Pete´s Messageboard... No ISO/BIOS requests!

Homepage Members Register Login Search Old board


Neuer Thread ...
More : [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11]


AuthorTopics » Book an abo for this threadClose Thread Move Thread Fix the thread Print view Delete this thread

VerGreeneyes 
Strong supporter
......

...

Status:Offline
Date registered: 26.04.2007
Post:89
Send Message
...   Created on 01.12.2007 - 16:47Jump to top Quote this post Report this post Edit Delete

A quick update and a glimpse of what I've been working on ...

Code:
1:
2:
3:
4:
5:
6:
7:
8:
9:
10:
11:
12:
13:
14:
15:
16:
17:
18:
19:
20:
21:
22:
23:
24:
25:
26:
27:
28:
29:
30:
31:
32:
33:
34:
35:
36:
37:
38:
39:
40:
41:
42:
43:
44:
45:
46:
47:
48:
49:
50:
51:
52:
53:
54:
55:
56:
57:
58:
59:
60:
61:
62:
63:
64:
65:
66:
67:
68:
69:
70:
71:
72:
73:
74:
75:
76:
77:
78:
79:
80:
81:
82:
83:
84:
85:
86:
87:
88:
89:
90:
91:
92:
93:
94:
95:
96:
97:
98:
99:
100:
101:
102:
103:
104:
105:
106:
107:
108:
109:
110:
111:
 
