\r
#include "image_kernel.h"\r
\r
+#include "image_shader.h"\r
+\r
+#include "blending_glsl.h"\r
+#include "../gpu/shader.h"\r
+#include "../gpu/device_buffer.h"\r
+#include "../gpu/ogl_device.h"\r
+\r
#include <common/exception/exceptions.h>\r
#include <common/gl/gl_check.h>\r
+#include <common/env.h>\r
\r
#include <core/video_format.h>\r
#include <core/producer/frame/pixel_format.h>\r
-#include <core/producer/frame/image_transform.h>\r
+#include <core/producer/frame/frame_transform.h>\r
\r
-#include <Glee.h>\r
+#include <GL/glew.h>\r
\r
#include <boost/noncopyable.hpp>\r
\r
#include <unordered_map>\r
\r
-namespace caspar { namespace mixer {\r
-\r
-class shader_program : boost::noncopyable\r
-{\r
- GLuint program_;\r
-public:\r
-\r
- shader_program() : program_(0) {}\r
- shader_program(shader_program&& other) : program_(other.program_){other.program_ = 0;}\r
- shader_program(const std::string& vertex_source_str, const std::string& fragment_source_str) : program_(0)\r
- {\r
- GLint success;\r
- \r
- const char* vertex_source = vertex_source_str.c_str();\r
- \r
- auto vertex_shader = glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB);\r
- \r
- GL(glShaderSourceARB(vertex_shader, 1, &vertex_source, NULL));\r
- GL(glCompileShaderARB(vertex_shader));\r
-\r
- GL(glGetObjectParameterivARB(vertex_shader, GL_OBJECT_COMPILE_STATUS_ARB, &success));\r
- if (success == GL_FALSE)\r
- {\r
- char info[2048];\r
- GL(glGetInfoLogARB(vertex_shader, sizeof(info), 0, info));\r
- GL(glDeleteObjectARB(vertex_shader));\r
- std::stringstream str;\r
- str << "Failed to compile vertex shader:" << std::endl << info << std::endl;\r
- BOOST_THROW_EXCEPTION(gl::gl_error() << msg_info(str.str()));\r
- }\r
- \r
- const char* fragment_source = fragment_source_str.c_str();\r
- \r
- auto fragmemt_shader = glCreateShaderObjectARB(GL_FRAGMENT_SHADER_ARB);\r
- \r
- GL(glShaderSourceARB(fragmemt_shader, 1, &fragment_source, NULL));\r
- GL(glCompileShaderARB(fragmemt_shader));\r
-\r
- GL(glGetObjectParameterivARB(fragmemt_shader, GL_OBJECT_COMPILE_STATUS_ARB, &success));\r
- if (success == GL_FALSE)\r
- {\r
- char info[2048];\r
- GL(glGetInfoLogARB(fragmemt_shader, sizeof(info), 0, info));\r
- GL(glDeleteObjectARB(fragmemt_shader));\r
- std::stringstream str;\r
- str << "Failed to compile fragment shader:" << std::endl << info << std::endl;\r
- BOOST_THROW_EXCEPTION(gl::gl_error() << msg_info(str.str()));\r
- }\r
- \r
- program_ = glCreateProgramObjectARB();\r
- \r
- GL(glAttachObjectARB(program_, vertex_shader));\r
- GL(glAttachObjectARB(program_, fragmemt_shader));\r
-\r
- GL(glLinkProgramARB(program_));\r
- \r
- GL(glDeleteObjectARB(vertex_shader));\r
- GL(glDeleteObjectARB(fragmemt_shader));\r
-\r
- GL(glGetObjectParameterivARB(program_, GL_OBJECT_LINK_STATUS_ARB, &success));\r
- if (success == GL_FALSE)\r
