From f98487cc9fac651af80577fddfcb6b100853a1f0 Mon Sep 17 00:00:00 2001 From: fogal1 Date: Sat, 1 Aug 2009 17:14:07 +0000 Subject: [PATCH] Volume plot, Tuvok: updates for new API. Second pass integration. This code is a lot simpler, actually works on my Mac, and (most importantly :) can render in parallel. Still a bit of work left. Some minor viewing issues, then 2D transfer functions, and the shaders need some simple updates. Unfortunately there's a lot of shaders to do this for, so we might fix shaders ad hoc until some time is reserved for infrastructure work in Tuvok. git-svn-id: http://visit.ilight.com/svn/visit/trunk/src@8020 18c085ea-50e0-402c-830e-de6fd14e8384 --- plots/Volume/avtOpenGLTuvokVolumeRenderer.C | 265 +++++++++----------- 1 file changed, 120 insertions(+), 145 deletions(-) diff --git a/plots/Volume/avtOpenGLTuvokVolumeRenderer.C b/plots/Volume/avtOpenGLTuvokVolumeRenderer.C index d134fc4a90..baf59b9f16 100644 --- a/plots/Volume/avtOpenGLTuvokVolumeRenderer.C +++ b/plots/Volume/avtOpenGLTuvokVolumeRenderer.C @@ -1,6 +1,6 @@ /***************************************************************************** * -* Copyright (c) 2000 - 2008, Lawrence Livermore National Security, LLC +* Copyright (c) 2000 - 2009, Lawrence Livermore National Security, LLC * Produced at the Lawrence Livermore National Laboratory * LLNL-CODE-400142 * All rights reserved. @@ -45,7 +45,6 @@ #include "avtOpenGLTuvokVolumeRenderer.h" -#include #include #include #include @@ -54,16 +53,21 @@ #include #include -#include -#include -#include - +#include +#include +#include +#include + +#include +#include +#include #include -#include -#include +#include +#include #include +#include +#include #include -#include // Don't warn if a function is unused. Useful for keeping a function static // even if it's only used while debugging. @@ -75,19 +79,14 @@ #endif static AbstrRenderer* create_renderer(const VolumeAttributes &); -static VolumeDatasetInfo *create_dataset_info(vtkRectilinearGrid *); +static tuvok::UnbrickedDSMetadata *create_dataset_info(vtkRectilinearGrid *); FQN_UNUSED static void debug_vtk_array(vtkDataArray *); -FQN_UNUSED static void debug_transfer_function(const std::vector& rgba); FQN_UNUSED static void debug_view(const avtViewInfo &); -static void initialize_glew(); static void tuvok_set_data(AbstrRenderer *, vtkRectilinearGrid *, vtkDataArray *, float *, size_t); static void tuvok_set_transfer_fqn(AbstrRenderer &, const VolumeAttributes &); static void tuvok_set_view(AbstrRenderer &, const avtViewInfo &); -static std::vector float_to_8bit(float *, size_t); - -static bool glew_initialized = false; // **************************************************************************** // Method: avtOpenGLTuvokVolumeRenderer::avtOpenGLTuvokVolumeRenderer @@ -103,14 +102,17 @@ static bool glew_initialized = false; // Connect the appropriate type of debug output. // NULL out our renderer, until we know what kind of one to make. // +// Tom Fogal, Tue Jul 7 12:01:34 MDT 2009 +// Use initializer list. +// // **************************************************************************** avtOpenGLTuvokVolumeRenderer::avtOpenGLTuvokVolumeRenderer() + : renderer(NULL) { Controller::Instance().AddDebugOut(new VisItDebugOut()); // enable tuvok logging output -- very slow, do not leave enabled! Controller::Debug::Out().SetOutput(true, true, true, true); - this->renderer = NULL; } // **************************************************************************** @@ -163,6 +165,10 @@ avtOpenGLTuvokVolumeRenderer::~avtOpenGLTuvokVolumeRenderer() // Tom Fogal, Thu Mar 19 00:14:16 MST 2009 // First pass at an implementation; camera settings are a bit off right now. // +// Tom Fogal, Sun Jul 26 15:22:47 MDT 2009 +// Second pass; take most of Brad's code and work it into the +// UnbrickedDataset methodology of setting data. +// // **************************************************************************** #if 0 @@ -176,7 +182,7 @@ avtOpenGLTuvokVolumeRenderer::Render( const