sigblips: Looks like it is just me today so I'm going to quickly mention a couple items for the setiQuest forum readers. Hi Dave.
[11:30am] sigblips: I wonder if the USA's daylight savings time is mixing some people up on when the community IRC meetings start?
[11:30am] sigblips: What am I working on? I've been busy implementing SIMD optimizations to http://baudline.com making it significantly faster. Woohoo faster! This is fun.
[11:33am] sigblips: SETI Live launched a new revision of its website today. Lot's of nice minor improvements but nothing that addresses SETI Live's fundamental problems. http://blog.setilive.org/2012/11/09/improved-site-on-its-way/
[11:36am] sigblips: Followups are 4 months late but are said to finally be operational. To find out we'll have to wait for later tonight when the ATA - SETI Live stream goes on-line.
[11:39am] sigblips: The Uwingu "Connect to the Sky!" beta went on-line this past week and their secret plan for generating money for scientific space activities is to sell exoplanet naming rights. http://uwingu.com
[11:40am] sigblips: A couple months ago Uwingu had a successful Kickstarter launch that generated about $80K of which about $2.5K went to the SETI Institute.
[11:42am] sigblips: Uwingu used the $75K in Kickerstarter funding to build their website!? I guess I'm in the wrong line of work. :(
[11:46am] sigblips: Exoplanets are a dime a dozen, figuratively, so I'm not really interested in buying Planet SigBlips but I would pay good money for the naming rights to Pulsar SigBlips! It's all about scarcity and apparent value.
[11:50am] sigblips: That and the fact that Pulsars are far cooler than Exoplanets.
[11:52am] sigblips: Have you voted in the latest setiQuest poll? http://setiquest.org/forum/topic/what-your-favorite-setiquest-data-signa...
[11:56am] sigblips: All this talk of Uwingu and funding makes me wonder how much SETI Live cost to create? Any guesses?
[11:56am] sigblips: SETI Live is a lot more complicated than a website for naming exoplanets.
[12:00pm] sigblips: Things are expensive to create. I have zero insider information on this but my guess is that the cost for SETI Live is in the $400K ballpark.
[12:01pm] sigblips: That concludes this week's setiQuest community IRC meeting. See you next week. Maybe. Maybe not. It's all up to you.
" I've been busy implementing SIMD optimizations to http://baudline.com making it significantly faster. Woohoo faster! This is fun"
Now isn't this exactly what I told you just over a year ago when we were discussing the Radon Transform? I thought you were all against using these new fangled methods of increasing performance!
Only joking, I must admit that when we saw the advantages we got on our real time image processing algorithms, we, like you were absolutely blown away, we started off using the MMX registers, which wasn't too bad, as pixels were always 8 bit bytes, and very rarely negative; however the SIMD array really gave us an edge, much more logical to use as well. I must admit that I was very lucky to have a guy in my team that knew the structure of the Intel chips intimately, and was able to program them right down at the assembler level (goes back to what you were saying about optimizing the algorithm before throwing in the smart technology). I understand that they have made using them significantly easier these days by providing access to them via C, I wonder how much you lose in performance by working at the higher level?
I kind of envy you in a nostalgic way, although I must admit that working at that low level is well beyond me these days
The SIMD stuff I'm optimizing in baudline is floating point. MMX is integer and came before SSE which is floating point. There have been 6 op code SSE extensions over the years. Next year's Intel CPUs will have yet another new SIMD extension called AVX2. As you can imagine this makes supporting multiple chip architectures a pain.
SIMD can be programmed in C with intrinsics that are basically in-line assembly instructions. There are autovectorizing compilers that will generate SIMD instructions automatically. The gcc compiler is one of them but it's not that good at autovectorizing. I've heard Intel's compiler is better but I haven't tried it yet.
Performance-wise there are a whole lot of gotchas. Things like memory address alignment, caching issues, all the different extension versions, a huge amount of performance variation between CPU architectures, and then there are AMD chips which are compatible but ...
Baudline also supports multiple operating systems (FreeBSD, Linux, Mac OS X, Solaris) on multiple CPU architectures (x86, x86_64, PowerPC, SPARC, Z-series) which makes this all the more complicated. Then there are the minor variations between operating system distributions/versions and different CPU revisions/models/manufacturers. It is one huge can of worms. I wish I could target just one CPU on one operating system but I can't.
I've learned that getting solid consistent performance from SIMD is a lot of work and it's not easy.
I don't believe you. Those $75K and $400K prices for building a website are ridiculous. Explain how you came up with those nonsensical numbers.
Also what do you mean by "exoplanets are a dime a dozen" ????????
I agree that those numbers are bigger than you would expect. Things are expensive to create.
The Uwingu $75K number is easy to explain. On their fundraising page they said they wanted $75K so that they could launch their first product. Check it out http://www.indiegogo.com/projects/180221
I said that the SETI Live $400K number is a ball park guess. I based my estimate on a basic labor calculation (#people * duration * $hourly rate) plus the cost for several business trips plus a small amount for computing resources. Both the SETI Institute and the Zooniverse host their respective parts of SETI Live on AWS but comparatively this cost is small. I looked at the http://setilive.org/#/about/team page for the number of people working on the project. That team page count is actually low because I know that there are several people who worked on it from the SETI Institute side that are not on that list. Work began on this project more than one year ago. http://setiquest.org/forum/topic/community-meeting-2011-08-23 Most of the SETI Live team did not work on it full-time. I had to fudge for the hourly rate. Guess and plug in your own numbers. What do you get?
Astronomers predict that there are 160 billion exoplanets in our galaxy. That is a lot of exoplanets. They really aren't a dime a dozen because that would be less than a penny each. Uwingu is selling them for $0.99 each. It was a joke. Get it?
Correction: Uwingu used the Indiegogo service for their fundraising and not Kickstarter.
I don't get your joke.
You are not funny.