- {\r
- char info[2048];\r
- GL(glGetInfoLogARB(program_, sizeof(info), 0, info));\r
- GL(glDeleteObjectARB(program_));\r
- std::stringstream str;\r
- str << "Failed to link shader program:" << std::endl << info << std::endl;\r
- BOOST_THROW_EXCEPTION(gl::gl_error() << msg_info(str.str()));\r
- }\r
- GL(glUseProgramObjectARB(program_));\r
- glUniform1i(glGetUniformLocation(program_, "plane[0]"), 0);\r
- glUniform1i(glGetUniformLocation(program_, "plane[1]"), 1);\r
- glUniform1i(glGetUniformLocation(program_, "plane[2]"), 2);\r
- glUniform1i(glGetUniformLocation(program_, "plane[3]"), 3);\r
- }\r
-\r
- GLint get_location(const char* name)\r
- {\r
- GLint loc = glGetUniformLocation(program_, name);\r
- return loc;\r
- }\r
-\r
- shader_program& operator=(shader_program&& other) \r
- {\r
- program_ = other.program_; \r
- other.program_ = 0; \r
- return *this;\r
- }\r
-\r
- ~shader_program()\r
- {\r
- glDeleteProgram(program_);\r
- }\r
-\r
- void use()\r
- { \r
- GL(glUseProgramObjectARB(program_)); \r
- }\r
-};\r
-\r
-GLubyte progressive_pattern[] = {\r
- 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,\r
- 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,\r
- 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,\r
- 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xFF, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff};\r
+namespace caspar { namespace core {\r
\r
GLubyte upper_pattern[] = {\r
0xff, 0xff, 0xff, 0xff, 0x00, 0x00, 0x00, 0x00, 0xff, 0xff, 0xff, 0xff, 0x00, 0x00, 0x00, 0x00, 0xff, 0xff, 0xff, 0xff, 0x00, 0x00, 0x00, 0x00, 0xff, 0xff, 0xff, 0xff, 0x00, 0x00, 0x00, 0x00,\r
\r
struct image_kernel::implementation : boost::noncopyable\r
{ \r
- std::unordered_map<core::pixel_format::type, shader_program> shaders_;\r
-\r
-public:\r
- std::unordered_map<core::pixel_format::type, shader_program>& shaders()\r
+ std::shared_ptr<shader> shader_;\r
+ bool blend_modes_;\r
+ \r
+ void draw(ogl_device& ogl, draw_params&& params)\r
{\r
- GL(glEnable(GL_POLYGON_STIPPLE));\r
- GL(glEnable(GL_BLEND));\r
- GL(glBlendFuncSeparate(GL_SRC_ALPHA, GL_ONE_MINUS_SRC_ALPHA, GL_ONE, GL_ONE));\r
+ static const double epsilon = 0.001;\r
\r
- if(shaders_.empty())\r
+ CASPAR_ASSERT(params.pix_desc.planes.size() == params.textures.size());\r
+\r
+ if(params.textures.empty() || !params.background)\r
+ return;\r
+\r
+ if(params.transform.opacity < epsilon)\r
+ return;\r
+ \r
+ if(!std::all_of(params.textures.begin(), params.textures.end(), std::mem_fn(&device_buffer::ready)))\r
{\r
- std::string common_vertex = \r
- "void main() "\r
- "{ "\r
- " gl_TexCoord[0] = gl_MultiTexCoord0; "\r
- " gl_TexCoord[1] = gl_MultiTexCoord1; "\r
- " gl_FrontColor = gl_Color; "\r
- " gl_Position = ftransform(); "\r
- "} ";\r
-\r
- std::string common_fragment = \r
- "uniform sampler2D plane[4]; "\r
