1&fc convert j double number to single precision bit pattern because single precision is used in kernels. Double precision is not well supported in low-end GPU. If 1&fc is omitted, GPU will interpret the buffer of double precision as single precision bit patterns.
On Fri, Feb 28, 2020, 9:52 PM Brian Schott <schott.br...@gmail.com> wrote: > Bill, especially (Thomas, Raul, and others), > > Your comment regarding trying a simpler figure and simplifying the color > has led me to consider several alternatives. Below I show a revised script > which uses a tetrahedron (4 faces) and some color changes. I think the > tetrahedron is mostly working for me but I am really confused about the > colors. > > I apologize for the length of this email, but I have many questions. > > As an aside, I have discovered an error in the shader.ijs in the qtdemo's. > The last 0 in the following line should be a 1 because it's the alpha > value. (Coincidentally, there is another line in shader.ijs which also > addresses ClearColor, but it is ignored, I think, and it separates the > integers with ;'s .) Correcting that alpha value creates a black > background, which can be effective. > > glClearColor 0 0 0 0 > > Btw, I have left 2 glClearColor's in the script so it's easy to swap them: > white 1 1 1 1 and > black 0 0 0 1 . > > The major changes I have made to the script are the following. > > NB. changed cube to a mirrored sphenoid(?) > NB. (eg, see > https://en.wikipedia.org/wiki/Architectonic_and_catoptric_tessellation) > > rt=: (|.@:{)`[`]} NB. rotates vertices of xth face for winding order > vertexData=: 1&fc ,1 rt 3 rt 4 3 3$ %&10, 0&". ;._2 [ 0 : 0 > _3 _9 _4 > 3 _9 _4 > 0 0 2 > 0 8 _4 > ) > > I discovered that the verb `rt` (or something like it) is required to make > the tetrahedron work because of face winding issues. The original cube was > adjusted another way, apparently. > With `vertexData` defined either as above OR without `1&fc` prepended I get > the SAME --mostly desirable-- results. The inclusion and exclusion of > `1&fc` giving the same result is confusing, by itself, and begs the > question, "Why is `1&fc` used?" > > I said "mostly desirable" above, instead of "desirable" because of the next > 2 issues. > > v0) The 10 in the first line of the definition of vertexData is used to > simplify data entry and I have also used it to experiment with scaling the > size of the tetrahedron. This has led to another surprise, that no scaling > seems to result from increasing the 10 to 80 or so. Why? > > v1) I have been disillusioned by the face coloring of the tetrahedron, even > in the simplest situations because when the tetrahedron is rotated in > certain directions, the black or white background seems to bleed through > one of the edges. What causes that? > > Now, moving on to coloring. I have no idea how the face COLORing works > using almost any version of `colorData` below. I can generalize though on > one thing: if I leave out the `1&fc`, no colors work on ANY version > of `colorData` . > > The first version below seems to give a complete tetrahedron except for the > background color bleeding through one edge. And as mentioned earlier, it > only works with vertices rotated with the rt verb. > > colorData=: 1&fc ,1 rt 3 rt 4 3 3$ , 0&". ;._2 [ 0 : 0 > 0 0 1 > 1 1 1 > 1 0 0 > 0 1 0 > ) > > f0) If the 1's in the second row are changed to 0.9's, the background color > no longer bleeds in. Why? > > f1) (This approach does not seem to work at all, and I have a replacement > for it that works a little, but will send it in a new message) I want 2 of > the faces in the tetrahedron