Help?

[3dickulus]

@M Benesi: correct Kn2, with #define KN_VOLUMETRIC, a script should compile in milliseconds, as low as 30 for simple scripts and up to a few seconds (on first run or major changes) for more complex stuff, even on my GeForce 8800GT (R.I.P)

@Crist-JRoger: attached is the script that will be in the Examples/Benesi folder 

[Crist-JRoger]

3dickulus, thank you very much! Now i see that it`s geometry bug in BenesiPineFoldDE.frag, but not a godray effect 

[3dickulus]

geometry bug? you mean a piece of the fractal is in the way? or is it in the math?

[M Benesi]

@M Benesi: correct Kn2, with #define KN_VOLUMETRIC, a script should compile in milliseconds, as low as 30 for simple scripts and up to a few seconds (on first run or major changes) for more complex stuff, even on my GeForce 8800GT (R.I.P)

 You "theoretically" have 504 gigaflops... I have 90.2.  So the one compiled in about 27 seconds (so it would have been about 5 seconds on yours, if we're both approaching the limits, I don't OC this one...), and normally takes about 3.872.  It was probably a first run... and maybe I was doing other stuff in the background.  It was... excessive.  


  Anyway, M3D is a bit quicker... unless I make the window in Fragmentarium realllly tiny.  I test formulas in Fragmentarium... then I struggle to port them to M3D, since I'm learning assembly...slowly.  

  I've been working on creating a new tetrahedral transform.  See how that ends up looking.  Still haven't figured out height scaling... attempt to do that tomorrow.  Attempt being the key word.  After that, since the Pine Tree seems to be aligned with a stellated octahedron, someone has to do the hard transform...

  Luca helped me with the code for the spherical transform, and I did the unspherical transform.  Definitely a nice little touch.  If you switch between coordinate systems when you're doing stuff... you can change things a bit:


  I might have altered the parameter file for the above image.. but it's a start. 

[3dickulus]

It is very cool to watch this code and formula evolve, amazing work M Benesi !

btw, my current nVidia GTX760 is 2258 GFLOPS, a bargain @ $249CDN
       above I was referring to my old card that Fragmentarium + FuzzBunnies = cooked

Would assembler output of fragment code be helpful when migrating from frags to M3D? when you compile Fragmentarium from source you can set a #define in DisplayWidget.h that will enable "Shader Asm" entry under the "Edit" menu. This is only for nVidia cards that handle at least GLv4+. It shows the shader as GPU assembler code, I know, it's not CPU asm but it may give some insight as to the structure, flow and perhaps some of the math calculations.

F v1.0.13 may also have a few bugs re: shader compilation and log widget management, Patryk also pointed out a filename extension bug that has since been fixed. ( ty Patryk  )

[M Benesi]

  Thanks!

  Your card might be a teeeeenny bit faster than mine...    Don't know if my 8 year old 8600mobile can handle GLv4... "Full OpenGL® support, including OpenGL 2.1"  hahahaha....

  I do have access to another computer, with I think an 840m, so I should be able to compile on that and grab the ASM.  It's worth a gander, at least for the fun of doing a build.  Good project for a bit later.  Had some math ideas today...

[3dickulus]

don't grab the source code until you are ready to compile as there may be fixes or features between now and then and my Win box seems to be acting normal? so I may provide a win exe soon.

[M Benesi]

k... I am currently having a major duhh moment with some math.   So.. no rushes! 

[DarkBeam]

Given I don't understand how your spheric works (explanation? )
I think a tetra to sphere should be like:
p.xyz *= ((1+disttotetra)/sqrt(1+x*x+y*y+z*z))
disttotetra should be the minimum distance of four axis arranged in a tetrahedron like min(x;a*x+b*y;-a*x+b*y;...) now find the rest

[M Benesi]

Given I don't understand how your spheric works (explanation? )

 Square part (not whole cube!) that "squares the circle" on the x axis (click to enbiggen):


 The rest can be worked out from the inverse functions page on Wikipedia, or you can do it geometrically with the Pythagorean theorum...  

  (theta= atan2(y/x) if you're doing the inverse function thing!  I can explain it further later, if you want)

[DarkBeam]

You can open a new thread please do... it is very interesting

[M Benesi]

k...  a bit later.  In what section should we put the thread??

  Should the thread by titled "Explanation of transforms"...

 I also have to do a sphere to conic transform today for part of the tetrahedral transform! 

[DarkBeam]

BenesiSpheric is a good title
I can move it later somewhere

[M Benesi]

hahhaha...k.   

[M Benesi]

  Still stuck on some math Luca.. getting to the stuff you requested.  This can be a general help thread... with mods (like you... ) splitting things off into appropriate sub threads as needed???

  My list is:
 1) conic section for tetrahedral
 2) compress [T1, uncircle, pine tree, circle] into one formula (so it doesn't take up so many tabs)
 3) add "every nth iteration" option to that formula... maybe...
 4) explain circle/square math more
 5) clean garage (lol... should be relaxing)
 6) write out math for IQ z^2 algebraic version, and... remind me if I forget stuff.

  6)  //done   Seems same as pine tree, on a different axis, and x and z components are not switched... 

  


  Was it IQ bulb??  Tired.. have to look again tomorrow, but I think it's:
UPDATE... fix math that I thought of after I turned off computer:

Code:

sq_r := sqrt(x*x+y*y+z*z)
sq_xz := sqrt(x*x+z*z)
r:= sq_r ^ Power

theta  = atan2( sq_xz , y) * P
zangle = atan2(x,z)  * zAP

x := sin(zangle)*sin(theta)*r+Cx   =====    4xyz /sqrt (x^2 + z^2)        //divide by sqrt(x^2+z^2)  not multiply!
y := cos(theta)            *r+Cy   ======     y^2 - x^2 + z^2
z := sin(theta)*cos(zangle)*r+Cz   ====     2y * (x^2 - y^2)  / sqrt (y^2 + z^2)     //same here!!!!