- "uniform float gain; "\r
- "uniform bool HD; "\r
- "uniform bool has_separate_key; "\r
- \r
- // NOTE: YCbCr, ITU-R, http://www.intersil.com/data/an/an9717.pdf \r
- // TODO: Support for more yuv formats might be needed. \r
- "vec4 ycbcra_to_bgra_sd(float y, float cb, float cr, float a) "\r
- "{ "\r
- " cb -= 0.5; "\r
- " cr -= 0.5; "\r
- " y = 1.164*(y-0.0625); "\r
- " "\r
- " vec4 color; "\r
- " color.r = y + 1.596 * cr; "\r
- " color.g = y - 0.813 * cr - 0.391 * cb; "\r
- " color.b = y + 2.018 * cb; "\r
- " color.a = a; "\r
- " "\r
- " return color; "\r
- "} " \r
- " "\r
-\r
- "vec4 ycbcra_to_bgra_hd(float y, float cb, float cr, float a) "\r
- "{ "\r
- " cb -= 0.5; "\r
- " cr -= 0.5; "\r
- " y = 1.164*(y-0.0625); "\r
- " "\r
- " vec4 color; "\r
- " color.r = y + 1.793 * cr; "\r
- " color.g = y - 0.534 * cr - 0.213 * cb; "\r
- " color.b = y + 2.115 * cb; "\r
- " color.a = a; "\r
- " "\r
- " return color; "\r
- "} " \r
- " ";\r
+ CASPAR_LOG(warning) << L"[image_mixer] Performance warning. Host to device transfer not complete, GPU will be stalled";\r
+ ogl.yield(); // Try to give it some more time.\r
+ } \r
+ \r
+ // Bind textures\r
+\r
+ for(size_t n = 0; n < params.textures.size(); ++n)\r
+ params.textures[n]->bind(n);\r
+\r
+ if(params.local_key)\r
+ params.local_key->bind(texture_id::local_key);\r
+ \r
+ if(params.layer_key)\r
+ params.layer_key->bind(texture_id::layer_key);\r
\r
- shaders_[core::pixel_format::gray] = shader_program(common_vertex, common_fragment +\r
-\r
- "void main() "\r
- "{ "\r
- " vec4 rgba = vec4(texture2D(plane[0], gl_TexCoord[0].st).rrr, 1.0); "\r
- " if(has_separate_key) "\r
- " rgba.a = texture2D(plane[3], gl_TexCoord[1].st).r; "\r
- " gl_FragColor = rgba * gain; "\r
- "} ");\r
-\r
- shaders_[core::pixel_format::abgr] = shader_program(common_vertex, common_fragment +\r
-\r
- "void main() "\r
- "{ "\r
- " vec4 abgr = texture2D(plane[0], gl_TexCoord[0].st); "\r
- " if(has_separate_key) "\r
- " abgr.b = texture2D(plane[3], gl_TexCoord[1].st).r; "\r
- " gl_FragColor = abgr.argb * gain; "\r
- "} ");\r
+ // Setup shader\r
+\r
+ if(!shader_)\r
+ shader_ = get_image_shader(ogl, blend_modes_);\r
+ \r
+ ogl.use(*shader_);\r
+\r
+ shader_->set("plane[0]", texture_id::plane0);\r
+ shader_->set("plane[1]", texture_id::plane1);\r
+ shader_->set("plane[2]", texture_id::plane2);\r
+ shader_->set("plane[3]", texture_id::plane3);\r
+ shader_->set("local_key", texture_id::local_key);\r
+ shader_->set("layer_key", texture_id::layer_key);\r
+ shader_->set("is_hd", params.pix_desc.planes.at(0).height > 700 ? 1 : 0);\r
+ shader_->set("has_local_key", params.local_key);\r
+ shader_->set("has_layer_key", params.layer_key);\r
+ shader_->set("pixel_format", params.pix_desc.pix_fmt); \r
+ shader_->set("opacity", params.transform.is_key ? 1.0 : params.transform.opacity); \r
+ \r