x && c4c8 && !c4c5 && !c3c4 ||
x && c4c8 && !c4c5 && !c7d2 ||
x && c4c8 && c5c7 && !c4c5 ||
x && !c4c8 && !c5c7 && !c4c5 && !c3c7 ||
x && !c4c8 && !c5c7 && !c4c5 && !c7c8 ||
x && !c4c8 && !c5c7 && !c4c5 && !c0c4 ||
x && !c4c8 && !c5c7 && !c4c6 && !c3c7 ||
x && !c4c8 && !c5c7 && !c4c6 && !c7c8 ||
x && !c4c8 && !c5c7 && !c4c6 && !c0c4 ||
x && !c4c8 && !c5c7 && !c3c7 && !c7d0 ||
x && !c4c8 && !c5c7 && !c7d0 && !c7c8 ||
x && !c4c8 && !c5c7 && !c7d0 && !c0c4 ||
x && !c4c8 && !c4c5 && !c3c7 && !c6c7 ||
x && !c4c8 && !c4c5 && !c3c7 && !c2c4 ||
x && !c4c8 && !c4c5 && !c7c8 && !c6c7 ||
x && !c4c8 && !c4c5 && !c7c8 && !c2c4 ||
x && !c4c8 && !c4c5 && !c0c4 && !c6c7 ||
x && !c4c8 && !c4c5 && !c0c4 && !c2c4 ||
x && !c4c8 && !c4c6 && !c3c7 && !c6c7 ||
x && !c4c8 && !c4c6 && !c3c7 && !c2c4 ||
x && !c4c8 && !c4c6 && !c7c8 && !c6c7 ||
x && !c4c8 && !c4c6 && !c7c8 && !c2c4 ||
x && !c4c8 && !c4c6 && !c0c4 && !c6c7 ||
x && !c4c8 && !c4c6 && !c0c4 && !c2c4 ||
x && !c4c8 && !c3c7 && !c7d0 && !c6c7 ||
x && !c4c8 && !c3c7 && !c7d0 && !c2c4 ||
x && !c4c8 && !c7d0 && !c7c8 && !c6c7 ||
x && !c4c8 && !c7d0 && !c7c8 && !c2c4 ||
x && !c4c8 && !c7d0 && !c0c4 && !c6c7 ||
x && !c4c8 && !c7d0 && !c0c4 && !c2c4 ||
x && !c5c7 && !c4c5 && !c3c7 && !c3c4 ||
x && !c5c7 && !c4c5 && !c3c7 && !c7d2 ||
x && !c5c7 && !c4c5 && !c7c8 && !c3c4 ||
x && !c5c7 && !c4c5 && !c7c8 && !c7d2 ||
x && !c5c7 && !c4c5 && !c0c4 && !c3c4 ||
x && !c5c7 && !c4c5 && !c0c4 && !c7d2 ||
x && !c5c7 && !c4c6 && !c3c7 && !c3c4 ||
x && !c5c7 && !c4c6 && !c3c7 && !c7d2 ||
x && !c5c7 && !c4c6 && !c7c8 && !c3c4 ||
x && !c5c7 && !c4c6 && !c7c8 && !c7d2 ||
x && !c5c7 && !c4c6 && !c0c4 && !c3c4 ||
x && !c5c7 && !c4c6 && !c0c4 && !c7d2 ||
x && !c5c7 && !c3c7 && !c7d0 && !c3c4 ||
x && !c5c7 && !c3c7 && !c7d0 && !c7d2 ||
x && !c5c7 && !c7d0 && !c7c8 && !c3c4 ||
x && !c5c7 && !c7d0 && !c7c8 && !c7d2 ||
x && !c5c7 && !c7d0 && !c0c4 && !c3c4 ||
x && !c5c7 && !c7d0 && !c0c4 && !c7d2 ||
x && !c4c5 && !c3c7 && !c3c4 && !c6c7 ||
x && !c4c5 && !c3c7 && !c3c4 && !c2c4 ||
x && !c4c5 && !c3c7 && !c7d2 && !c6c7 ||
x && !c4c5 && !c3c7 && !c7d2 && !c2c4 ||
x && !c4c5 && !c7c8 && !c3c4 && !c6c7 ||
x && !c4c5 && !c7c8 && !c3c4 && !c2c4 ||
x && !c4c5 && !c7c8 && !c7d2 && !c6c7 ||
x && !c4c5 && !c7c8 && !c7d2 && !c2c4 ||
x && !c4c5 && !c0c4 && !c3c4 && !c6c7 ||
x && !c4c5 && !c0c4 && !c3c4 && !c2c4 ||
x && !c4c5 && !c0c4 && !c7d2 && !c6c7 ||
x && !c4c5 && !c0c4 && !c7d2 && !c2c4 ||
x && !c4c8 && !c5c7 && c3c7 && !c7c8 ||
x && !c4c8 && !c5c7 && c3c7 && !c0c4 ||
x && !c4c8 && c3c7 && !c7c8 && !c6c7 ||
x && !c4c8 && c3c7 && !c7c8 && !c2c4 ||
x && !c4c8 && c3c7 && !c0c4 && !c6c7 ||
x && !c4c8 && c3c7 && !c0c4 && !c2c4 ||
x && !c5c7 && c3c7 && !c7c8 && !c3c4 ||
x && !c5c7 && c3c7 && !c7c8 && !c7d2 ||
x && !c5c7 && c3c7 && !c0c4 && !c3c4 ||
x && !c5c7 && c3c7 && !c0c4 && !c7d2 ||
x && !c4c8 && !c5c7 && !c4c5 && c4c6 ||
x && !c4c8 && !c5c7 && c4c6 && !c7d0 ||
x && !c4c8 && !c4c5 && c4c6 && !c6c7 ||
x && !c4c8 && !c4c5 && c4c6 && !c2c4 ||
x && !c4c8 && c4c6 && !c7d0 && !c6c7 ||
x && !c4c8 && c4c6 && !c7d0 && !c2c4 ||
x && !c5c7 && !c4c5 && c4c6 && !c3c4 ||
x && !c5c7 && !c4c5 && c4c6 && !c7d2 ||
x && !c5c7 && c4c6 && !c7d0 && !c3c4 ||
x && !c5c7 && c4c6 && !c7d0 && !c7d2 ||
x && !c4c5 && c4c6 && !c3c4 && !c6c7 ||
x && !c4c5 && c4c6 && !c3c4 && !c2c4 ||
x && !c4c5 && c4c6 && !c7d2 && !c6c7 ||
x && !c4c5 && c4c6 && !c7d2 && !c2c4 ||
x && !c4c8 && !c5c7 && c4c6 && c3c7 ||
x && !c4c8 && c4c6 && c3c7 && !c6c7 ||
x && !c4c8 && c4c6 && c3c7 && !c2c4 ||
x && !c5c7 && c4c6 && c3c7 && !c3c4 ||
x && !c5c7 && c4c6 && c3c7 && !c7d2 ||
x && !c4c8 && c5c7 && !c3c7 && !c6c7 ||
x && !c4c8 && c5c7 && !c3c7 && !c2c4 ||
x && !c4c8 && c5c7 && !c7c8 && !c6c7 ||
x && !c4c8 && c5c7 && !c7c8 && !c2c4 ||
x && !c4c8 && c5c7 && !c0c4 && !c6c7 ||
x && !c4c8 && c5c7 && !c0c4 && !c2c4 ||
x && c5c7 && !c4c5 && !c3c7 && !c6c7 ||
x && c5c7 && !c4c5 && !c3c7 && !c2c4 ||
x && c5c7 && !c4c5 && !c7c8 && !c6c7 ||
x && c5c7 && !c4c5 && !c7c8 && !c2c4 ||
x && c5c7 && !c4c5 && !c0c4 && !c6c7 ||
x && c5c7 && !c4c5 && !c0c4 && !c2c4 ||
x && !c4c8 && c5c7 && c4c6 && !c6c7 ||
x && !c4c8 && c5c7 && c4c6 && !c2c4 ||
x && c5c7 && !c4c5 && c4c6 && !c6c7 ||
x && c5c7 && !c4c5 && c4c6 && !c2c4 ||
x && c4c8 && !c5c7 && !c4c6 && !c3c4 ||
x && c4c8 && !c5c7 && !c4c6 && !c7d2 ||
x && c4c8 && !c5c7 && !c7d0 && !c3c4 ||
x && c4c8 && !c5c7 && !c7d0 && !c7d2 ||
x && c4c8 && !c5c7 && c3c7 && !c3c4 ||
x && c4c8 && !c5c7 && c3c7 && !c7d2

