<div dir="ltr">Yes, the texture sample is important. It may be affecting the timing of the uniform loads via sampler.<div> </div><div>If I remove use of the texture, or replace it with vec3(1.0), the problem is very hard to detect. </div><div> </div><div>I'm using white to make it very clear where the incorrect results creep in.</div><div> </div><div>That being said, I'm sure there are other ways to get the bug to show up..</div><div> </div><div>-C</div></div><div class="gmail_extra"> <div class="gmail_quote">On Tue, Jul 1, 2014 at 3:36 PM, Chris Forbes <<a href="mailto:chrisf@ijw.co.nz" target="_blank">chrisf@ijw.co.nz</a>> wrote: <blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">The texture actually seems unnecessary -- are you using that to defeat some optimization? <div><div class="h5"> On Wed, Jul 2, 2014 at 9:28 AM, Cody Northrop <<a href="mailto:cody@lunarg.com">cody@lunarg.com</a>> wrote: > On Tue, Jul 1, 2014 at 3:05 PM, Kenneth Graunke <<a href="mailto:kenneth@whitecape.org">kenneth@whitecape.org</a>> > wrote: >> >> On Tuesday, July 01, 2014 12:53:18 PM Cody Northrop wrote: >> > Add a test that ensures fragment discard while loading from >> > uniform buffer objects works correctly. At time of submission, >> > this test fails on i965. >> > >> > Test will work with the following patch, sent to mesa-dev: >> > >> > i965/fs: Update discard jump to preserve uniform loads via sampler. >> > >> > Signed-off-by: Cody Northrop <<a href="mailto:cody@lunarg.com">cody@lunarg.com</a>> >> > Reviewed-by: Courtney Goeltzenleuchter <<a href="mailto:courtney@lunarg.com">courtney@lunarg.com</a>> >> > --- >> > tests/all.py | 1 + >> > .../arb_uniform_buffer_object/CMakeLists.gl.txt | 1 + >> > .../arb_uniform_buffer_object/rendering-discard.c | 195 >> +++++++++++++++++++++ >> > 3 files changed, 197 insertions(+) >> > create mode 100644 >> > tests/spec/arb_uniform_buffer_object/rendering-discard.c >> > >> > diff --git a/tests/all.py b/tests/all.py >> > index 17d5d9b..04ae9e5 100644 >> > --- a/tests/all.py >> > +++ b/tests/all.py >> > @@ -2965,6 +2965,7 @@ arb_uniform_buffer_object['negative- >> getactiveuniformblockiv'] = concurrent_test( >> > arb_uniform_buffer_object['negative-getactiveuniformsiv'] = >> concurrent_test('arb_uniform_buffer_object-negative-getactiveuniformsiv') >> > arb_uniform_buffer_object['referenced-by-shader'] = >> concurrent_test('arb_uniform_buffer_object-referenced-by-shader') >> > arb_uniform_buffer_object['rendering'] = >> concurrent_test('arb_uniform_buffer_object-rendering') >> > +arb_uniform_buffer_object['rendering-discard'] = >> concurrent_test('arb_uniform_buffer_object-rendering-discard') >> > arb_uniform_buffer_object['row-major'] = >> concurrent_test('arb_uniform_buffer_object-row-major') >> > arb_uniform_buffer_object['uniformblockbinding'] = >> concurrent_test('arb_uniform_buffer_object-uniformblockbinding') >> > >> > diff --git a/tests/spec/arb_uniform_buffer_object/CMakeLists.gl.txt >> b/tests/spec/arb_uniform_buffer_object/CMakeLists.gl.txt >> > index 7d65e2d..6bc3976 100644 >> > --- a/tests/spec/arb_uniform_buffer_object/CMakeLists.gl.txt >> > +++ b/tests/spec/arb_uniform_buffer_object/CMakeLists.gl.txt >> > @@ -39,6 +39,7 @@ piglit_add_executable >> > (arb_uniform_buffer_object-negative- >> getactiveuniformblocki >> > piglit_add_executable (arb_uniform_buffer_object-negative- >> getactiveuniformsiv negative-getactiveuniformsiv.c) >> > piglit_add_executable (arb_uniform_buffer_object-referenced-by-shader >> referenced-by-shader.c) >> > piglit_add_executable (arb_uniform_buffer_object-rendering rendering.c) >> > +piglit_add_executable (arb_uniform_buffer_object-rendering-discard >> rendering-discard.c) >> > piglit_add_executable (arb_uniform_buffer_object-row-major row-major.c) >> > piglit_add_executable (arb_uniform_buffer_object-uniformblockbinding >> uniformblockbinding.c) >> > >> > diff --git a/tests/spec/arb_uniform_buffer_object/rendering-discard.c >> b/tests/spec/arb_uniform_buffer_object/rendering-discard.c >> > new file mode 100644 >> > index 0000000..cf3624d >> > --- /dev/null >> > +++ b/tests/spec/arb_uniform_buffer_object/rendering-discard.c >> > @@ -0,0 +1,195 @@ >> > +/* >> > + * Copyright (c) 2014 LunarG, Inc. >> > + * >> > + * Permission is hereby granted, free of charge, to any person >> > obtaining a >> > + * copy of this software and associated documentation files (the >> "Software"), >> > + * to deal in the Software without restriction, including