+ // Setup blend_func\r
+ \r
+ if(params.transform.is_key)\r
+ params.blend_mode = blend_mode::normal;\r
+\r
+ if(blend_modes_)\r
+ {\r
+ params.background->bind(6);\r
+\r
+ shader_->set("background", texture_id::background);\r
+ shader_->set("blend_mode", params.blend_mode);\r
+ }\r
+ else\r
+ {\r
+ ogl.blend_func_separate(GL_ONE, GL_ONE_MINUS_SRC_ALPHA, GL_ONE, GL_ONE_MINUS_SRC_ALPHA); \r
+ }\r
+\r
+ // Setup image-adjustements\r
+ \r
+ if(params.transform.levels.min_input > epsilon ||\r
+ params.transform.levels.max_input < 1.0-epsilon ||\r
+ params.transform.levels.min_output > epsilon ||\r
+ params.transform.levels.max_output < 1.0-epsilon ||\r
+ std::abs(params.transform.levels.gamma - 1.0) > epsilon)\r
+ {\r
+ shader_->set("levels", true); \r
+ shader_->set("min_input", params.transform.levels.min_input); \r
+ shader_->set("max_input", params.transform.levels.max_input);\r
+ shader_->set("min_output", params.transform.levels.min_output);\r
+ shader_->set("max_output", params.transform.levels.max_output);\r
+ shader_->set("gamma", params.transform.levels.gamma);\r
+ }\r
+ else\r
+ shader_->set("levels", false); \r
+\r
+ if(std::abs(params.transform.brightness - 1.0) > epsilon ||\r
+ std::abs(params.transform.saturation - 1.0) > epsilon ||\r
+ std::abs(params.transform.contrast - 1.0) > epsilon)\r
+ {\r
+ shader_->set("csb", true); \r
+ \r
+ shader_->set("brt", params.transform.brightness); \r
+ shader_->set("sat", params.transform.saturation);\r
+ shader_->set("con", params.transform.contrast);\r
+ }\r
+ else\r
+ shader_->set("csb", false); \r
+ \r
+ // Setup interlacing\r
+\r
+ if(params.transform.field_mode == core::field_mode::progressive) \r
+ ogl.disable(GL_POLYGON_STIPPLE); \r
+ else \r
+ {\r
+ ogl.enable(GL_POLYGON_STIPPLE);\r
+\r
+ if(params.transform.field_mode == core::field_mode::upper)\r
+ ogl.stipple_pattern(upper_pattern);\r
+ else if(params.transform.field_mode == core::field_mode::lower)\r
+ ogl.stipple_pattern(lower_pattern);\r
+ }\r
+\r
+ // Setup drawing area\r
\r
- shaders_[core::pixel_format::argb]= shader_program(common_vertex, common_fragment +\r
-\r
- "void main() " \r
- "{ "\r
- " vec4 argb = texture2D(plane[0], gl_TexCoord[0].st); "\r
- " if(has_separate_key) "\r
- " argb.b = texture2D(plane[3], gl_TexCoord[1].st).r; "\r
- " gl_FragColor = argb.grab * gl_Color * gain; "\r
- "} ");\r
+ ogl.viewport(0, 0, params.background->width(), params.background->height());\r
+ \r
+ auto m_p = params.transform.clip_translation;\r
+ auto m_s = params.transform.clip_scale;\r
+\r
+ bool scissor = m_p[0] > std::numeric_limits<double>::epsilon() || m_p[1] > std::numeric_limits<double>::epsilon() ||\r
+ m_s[0] < (1.0 - std::numeric_limits<double>::epsilon()) || m_s[1] < (1.0 - std::numeric_limits<double>::epsilon());\r
+\r
+ if(scissor)\r
+ {\r
+ double w = static_cast<double>(params.background->width());\r
