livio

I'm getting in touch with the "how-to-have-my-render-back" matter. Avoiding the sketch method because of poor performance, I'm testing the "gl.render (to_texture) > gl.texture > jit.matrix" method and, pushing forward, the "gl.texture (begin_capture)" method.

(that I hacked from jit.gl.render.radialblur.pat)

- Why do I have to interpose a jit.gl.slab to get my texture pass to jit.matrix?

- The begin_capture make the gl.texture output the framebuffer, but there are digital glitches (not that bad though). Do I miss something?

hello!
I'm getting in touch with the "how-to-have-my-render-back" matter. Avoiding the sketch method because of poor performance, I'm testing the "gl.render (to_texture) > gl.texture > jit.matrix" method and, pushing forward, the "gl.texture (begin_capture)" method.

Pls check the patch here after 
(that I hacked from jit.gl.render.radialblur.pat)

Here's the questions:
- Why do I have to interpose a jit.gl.slab to get my texture pass to jit.matrix?
- The begin_capture make the gl.texture output the framebuffer, but there are digital glitches (not that bad though). Do I miss something? 

to_texture-capture-misteries

I wouldn't go this route. begin_capture/end_capture will cause

headaches from a aptcher because of timing. It's poossible to use,

but I wouldn't recommend it. Other object might interfere and cause

errors as well. What's wrong with the sketch approach exactly? Why

do you say it's slow? It's the best patcher method for RTT IMHO.

> I'm getting in touch with the "how-to-have-my-render-back" matter. Avoiding the sketch method because of poor performance, I'm testing the "gl.render (to_texture) > gl.texture > jit.matrix" method and, pushing forward, the "gl.texture (begin_capture)" method.

> (that I hacked from jit.gl.render.radialblur.pat)

> - Why do I have to interpose a jit.gl.slab to get my texture pass to jit.matrix?

> - The begin_capture make the gl.texture output the framebuffer, but there are digital glitches (not that bad though). Do I miss something?

> #P hidden message 332 103 23 196617 0.4;

> #P user jit.pwindow 887 328 130 130 0 1 0 0 1 0;

> #P newex 888 307 161 196617 jit.matrix 4 char 256 256;

> #P message 1345 191 65 196617 end_capture;

> #P message 1322 171 73 196617 begin_capture;

> #P user jit.pwindow 1321 334 130 130 0 1 0 0 1 0;

> #P newex 1322 313 129 196617 jit.matrix 4 char 256 256;

> #P newex 1322 270 85 196617 jit.gl.slab blurry;

> #P newex 1322 217 231 196617 jit.gl.texture blurry @name other @dim 256 256;

> #P user jit.pwindow 1072 328 130 130 0 1 0 0 1 0;

> #P newex 1073 307 129 196617 jit.matrix 4 char 256 256;

> #P newex 1073 264 85 196617 jit.gl.slab blurry;

> #P message 373 229 189 196617 usetexture feedback , to_texture , erase;

> #P newex 1073 217 246 196617 jit.gl.texture blurry @name feedback @dim 256 256;

> #P hidden newex 271 123 50 196617 t b b;

> #P comment 789 626 100 196617 render to the named target , in this case the pwindow.;

> #P comment 65 624 379 196617 init stuff for jit.render: set draw colors and create named textures. the feedback texture is just a little smaller than the pwindow draw target.;

> #P flonum 373 333 35 9 0 0 0 3 0 0 0 221 221 221 222 222 222 0 0 0;

I wouldn't go this route.  begin_capture/end_capture will cause
headaches from a aptcher because of timing.  It's poossible to use,
but I wouldn't recommend it.  Other object might interfere and cause
errors as well.  What's wrong with the sketch approach exactly?  Why
do you say it's slow? It's the best patcher method for RTT IMHO.

