rendering-in-cgi/Assignments/Assignment3/application_integrator.cpp
2024-06-27 14:05:57 +02:00

191 lines
6.3 KiB
C++

#include "application_integrator.h"
#include "algorithms/parallel_for.h"
#include "imgui.h"
#include "math/vec3fa.h"
#include "random_sampler.hpp"
#include "random_sampler_wrapper.hpp"
#include "sampler.h"
#include "tasking/taskschedulerinternal.h"
#include <atomic>
#include <mutex>
ApplicationIntegrator::ApplicationIntegrator(int argc, char **argv,
const std::string &name)
: Application(argc, argv, name) {
resetRender();
}
void ApplicationIntegrator::drawGUI() {
bool bDirty = false;
if (ImGui::SliderInt("chain lengths", &chain_lengths, 1, 200000))
resetRender();
if (ImGui::SliderFloat("small step size", &small_step_size, 0.0, 0.1))
resetRender();
if (ImGui::SliderFloat("large step probability", &large_step_probability, 0.0,
1.0))
resetRender();
if (ImGui::Checkbox("Metropolis", &bMetropolis)) {
resetRender();
}
if (bDirty) {
resetRender();
}
}
inline float luminance(Vec3fa v) {
return 0.2126f * v.x + 0.7152f * v.y + 0.0722f * v.z;
}
void ApplicationIntegrator::resetRender() {
Application::resetRender();
luminance_count.store(0, std::memory_order_relaxed);
std::lock_guard<std::mutex> guard(large_step_global_luminance_mutex);
large_step_global_luminance = 0.0;
frame_count = 0;
for (size_t i = 0; i < NUM_CHAINS; i++) {
chains[i] = MLTRandomSampler(small_step_size, large_step_probability);
chains[i].init(i);
}
if (bMetropolis) {
data.film.count = false;
} else {
data.film.count = true;
data.film.scalar = 1.0;
}
}
void ApplicationIntegrator::render(int *pixels, int width, int height,
float time, const ISPCCamera &camera) {
deviceRender(camera);
if (!bMetropolis) {
mcRender(pixels, width, height, time, camera);
} else {
mltRender(pixels, width, height, time, camera);
}
}
void ApplicationIntegrator::mltRender(int *pixels, int width, int height,
float time, const ISPCCamera &camera) {
// data.film.scalar = ... use it for setting up the correct normalization
// coefficient
//
//
// you may want to use Distribution1D for the bootstrap
// d = Distribution1D(float* bis_values, num_bins)
// float integral = d.funcInt;
// int index_of_the_sampled_bin = d.SampleDiscrete(rng.get1D());
parallel_for(size_t(0), size_t(NUM_CHAINS), [&](const range<size_t> &range) {
const int threadIndex = (int)TaskScheduler::threadIndex();
float large_luminance_outer_sum = 0.0;
for (size_t i = range.begin(); i < range.end(); i++) {
float large_luminance_sum = 0.0;
for (size_t j = 0; j < chain_lengths; j++) {
chains[i].new_ray(RandomSampler_get1D(chains[i].sampler) < large_step_probability);
float x = chains[i].get1D() * width;
float y = chains[i].get1D() * height;
int x_pixel = x;
int y_pixel = y;
Vec3f f = renderPixel(x, y, camera, g_stats[threadIndex], chains[i]);
float l = luminance(f);
if (chains[i].is_large_step()) {
large_luminance_sum += l;
luminance_count.fetch_add(1, std::memory_order_relaxed);
}
if (l > 0.0) {
data.film.addSplat(x_pixel, y_pixel, (f / l) * std::fmin(l / last_l[i], 1.0));
}
if ((last_l[i] == 0.0 && l > 0.0) || (last_l[i] > 0.0 && RandomSampler_get1D(chains[i].sampler) < l / last_l[i])) {
// if (i == 0)
// printf("%d accept %d %f %f %f\n", i, chains[i].is_large_step(), last_l[i], l, l / last_l[i]);
chains[i].accept();
last_l[i] = l;
}
}
large_luminance_outer_sum += large_luminance_sum;
}
std::lock_guard<std::mutex> guard(large_step_global_luminance_mutex);
large_step_global_luminance += large_luminance_outer_sum;
});
frame_count++;
std::lock_guard<std::mutex> guard(large_step_global_luminance_mutex);
data.film.scalar = (float)(width * height) * large_step_global_luminance / ((NUM_CHAINS * frame_count * chain_lengths) * luminance_count.load(std::memory_order_relaxed));
printf("%zu, %f, %f, %zu, %d\n", frame_count, data.film.scalar, large_step_global_luminance / luminance_count.load(std::memory_order_relaxed), luminance_count.load(std::memory_order_relaxed), NUM_CHAINS);
}
void ApplicationIntegrator::mcRender(int *pixels, int width, int height,
float time, const ISPCCamera &camera) {
const int numTilesX = (width + TILE_SIZE_X - 1) / TILE_SIZE_X;
const int numTilesY = (height + TILE_SIZE_Y - 1) / TILE_SIZE_Y;
parallel_for(size_t(0), size_t(numTilesX * numTilesY),
[&](const range<size_t> &range) {
const int threadIndex = (int)TaskScheduler::threadIndex();
for (size_t i = range.begin(); i < range.end(); i++)
renderTile((int)i, threadIndex, pixels, width, height, time,
camera, numTilesX, numTilesY);
});
}
/* renders a single screen tile */
void ApplicationIntegrator::mcRenderTile(int taskIndex, int threadIndex,
int *pixels, const unsigned int width,
const unsigned int height, const float time,
const ISPCCamera &camera,
const int numTilesX,
const int numTilesY) {
const unsigned int tileY = taskIndex / numTilesX;
const unsigned int tileX = taskIndex - tileY * numTilesX;
const unsigned int x0 = tileX * TILE_SIZE_X;
const unsigned int x1 = min(x0 + TILE_SIZE_X, width);
const unsigned int y0 = tileY * TILE_SIZE_Y;
const unsigned int y1 = min(y0 + TILE_SIZE_Y, height);
for (unsigned int y = y0; y < y1; y++)
for (unsigned int x = x0; x < x1; x++) {
RandomSamplerWrapper sampler;
Vec3fa L = Vec3fa(0.0f);
for (int i = 0; i < data.spp; i++) {
sampler.init(x, y, (data.frame_count) * data.spp + i);
/* calculate pixel color */
float fx = x + sampler.get1D();
float fy = y + sampler.get1D();
L = L + renderPixel(fx, fy, camera, g_stats[threadIndex], sampler);
}
L = L / (float)data.spp;
/* write color to framebuffer */
data.film.addSplat(x, y, L);
}
}