That's another way to write the following:
Code:
1:
2:
3:
4:
5:
 
bool a = (c4==c5 && c4==c6 && c3!=c7 && c7==d0);
bool b = (c3==c7 && c7==c8 && c4!=c6 && c0==c4);
bool c = (c3==c4 && c7==d2 || a);
bool d = (c6==c7 && c2==c4 || b);
x && (c4==c8 && (c5==c7 && c4!=c5 || c5!=c7 && !c) || c4!=c8 && (c5==c7 && !d || c5!=c7 && !a && !b))

Obviously it's a lot longer, but each of those lines are as short as they can be - it's just a matter of choosing all the needed ones (they overlap) then rewriting them with () in appropriate places ... and I should get the shortest possible way of writing this conditional. Unfortunately choosing the needed ones is either slow or uses a -lot- of memory (more than my system has) so I'm still working that part out.

guest ...
Real addict
.........

...

Status:Offline
Date registered: 30.07.2004
Post:856
Send Message
...   Created on 02.12.2007 - 20:53Jump to top Quote this post Report this post Edit Delete

Looks like a good start.




VerGreeneyes 
Strong supporter
......

...

Status:Offline
Date registered: 26.04.2007
Post:89
Send Message
...   Created on 05.12.2007 - 02:48Jump to top Quote this post Report this post Edit Delete

guest schrieb
    Looks like a good start.


Thanks. Now instead of solving the rest of the problem I've taken a bit of a detour to optimise the first part. This because I have no easy way to input those results for the second part to work with, so any test I can do includes solving the first part. Using arrays rather than linked lists and basically going from C++ to C I've managed to get calculation time for this particular problem down to about 390 seconds, or six and a half minutes. Compare this to the ~40 minutes I got at first and the 25 minutes I eventually got the linked lists approach down to - and imagine that a more difficult problem might now take say one day rather than four to solve. Now that this is done (working out the details was annoyingly difficult) I can focus on finishing this thing.