On 5/10/07, livio  wrote:
>
> hello!
> I'm getting in touch with the "how-to-have-my-render-back" matter. Avoiding the sketch method because of poor performance, I'm testing the "gl.render (to_texture) > gl.texture > jit.matrix" method and, pushing forward, the "gl.texture (begin_capture)" method.
>
> Pls check the patch here after
> (that I hacked from jit.gl.render.radialblur.pat)
>
> Here's the questions:
> - Why do I have to interpose a jit.gl.slab to get my texture pass to jit.matrix?
> - The begin_capture make the gl.texture output the framebuffer, but there are digital glitches (not that bad though). Do I miss something?
>
> Cheers,
> Giorgio
>
> --
> oivil.eu
>
>
> #P window setfont "Sans Serif" 20.;
> #P window linecount 1;
> #P comment 1205 428 21 196628 !;
> #P comment 1017 428 21 196628 ?;
> #P comment 1454 432 21 196628 ?;
> #P comment 1288 166 21 196628 >;
> #P window setfont "Sans Serif" 9.;
> #P hidden message 332 103 23 196617 0.4;
> #P user jit.pwindow 887 328 130 130 0 1 0 0 1 0;
> #P newex 888 307 161 196617 jit.matrix 4 char 256 256;
> #P message 1345 191 65 196617 end_capture;
> #P message 1322 171 73 196617 begin_capture;
> #P user jit.pwindow 1321 334 130 130 0 1 0 0 1 0;
> #P newex 1322 313 129 196617 jit.matrix 4 char 256 256;
> #P newex 1322 270 85 196617 jit.gl.slab blurry;
> #P newex 1322 217 231 196617 jit.gl.texture blurry @name other @dim 256 256;
> #P user jit.pwindow 1072 328 130 130 0 1 0 0 1 0;
> #P newex 1073 307 129 196617 jit.matrix 4 char 256 256;
> #P newex 1073 264 85 196617 jit.gl.slab blurry;
> #P message 373 229 189 196617 usetexture feedback , to_texture , erase;
> #P newex 1073 217 246 196617 jit.gl.texture blurry @name feedback @dim 256 256;
> #P newex 871 694 50 196617 route init;
> #P hidden newex 271 123 50 196617 t b b;
> #P window linecount 3;
> #P comment 789 626 100 196617 render to the named target , in this case the pwindow.;
> #P window linecount 2;
> #P comment 65 624 379 196617 init stuff for jit.render: set draw colors and create named textures. the feedback texture is just a little smaller than the pwindow draw target.;
> #P window linecount 1;
> #P message 373 576 50 196617 bang;
> #P hidden message 74 563 14 196617 1;
> #P flonum 373 333 35 9 0 0 0 3 0 0 0 221 221 221 222 222 222 0 0 0;
> #P message 373 353 97 196617 rotate $1 0. 0. 1.;
> #P message 373 303 99 196617 quads;
> #P message 373 499 99 196617 tri_grid;
> #P message 373 276 126 196617 lookat 0 0 0 , camera 0 0 2;
> #P window linecount 3;
> #P message 373 433 129 196617 lookat 0 0 0 , camera 2 2 2 , rotate 0. 1. 1. 1. , usetexture movietexture;
> #P window linecount 5;
> #P hidden message 570 91 98 196617 plane 0 , -1 1 1 -1 , plane 1 , 1 1 -1 -1 , plane 2 , 1 1 1 1 , plane 3 , 0. 1. 1. 0. , plane 4 , 1. 1. 0. 0.;
> #P window linecount 2;
> #P newex 373 384 98 196617 jit.matrix square 5 float32 1 4;
> #P window linecount 1;
> #P hidden newex 570 160 98 196617 jit.fill square;
> #P message 55 276 140 196617 name blurry , size 256 256;
> #P user jit.pwindow 54 296 258 258 0 1 0 0 1 0;
> #X name blurry;
> #N vpatcher 911 83 1697 508;
> #P window setfont "Sans Serif" 9.;
> #P window linecount 1;
> #P newex 620 141 98 196617 jit.fill small5;
> #P window linecount 2;
> #P newex 611 215 98 196617 jit.matrix small5 5 float32 2 2;
> #P newex 502 215 98 196617 jit.matrix small4 5 float32 2 2;
> #P newex 393 215 98 196617 jit.matrix small3 5 float32 2 2;
> #P newex 284 215 98 196617 jit.matrix small2 5 float32 2 2;
> #P newex 175 215 98 196617 jit.matrix small1 5 float32 2 2;
> #P window linecount 1;
> #P newex 368 31 50 196617 loadbang;
> #P newex 516 141 98 196617 jit.fill small4;
> #P newex 412 141 98 196617 jit.fill small3;
> #P newex 308 141 98 196617 jit.fill small2;
> #P newex 204 141 98 196617 jit.fill small1;
> #P outlet 66 349 15 0;
> #P inlet 66 46 15 0;
> #P window linecount 5;
> #P message 620 72 101 196617 plane 2 , -1 1 -1 1 , plane 0 , 1 1 -1 -1 , plane 1 , -1 -1 -1 -1 , plane 3 , 0. 1. 0. 1. , plane 4 , 0. 0. 1. 1.;
> #P message 516 72 98 196617 plane 2 , -1 1 -1 1 , plane 0 , 1 1 -1 -1 , plane 1 , 1 1 1 1 , plane 3 , 0. 1. 0. 1. , plane 4 , 0. 0. 1. 1.;
> #P message 412 72 101 196617 plane 2 , -1 1 -1 1 , plane 1 , 1 1 -1 -1 , plane 0 , -1 -1 -1 -1 , plane 3 , 0. 1. 0. 1. , plane 4 , 0. 0. 1. 1.;