to have the same color and each of the other 2 > to have a different color (3 colors altogether). I devised the following > NONWORKING scheme of colors to accomplish this result. By nonworking I mean > only the background color shows in the graphics pane, no tetrahedron. Why? > > colorData=: 1&fc ,_3]\3#"2] 0&". ;._2 [ 0 : 0 > 0 0 1 > 1 1 1 > 1 0 0 > 0 1 0 > ) > > To demonstrate the result above consider the following result which makes > each of the 4 faces, represented by the boxes' contents below, has 3 > identical rows. > ;/_3]\3#"2] 0&". ;._2 [ 0 : 0 > 0 0 1 > 1 1 1 > 1 0 0 > 0 1 0 > ) > +-----+-----+-----+-----+ > |0 0 1|1 1 1|1 0 0|0 1 0| > |0 0 1|1 1 1|1 0 0|0 1 0| > |0 0 1|1 1 1|1 0 0|0 1 0| > +-----+-----+-----+-----+ > > > > ***************my revised slider.ijs script is below************ > > cocurrent 'demoshader' > > mp=: (+/ . *)"2 > rt=: (|.@:{)`[`]}NB. rotates vertices of xth face for winding order > sprog=: 0 > GLSL=: 0 > STOP=: 0 > > A=: 0 : 0 > pc a; > minwh 300 300;cc g opengl flush; > rem form end; > ) > > a_run=: 3 : 0 > NB. if. -. checkrequire_qtdemo_ 'gles';'api/gles' do. return. end. > require 'gl2' > coinsert 'jgl2' > require 'gles' > coinsert 'jgles' > require '~addons/ide/qt/opengl.ijs' > coinsert 'qtopengl' > STEPS=: 100 > R=: 20 30 0 > EYE=: 0 0 1 > LR=: UD=: IO=: 0 > UP=: 0 1 0 > wd A > HD=: ".wd'qhwndc g' > wd 'ptimer 100' > wd 'pshow' > ) > > a_g_initialize=: 3 : 0 > if. p=. >@{. glGetString GL_VERSION do. smoutput 'GL_VERSION: ', memr 0 _1 > 2,~ p end. > if. 0=p do. smoutput 'cannot retrieve GL_VERSION' return. end. > if. p=. >@{. glGetString GL_VENDOR do. smoutput 'GL_VENDOR: ', memr 0 _1 > 2,~ p end. > if. p=. >@{. glGetString GL_RENDERER do. smoutput 'GL_RENDERER: ', memr 0 > _1 2,~ p end. > if. p=. >@{. glGetString GL_SHADING_LANGUAGE_VERSION do. smoutput > 'GL_SHADING_LANGUAGE_VERSION: ', memr 0 _1 2,~ p end. > GLSL=: wglGLSL'' > > wglPROC'' > sprog=: 0 > if. GLSL>120 do. > vsrc=. vsrc2 > fsrc=. fsrc2 > else. > vsrc=. vsrc1 > fsrc=. fsrc1 > if. 0=GLES_VERSION_jgles_ do. > vsrc=. vsrc,~ '#define lowp', LF, '#define mediump', LF, '#define > highp', LF > fsrc=. fsrc,~ '#define lowp', LF, '#define mediump', LF, '#define > highp', LF > end. > end. > vsrc=. '#version ',(":GLSL),((GLSL>:300)#(*GLES_VERSION){::' core';' > es'),LF,vsrc > fsrc=. '#version ',(":GLSL),((GLSL>:300)#(*GLES_VERSION){::' core';' > es'),LF,fsrc > if.(GLSL>:300)*.0~:GLES_VERSION_jgles_ do. > fsrc=. ('void main';'out vec4 gl_FragColor;',LF,'void main') > stringreplace fsrc > end. > smoutput vsrc > smoutput fsrc > 'err program'=. gl_makeprogram vsrc;fsrc > if. #err do. smoutput err return. end. > > vertexAttr=: >@{. glGetAttribLocation program;'vertex' > assert. _1~: vertexAttr > colorAttr=: >@{. glGetAttribLocation program;'color' > assert. _1~: colorAttr > mvpUni=: >@{. glGetUniformLocation program;'mvp' > assert. _1~: mvpUni > > glGenBuffers 2;vbo=: 2#_1 > glBindBuffer GL_ARRAY_BUFFER; {.vbo > glBufferData GL_ARRAY_BUFFER; (#vertexData); (symdat <'vertexData'); > GL_STATIC_DRAW > glBindBuffer GL_ARRAY_BUFFER; {:vbo > glBufferData GL_ARRAY_BUFFER; (#colorData); (symdat <'colorData'); > GL_STATIC_DRAW > glBindBuffer GL_ARRAY_BUFFER; 0 > > sprog=: program > > NB. glClearColor 0; 0; 1; 1 > ) > > a_g_char=: 3 : 0 > if. 0=#sysdata do. return. end. > R=: 360 | R + 2 * 'xyz' = 0 { sysdata > k=. 