+ double h = static_cast<double>(params.background->height());\r
\r
- shaders_[core::pixel_format::bgra]= shader_program(common_vertex, common_fragment +\r
-\r
- "void main() "\r
- "{ "\r
- " vec4 bgra = texture2D(plane[0], gl_TexCoord[0].st); "\r
- " if(has_separate_key) "\r
- " bgra.a = texture2D(plane[3], gl_TexCoord[1].st).r; "\r
- " gl_FragColor = bgra.rgba * gl_Color * gain; "\r
- "} ");\r
+ ogl.enable(GL_SCISSOR_TEST);\r
+ ogl.scissor(static_cast<size_t>(m_p[0]*w), static_cast<size_t>(m_p[1]*h), static_cast<size_t>(m_s[0]*w), static_cast<size_t>(m_s[1]*h));\r
+ }\r
+\r
+ auto f_p = params.transform.fill_translation;\r
+ auto f_s = params.transform.fill_scale;\r
\r
- shaders_[core::pixel_format::rgba] = shader_program(common_vertex, common_fragment +\r
-\r
- "void main() "\r
- "{ "\r
- " vec4 rgba = texture2D(plane[0], gl_TexCoord[0].st); "\r
- " if(has_separate_key) "\r
- " rgba.a = texture2D(plane[3], gl_TexCoord[1].st).r; "\r
- " gl_FragColor = rgba.bgra * gl_Color * gain; "\r
- "} ");\r
+ // Set render target\r
\r
- shaders_[core::pixel_format::ycbcr] = shader_program(common_vertex, common_fragment +\r
-\r
- "void main() "\r
- "{ "\r
- " float y = texture2D(plane[0], gl_TexCoord[0].st).r; "\r
- " float cb = texture2D(plane[1], gl_TexCoord[0].st).r; "\r
- " float cr = texture2D(plane[2], gl_TexCoord[0].st).r; "\r
- " float a = 1.0; " \r
- " if(has_separate_key) "\r
- " a = texture2D(plane[3], gl_TexCoord[1].st).r; "\r
- " if(HD) "\r
- " gl_FragColor = ycbcra_to_bgra_hd(y, cb, cr, a) * gl_Color * gain;"\r
- " else "\r
- " gl_FragColor = ycbcra_to_bgra_sd(y, cb, cr, a) * gl_Color * gain;"\r
- "} ");\r
+ ogl.attach(*params.background);\r
\r
- shaders_[core::pixel_format::ycbcra] = shader_program(common_vertex, common_fragment +\r
-\r
- "void main() "\r
- "{ "\r
- " float y = texture2D(plane[0], gl_TexCoord[0].st).r; "\r
- " float cb = texture2D(plane[1], gl_TexCoord[0].st).r; "\r
- " float cr = texture2D(plane[2], gl_TexCoord[0].st).r; "\r
- " float a = texture2D(plane[3], gl_TexCoord[0].st).r; "\r
- " if(HD) "\r
- " gl_FragColor = ycbcra_to_bgra_hd(y, cb, cr, a) * gl_Color * gain;"\r
- " else "\r
- " gl_FragColor = ycbcra_to_bgra_sd(y, cb, cr, a) * gl_Color * gain;"\r
- "} ");\r
+ // Draw\r
+\r
+ glBegin(GL_QUADS);\r
+ glMultiTexCoord2d(GL_TEXTURE0, 0.0, 0.0); glMultiTexCoord2d(GL_TEXTURE1, f_p[0] , f_p[1] ); glVertex2d( f_p[0] *2.0-1.0, f_p[1] *2.0-1.0);\r
+ glMultiTexCoord2d(GL_TEXTURE0, 1.0, 0.0); glMultiTexCoord2d(GL_TEXTURE1, (f_p[0]+f_s[0]), f_p[1] ); glVertex2d((f_p[0]+f_s[0])*2.0-1.0, f_p[1] *2.0-1.0);\r
+ glMultiTexCoord2d(GL_TEXTURE0, 1.0, 1.0); glMultiTexCoord2d(GL_TEXTURE1, (f_p[0]+f_s[0]), (f_p[1]+f_s[1])); glVertex2d((f_p[0]+f_s[0])*2.0-1.0, (f_p[1]+f_s[1])*2.0-1.0);\r