without >> limitation >> > + * the rights to use, copy, modify, merge, publish, distribute, >> > sublicense, >> > + * and/or sell copies of the Software, and to permit persons to whom >> > the >> > + * Software is furnished to do so, subject to the following conditions: >> > + * >> > + * The above copyright notice and this permission notice (including the >> next >> > + * paragraph) shall be included in all copies or substantial portions >> > of >> the >> > + * Software. >> > + * >> > + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, >> > EXPRESS >> OR >> > + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF >> > MERCHANTABILITY, >> > + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT >> > SHALL >> > + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR >> OTHER >> > + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, >> > ARISING >> > + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER >> > + * DEALINGS IN THE SOFTWARE. >> > + */ >> > + >> > +/** @file rendering-discard.c >> > + * >> > + * Test rendering with UBOs in the presence of discard. Draw a single >> square >> > + * with a fragment shader that conditionally discards along a boundary >> > that >> can >> > + * cause problems when rendering multiple fragment quads at once. Test >> should >> > + * render a single color, no texture should bleed through. >> > + */ >> > + >> > +#include "piglit-util-gl-common.h" >> > + >> > +PIGLIT_GL_TEST_CONFIG_BEGIN >> > + >> > + config.supports_gl_core_version = 32; >> > + config.supports_gl_compat_version = 32; >> > + config.window_visual = PIGLIT_GL_VISUAL_DOUBLE | >> PIGLIT_GL_VISUAL_RGBA; >> > + >> > +PIGLIT_GL_TEST_CONFIG_END >> > + >> > +static const char vertex_shader_text[] = >> > + "#version 140\n" >> > + "#extension GL_ARB_explicit_attrib_location: require\n" >> > + "layout(location = 0) in vec4 piglit_vertex;\n" >> > + "void main()\n" >> > + "{\n" >> > + " gl_Position = piglit_vertex;\n" >> > + "}\n" >> > + ; >> > + >> > +static const char fragment_shader_text[] = >> > + "#version 140\n" >> > + "#extension GL_ARB_shading_language_420pack : enable\n" >> > + "#extension GL_ARB_uniform_buffer_object : enable\n" >> > + "\n" >> > + "layout(std140, binding = 0) uniform globals \n" >> > + "{\n" >> > + " vec3 global0; \n" >> > + " vec3 global1; \n" >> >> These two UBO fields appear to be unused - are they necessary? > > > Hmmm, they were in the original test case, but the test appears to still > fail if I shrink the buffer back down to size. I'll reduce it to just one > vec4. > >> >> >> > + " vec3 global2; \n" >> > + " float global3; \n" >> > + "}\n;" >> > + "\n" >> > + "layout(binding = 0) uniform sampler2D tex2;\n" >> > + "\n" >> > + "void main()\n" >> > + "{\n" >> > + " vec3 foo = texture(tex2, vec2(0.5)).xyz ;\n" >> > + "" >> > + " // This condition will discard fragments to the left \n" >> > + " // of a diagonal line. It is designed to partially \n" >> > + " // discard the contents of SIMD registers. \n" >> > + " if ((gl_FragCoord.x - gl_FragCoord.y) < 0.0) \n" >> > + " discard; \n" >> > + "" >> > + " gl_FragColor = vec4(foo * global2 * global3, 1.0 );\n" >> > + "}\n" >> > + ; >> > + >> > +/* Note that the expected color is set up to both match the clear >> > color, >> > + * but also be multiplied in the fragment shader to generate the same >> > + * value. The multiplication is critical to testing behavior on some >> > + * backends, so it should remain in the shader. >> > + */ >> > +const float expected_color[4] = { 0.2, 0.2, 0.2, 1.0 }; >> > + >> > +/* >> > + * Create a single-color image. >> > + */ >> > +static GLubyte * >> > +create_image(GLint w, GLint h, const GLubyte color[4]) >> > +{ >> > + GLubyte *buf = (GLubyte *) malloc(w * h * 4); >> > + int i; >> > + for (i = 0; i < w * h; i++) { >> > + buf[i*4+0] = color[0]; >> > + buf[i*4+1] = color[1]; >> > + buf[i*4+2] = color[2]; >> > + buf[i*4+3] = color[3]; >> > + } >> > + return buf; >> > +} >> > + >> > +static void >> > +setup_texture(void) >> > +{ >> > + GLubyte colors[4] = {255, 255, 255, 255}; >> > + GLuint tex1; >> > + GLint width = 128, height = 64, levels = 1; >> > + GLint level = 0; >> > + GLubyte *colorBuf = create_image(width, height, colors); >> > + >> > + /* Set up a texture to pull from, just fill it with solid white >> > */ >> > + glGenTextures(1, &tex1); >> > + glBindTexture(GL_TEXTURE_2D, tex1); >> > + glTexStorage2D(GL_TEXTURE_2D, levels, GL_RGBA8, width, height); >> > + piglit_check_gl_error(GL_NO_ERROR); >> > + glTexSubImage2D(GL_TEXTURE_2D, level, 