Added basic support for OpenGL 4.3 Compute Shaders

* Adapted cgt::Shader to be able to represent and store CS objects
* cgt::ImageUnit manages the GPU image units in use
* GpuCapabilities has been updated for some related queries
* GLTextureFormatTraits simplify lookup of relevant info for a known internalFormat of a texture

core/datastructures/imagerepresentationconversioncore.h

@@ -98,7 +98,7 @@ namespace campvis {
                 // converting from GL representation
                 GenericImageRepresentationLocal<BASETYPE, NUMCHANNELS>* toReturn = nullptr;
                 GLJobProc.enqueueJobBlocking([&]() {
-                    if (cgt::Texture::calcMatchingDataType(tester->getTexture()->getInternalFormat()) != TypeTraits<BASETYPE, NUMCHANNELS>::glDataType)
+                    if (cgt::Texture::calcMatchingPixelDataType(tester->getTexture()->getInternalFormat()) != TypeTraits<BASETYPE, NUMCHANNELS>::glDataType)
                         LDEBUGC("CAMPVis.core.datastructures.GenericLocalConversion", "Performing conversion between data types, you may lose information or the resulting data may show other unexpected features.");
 
                     WeaklyTypedPointer wtp = tester->getWeaklyTypedPointerConvert(TypeTraits<BASETYPE, NUMCHANNELS>::glDataType);

core/datastructures/imagerepresentationgl.cpp

@@ -201,15 +201,15 @@ namespace campvis {
     }
 
     const WeaklyTypedPointer ImageRepresentationGL::getWeaklyTypedPointerCopy() const {
-        GLint format = cgt::Texture::calcMatchingFormat(_texture->getInternalFormat());
-        GLenum dataType = cgt::Texture::calcMatchingDataType(_texture->getInternalFormat());
+        GLint format = cgt::Texture::calcMatchingPixelFormat(_texture->getInternalFormat());
+        GLenum dataType = cgt::Texture::calcMatchingPixelDataType(_texture->getInternalFormat());
         GLubyte* data = _texture->downloadTextureToBuffer(format, dataType);
 
         return WeaklyTypedPointer(WeaklyTypedPointer::baseType(dataType), _texture->getNumChannels(), data);
     }
 
     const WeaklyTypedPointer ImageRepresentationGL::getWeaklyTypedPointerConvert(GLenum dataType) const {
-        GLint format = cgt::Texture::calcMatchingFormat(_texture->getInternalFormat());
+        GLint format = cgt::Texture::calcMatchingPixelFormat(_texture->getInternalFormat());
         GLubyte* data = _texture->downloadTextureToBuffer(format, dataType);
 
         return WeaklyTypedPointer(WeaklyTypedPointer::baseType(dataType), _texture->getNumChannels(), data);

ext/cgt/CMakeLists.txt

@@ -29,8 +29,10 @@ SET(CGT_SOURCES
 	frustum.cpp
 	glcanvas.cpp
 	glcontextmanager.cpp
+	gltextureformattraits.cpp
 	gpucapabilities.cpp
 	gpucapabilitieswindows.cpp
+	imageunit.cpp
 	init.cpp
 	logmanager.cpp
 	job.cpp

