So first things first, I am going to be in Redwood City, CA working for EA, for the EA studio over there. Yes I have heard all the bad press about working at EA, but I wanted not to be in New York City, and being close to San Francisco will be nice. In the course of my interviewing, I found (and then rejected) a job at Massive Inc, a company now owned by those kind folks at Microsoft. Either ways, I feel that the idea behind what is Massive Incorporated is amazingly cool.

So here’s how you have your Mac throw up a kernel panic. Start up SubEthaEdit, the new version, watch something using QuickTime, quit SubEthaEdit, and watch your kernel go down, or should this fail, do it about ten times, and it will. This has happened to at least one other person I know, and might be Intel specific. Funny kernel panic that, here’s a clue as to where it happens:

      Kernel loadable modules in backtrace (with dependencies):
         com.apple.iokit.AppleYukon(1.0.9b3)@0x9b2000
            dependency: com.apple.iokit.IONetworkingFamily(1.5.1)@0x848000
            dependency: com.apple.iokit.IOPCIFamily(2.2)@0x59b000
            dependency: com.apple.iokit.IOACPIFamily(1.2.0)@0x637000
         com.apple.iokit.IONetworkingFamily(1.5.1)@0x848000

Any of the kind folks from Apple who previously spoke about how fearful they are of including DTrace in Leopard want to talk about the possible ease with which one could try and solve this, and someone want to comment on the b in that version number.

So, anywhoo, I was googling for other people who had this triggered on them, but I didn’t quite find anyone who had spoken of this. Instead I found a solution to my question about getting outputs and inputs connected, so I think I am moving away from trying to code that anymore. Anyways, should anyone be looking for how to code these things, as opposed to downloading code, you want to ignore CoreAudio, and use parts of the IOAudioDevices, engines and such.

Ze Panda

PS: Anyone want to advice me on fun places to go for in San Francisco.

Watch this, it is amusing, despite obvious similarities to lives of too many people in any cs departments.

Ze Panda

PS: I am surprise NVidia let me into there CUDA pre-release program
PPS: Documentation on ATI shaders, especially things related to how to build them on Macs is fairly limited, this seems to be a recurring theme.

So anyone know why allocation of largish lists seems to take so long in Python, like time measurable in seconds for largish sizes (about ten million or so). Like if they were doing it the obvious way (taking into account the fact that someone might actually be doing something like [0 for x in xrange (0, maxwt)]), they would for one realize that someone’s just about to allocate maxwt-1 elements (how about being able to specify this, I don’t mind hinting that this is going to happen, or even giving it to them), and be using something which allocates increasingly larger amounts of memory as you run out of space. It is an amortized O(1) operation if you keep doubling your memory heap, and seeing as the python allocator runs out of memory soon enough anyways, there seems little danger that this would lead to too many bad things. Or their could be an easier way to pass this information to the execution environment. Either ways, lack of fast array allocation has me disappointed.

And before supporters of OCaml (the other language I like), comment on this, OCaml arrays are in my opinion too constrained by the maximum allowed size, and BigArrays need a better interface.

Of course I could just use C for code which builds partial tables for large problems using DP, but umm I wanted to use the freedom offered by a grad seminar.

Ze Panda

Turns out some people over in the cold land which lies to the north of my current location, and which certain friends of mine call home, have decided that they have a commercially viable quantum computer that they plan on selling soon. No specs seem to be easily visible on the website, so questions about whether this merely reflects their ability to multiply two numbers with some high probability of success, or an actually viable computer, currently abound. A quick search through Google Scholar yields nothing (no research papers, I am disappointed, no specs, and no papers make me unhappy), however the company did file for a bunch of Quantum Computing related patents in 2000 and 2004, though none seem to have been filed post January 2005, which leads to obvious questions about what in the world happened in the last month, and observations about the awesomeness of Google Patents.

Scientific American seems to have a story about this, painting a less rosy, more realistic picture, and much like everything else that can be seen, Flickr has an image.

Scientific American also claims that it won’t be breaking any encryption schemes, making this entirely worthless. Also the guy in the Scientific American article seems to accept the fact that it probably won’t be viable at larger scales, and might not scale up, and might not be the commercial success everyone wants it to be. Wait, a company announces a commercially viable quantum computer that might not work, and does so by solving a Sudoku (simple enough thanks to all the constraints and the magic of constraint programming)... Bah, this has been a disappointing discovery, I think I will go back to approximating my NP-complete problems.

Ze Panda

See the last entry, then see this
Ze Panda