|
Members in Shoutbox
None.
Shoutbox Search
Shoutbox Commands
/w [name] > Whisper
/r > Reply to last whisper /me > Marks as action Shoutbox Information
Moderators may delete any and all shouts at will.
|
Global Shoutbox
Please log in to shout.
[2014-9-28. : 4:57 pm]  MasterJohnny -- trgktrgk shouted: For calculating A^k, use http://en.wikipedia.org/wiki/Eigendecomposition_of_a_matrix does nonsymmetric imply that i can always get a diagonalizable matrix?[2014-9-28. : 4:56 pm]  Moose -- Yeah, or e^A, which can be very useful for solving certain systems of differential equations[2014-9-28. : 4:55 pm]  jjf28 -- Mini Moose 2707Mini Moose 2707 shouted: Yeah, cosine of a matrix, everything you thought you knew was wrong I thought I knew everything I thought I knew was wrong[2014-9-28. : 4:52 pm] trgk -- A = Q T (Q^-1) . so A^k = Q (T^k) (Q^-1). Since T is diagonal matrix, it can be calculated very fast[2014-9-28. : 4:51 pm] trgk -- For calculating A^k, use http://en.wikipedia.org/wiki/Eigendecomposition_of_a_matrix[2014-9-28. : 4:36 pm] Moose -- That has a link to http://en.wikipedia.org/wiki/Coppersmith%E2%80%93Winograd_algorithm which is faster. That article mentions something faster developed in 2010.[2014-9-28. : 4:26 pm] MasterJohnny -- my cs friend said that I could do http://en.wikipedia.org/wiki/Strassen_algorithm to get A^k faster[2014-9-28. : 4:25 pm] MasterJohnny -- sigsaucysigsaucy shouted: either solve for A inverse, this might take longer but it will let you solve Ax =b quickly for multiple b easily So I would multiply A inverse k-1 times?[2014-9-28. : 4:22 pm] jjf28 -- should prolly use the stack trick to get that unit spawn correct... but it would slightly slow units under it[2014-9-28. : 4:20 pm] jjf28 -- Mini Moose 2707Mini Moose 2707 shouted: You could use the high ground built into locations themselves indeed, using locations med air/high air (whether you're over med/high ground) can help; chkdraft's "show tile elevations" can show you which is which. You can also make a pixel location over the flyer and see if you can spawn a unit that stays in said location to check for unwalkable area[2014-9-28. : 4:03 pm] Roy -- You could also find refurbished or used laptops that meet those requirements.[2014-9-28. : 2:49 pm] LoveLess -- JackJack shouted: http://devilesk.com/ is this the evildesk we all know and love to hate? Yes, he is an avid DOTA2 player.[2014-9-28. : 2:23 pm] FlameViper -- Does anyone know how to detect terrain? I want to order a flying unit to stop if it tries going over the wall.[2014-9-28. : 11:01 am] Generalpie -- JackJack shouted: http://devilesk.com/ is this the evildesk we all know and love to hate? Steam Group: "Mikelat's club of extraordinary cool people"[2014-9-28. : 7:57 am] Jack -- http://devilesk.com/ is this the evildesk we all know and love to hate?[2014-9-28. : 5:42 am] Excalibur -- http://www.staredit.net/topic/16571/ Made a thread for my newest project.[2014-9-28. : 5:00 am] jjf28 -- JackJack shouted: jjf28 lies, that's not the only reason and it's extremely difficult/impossible to get smaller with silicon pretty much, we have higher-power-processors ready to go if the cooling made financial sense; i'd call getting smaller parts a near-horizon problem rather than something holding us back right now[2014-9-28. : 4:56 am] Jack -- Heat and leakage are also a problem it's true, I should have mentioned that[2014-9-28. : 4:54 am] Jack -- jjf28jjf28 shouted: Jack lies, it's due to the power/heat wall lies, that's not the only reason and it's extremely difficult/impossible to get smaller with silicon[2014-9-28. : 4:41 am] jjf28 -- apparently someone did it: http://www.overclockers.com/how-to-build-a-refrigerator-cpu-cooler/[2014-9-28. : 4:41 am] jjf28 --  I joking thought "why don't we make the processor air-tight