avtVolumeRendererImplementation::RenderProperties &props, const avtVolumeRendererImplementation::VolumeData &volume) { - initialize_glew(); + avt::glew::initialize(); if(NULL == this->renderer) { @@ -244,7 +250,13 @@ avtOpenGLTuvokVolumeRenderer::Render( const avtVolumeRendererImplementation::RenderProperties &props, const avtVolumeRendererImplementation::VolumeData &volume) { - initialize_glew(); + avt::glew::initialize(); + + // Bit of a hack. Tuvok changes some sort of texture state, which is doing + // Bad Things (tm) to the rest of VisIt's rendering code. We'll need to + // track down and fix Tuvok at some point, but in the meantime pushing and + // popping our texture state will workaround the issue. + glPushAttrib(GL_TEXTURE_BIT); if(NULL == this->renderer) { @@ -259,7 +271,40 @@ avtOpenGLTuvokVolumeRenderer::Render( tuvok_set_view(*this->renderer, props.view); + this->renderer->SetGlobalBBox(true); + this->renderer->SetLocalBBox(true); + this->renderer->SetRenderCoordArrows(true); + + // Resize the renderer's buffer so it matches the window size. + this->renderer->Resize(UINTVECTOR2(props.windowSize[0],props.windowSize[1])); + + // Set the background color + FLOATVECTOR3 bg[2]; + bg[0] = FLOATVECTOR3( + props.backgroundColor[0], + props.backgroundColor[1], + props.backgroundColor[2]); + bg[1] = FLOATVECTOR3( + props.backgroundColor[0], + props.backgroundColor[1], + props.backgroundColor[2]); + this->renderer->SetBackgroundColors(bg); + +#if 0 + // Works well in the UI, but messes up testing via the CLI. Hrm. + GLfloat rmat[16]; + glGetFloatv(GL_MODELVIEW_MATRIX, rmat); + // Null out the translation components + rmat[12] = 0.f; + rmat[13] = 0.f; + rmat[14] = 0.f; + rmat[15] = 1.f; + this->renderer->SetRotation(rmat); +#endif + this->renderer->Paint(); + + glPopAttrib(); // fix texture state; see comment above PushAttrib. } #endif @@ -286,6 +331,9 @@ avtOpenGLTuvokVolumeRenderer::Render( // Rename so statics follow a consistent + distinct convention. Make a more // compatible renderer. // +// Tom Fogal, Fri May 1 18:11:20 MDT 2009 +// Updated for revised Tuvok API. Use RuntimeSettings to lookup shader dir. +// // **************************************************************************** static AbstrRenderer * create_renderer(const VolumeAttributes &) @@ -297,33 +345,40 @@ create_renderer(const VolumeAttributes &) // switch downsample to true if OpenGL crashes for you. const bool downsample = false; const bool disable_border = false; - const bool simple = true; + const bool no_clip_planes = false; // mac long shader workaround. + // Don't assume bit width relations between dataset and TFqns. + const bool bias_tfqn_scaling = true; MasterController &mc = Controller::Instance(); - AbstrRenderer *ren = mc.RequestNewVolumerenderer( + AbstrRenderer *ren = mc.RequestNewVolumeRenderer( MasterController::OPENGL_SBVR, use_only_PoT_textures, downsample, - disable_border, simple); - + disable_border, no_clip_planes, bias_tfqn_scaling + ); // We need to know where Tuvok stores its shaders, since it must load them - // at runtime. They're in our source tree, but of course we can't even - // assume that's present. We'll need to coordinate with copying the - // shaders around at install time, but for now we'll just use an - // environment variable. - if(!Environment::exists("TUVOK_SHADER_DIR")) - { - EXCEPTION1(ImproperUseException, - "Don't know where to find Tuvok Shaders! Please set the " - "TUVOK_SHADER_DIR environment variable."); + // at runtime. They should be placed relative to the VisIt binary, but use + // a RuntimeSetting to allow overrides. + const std::string shader_dir = RuntimeSetting::lookups("tuvok-shader-dir"); + debug5 << "Adding shader path: " << shader_dir << std::endl; + { // Make sure the shader path makes sense. + VisItStat_t statbuf; // ignored, just want the return val. + if(VisItStat(shader_dir.c_str(), &statbuf) != 0) + { + std::ostringstream dir_error; + dir_error << "Tuvok cannot find its shaders in '" << shader_dir << "'" + << "! Try using the --tuvoks-shaders command line option, " + << "or setting the VISIT_TUVOK_SHADER_DIR environment " + << "variable."; + EXCEPTION1(ImproperUseException, dir_error.str().c_str()); + } } - ren->AddShaderPath(Environment::get("TUVOK_SHADER_DIR").c_str()); - ren->SetDataSet(new CoreVolume()); + ren->AddShaderPath(shader_dir.c_str()); + ren->SetDataset(new tuvok::UnbrickedDataset()); + ren->SetGlobalBBox(true); + // Tuvok needs to know how big to make its FBOs. We'll resize it when we + // actually render, but make sure we have something for now. + ren->Resize(UINTVECTOR2(300,300)); ren->Initialize(); - // Tuvok needs to know how big to make its FBOs. We'll need to modify - // VisIt to somehow pass the information of the view window size down to - // here, but we need something for now. Make it huge to ensure our window - // doesn't exceed the FBO size. - ren->Resize(UINTVECTOR2(1200,1200)); // HACK! ren->SetRendermode(AbstrRenderer::RM_1DTRANS); ren->SetUseLighting(false); ren->SetBlendPrecision(AbstrRenderer::BP_8BIT); @@ -340,8 +395,7 @@ create_renderer(const VolumeAttributes &) // // Purpose: Translates our/vtk's metadata into the object Tuvok wants it as. // -// Returns: An object which should be given to a VolumeDataset to replace its -// existing metadata. +// Returns: An object representing metadata as tuvok understands it. // // Programmer: Tom Fogal // Creation: Fri Mar 6 11:23:52 MST 2009 @@ -351,11 +405,14 @@ create_renderer(const VolumeAttributes &) // Tom Fogal, Wed Mar 18 21:47:31 MST 2009 // Rename so statics follow a consistent + distinct convention. // +// Tom Fogal, Fri May 1 20:23:13 MDT 2009 +// Updated for revised Tuvok API. +// // **************************************************************************** -static VolumeDatasetInfo * +static tuvok::UnbrickedDSMetadata * create_dataset_info(vtkRectilinearGrid *grid) { - CoreVolumeInfo *vds_info = new CoreVolumeInfo(); + tuvok::UnbrickedDSMetadata *vds_info = new tuvok::UnbrickedDSMetadata(); { int dims[3]; grid->GetDimensions(dims); @@ -370,37 +427,6 @@ create_dataset_info(vtkRectilinearGrid *grid) return vds_info; } -// **************************************************************************** -// Function: initialize_glew -// -// Purpose: Does a one-time GLEW initialization. -// -// Programmer: Tom Fogal -// Creation: Fri Mar 6 14:03:22 MST 2009 -// -// Modifications: -// -// **************************************************************************** -static void -initialize_glew() -{ - if(!glew_initialized) { -#if 1 - GLenum err = glewInit(); -#else - // Lets one use HW rendering in serial w/ -nowin. - GLenum err = glewInitLibrary("/usr/lib/libGL.so", - GLEW_NAME_CONVENTION_GL); -#endif - if(GLEW_OK != err) { - debug1 << "GLEW initialization failed: " << glewGetErrorString(err) - << std::endl; - } else { - glew_initialized = true; - } - } -} - /// prints VTK array information to the debug stream. static void debug_vtk_array(vtkDataArray *arr) @@ -427,22 +453,6 @@ lerp(in value, in imin, in imax, out omin, out omax) return ret; } -static void -debug_transfer_function(const std::vector& rgba) -{ -#define UC(x) static_cast(x) - for(size_t i=0; i < rgba.size(); i+=4) - { - debug5 << "tf(" << setw(3) << i/4 << "): " - << setw(5) << setprecision(3) << setfill(' ') - << setiosflags(std::ios_base::right) - << lerp(rgba[i+0], UC(0),UC(255), 0.f,1.f) << ", " - << lerp(rgba[i+1], UC(0),UC(255), 0.f,1.f) << ", " - << lerp(rgba[i+2], UC(0),UC(255), 0.f,1.f) << ", " - << lerp(rgba[i+3], UC(0),UC(255), 0.f,1.f) << std::endl; - } -} - static void dbg_4x4_matrix(const char *pfx, const GLfloat m[16]) { @@ -496,33 +506,28 @@ debug_view(const avtViewInfo &v) // // Modifications: // +// Tom Fogal, Fri May 1 20:25:02 MDT 2009 +// Updated for revised Tuvok API. Get rid of 8bit case, no longer relevant. +// // **************************************************************************** static void tuvok_set_data(AbstrRenderer *ren, vtkRectilinearGrid *grid, vtkDataArray *data, float *gr_mag, size_t n_mag) { // base class doesn't have these `Set' methods; we can only set this kind - // of thing manually from the subclass we'll be using (for now). - CoreVolume &vol = dynamic_cast(ren->GetDataSet()); - - vol.SetInfo(create_dataset_info(grid)); + // of thing manually from the subclass we'll be using. + tuvok::UnbrickedDataset &vol = dynamic_cast + (ren->GetDataset()); + vol.SetMetadata(create_dataset_info(grid)); vol.SetGradientMagnitude(gr_mag, n_mag); debug_vtk_array(data); // This will break if we're not given float data, but that's guaranteed // for now! -#if 0 + // We could also send down unsigned char data at this point. Anything else + // would require some simple but non-zero amount of work inside Tuvok. vol.SetData(static_cast(data->GetVoidPointer(0)), data->GetNumberOfTuples()); -#else - debug5 << "Converting to 32bit data to 8bit data." << std::endl; - std::vector eight_bit_data; - eight_bit_data = float_to_8bit( - static_cast(data->GetVoidPointer(0)), - data->GetNumberOfTuples() - ); - vol.SetData(&eight_bit_data.at(0), data->GetNumberOfTuples()); -#endif } // **************************************************************************** @@ -535,6 +540,9 @@ tuvok_set_data(AbstrRenderer *ren, vtkRectilinearGrid *grid, // // Modifications: // +// Tom Fogal, Sun Jul 26 15:30:33 MDT 2009 +// New, simpler API. Disable (very) verbose debugging. +// // **************************************************************************** static void tuvok_set_transfer_fqn(AbstrRenderer &ren, const VolumeAttributes &atts) @@ -542,11 +550,7 @@ tuvok_set_transfer_fqn(AbstrRenderer &ren, const VolumeAttributes &atts) std::vector rgba(256*4); atts.GetTransferFunction(&rgba.at(0)); - // TF seems to be working; don't spam the logs right now. - //debug_transfer_function(rgba); - - TransferFunction1D *tf = ren.Get1DTrans(); - tf->Set(rgba); + ren.Set1DTrans(rgba); // Ensure tuvok knows that it must re-upload the TF to the GPU. ren.Changed1DTrans(); } @@ -584,9 +588,6 @@ tuvok_set_view(AbstrRenderer &ren, const avtViewInfo &v) eye[1] = Eye[1]; eye[2] = Eye[2]; - // This API is going to change, as you might guess from the code. But the - // information given will remain the same; only how we grab the object && - // the function name will differ. // The arguments this API expects are: // field of view // near plane Z value @@ -595,45 +596,19 @@ tuvok_set_view(AbstrRenderer &ren, const avtViewInfo &v) // reference point in world coordinates // view up vector // The simplest way to think of this data is `the stuff one would pass to - // "gluLookAt" and "gluPerspective"' in any other app. VisIt's notion of a - // camera seems to be a bit different ... - ImmediateGLSBVR& glren = - dynamic_cast(ren); - glren.Hack( + // "gluLookAt" and "gluPerspective"' in any other app. + ren.SetViewParameters( static_cast(v.viewAngle), static_cast(v.nearPlane), static_cast(v.farPlane), eye, ref, vup ); -} - -// **************************************************************************** -// Function: float_to_8bit -// -// Purpose: Temporary hack to convert an FP dataset to an 8bit dataset. -// Does so in an absolutely terrible way. -// -// Programmer: Tom Fogal -// Creation: Wed Mar 25 12:33:53 MST 2009 -// -// Modifications: -// -// **************************************************************************** -struct lerpf_8 : std::unary_function { - unsigned char operator()(float f) const { - return lerp(f, std::numeric_limits::min(), - std::numeric_limits::max(), - std::numeric_limits::min(), - std::numeric_limits::max()); - } -}; - -static std::vector -float_to_8bit(float *data, size_t v) -{ - std::vector ret(v); - std::transform(data, data + v, ret.begin(), lerpf_8()); - return ret; +#if 1 + eye[0] = 0.f; eye[1] = 0.f; eye[2] = 1.6f; + ref[0] = 0.f; ref[1] = 0.f; ref[2] = 0.f; + vup[0] = 0.f; vup[1] = 1.f; vup[2] = 0.f; + ren.SetViewParameters(50.0f, 0.1f, 100.0f, eye,ref,vup); +#endif } #endif // USE_TUVOK -- 2.24.1