0, 0, width, height, >> > GL_RGBA, >> GL_UNSIGNED_BYTE, colorBuf); >> > + glActiveTexture(GL_TEXTURE0); >> > + glBindTexture(GL_TEXTURE_2D, tex1); >> > +} >> > + >> > +static void >> > +setup_uniform_buffer(void) >> > +{ >> > + GLuint uBuffer; >> > + GLfloat bufData[12] = { >> > + expected_color[0], expected_color[1], expected_color[2], >> expected_color[3], >> > + expected_color[0], expected_color[1], expected_color[2], >> expected_color[3], >> > + expected_color[0], expected_color[1], expected_color[2], >> expected_color[3], >> > + }; >> > + >> > + /* Set up a uniform buffer with data that, when processed by >> > the >> > + * fragment shader, should draw the same color as the clear >> > color. >> > + * This normally "bad" test behavior is designed to make it >> > easier >> > + * to see that any unwanted color has entered along the >> > boundary >> > + * condition. In this case, it will usually be the contents of >> > the >> > + * texture, but it can also be random data. >> > + * Note that the size of the buffer and the offsets of the >> > values >> > + * being loaded are important to trigger specific behavior in >> > some >> > + * backends that optimize loading of uniform values. >> > + */ >> > + glGenBuffers(1, &uBuffer); >> > + glBindBuffer(GL_UNIFORM_BUFFER, uBuffer); >> > + glBufferData(GL_UNIFORM_BUFFER, sizeof(bufData), NULL, >> GL_STATIC_DRAW); >> > + glBindBufferBase(GL_UNIFORM_BUFFER, 0, uBuffer); >> > + glBufferSubData(GL_UNIFORM_BUFFER, 0, sizeof(bufData), >> > bufData); >> > +} >> > + >> > +void >> > +piglit_init(int argc, char **argv) >> > +{ >> > + GLint prog = 0; >> > + >> > + piglit_require_extension("GL_ARB_uniform_buffer_object"); >> > + piglit_require_extension("GL_ARB_shading_language_420pack"); >> > + piglit_require_extension("GL_ARB_explicit_attrib_location"); >> >> As written, this test also requires GL_ARB_texture_storage. > > > Thanks, I'll add that. Did I miss some debug spew that could have caught > this? > >> >> I didn't review the test too thoroughly, but it looks reasonable enough. >> With >> the extra extension check added, this would get: >> >> Acked-by: Kenneth Graunke <<a href="mailto:kenneth@whitecape.org">kenneth@whitecape.org</a>> >> >> (perhaps other people will want to review it, though) >> >> Thanks for doing this. > > > I'll wait a bit before resubmitting, see if any other comments are made. > >> >> >> > + >> > + prog = piglit_build_simple_program(vertex_shader_text, >> fragment_shader_text); >> > + assert(prog); >> > + glUseProgram(prog); >> > + >> > + setup_texture(); >> > + setup_uniform_buffer(); >> > + >> > + glClearColor(expected_color[0], expected_color[1], >> > expected_color[2], >> expected_color[3]); >> > +} >> > + >> > + >> > +enum piglit_result >> > +piglit_display(void) >> > +{ >> > + int probe = 0; >> > + >> > + glViewport(0, 0, piglit_width, piglit_height); >> > + >> > + glClear(GL_COLOR_BUFFER_BIT); >> > + >> > + if (!piglit_check_gl_error(GL_NO_ERROR)) >> > + return PIGLIT_FAIL; >> > + >> > + piglit_draw_rect(-1, -1, 2, 2); >> > + >> > + /* Probe a rect so we don't have to guess at which pixel is >> > + * incorrect. Just a small area in the corner should be >> > sufficient, >> as >> > + * long as the boundary condition exists within it. The expected >> value >> > + * should match both fragment shader output and the clear color. >> > + */ >> > + probe = piglit_probe_rect_rgba(0, 0, piglit_width/4, >> > piglit_height/4, >> expected_color); >> > + >> > + piglit_present_results(); >> > + >> > + return probe == 1 ? PIGLIT_PASS : PIGLIT_FAIL; >> > +} > > > > > -- > Cody Northrop > Graphics Software Engineer > LunarG, Inc.- 3D Driver Innovations > Email: <a href="mailto:cody@lunarg.com">cody@lunarg.com</a> > Website: <a href="http://www.lunarg.com" target="_blank">http://www.lunarg.com</a> > </div></div>> _______________________________________________ > Piglit mailing list > <a href="mailto:Piglit@lists.freedesktop.org">Piglit@lists.freedesktop.org</a> > <a href="http://lists.freedesktop.org/mailman/listinfo/piglit" target="_blank">http://lists.freedesktop.org/mailman/listinfo/piglit</a> > </blockquote></div> <div> </div>-- <div dir="ltr"><div> Cody Northrop Graphics Software Engineer LunarG, Inc.- 3D Driver Innovations Email: <a href="mailto:cody@lunarg.com" target="_blank">cody@lunarg.com</a> Website: <a href="http://www.lunarg.com/" target="_blank">http://www.lunarg.com</a></div></div> </div>