VerGreeneyes 
Strong supporter
......

...

Status:Offline
Date registered: 26.04.2007
Post:89
Send Message
...   Created on 11.12.2007 - 04:10Jump to top Quote this post Report this post Edit Delete

Hey guys, finally some results for you Here's the new optimised shader, which gets ~80 fps on my laptop. (that's twice as fast as the version this thread started out with!) I can't make any guarantees that you'll see the same speedup, for instance my desktop's GF6600 hates some of the optimisations my laptop loves - but since the laptop uses a newer card I'm optimising it for that. Anyway, give it a try and let me know what you think

vertex file:

Code:
1:
2:
3:
4:
5:
 
void main()
{ gl_Position = ftransform();
  gl_TexCoord[0] = gl_MultiTexCoord0;
}

fragment file:
Code:
1:
2:
3:
4:
5:
6:
7:
8:
9:
10:
11:
12:
13:
14:
15:
16:
17:
18:
19:
20:
21:
22:
23:
24:
25:
26:
27:
28:
29:
30:
31:
32:
33:
34:
35:
36:
37:
38:
39:
40:
41:
42:
43:
44:
45:
46:
47:
48:
49:
50:
51:
52:
53:
54:
55:
56:
57:
58:
59:
60:
61:
62:
63:
64:
65:
66:
67:
68:
69:
70:
71:
 
/*
  2xSaI GLSL shader
           - Copyright (C) 2007 guest(r) - guest.r@gmail.com, Ver - ver.greeneyes@gmail.com
           - License: GNU-GPL

   The 2xSaI algorithm
           - Copyright (c) 1999-2001 by Derek Liauw Kie Fa.
*/

const vec4 dt = vec4(16777216.,65536.,256.,1.);
uniform vec4 OGL2Size,OGL2InvSize;
uniform sampler2D OGL2Texture;

vec4 TWOxSaI();

void main()
{ gl_FragColor = TWOxSaI();
}

float reduce(vec4 colour)
{ return dot(colour,dt);
}

vec4 TWOxSaI()
{ vec4 rValue;
  vec2 OGL2Pos = gl_TexCoord[0].xy*OGL2Size.xy,
     fp = fract(OGL2Pos), dx, dy,
     pC4 = floor(OGL2Pos)/OGL2Size.xy;
  if(fp.x<.5 || fp.y>=.5)
  { dx = vec2(OGL2InvSize.x,0.);
    dy = vec2(0.,OGL2InvSize.y);
  }
  else
  { dx = vec2(0.,OGL2InvSize.y);
    dy = vec2(OGL2InvSize.x,0.);
  }
  vec4 C0 = texture2D(OGL2Texture,pC4-   dx-   dy),
       C1 = texture2D(OGL2Texture,pC4      -   dy),
       C2 = texture2D(OGL2Texture,pC4+   dx-   dy),
       C3 = texture2D(OGL2Texture,pC4-   dx      ),
       C4 = texture2D(OGL2Texture,pC4            ),
       C5 = texture2D(OGL2Texture,pC4+   dx      ),
       C6 = texture2D(OGL2Texture,pC4-   dx+   dy),
       C7 = texture2D(OGL2Texture,pC4      +   dy),
       C8 = texture2D(OGL2Texture,pC4+   dx+   dy),
       D0 = texture2D(OGL2Texture,pC4-   dx+2.*dy),
       D1 = texture2D(OGL2Texture,pC4      +2.*dy),
       D2 = texture2D(OGL2Texture,pC4+   dx+2.*dy),
       D4 = texture2D(OGL2Texture,pC4+2.*dx      ),
       D5 = texture2D(OGL2Texture,pC4+2.*dx+   dy),
       p10,p11;
  float c0 = reduce(C0),c1 = reduce(C1),c2 = reduce(C2),c3 = reduce(C3),
        c4 = reduce(C4),c5 = reduce(C5),c6 = reduce(C6),c7 = reduce(C7),
        c8 = reduce(C8),d0 = reduce(D0),d1 = reduce(D1),d2 = reduce(D2),
        d4 = reduce(D4),d5 = reduce(D5);
  float r = float(c5==d4 && c2==c5) + float(c5==c6 && c5==d1) + float(c3!=c4 && c1!=c4 && c3==c5 && c1==c5) + float(c4!=d5 && c4!=d2 && c5==d5 && c5==d2)
          -(float(c3==c4 && c1==c4) + float(c4==d5 && c4==d2) + float(c5!=d4 && c2!=c5 && c4==d4 && c2==c4) + float(c5!=c6 && c5!=d1 && c4==c6 && c4==d1));
  bool a = c4==c5 && c4==c6 && c3!=c7 && c7==d0 || c4==c8 && c3==c4 && c7==d2;
  bool b = c3==c7 && c7==c8 && c4!=c6 && c0==c4 || c5==c7 && c6==c7 && c2==c4;
  bool x = fp.x<.5 && fp.y>=.5 || fp.x>=.5 && fp.y<.5;
  bool y = fp.x>=.5 && fp.y>=.5;