ext/cgt/gltextureformattraits.h

+/**********************************************************************
+*                                                                    *
+* cgt - CAMP Graphics Toolbox, Copyright (C) 2012-2015               *
+*     Chair for Computer Aided Medical Procedures                    *
+*     Technische Universitaet Muenchen, Germany.                     *
+*     <http://campar.in.tum.de/>                                     *
+*                                                                    *
+* forked from tgt - Tiny Graphics Toolbox, Copyright (C) 2006-2011   *
+*     Visualization and Computer Graphics Group, Department of       *
+*     Computer Science, University of Muenster, Germany.             *
+*     <http://viscg.uni-muenster.de>                                 *
+*                                                                    *
+* This file is part of the cgt library. This library is free         *
+* software; you can redistribute it and/or modify it under the terms *
+* of the GNU Lesser General Public License version 2.1 as published  *
+* by the Free Software Foundation.                                   *
+*                                                                    *
+* This library is distributed in the hope that it will be useful,    *
+* but WITHOUT ANY WARRANTY; without even the implied warranty of     *
+* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the       *
+* GNU Lesser General Public License for more details.                *
+*                                                                    *
+* You should have received a copy of the GNU Lesser General Public   *
+* License in the file "LICENSE.txt" along with this library.         *
+* If not, see <http://www.gnu.org/licenses/>.                        *
+*                                                                    *
+**********************************************************************/
+
+#pragma once
+
+#include <GL/glew.h>
+
+#include <string>
+
+namespace cgt {
+
+
+class GLTextureFormatTraits {
+
+public:
+    /// Constructor for unknown format
+    GLTextureFormatTraits(GLint internalFormat__)
+        : known_{ false }
+        , internalFormatName_{ "Unknown" }
+        , bpp_{ 0 }
+        , channels_{ 0 }
+        , internalFormat_{ internalFormat__ }
+        , pixelFormat_{ -1 }
+        , pixelDataType_{ -1 }
+        , glslFormatQualifier_{ "Unknown" }
+    {};
+
+    /// Constructor for known formats
+    GLTextureFormatTraits(GLint internalFormat__, const char* internalFormatName__, int bpp__, int channels__, GLint pixelFormat__, GLint pixelDataType__, const char* glslFormatQualifier__)
+        : internalFormatName_{ internalFormatName__ }
+        , known_{ true }
+        , bpp_{ bpp__ }
+        , channels_{ channels__ }
+        , internalFormat_{ internalFormat__ }
+        , pixelFormat_{ pixelFormat__ }
+        , pixelDataType_{ pixelDataType__ }
+        , glslFormatQualifier_{ glslFormatQualifier__ }
+    {};
+
+    static GLTextureFormatTraits get(GLint internalFormat);
+
+    /**
+     * Whether the specified GLint is a valid internal format
+     */
+    bool known() const { return known_; };
+
+    /**
+     * The name of the macro as a string
+     */
+    std::string internalFormatName() const { return internalFormatName_; } ;
+
+    /**
+     *	Number of bytes per pixel
+     */
+     int bpp() const { return bpp_; };
+
+    /**
+     *	Number of channels (1-4)
+     */
+     int channels() const { return channels_; };
+
+    /**
+     *	GLint of the internal format
+     */
+    GLint internalFormat() const { return internalFormat_; };
+
+    /**
+     *	Matching pixel format (GL_RED, GL_RG, GL_RGB, GL_BGR, GL_RGBA, GL_BGRA, GL_RED_INTEGER, GL_RG_INTEGER, GL_RGB_INTEGER, GL_BGR_INTEGER, GL_RGBA_INTEGER, GL_BGRA_INTEGER, GL_STENCIL_INDEX, GL_DEPTH_COMPONENT, GL_DEPTH_STENCIL)
+     */
+    GLint pixelFormat() const { return pixelFormat_; };
+
+    /**
+     *	Matching pixel data type (GL_UNSIGNED_BYTE, GL_BYTE, GL_UNSIGNED_SHORT, GL_SHORT, GL_UNSIGNED_INT, GL_INT, GL_FLOAT, GL_UNSIGNED_BYTE_3_3_2, GL_UNSIGNED_BYTE_2_3_3_REV, GL_UNSIGNED_SHORT_5_6_5, GL_UNSIGNED_SHORT_5_6_5_REV, GL_UNSIGNED_SHORT_4_4_4_4, GL_UNSIGNED_SHORT_4_4_4_4_REV, GL_UNSIGNED_SHORT_5_5_5_1, GL_UNSIGNED_SHORT_1_5_5_5_REV, GL_UNSIGNED_INT_8_8_8_8, GL_UNSIGNED_INT_8_8_8_8_REV, GL_UNSIGNED_INT_10_10_10_2, and GL_UNSIGNED_INT_2_10_10_10_REV)
+     */
+    GLint pixelDataType() const { return pixelDataType_; };
+
+    /**
+     *	Matching write format qualifier for use in glsl with image read/write operations. (rgba32f, r16, r8u, ...)
+     */
+    std::string glslFormatQualifier() const { return glslFormatQualifier_; };
+
+private:
+    bool known_;
+    const std::string internalFormatName_;
+    const int bpp_;
+    const int channels_;
+    const GLint internalFormat_;
+    const GLint pixelFormat_;
+    const GLint pixelDataType_;
+    const std::string glslFormatQualifier_;
+
+};
+
+
+}; // namspace cgt
\ No newline at end of file

ext/cgt/gpucapabilities.cpp

@@ -107,6 +107,11 @@ bool GpuCapabilities::areShadersSupported() {
     return shaderSupport_;
 }
 