> #P message 308 72 98 196617 plane 2 , -1 1 -1 1 , plane 1 , 1 1 -1 -1 , plane 0 , 1 1 1 1 , plane 3 , 0. 1. 0. 1. , plane 4 , 0. 0. 1. 1.;
> #P window linecount 1;
> #P newex 66 179 555 196617 t b b b b b b;
> #P window linecount 5;
> #P message 204 72 101 196617 plane 0 , -1 1 -1 1 , plane 1 , 1 1 -1 -1 , plane 2 , -1 -1 -1 -1 , plane 3 , 0. 1. 0. 1. , plane 4 , 0. 0. 1. 1.;
> #P message 100 72 98 196617 plane 0 , -1 1 -1 1 , plane 1 , 1 1 -1 -1 , plane 2 , 1 1 1 1 , plane 3 , 0. 1. 0. 1. , plane 4 , 0. 0. 1. 1.;
> #P window linecount 2;
> #P newex 66 215 98 196617 jit.matrix small0 5 float32 2 2;
> #P window linecount 1;
> #P newex 100 141 98 196617 jit.fill small0;
> #P window linecount 3;
> #P newex 66 270 99 196617 jit.matrix geom 5 float32 8 8 @interp 1 @thru 1;
> #P window linecount 7;
> #P comment 1 101 79 196617 a small 2x2x5 matrix is created "by hand" , then interpolated to create geometry.;
> #P connect 11 0 6 0;
> #P connect 6 0 3 0;
> #P fasten 22 0 1 0 616 253 71 253;
> #P fasten 21 0 1 0 507 253 71 253;
> #P fasten 20 0 1 0 398 253 71 253;
> #P fasten 19 0 1 0 289 253 71 253;
> #P fasten 18 0 1 0 180 253 71 253;
> #P fasten 3 0 1 0 71 253 71 253;
> #P connect 1 0 12 0;
> #P connect 17 0 4 0;
> #P fasten 4 0 2 0 105 146 105 146;
> #P connect 17 0 5 0;
> #P connect 5 0 13 0;
> #P connect 6 2 19 0;
> #P connect 17 0 7 0;
> #P connect 7 0 14 0;
> #P connect 17 0 8 0;
> #P connect 8 0 15 0;
> #P connect 6 4 21 0;
> #P connect 17 0 9 0;
> #P connect 9 0 16 0;
> #P connect 17 0 10 0;
> #P connect 10 0 23 0;
> #P pop;
> #P newobj 373 539 67 196617 p three_sides;
> #P hidden message 150 563 50 196617 89.;
> #P hidden newex 299 72 50 196617 loadbang;
> #P toggle 56 563 15 0;
> #P message 56 582 50 196617 interp $1;
> #P message 111 583 70 196617 lens_angle $1;
> #P flonum 110 563 35 9 0 0 0 3 0 0 0 221 221 221 222 222 222 0 0 0;
> #P newex 373 147 144 196617 prepend texture movietexture;
> #P newex 54 157 162 196617 jit.matrix 4 char 64 64 @interp 1;
> #P toggle 54 71 15 0;
> #P newex 54 91 55 196617 metro 50;
> #P newex 54 127 109 196617 jit.noise 4 char 64 64;
> #P newex 54 199 183 196617 t b b b b b b b b b;
> #P message 56 649 399 196617 color 1. 1. 1. 1. , erase_color 0. 0. 0. 1. , depth_enable 0 , texture movietexture 64 64;
> #P newex 772 671 109 196617 jit.gl.render blurry;
> #P window setfont "Sans Serif" 20.;
> #P comment 1423 166 21 196628
> #P connect 6 0 5 0;
> #P fasten 5 0 4 0 59 118 59 118;
> #P fasten 4 0 7 0 59 150 59 150;
> #P fasten 7 0 3 0 59 195 59 195;
> #P hidden connect 31 1 17 0;
> #P connect 17 0 16 0;
> #P hidden connect 27 0 12 0;
> #P connect 12 0 11 0;
> #P fasten 32 0 2 0 876 719 41 719 41 642 61 642;
> #P hidden connect 13 0 27 0;
> #P hidden connect 14 0 9 0;
> #P connect 9 0 10 0;
> #P hidden connect 13 0 14 0;
> #P hidden connect 13 0 31 0;
> #P hidden connect 13 0 46 0;
> #P fasten 7 0 8 0 59 186 363 186 363 137 378 137;
> #P fasten 3 8 34 0 227 224 378 224;
> #P fasten 3 7 22 0 206 261 378 261;
> #P fasten 3 6 24 0 185 234 358 234 358 300 378 300;
> #P hidden connect 46 0 26 0;
> #P fasten 3 5 26 0 164 239 356 239 356 325 378 325;
> #P connect 26 0 25 0;
> #P fasten 3 4 19 0 143 242 354 242 354 380 378 380;
> #P fasten 3 3 21 0 122 270 378 270;
> #P fasten 3 2 23 0 101 251 378 251;
> #P fasten 3 1 15 0 80 256 348 256 348 531 378 531;
> #P fasten 3 0 28 0 59 263 346 263 346 567 378 567;
> #P hidden connect 13 0 20 0;
> #P hidden fasten 20 0 18 0 575 165 575 165;
> #P fasten 19 0 1 0 378 417 777 417;
> #P fasten 34 0 1 0 378 249 777 249;
> #P fasten 28 0 1 0 378 598 777 598;
> #P fasten 25 0 1 0 378 373 777 373;
> #P fasten 24 0 1 0 378 322 777 322;
> #P fasten 23 0 1 0 378 522 777 522;
> #P fasten 22 0 1 0 378 297 777 297;
> #P fasten 21 0 1 0 378 475 777 475;
> #P fasten 15 0 1 0 378 561 777 561;
> #P hidden fasten 8 0 1 0 378 170 777 170;
> #P fasten 10 0 1 0 116 603 777 603;
> #P fasten 2 0 1 0 61 667 777 667;
> #P fasten 11 0 1 0 61 606 777 606;
> #P connect 1 1 32 0;
> #P connect 33 0 44 0;
> #P connect 44 0 45 0;
> #P connect 3 7 33 0;
> #P connect 33 0 35 0;
> #P connect 35 0 36 0;
> #P connect 36 0 37 0;
> #P connect 3 7 38 0;
> #P connect 42 0 38 0;
> #P connect 43 0 38 0;
> #P connect 38 0 39 0;
> #P connect 39 0 40 0;
> #P connect 40 0 41 0;
> #P window clipboard copycount 51;
>
>