  rValue = C4;
  if(x && c4!=c8 && b) rValue = C7;
  if(!(!x || c4!=c8 && b || c5!=c7 && a || c4==c8 && c5==c7 && c4==c5)) rValue = .5*(C4+C7);
  if(y && c5==c7 && (c4!=c8 || c4!=c5 && r< 0.)) rValue = C5;
  if(y && (c4!=c8 && c5!=c7 || c4==c8 && c5==c7 && c4!=c5 && r==0.)) rValue = .25*(C4+C5+C7+C8);

  return rValue;
}


[Dieser Beitrag wurde am 11.12.2007 - 04:59 von VerGreeneyes aktualisiert]




VerGreeneyes 
Strong supporter
......

...

Status:Offline
Date registered: 26.04.2007
Post:89
Send Message
...   Created on 11.12.2007 - 05:33Jump to top Quote this post Report this post Edit Delete

By the way, guest(r), for range testing rValue = C4 has to be made independent from the rest, so we should use the range-testing equivalent of the following:

Code:
1:
 
if(!(y && (c4!=c8 && (y || x && (b || c5==c7)) || c5==c7 && c4!=c5 && r<=0.) || x && (c5==c7 && c4!=c5 || c5!=c7 && !a))) rValue = C4;

I believe this is the shortest way to write it, but if you can figure out a shorter way it'd be appreciated, as it's still rather messy...




guest ...
Real addict
.........

...

Status:Offline
Date registered: 30.07.2004
Post:856
Send Message
...   Created on 11.12.2007 - 22:09Jump to top Quote this post Report this post Edit Delete

Seems you found a version which fits on some NV cards.
But i like the logical transparence of the original algorithm better.




VerGreeneyes 
Strong supporter
......

...

Status:Offline
Date registered: 26.04.2007
Post:89
Send Message
...   Created on 12.12.2007 - 00:51Jump to top Quote this post Report this post Edit Delete

Well I can't say how it'll work on ATI cards, but it should be faster. (feedback would be appreciated!) To be honest I don't think the original version was any clearer than this one :P The advantage, I think, is that it has just one, independent case for each action, and seeing the patterns that lead to the booleans a and b was pretty cool. Of course I could convert it back to the previous form and it'd still be simpler, I think.




guest ...
Real addict
.........

...

Status:Offline
Date registered: 30.07.2004
Post:856
Send Message
...   Created on 12.12.2007 - 18:58Jump to top Quote this post Report this post Edit Delete

Naa, this works much slower on x1950. Like 185 vs. 135 FPS compared with your previous version. A simpler and slower logical/branch shader unit might be the cause.