+bool GpuCapabilities::areComputeShadersSupported()
+{
+    return computeShaderSupport_;
+}
+
 bool GpuCapabilities::areShadersSupportedARB() {
     return shaderSupportARB_;
 }
@@ -139,6 +144,11 @@ int GpuCapabilities::getNumTextureUnits() {
     return numTextureUnits_;
 }
 
+int GpuCapabilities::getNumImageUnits()
+{
+    return numImageUnits_;
+}
+
 bool GpuCapabilities::isNpotSupported() {
     return npotTextures_;
 }
@@ -291,6 +301,7 @@ void GpuCapabilities::detectCapabilities() {
     shaderSupport_ = (glVersion_ >= GlVersion::CGT_GL_VERSION_2_0);
     shaderSupportARB_ = (isExtensionSupported("GL_ARB_vertex_program") &&
                          isExtensionSupported("GL_ARB_fragment_program"));
+    computeShaderSupport_ = (glVersion_ >= GlVersion::CGT_GL_VERSION_4_3);
 
     if (!shaderVersion_.parseVersionString(glslVersionString_)) {
         LERROR("Malformed GLSL version string: " << glslVersionString_);
@@ -336,6 +347,8 @@ void GpuCapabilities::detectCapabilities() {
     if (numTextureUnits_ < 0)
         glGetIntegerv(GL_MAX_TEXTURE_UNITS, (GLint *) &numTextureUnits_);
 
+    glGetIntegerv(GL_MAX_IMAGE_UNITS, (GLint*)&numImageUnits_);
+
     npotTextures_ = (isExtensionSupported("GL_ARB_texture_non_power_of_two"));
 #if defined(__APPLE__) || defined(__linux__)
     // No NPOT support on older ATI Macs and Linux.

ext/cgt/gpucapabilities.h

@@ -258,13 +258,19 @@ public:
     std::string getShadingLanguageVersionString();
 
     /**
-     * Returns wether shaders are supported, which
+     * Returns whether shaders are supported, which
      * is true for OpenGL version 2.0 or later.
      */
     bool areShadersSupported();
 
     /**
-     * Returns wether the ARB shader extensions
+    * Returns whether shaders are supported, which
+    * is true for OpenGL version 4.3 or later.
+    */
+    bool areComputeShadersSupported();
+    
+    /**
+     * Returns whether the ARB shader extensions
      * are supported (GL_ARB_vertex_program and
      * GL_ARB_fragment_program).
      *
@@ -326,6 +332,12 @@ public:
      */
     int getNumTextureUnits();
 
+    /**
+    * Returns the number of image units
+    * provided by the GPU.
+    */
+    int getNumImageUnits();
+
     /**
      * Returns wether non-power-of-two textures
      * are supported (extension GL_ARB_texture_non_power_of_two).
@@ -429,6 +441,7 @@ private:
 
     bool shaderSupport_;
     bool shaderSupportARB_;
+    bool computeShaderSupport_;
     GlVersion shaderVersion_;
     ShaderModel shaderModel_;
 
@@ -436,6 +449,7 @@ private:
     bool texturing3D_;
     int max3DTexSize_;
     int numTextureUnits_;
+    int numImageUnits_;
     bool npotTextures_;
     bool textureRectangles_;
     bool anisotropicFiltering_;