and have a mini-fridge around it[2014-9-28. : 4:28 am] jjf28 -- if processing had only doubled two more times we'd have workable high-density 3d libraries for affordable laptops[2014-9-28. : 4:16 am] jjf28 -- JackJack shouted: Desktop processing power has plateaued, due in part to the difficulty of making smaller silicon dies. While there are still advances to be made in the areas of superconductors and quantum computing, many of the speed increases have come from parallel processing and CPU optimizations in the past few years, rather than a simple increase of the number of transistors on a CPU. lies, it's due to the power/heat wall[2014-9-28. : 2:51 am] O)FaRTy1billion[MM] -- I just always forget about the precomputed data because I just throw it in an MPQ that gets shoved on to the EXE, load it in to a variable, and never think about it again  [2014-9-28. : 2:50 am] O)FaRTy1billion[MM] -- Oh, wait, I do in fact do a bit of optimization ... TinyMap has a bunch of precomputed data and stuff, and other programs of mine will also load or generate look-up tables at startup.[2014-9-28. : 2:45 am] O)FaRTy1billion[MM] -- Because I don't feel like I did anything to optimize it, I just did the necessary steps for doing what needs to be done. ;o[2014-9-28. : 2:45 am] O)FaRTy1billion[MM] -- I remember saving with like Uberation or tools way before that and it'd take forever and have to have a loading bar.[2014-9-28. : 2:45 am] O)FaRTy1billion[MM] -- It's really fast, and it being so fast has always made me really question why and how old programs took so long to load/process. xD[2014-9-28. : 2:44 am] Moose -- Indeed, but I wouldn't care if something like TinyMap took 2.5 seconds instead of 2 seconds to compress a map. (I can't even guess an accurate time because it's sufficiently fast)[2014-9-28. : 2:44 am] O)FaRTy1billion[MM] -- Optimizing while still using modern hardware means you get to do even more polies per frame![2014-9-28. : 2:43 am] O)FaRTy1billion[MM] -- Mini Moose 2707Mini Moose 2707 shouted: I feel as though the increasing power of hardware has made many optimizations less useful. Less useful, except it still has its place when you're doing like millions of matrix calculations per frame. xD[2014-9-28. : 2:40 am] sigsaucy -- or you can do fancy stuff like decomposing A into a lower and upper triangular matrix which makes it easier to solve[2014-9-28. : 2:38 am] Jack -- Because of this, optimizations are becoming more important, particularly for big number-crunching programs.[2014-9-28. : 2:38 am] sigsaucy -- this is probably easier, but if you want to solve many Ax =b problems, its more efficient just to solve for A inverse once[2014-9-28. : 2:38 am] Jack -- Desktop processing power has plateaued, due in part to the difficulty of making smaller silicon dies. While there are still advances to be made in the areas of superconductors and quantum computing, many of the speed increases have come from parallel processing and CPU optimizations in the past few years, rather than a simple increase of the number of transistors on a CPU.[2014-9-28. : 2:37 am] sigsaucy -- or just do the way people learn in basic algebra where you add the two equations and substitute and manipulate[2014-9-28. : 2:37 am] sigsaucy -- either solve for A inverse, this might take longer but it will let you solve Ax =b quickly for multiple b easily[2014-9-28. : 2:35 am] Moose -- Oh well, I guess this is why they'll pay fat stacks to low latency programmers who want to game fractional seconds on the stock markets trollololol[2014-9-28. : 2:35 am] Moose -- Though Project Euler was at least good for that. There were a lot of things I knew how to solve, it was just a matter of getting an algorithm that was actually fast enough. |
moore's law
Members Online: Ultraviolet
Ultraviolet