VerGreeneyes 
Strong supporter
......

...

Status:Offline
Date registered: 26.04.2007
Post:89
Send Message
...   Created on 13.12.2007 - 00:55Jump to top Quote this post Report this post Edit Delete

Interesting. I'll see if I can convert it back to the form with more branching, see how it looks.




VerGreeneyes 
Strong supporter
......

...

Status:Offline
Date registered: 26.04.2007
Post:89
Send Message
...   Created on 14.12.2007 - 01:41Jump to top Quote this post Report this post Edit Delete

Hah, converting it back to the original structure I get more-or-less the same version as what you have on the front page. I did change it to make no unnecessary writes (C4) though, and I think it looks quite clean, so have a look ^_^

fragment file:

Code:
1:
2:
3:
4:
5:
6:
7:
8:
9:
10:
11:
12:
13:
14:
15:
16:
17:
18:
19:
20:
21:
22:
23:
24:
25:
26:
27:
28:
29:
30:
31:
32:
33:
34:
35:
36:
37:
38:
39:
40:
41:
42:
43:
44:
45:
46:
47:
48:
49:
50:
51:
52:
53:
54:
55:
56:
57:
58:
59:
60:
61:
62:
63:
64:
65:
66:
67:
68:
69:
70:
71:
72:
73:
74:
75:
76:
77:
78:
79:
80:
81:
82:
83:
84:
85:
86:
87:
 
/*
  2xSaI GLSL shader
           - Copyright (C) 2007 guest(r) - guest.r@gmail.com, Ver - ver.greeneyes@gmail.com
           - License: GNU-GPL

   The 2xSaI algorithm
           - Copyright (c) 1999-2001 by Derek Liauw Kie Fa.
*/

const vec4 dt = vec4(16777216.,65536.,256.,1.);
uniform vec4 OGL2Size,OGL2InvSize;
uniform sampler2D OGL2Texture;

void TWOxSaI(out vec4 colour);

void main()
{ TWOxSaI(gl_FragColor);
}

float GET_RESULT(float A, float B, float C, float D)
{ return float(A!=C && A!=D && B==C && B==D) - float(A==C && A==D);
}

float reduce(vec4 colour)
{ return dot(colour,dt);
}

void TWOxSaI(out vec4 colour)
{ vec2 OGL2Pos = gl_TexCoord[0].xy*OGL2Size.xy,
       fp = fract(OGL2Pos), dx, dy,
       pC4 = floor(OGL2Pos)/OGL2Size.xy;
  if(fp.x<.5 || fp.y>=.5)
  { dx = vec2(OGL2InvSize.x,0.);
    dy = vec2(0.,OGL2InvSize.y);
  }
  else
  { dx = vec2(0.,OGL2InvSize.y);
    dy = vec2(OGL2InvSize.x,0.);
  }
  vec4 C0 = texture2D(OGL2Texture,pC4-   dx-   dy),
       C1 = texture2D(OGL2Texture,pC4      -   dy),
       C2 = texture2D(OGL2Texture,pC4+   dx-   dy),
       C3 = texture2D(OGL2Texture,pC4-   dx      ),
       C4 = texture2D(OGL2Texture,pC4            ),
       C5 = texture2D(OGL2Texture,pC4+   dx      ),
       C6 = texture2D(OGL2Texture,pC4-   dx+   dy),
       C7 = texture2D(OGL2Texture,pC4      +   dy),
       C8 = texture2D(OGL2Texture,pC4+   dx+   dy),
       D0 = texture2D(OGL2Texture,pC4-   dx+2.*dy),
       D1 = texture2D(OGL2Texture,pC4      +2.*dy),
       D2 = texture2D(OGL2Texture,pC4+   dx+2.*dy),
       D4 = texture2D(OGL2Texture,pC4+2.*dx      ),
       D5 = texture2D(OGL2Texture,pC4+2.*dx+   dy);
  float c0 = reduce(C0),c1 = reduce(C1),c2 = reduce(C2),c3 = reduce(C3),
        c4 = reduce(C4),c5 = reduce(C5),c6 = reduce(C6),c7 = reduce(C7),
        c8 = reduce(C8),d0 = reduce(D0),d1 = reduce(D1),d2 = reduce(D2),
        d4 = reduce(D4),d5 = reduce(D5),r;
  bool x = fp.x<.5 && fp.y>=.5 || fp.x>=.5 && fp.y<.5;
  bool y = fp.x>=.5 && fp.y>=.5;