ext/cgt/imageunit.cpp

+#include "imageunit.h"
+#include "imageunit.h"
+/**********************************************************************
+ *                                                                    *
+ * cgt - CAMP Graphics Toolbox, Copyright (C) 2012-2015               *
+ *     Chair for Computer Aided Medical Procedures                    *
+ *     Technische Universitaet Muenchen, Germany.                     *
+ *     <http://campar.in.tum.de/>                                     *
+ *                                                                    *
+ * forked from tgt - Tiny Graphics Toolbox, Copyright (C) 2006-2011   *
+ *     Visualization and Computer Graphics Group, Department of       *
+ *     Computer Science, University of Muenster, Germany.             *
+ *     <http://viscg.uni-muenster.de>                                 *
+ *                                                                    *
+ * This file is part of the cgt library. This library is free         *
+ * software; you can redistribute it and/or modify it under the terms *
+ * of the GNU Lesser General Public License version 2.1 as published  *
+ * by the Free Software Foundation.                                   *
+ *                                                                    *
+ * This library is distributed in the hope that it will be useful,    *
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of     *
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the       *
+ * GNU Lesser General Public License for more details.                *
+ *                                                                    *
+ * You should have received a copy of the GNU Lesser General Public   *
+ * License in the file "LICENSE.txt" along with this library.         *
+ * If not, see <http://www.gnu.org/licenses/>.                        *
+ *                                                                    *
+ **********************************************************************/
+
+#include "cgt/imageunit.h"
+#include "cgt/gpucapabilities.h"
+#include "cgt/assert.h"
+
+namespace cgt {
+
+bool ImageUnit::initialized_ = false;
+unsigned short ImageUnit::totalActive_ = 0;
+unsigned short ImageUnit::maxTexUnits_ = 0;
+unsigned short ImageUnit::numKeptUnits_ = 0;
+std::vector<bool> ImageUnit::busyUnits_ = std::vector<bool>();
+tbb::spin_mutex ImageUnit::mutex_ = tbb::spin_mutex();
+
+ImageUnit::ImageUnit(bool keep)
+    : number_(0)
+    , assigned_(false)
+{
+    if (!initialized_)
+        init();
+}
+
+void ImageUnit::init() {
+    tbb::spin_mutex::scoped_lock lock(mutex_);
+    if (initialized_)
+        return;
+
+    maxTexUnits_ = GpuCaps.getNumImageUnits();
+    busyUnits_ = std::vector<bool>(maxTexUnits_, false);
+    numKeptUnits_ = 0;
+    initialized_ = true;
+}
+
+ImageUnit::~ImageUnit() {
+    if(assigned_)
+        releaseUnit();
+}
+
+
+ImageUnit::ImageUnit(ImageUnit && other)
+{
+    *this = std::move(other);
+}
+
+ImageUnit & ImageUnit::operator=(ImageUnit && rhs)
+{
+    if (this == &rhs)
+        return *this;
+
+    if (assigned_)
+        releaseUnit();
+
+    assigned_ = rhs.assigned_;
+    number_ = rhs.number_;
+    rhs.assigned_ = false;
+    rhs.number_ = 0;
+
+    return *this;
+}
+
+GLint ImageUnit::getUnitNumber() const {
+    if (!assigned_)
+        assignUnit();
+    return number_;
+}
+
+void ImageUnit::cleanup() {
+    tbb::spin_mutex::scoped_lock lock(mutex_);
+
+    for (size_t i = 0; i < maxTexUnits_; i++) {
+        if (busyUnits_.at(i))
+            busyUnits_.at(i) = false;
+    }
+    totalActive_ = 0;
+    numKeptUnits_ = 0;
+}
+
+bool ImageUnit::unused() {
+    return (totalActive_ == 0);
+}
+
+unsigned short ImageUnit::numLocalActive() {
+    return (totalActive_ - numKeptUnits_);
+}
+
+void ImageUnit::assignUnit() const {
+    tbb::spin_mutex::scoped_lock lock(mutex_);
+    cgtAssert(totalActive_ <= maxTexUnits_, "No more texture units available");
+
+    assigned_ = true;
+
+    for (size_t i = 0; i < maxTexUnits_; i++) {
+        if (!busyUnits_.at(i)) {
+            number_ = (GLint)i;
+            busyUnits_.at(i) = true;
+            totalActive_++;
+            break;
+        }
+    }
+}
+
+void ImageUnit::releaseUnit()
+{
+    tbb::spin_mutex::scoped_lock lock(mutex_);
+
+    busyUnits_.at(number_) = false;
+    totalActive_--;
+}
+
+
+} // namespace