On May 10, 2007, at 9:20 AM, Wesley Smith wrote:

> What's wrong with the sketch approach exactly? Why

> do you say it's slow? It's the best patcher method for RTT IMHO.

I think he means readpixels strategy, rather than using sketch to 

On May 10, 2007, at 9:20 AM, Wesley Smith wrote:
> What's wrong with the sketch approach exactly?  Why
> do you say it's slow? It's the best patcher method for RTT IMHO.

I think he means readpixels strategy, rather than using sketch to  
sequence the capture state.

It's good to know there someone listening out there..

Yes, I meant the sketch glreadpixels strategy. Nothing against the gl.sketch objet itself :)

Ok let's say the begin_capture won't never work.. I can't see why but let's go back to the to_texture trick.

It works now at different resoultion, for me the solution was to instantiate the context window at the same size of the texture and the matrix that will read it back. 

Still I don't understand why I should put a jit.slab in between jit.texture and jit.matrix..

Btw, any hint about depth sorting? I found this comment in the forum archive :

"If you still want tot retain depth cues with depth_enable = 0 and

blend_enable = 1, you have to do back face culling and depth sorting when you draw. i.e. draw from the back forward."

Hello,
It's good to know there someone listening out there..
Yes, I meant the sketch glreadpixels strategy. Nothing against the gl.sketch objet itself :)
Ok let's say the begin_capture won't never work.. I can't see why but let's go back to the to_texture trick.
It works now at different resoultion, for me the solution was to instantiate the context window at the same size of the texture and the matrix that will read it back. 
Still I don't understand why I should put a jit.slab in between  jit.texture and jit.matrix..
Btw, any hint about depth sorting? I found this comment in the forum archive :
"If you still want tot retain depth cues with depth_enable = 0 and
blend_enable = 1, you have to do back face culling and depth sorting when you draw. i.e. draw from the back forward."
how could this be structured?
Thanks a lot.
Cordially,
giorgio


to_texture & capture misteries