  colour = C4;
  if(c4==c8)
  { if(c5==c7)
    { if(c4!=c5)
      { if(x) colour = .5*(C4+C7);
        if(y)
        { r = GET_RESULT(c4,c5,c3,c1) + GET_RESULT(c4,c5,d5,d2)
            - GET_RESULT(c5,c4,d4,c2) - GET_RESULT(c5,c4,c6,d1);
          if(r< 0.) colour = C5;
          if(r==0.) colour = .25*(C4+C5+C7+C8);
        }
      }
    }else if(x && !(c3==c4 && c7==d2 || c4==c5 && c4==c6 && c3!=c7 && c7==d0)) colour = .5*(C4+C7);
  }else
  { if(c5==c7)
    { if(y) colour = C5;
      if(x) colour = (c2==c4 && c6==c7 || c3==c7 && c7==c8 && c4!=c6 && c0==c4) ? C7 : .5*(C4+C7);
    }else
    { if(y) colour = .25*(C4+C5+C7+C8);
      if(x)
      { if(c3==c7 && c7==c8 && c4!=c6 && c0==c4) colour = C7;
        else if(!(c4==c5 && c4==c6 && c3!=c7 && c7==d0)) colour = .5*(C4+C7);
      }
    }
  }
}


Edit: out of interest, could you compare the speed of the above to the speed of the following?

fragment file:
Code:
1:
2:
3:
4:
5:
6:
7:
8:
9:
10:
11:
12:
13:
14:
15:
16:
17:
18:
19:
20:
21:
22:
23:
24:
25:
26:
27:
28:
29:
30:
31:
32:
33:
34:
35:
36:
37:
38:
39:
40:
41:
42:
43:
44:
45:
46:
47:
48:
49:
50:
51:
52:
53:
54:
55:
56:
57:
58:
59:
60:
61:
62:
63:
64:
65:
66:
67:
68:
69:
70:
71:
72:
73:
74:
75:
76:
77:
78:
79:
80:
81:
82:
83:
84:
 
/*
  2xSaI GLSL shader
           - Copyright (C) 2007 guest(r) - guest.r@gmail.com, Ver - ver.greeneyes@gmail.com
           - License: GNU-GPL

   The 2xSaI algorithm
           - Copyright (c) 1999-2001 by Derek Liauw Kie Fa.
*/

const vec4 dt = vec4(16777216.,65536.,256.,1.);
uniform vec4 OGL2Size,OGL2InvSize;
uniform sampler2D OGL2Texture;

void TWOxSaI(out vec4 colour);

void main()
{ TWOxSaI(gl_FragColor);
}

float GET_RESULT(float A, float B, float C, float D)
{ return float(A!=C && A!=D && B==C && B==D) - float(A==C && A==D);
}

