Whoa. In my apache logs:
18.104.22.168 - - [23/Feb/2011:00:35:47 -0600] "GET /shortlog/month HTTP/1.1" 200 29875 "http://www.google.com/search?hl=en&source=hp&q=nick+peach+hbgary&aq=f&aqi=&aql=&oq=" "Mozilla/4.0 (compatible; MSIE 7.0; Windows NT 5.1; .NET CLR 1.1.4322; .NET CLR 1.0.3705; .NET CLR 2.0.50727; .NET CLR 3.0.04506.30; InfoPath.2; MS-RTC S; .NET CLR 3.0.4506.2152; .NET CLR 3.5.30729; MS-RTC LM 8)"
Again: no good reason I'm that result. Who was searching, anyway?
(03:24:23) zarvox@rabbit ~ $ host 22.214.171.124 10.64.159.171.in-addr.arpa domain name pointer spxysfo1.bankofamerica.com.
I'm a little creeped out. I would guess that either 1) someone inside BofA is looking for info about their contact with HBGary, or 2) someone inside BofA is looking to see what the Internet currently says about them.
Initial benchmarks showed my approach to be about half as fast on my laptop as just solving for all roots explicitly with the BLAS eigensolver. On Kenghao's machine, it was even worse - my code was slower by a factor of four.
Then we tried it on a Nexus One, and my code was 37% faster than the eigensolver. Since this is intended for mobile phones, eventually, this is promising. Later, I remembered that I hadn't enabled compiler optimizations, and even on my computer, performance shot to within 20% of that of BLAS. I still have some ideas on how to make it even faster.