Posted 759 days ago by Shane Reilly
Update: The speed boost is most apparent for larger blurs. The demo as-is uses blurs that are only a few pixels in diameter so there is, unfortunately, no speed-up currently for the screensaver under OpenGL. It actually runs a few percent slower than my current implementation. There may be a way to do a hybrid approach that will combine the benefits of both approaches, but it does not seem likely since the bottleneck seems to be the transmission of data between the graphics card and the CPU (there are 50,000 blurs drawn per frame). This issue might theoretically be overcome by programming the graphics card directly.
The new implementation does, on the other hand, use substantially less memory, requires less code complexity, shows finer detail with greater color depth, and scales nicely for larger blurs, showing only modest degradation in speed as blur size increases significantly.
Posted 760 days ago by Shane Reilly
Will the main page of the website be up again soon as well?
Posted 760 days ago by Shane Reilly
I have done some in-depth research into using OpenGL for the star-sprites. It seems that a 4x speed-up is a conservative estimate at least on my newest machine. The resulting speed-up could be closer to 10 or 15 times the single-threaded CPU-driven algorithm.
By using "glBlendFunc(GL_SRC_ALPHA, GL_ONE);" I am able to create additive blending without the use of get and put pixel.
Once the new implementation is complete, I will see if it is possible to create a moving screensaver view. At the very least, I should be able to pan images across the screen for a more dynamic effect instead of relying on a slideshow effect.
Posted 778 days ago by Shane Reilly
I tried both of your demos and while they work, I don't get it I guess. Maybe shooting for being too complex? Suggestion would be to simplify the graphics representation down to something that will fit entirely in the window. I realize my other suggestion I posted in the wrong thread. But this one is more of the same especially after toying with the cube demo and the other one.
My thoughts are perhaps there should be an option for a way more simplified view of it that still looks cool that passerbyes would stop and want to watch. As it is they are probably not there yet. Just my thoughts. :)
It should be relatively easy to fit the demo in the screen by pressing F2 or maybe F4 (increase speed x 10 and x 1000 respectively) and then holding "<" - at least on relatively fast machines ~2ghz or higher.
Posted 778 days ago by Shane Reilly
The apps used to repair the broken links may be a little out-of-date. I will put up the fixes for non-US keyboards in the next week or so by recompiling the new source.
That seams to be a reply on the 6 months old message posted by Werkstatt
;) Do you live close to Alfa centauri or in an other time frame? :)
It was not meant as a reply to the keyboard issue. Many of the links to the demo applications were broken some time in the last few months. To answer your second question, I am on Mars time.
Posted 784 days ago by Shane Reilly
The apps used to repair the broken links may be a little out-of-date. I will put up the fixes for non-US keyboards in the next week or so by recompiling the new source.
Posted 785 days ago by Shane Reilly
I have looked into this a bit myself. I am looking forward to testing an OpenGL implementation particularly on my lower end machines.
Posted 785 days ago by Shane Reilly
I am happy to say that I am still unofficially with the project. This last semester has been hectic, but I look forward to getting the screensaver up and running in the near future. It is complete for Windows and Linux, but needs final integration. I have been researching a dozen methods of creating low-maintenance event loops and I believe I have a workable solution for integrating the final application.
This thread is more recent and contains the working demo in case anyone would like to try it out (the broken links have been fixed).
Posted 898 days ago by Shane Reilly
"ESA's GAIA will map ~100 million stars, right? Still a long way off from all of them (and I doubt they'll all be part of the milky way), but.."
-Emanuel
My original numbers were misinformed. Currently the largest catalogs I know of have roughly 1 billion relative star coordinates. NOMAD is one example. Also the Tycho II catalog does not contain parallax data which will be available in many upcoming catalogs such as the one you mentioned. The current estimates for the distances of Tycho II stars are not accurate enough to show decent distributions of star density when all of the stars are mapped. The Hipparcos catalog (released in 1997) does contain parallax data of 100,000 stars which allow accurate 3-dimensional coordinates of stars to be mapped.
For those that are interested, here is a simulation showing Tycho II estimations of 2 million stars made by "granthutchison" as a Celestia add-on. The original data can be found here.
showstars_win.zip
Instructions for use are found in the zip file. Star colors were not included in this version.
Posted 951 days ago by Shane Reilly
The immediate galaxy is mostly mapped if you only consider relative angles to the Earth. The part hiding behind the central halo is much more difficult to map because the star density is greater there. The distances of the stars are harder to find and mapping them is an ongoing process.
Countries including the US, UK and now China have all had plans for telescopes that are specifically designed for this task. I believe the most complete catalog of star distances is currently Tycho2 which maps 2.5 million stars out of the estimated 300 billion stars in the Milky Way.
MilkWay@Home uses data from the Sloan Digital Sky Survey to narrow in on the structure and evolution of streams in the galaxy. The SDSS maps galaxies and other very distant objects making it possible to map approximate positions of F stars in the Milky Way thousands of times more distant than those in Tycho2. The SDSS has currently mapped more than 6% of the sky (http://www.sdss.org/news/releases/20031028.powerspectrum.html).