UncleAl Posted November 11, 2009 Author Report Share Posted November 11, 2009 Never underestimate the disability of Microcrap software. ZIP safely arrived. Time to arrange and massage data. We'll see if wishing is better than having. Quote Link to comment Share on other sites More sharing options...
GAHD Posted November 13, 2009 Report Share Posted November 13, 2009 Al, next time this comes up get alex to make up a <8GB boot image(for my thumb drive) to just auto-run it all. I've got no problems donating my laptop's spare nightime cycles to give ya a hand. Specs: Q6600 (core 2 Quad, 2.4 GHz), 2 GB RAM (couldn't find 2 sticks of 4GB in laptop style...this may have changed recently) 1 TiB (3*RAID0) HDD, appox 600GB unused. Edit: Additional note, I use "fake RAID" so linux needs some tweaks to be able to write to it,best to have it auto-write to the thumbdrive, or auto-send to a remote FTP upon completion. EIther way, 4 cores, Intel or not, can crunch a lot of numbers for ya. Quote Link to comment Share on other sites More sharing options...
UncleAl Posted November 13, 2009 Author Report Share Posted November 13, 2009 I'll shortly have about 900K unique points for gamma-glycine. That stretches the limits of my fitting and plotting software. It will be sufficient to validate the one parameter (mx + B) fit versus calcuated values. Thank you! to everybodywho lent a hand (and silicon). Our published example is quartz. We listened to physicists re graph linearity vs. radius and ate a CPU-decade sampling 96K points out to 1.1 million A radius. Runtime/point increases as (radius)^2. Extreme sampling density is the key, and out to 5000 A radius is entirely adequate. Question authority. We'd like to redo quartz the productive way. I'll wang together source code and have it compiled into a static file. Validate by comparing outputs, set up timing runs. Anybody who wants to volunteer some spare hardware or overnight runs will be invited. Then... we stop. Know when you have the answer. The very first thing to do is to see what we've got before we get more. Gnuplot won't handle much more than 15K points a whack. That can be subverted with "multiplot" - but I've got to write an extended load file. Don't go away. "8^>) Quote Link to comment Share on other sites More sharing options...
alexander Posted November 13, 2009 Report Share Posted November 13, 2009 Al, next time this comes up get alex to make up a <8GB boot image(for my thumb drive) to just auto-run it all. I've got no problems donating my laptop's spare nightime cycles to give ya a hand. Specs: Q6600 (core 2 Quad, 2.4 GHz), 2 GB RAM (couldn't find 2 sticks of 4GB in laptop style...this may have changed recently) 1 TiB (3*RAID0) HDD, appox 600GB unused. Edit: Additional note, I use "fake RAID" so linux needs some tweaks to be able to write to it,best to have it auto-write to the thumbdrive, or auto-send to a remote FTP upon completion. EIther way, 4 cores, Intel or not, can crunch a lot of numbers for ya. Yes except that code is not threaded so it only uses one core :naughty: Quote Link to comment Share on other sites More sharing options...
Pyrotex Posted November 13, 2009 Report Share Posted November 13, 2009 You guys just blow me away. Fascinating. :naughty: Whatever happens, I'm on your side. Quote Link to comment Share on other sites More sharing options...
GAHD Posted November 14, 2009 Report Share Posted November 14, 2009 Yes except that code is not threaded so it only uses one core :)then FIX that :naughty: Quote Link to comment Share on other sites More sharing options...
UncleAl Posted November 19, 2009 Author Report Share Posted November 19, 2009 When you get to the bottom of a hole, stop digging. Project status: I've got 947K points total for glycine gamma-polymorph less 13,000 points that must be rerun (about 6 hrs). The 302K new points will be added to the original stack, duplicates (there should not be any, but "should" is not good enough) will be removed. Least squares fit for slope and intercept, then graph the points. If it mirrors theory, even better. gamma-Glycine is nice because compute time varies as atoms/volume and (radius^2). A very tightly packed crystal lattice is faster to compute to a given number of contained atoms. The published paper http://www.mazepath.com/uncleal/boojum.pdf recommended alpha-quartz for the parity Eotvos experiment challenging fashionable gravitation theory. It is commercially grown to high specs. Folks know how to shape it to lamba/20 accuracy, how to gild it (100 nm titanium tie-layer), and it doesn't outgas gas in hard vacuum. Quartz is crunching away in my garage for the next 110+ days. I've got two volunteers with cluster access, so that could substantially contact. It's been wonderful! Everybody was splendid. We are pushing the paper to have the experiment performed in (reluctant - Not Invented Here) academia. Only technical objections matter. I'll have a couple of Hypography-sharpened sticks to make the pigs squeal. I'll post a link to the graph when the data is fully arrived and massaged. Quote Link to comment Share on other sites More sharing options...
Recommended Posts
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.