ext/cgt/imageunit.h

+/**********************************************************************
+ *                                                                    *
+ * cgt - CAMP Graphics Toolbox, Copyright (C) 2012-2015               *
+ *     Chair for Computer Aided Medical Procedures                    *
+ *     Technische Universitaet Muenchen, Germany.                     *
+ *     <http://campar.in.tum.de/>                                     *
+ *                                                                    *
+ * forked from tgt - Tiny Graphics Toolbox, Copyright (C) 2006-2011   *
+ *     Visualization and Computer Graphics Group, Department of       *
+ *     Computer Science, University of Muenster, Germany.             *
+ *     <http://viscg.uni-muenster.de>                                 *
+ *                                                                    *
+ * This file is part of the cgt library. This library is free         *
+ * software; you can redistribute it and/or modify it under the terms *
+ * of the GNU Lesser General Public License version 2.1 as published  *
+ * by the Free Software Foundation.                                   *
+ *                                                                    *
+ * This library is distributed in the hope that it will be useful,    *
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of     *
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the       *
+ * GNU Lesser General Public License for more details.                *
+ *                                                                    *
+ * You should have received a copy of the GNU Lesser General Public   *
+ * License in the file "LICENSE.txt" along with this library.         *
+ * If not, see <http://www.gnu.org/licenses/>.                        *
+ *                                                                    *
+ **********************************************************************/
+
+#ifndef CGT_IMAGEUNIT_H
+#define CGT_IMAGEUNIT_H
+
+#include <string>
+
+#include <tbb/spin_mutex.h>
+#include "cgt/cgt_gl.h"
+#include "cgt/shadermanager.h"
+
+namespace cgt {
+
+/**
+ * OpenGL Image Unit
+ */
+class CGT_API ImageUnit {
+public:
+    ImageUnit(bool keep = false);
+    ~ImageUnit();
+
+    void releaseUnit();
+
+    // copy construction and assignment is not allowed
+    ImageUnit(const ImageUnit& other) = delete; 
+    ImageUnit& operator=(const ImageUnit& other) = delete;
+
+    // move construction and assignment is possible
+    ImageUnit(ImageUnit&& other);
+    ImageUnit& operator=(ImageUnit&& rhs);
+
+
+    GLint getUnitNumber() const;
+
+    static void cleanup();
+
+    static unsigned short numLocalActive();
+
+    /**
+     * Returns true if no texture units are currently assigned.
+     */
+    static bool unused();
+
+protected:
+    void assignUnit() const;
+    static void init();
+
+    mutable GLint number_;
+    mutable bool assigned_;
+
+    static bool initialized_;
+    static unsigned short totalActive_;
+    static unsigned short maxTexUnits_;
+    static unsigned short numKeptUnits_;
+    static std::vector<bool> busyUnits_;
+    static tbb::spin_mutex mutex_;
+};
+
+} // namespace cgt
+
+#endif // CGT_IMAGEUNIT_H

ext/cgt/shadermanager.cpp

@@ -32,6 +32,7 @@
 
 #include <iostream>
 #include <fstream>
+#include <memory>
 
 using std::string;
 
@@ -602,9 +603,9 @@ void Shader::loadSeparate(const string& vert_filename, const string& geom_filena
                           const string& frag_filename, const string& customHeader, const std::string& customGlslVersion)
                           throw (Exception)
 {
-    ShaderObject* frag = 0;
-    ShaderObject* vert = 0;
-    ShaderObject* geom = 0;
+    ShaderObject* frag = nullptr;
+    ShaderObject* vert = nullptr;
+    ShaderObject* geom = nullptr;
 
     if (!vert_filename.empty()) {
         vert = new ShaderObject(vert_filename, ShaderObject::VERTEX_SHADER);
@@ -752,6 +753,68 @@ void Shader::loadSeparate(const string& vert_filename, const string& geom_filena
     }
 }
 