0{sysdata > STEPS=: 200 <. STEPS + 's' = k > STEPS=: 3 >. STEPS - 'a' = k > LR=: LR - 0.01*'l'=k > LR=: LR + 0.01*'r'=k > gl_paint '' > ) > > a_timer=: 3 : 0 > if. 0=STOP do. return. end. > R=: 360 | R + 2 * 1 1 1 > gl_sel HD > gl_paint'' > ) > > a_g_paint=: 3 : 0 > if. 0=sprog do. return. end. > > wh=. gl_qwh'' > glClearColor 1 1 1 1 > glClearColor 0 0 0 1 NB. this is new > glClear GL_COLOR_BUFFER_BIT + GL_DEPTH_BUFFER_BIT > > glUseProgram sprog > glEnable GL_DEPTH_TEST > glEnable GL_CULL_FACE > > NB. matrix convention: current matrix on the left > NB. note pre-multiplication > > NB. model-view > mvp=: (gl_Rotate (0{R), 1 0 0 ) mp (gl_Rotate (1{R), 0 1 0) mp (gl_Rotate > (2{R), 0 0 1) mp (gl_Scale STEPS%100) mp (gl_Translate 0 0 _8) mp > glu_LookAt EYE,LR,UD,IO,UP > > NB. projection > mvp=: mvp mp gl_Perspective 30, (%/wh),1 10 > > NB. note GL_FALSE, no transpose > glUniformMatrix4fv mvpUni; 1; GL_FALSE; mvp > > glBindBuffer GL_ARRAY_BUFFER; {.vbo > glEnableVertexAttribArray vertexAttr > glVertexAttribPointer vertexAttr; 3; GL_FLOAT; 0; 0; 0 > > glBindBuffer GL_ARRAY_BUFFER; {:vbo > glEnableVertexAttribArray colorAttr > glVertexAttribPointer colorAttr; 3; GL_FLOAT; 0; 0; 0 > > glDrawArrays GL_TRIANGLES; 0; 12 NB. 36 was for cubes ******* > > glBindBuffer GL_ARRAY_BUFFER; 0 > glDisableVertexAttribArray colorAttr > glDisableVertexAttribArray vertexAttr > > glDisable GL_DEPTH_TEST > glDisable GL_CULL_FACE > > glUseProgram 0 > > gl_clear '' > gl_rgb 255 255 255 > gl_textcolor '' > gl_textxy 10 30 > gl_text 'keys: x y z a s l r F4' > gl_textxy 10 50 > gl_text 'scale: ',":STEPS%100 > gl_textxy 10 70 > gl_text 'angle: ',":R > gl_textxy 10 90 > if. 0=sprog do. return. end. > gl_text 'matrix:' > for_i. i.4 do. > gl_textxy 10, 105+i*15 > gl_text 6j2": i{mvp > end. > ) > > a_f4_fkey=: 3 : 0 > STOP=: -.STOP > NB. gl_paint '' > ) > > a_cancel=: a_close > > a_close=: 3 : 0 > STOP=: 0 > wd 'ptimer 0' > glDeleteBuffers ::0: 2; vbo > glDeleteProgram ::0: sprog > wd 'pclose' > ) > > NB. Mirrored sphenoid? > vertexData=: 1&fc ,1 rt 3 rt 4 3 3$ %&80, 0&". ;._2 [ 0 : 0 > _3 _9 _4 > 3 _9 _4 > 0 0 2 > 0 8 _4 > ) > > colorData=: 1&fc ,1 rt 3 rt 4 3 3$ , 0&". ;._2 [ 0 : 0 > 0 0 1 > 1 1 1 > 1 0 0 > 0 1 0 > ) > > > NB. ========================================================= > vsrc1=: 0 : 0 > attribute highp vec3 vertex; > attribute lowp vec3 color; > varying lowp vec4 v_color; > uniform mat4 mvp; > void main(void) > { > gl_Position = mvp * vec4(vertex,1.0); > v_color = vec4(color,1.0); > } > ) > > fsrc1=: 0 : 0 > varying lowp vec4 v_color; > void main(void) > { > gl_FragColor = v_color; > } > ) > > NB. ========================================================= > vsrc2=: 0 : 0 > in vec3 vertex; > in vec3 color; > out vec4 v_color; > uniform mat4 mvp; > void main(void) > { > gl_Position = mvp * vec4(vertex,1.0); > v_color = vec4(color,1.0); > } > ) > > fsrc2=: 0 : 0 > in vec4 v_color; > void main(void) > { > gl_FragColor = v_color; > } > ) > > NB. ========================================================= > a_run'' > > On Fri, Feb 14, 2020 at 7:19 PM bill lam <bbill....@gmail.com> wrote: > > > If you intended to run turtle graphics only, then drawing cube is over > > skilled. Drawing triangles is much simpler. eg a triangle with vertices > > 0 1 0, 1 _1 0, _1 _1 0 > > then use different scale, translate and rotate to place them into > different > > positions and orientations. > > > > Also only a constant solid color from fragment shader is sufficient. > > > > On Sat, Feb 15, 2020, 7:19 AM Brian Schott <schott.br...@gmail.com> > wrote: > > > > -- > (B=) > ---------------------------------------------------------------------- > For information about J forums see http://www.jsoftware.com/forums.htm > ---------------------------------------------------------------------- For information about J forums see http://www.jsoftware.com/forums.htm
