(By the way, if any of you happen to be looking for shared web hosting, consider buying from me at www.calindora.com. Why, you ask? I got tired of seeing hosting companies (probably dishonestly?) offering 600GB or 34TB or whatever amounts of disk space they’re claiming to offer these days. I feel bad for anyone who actually intends to try to use all that space.)

You can, as always, find this latest release at the Custom Title MOD web site. Please test this version and let me know if you have any problems, concerns, or suggestions. And if you find it useful, please consider donating. Thanks.