float reduce(vec4 colour)
{ return dot(colour,dt);
}

void TWOxSaI(out vec4 colour)
{ vec2 OGL2Pos = gl_TexCoord[0].xy*OGL2Size.xy,
       fp = fract(OGL2Pos), dx, dy,
       pC4 = floor(OGL2Pos)/OGL2Size.xy;
  if(fp.x<.5 || fp.y>=.5)
  { dx = vec2(OGL2InvSize.x,0.);
    dy = vec2(0.,OGL2InvSize.y);
  }
  else
  { dx = vec2(0.,OGL2InvSize.y);
    dy = vec2(OGL2InvSize.x,0.);
  }
  vec4 C0 = texture2D(OGL2Texture,pC4-   dx-   dy),
       C1 = texture2D(OGL2Texture,pC4      -   dy),
       C2 = texture2D(OGL2Texture,pC4+   dx-   dy),
       C3 = texture2D(OGL2Texture,pC4-   dx      ),
       C4 = texture2D(OGL2Texture,pC4            ),
       C5 = texture2D(OGL2Texture,pC4+   dx      ),
       C6 = texture2D(OGL2Texture,pC4-   dx+   dy),
       C7 = texture2D(OGL2Texture,pC4      +   dy),
       C8 = texture2D(OGL2Texture,pC4+   dx+   dy),
       D0 = texture2D(OGL2Texture,pC4-   dx+2.*dy),
       D1 = texture2D(OGL2Texture,pC4      +2.*dy),
       D2 = texture2D(OGL2Texture,pC4+   dx+2.*dy),
       D4 = texture2D(OGL2Texture,pC4+2.*dx      ),
       D5 = texture2D(OGL2Texture,pC4+2.*dx+   dy);
  float c0 = reduce(C0),c1 = reduce(C1),c2 = reduce(C2),c3 = reduce(C3),
        c4 = reduce(C4),c5 = reduce(C5),c6 = reduce(C6),c7 = reduce(C7),
        c8 = reduce(C8),d0 = reduce(D0),d1 = reduce(D1),d2 = reduce(D2),
        d4 = reduce(D4),d5 = reduce(D5),r;
  bool x = fp.x<.5 && fp.y>=.5 || fp.x>=.5 && fp.y<.5;
  bool y = fp.x>=.5 && fp.y>=.5;

  colour = C4;
  if(c4==c8 && c5==c7 && c4!=c5)
  { if(x) colour = .5*(C4+C7);
    if(y)
    { r = GET_RESULT(c4,c5,c3,c1) + GET_RESULT(c4,c5,d5,d2)
        - GET_RESULT(c5,c4,d4,c2) - GET_RESULT(c5,c4,c6,d1);
      if(r< 0.) colour = C5;
      if(r==0.) colour = .25*(C4+C5+C7+C8);
    }
  }
  if(c4==c8 && c5!=c7 && x && !(c3==c4 && c7==d2 || c4==c5 && c4==c6 && c3!=c7 && c7==d0)) colour = .5*(C4+C7);
  if(c4!=c8 && c5==c7)
  { if(y) colour = C5;
    if(x) colour = (c2==c4 && c6==c7 || c3==c7 && c7==c8 && c4!=c6 && c0==c4) ? C7 : .5*(C4+C7);
  }
  if(c4!=c8 && c5!=c7)
  { if(y) colour = .25*(C4+C5+C7+C8);
    if(x)
    { if(c3==c7 && c7==c8 && c4!=c6 && c0==c4) colour = C7;
      else if(!(c4==c5 && c4==c6 && c3!=c7 && c7==d0)) colour = .5*(C4+C7);
    }
  }
}


[Dieser Beitrag wurde am 14.12.2007 - 01:49 von VerGreeneyes aktualisiert]




More : [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11]

Similarly threads:
Topics Created by Replies Boardname
Cartoon GLSL shader guest 7 pete_bernert
Is there a OGL2 2xSaI Shader out there? GreenImp 0 pete_bernert
Scale2x Plus GLSL shader guest 0 pete_bernert
SuperEagle GLSL shader guest 0 pete_bernert
Super2xSaI GLSL shader guest 0 pete_bernert
Neuer Thread ...





Masthead

This forum is a free service of razyboard.com powered by:
Geizkragen Price Comparison. Top product in the price comparison: Krups Nespresso Essenza (XN2001)
Do you want a free forum in less than two minutes? Then click here!



Verwandte Suchbegriffe:
2xsai shader | glsl algorithm | 2xsai algorithm | www.c7d0 | scale2x shader | scale2x glsl | glsl shader code | glsl branch texture2d
blank