+ glMultiTexCoord2d(GL_TEXTURE0, 0.0, 1.0); glMultiTexCoord2d(GL_TEXTURE1, f_p[0] , (f_p[1]+f_s[1])); glVertex2d( f_p[0] *2.0-1.0, (f_p[1]+f_s[1])*2.0-1.0);\r
+ glEnd();\r
+\r
+ // Cleanup\r
+\r
+ ogl.disable(GL_SCISSOR_TEST); \r
+ \r
+ params.textures.clear();\r
+ ogl.yield(); // Return resources to pool as early as possible.\r
+\r
+ if(blend_modes_)\r
+ {\r
+ // http://www.opengl.org/registry/specs/NV/texture_barrier.txt\r
+ // This allows us to use framebuffer (background) both as source and target while blending.\r
+ glTextureBarrierNV(); \r
}\r
- return shaders_;\r
}\r
};\r
\r
image_kernel::image_kernel() : impl_(new implementation()){}\r
-\r
-void image_kernel::draw(size_t width, size_t height, const core::pixel_format_desc& pix_desc, const core::image_transform& transform, bool has_separate_key)\r
+void image_kernel::draw(ogl_device& ogl, draw_params&& params)\r
{\r
- impl_->shaders()[pix_desc.pix_fmt].use();\r
-\r
- GL(glUniform1f(impl_->shaders()[pix_desc.pix_fmt].get_location("gain"), static_cast<GLfloat>(transform.get_gain())));\r
- GL(glUniform1i(impl_->shaders()[pix_desc.pix_fmt].get_location("HD"), pix_desc.planes.at(0).height > 700 ? 1 : 0));\r
- GL(glUniform1i(impl_->shaders()[pix_desc.pix_fmt].get_location("has_separate_key"), has_separate_key ? 1 : 0));\r
-\r
- if(transform.get_mode() == core::video_mode::upper)\r
- glPolygonStipple(upper_pattern);\r
- else if(transform.get_mode() == core::video_mode::lower)\r
- glPolygonStipple(lower_pattern);\r
- else\r
- glPolygonStipple(progressive_pattern);\r
- \r
- GL(glColor4d(1.0, 1.0, 1.0, transform.get_opacity()));\r
- GL(glViewport(0, 0, width, height));\r
- \r
- auto m_p = transform.get_key_translation();\r
- auto m_s = transform.get_key_scale();\r
- double w = static_cast<double>(width);\r
- double h = static_cast<double>(height);\r
-\r
- GL(glEnable(GL_SCISSOR_TEST));\r
- GL(glScissor(static_cast<size_t>(m_p[0]*w), static_cast<size_t>(m_p[1]*h), static_cast<size_t>(m_s[0]*w), static_cast<size_t>(m_s[1]*h)));\r
- \r
- auto f_p = transform.get_fill_translation();\r
- auto f_s = transform.get_fill_scale();\r
- \r
- glBegin(GL_QUADS);\r
- glMultiTexCoord2d(GL_TEXTURE0, 0.0, 0.0); glMultiTexCoord2d(GL_TEXTURE1, f_p[0] , f_p[1] ); glVertex2d( f_p[0] *2.0-1.0, f_p[1] *2.0-1.0);\r
- glMultiTexCoord2d(GL_TEXTURE0, 1.0, 0.0); glMultiTexCoord2d(GL_TEXTURE1, (f_p[0]+f_s[0]), f_p[1] ); glVertex2d((f_p[0]+f_s[0])*2.0-1.0, f_p[1] *2.0-1.0);\r
- glMultiTexCoord2d(GL_TEXTURE0, 1.0, 1.0); glMultiTexCoord2d(GL_TEXTURE1, (f_p[0]+f_s[0]), (f_p[1]+f_s[1])); glVertex2d((f_p[0]+f_s[0])*2.0-1.0, (f_p[1]+f_s[1])*2.0-1.0);\r
- glMultiTexCoord2d(GL_TEXTURE0, 0.0, 1.0); glMultiTexCoord2d(GL_TEXTURE1, f_p[0] , (f_p[1]+f_s[1])); glVertex2d( f_p[0] *2.0-1.0, (f_p[1]+f_s[1])*2.0-1.0);\r
- glEnd();\r
- GL(glDisable(GL_SCISSOR_TEST)); \r
+ impl_->draw(ogl, std::move(params));\r
}\r
\r
}}
\ No newline at end of file