+void Shader::loadCompute(const string& compFilename, const string& customHeader, const string& customGlslVersion)
+    throw (Exception)
+{
+    ShaderObject *comp;
+
+    if (!compFilename.empty()) {
+        comp = new ShaderObject(compFilename, ShaderObject::COMPUTE_SHADER);
+
+        if (!customHeader.empty()) {
+            comp->setHeader(customHeader);
+        }
+        if (!customGlslVersion.empty()) {
+            comp->setCustomGlslVersion(customGlslVersion);
+        }
+        else {
+            comp->setCustomGlslVersion("430");
+        }
+
+        try {
+            comp->loadSourceFromFile(compFilename);
+        }
+        catch (const Exception& e) {
+            LDEBUG("Failed to load compute shader " << compFilename << ": " << e.what());
+            delete comp;
+            throw Exception("Failed to load vertex shader " + compFilename + ": " + e.what());
+        }
+
+        comp->uploadSource();
+
+        if (!comp->compileShader()) {
+            LERROR("Failed to compile Compute shader " << compFilename);
+            LERROR("Compiler Log: \n" << comp->getCompilerLog());
+            delete comp;
+            throw Exception("Failed to compile vertex shader: " + compFilename);
+        }
+    }
+     
+    // Attach ShaderObjects, dtor will take care of freeing them
+    if (comp)
+        attachObject(comp);
+    
+    if (!linkProgram()) {
+        LERROR("Failed to link shader (" << compFilename << ")");
+        LERROR(comp->filename_ << " Compute shader compiler log: \n" << comp->getCompilerLog());
+        detachObject(comp);
+
+        LERROR("Linker Log: \n" << getLinkerLog());
+        throw Exception("Failed to link shader (" + compFilename + ")");
+    }
+
+
+    if (comp && comp->getCompilerLog().size() > 1) {
+        LDEBUG("Compute shader compiler log for file '" << compFilename << "': \n" << comp->getCompilerLog());
+    }
+    
+    if (getLinkerLog().size() > 1) {
+        LDEBUG("Linker log for '" << compFilename << "': \n" << getLinkerLog());
+    }
+}
+
+
+
 GLint Shader::getUniformLocation(const string& name) {
     GLint l;
     l = glGetUniformLocation(id_, name.c_str());
@@ -1316,6 +1379,45 @@ Shader* ShaderManager::load(const std::string& vertFilename, const std::string&
     return loadWithCustomGlslVersion(vertFilename, geomFilename, fragFilename, customHeader, "");
 }
 
+Shader * ShaderManager::loadCompute(const std::string& compFilename, const std::string & customHeader, const std::string & customGlslVersion) throw(Exception)
+{
+    //LDEBUG("Loading files " << vertFilename << " and " << fragFilename);
+    if (!GpuCaps.areComputeShadersSupported()) {
+        LERROR("Compute Shaders are not supported (Requires OpenGL 4.3 capable device).");
+        throw Exception("Compute Shaders are not supported.");
+    }
+
+    // create a somewhat unique identifier for this shader triple
+    string identifier = compFilename;
+
+    if (isLoaded(identifier)) {
+        LDEBUG("Shader already loaded. Increase usage count.");
+        increaseUsage(identifier);
+        return get(identifier);
+    }
+
+    Shader* shdr = new Shader();
+
+    // searching in all paths for every shader
+    string completeFilename;
+    if (!compFilename.empty())
+        completeFilename = completePath(compFilename);
+
+    // loading and linking found shaders
+    try {
+        shdr->loadCompute(completeFilename, customHeader, customGlslVersion);
+
+        // register even when caching is disabled, needed for rebuildFromFile()
+        reg(shdr, identifier);
+
+        return shdr;
+    }
+    catch (const Exception& /*e*/) {
+        delete shdr;
+        throw;
+    }
+}
+
 Shader* ShaderManager::loadWithCustomGlslVersion(const std::string& vertFilename, const std::string& geomFilename, const std::string& fragFilename, const std::string& customHeader, const std::string& customGlslVersion) throw(Exception) {
     //LDEBUG("Loading files " << vertFilename << " and " << fragFilename);
     if (!GpuCaps.areShadersSupported()) {

ext/cgt/shadermanager.h

@@ -71,7 +71,8 @@ public:
     enum ShaderType {
         VERTEX_SHADER = GL_VERTEX_SHADER,
         FRAGMENT_SHADER = GL_FRAGMENT_SHADER,