Working thread impl

This commit is contained in:
Stephen Seo 2021-08-23 21:19:15 +09:00
parent 61e6cf17fd
commit ade12df080

View file

@ -3,6 +3,8 @@
#include <cmath>
#include <fstream>
#include <array>
#include <thread>
#include <mutex>
#include <glm/matrix.hpp>
#include <glm/gtc/matrix_transform.hpp>
@ -316,74 +318,107 @@ Ex02::RT::Image Ex02::RT::renderColorsWithSpheres(
lights[2].falloffStart = 3.0f;
lights[2].falloffEnd = 7.0f;
if(threadCount <= 1) {
for(unsigned int j = 0; j < outputHeight; ++j) {
float offsetY = ((float)j + 0.5f - ((float)outputHeight / 2.0f));
for(unsigned int i = 0; i < outputWidth; ++i) {
float offsetX = ((float)i + 0.5f - ((float)outputWidth / 2.0f));
glm::vec3 rayDir{
offsetX,
offsetY,
-(float)outputHeight * EX02_RAY_TRACER_VIEW_RATIO};
glm::vec3 rayDirUnit = rayDir / std::sqrt(
rayDir.x * rayDir.x
+ rayDir.y * rayDir.y
+ rayDir.z * rayDir.z);
const auto yIteration = [&spheres, &lights, &image, outputWidth, outputHeight, rayPos] (unsigned int j, std::mutex *mutex) {
float offsetY = ((float)j + 0.5f - ((float)outputHeight / 2.0f));
for(unsigned int i = 0; i < outputWidth; ++i) {
float offsetX = ((float)i + 0.5f - ((float)outputWidth / 2.0f));
glm::vec3 rayDir{
offsetX,
offsetY,
-(float)outputHeight * EX02_RAY_TRACER_VIEW_RATIO};
glm::vec3 rayDirUnit = rayDir / std::sqrt(
rayDir.x * rayDir.x
+ rayDir.y * rayDir.y
+ rayDir.z * rayDir.z);
// cast ray to all spheres, finding closest result
std::optional<std::tuple<glm::vec3, float, unsigned int>> closestResult;
for(unsigned int idx = 0; idx < spheres.size(); ++idx) {
auto result = spheres[idx].rayToSphere(rayPos, rayDirUnit);
if(result) {
float dist = Internal::distBetweenPositions(
rayPos, spheres[idx].pos);
if(closestResult) {
if(dist < std::get<1>(*closestResult)) {
closestResult = {{*result, dist, idx}};
}
} else {
// cast ray to all spheres, finding closest result
std::optional<std::tuple<glm::vec3, float, unsigned int>> closestResult;
for(unsigned int idx = 0; idx < spheres.size(); ++idx) {
auto result = spheres[idx].rayToSphere(rayPos, rayDirUnit);
if(result) {
float dist = Internal::distBetweenPositions(
rayPos, spheres[idx].pos);
if(closestResult) {
if(dist < std::get<1>(*closestResult)) {
closestResult = {{*result, dist, idx}};
}
} else {
closestResult = {{*result, dist, idx}};
}
}
}
if(!closestResult) {
if(!closestResult) {
continue;
}
// cast ray to each light checking if colliding with other
// spheres
for(const auto &light : lights) {
glm::vec3 toLight = light.pos - std::get<0>(*closestResult);
glm::vec3 toLightUnit = toLight / std::sqrt(
toLight.x * toLight.x
+ toLight.y * toLight.y
+ toLight.z * toLight.z);
bool isBlocked = false;
for(unsigned int idx = 0; idx < spheres.size(); ++idx) {
if(idx == std::get<2>(*closestResult)) {
continue;
}
auto result = spheres[idx].rayToSphere(
std::get<0>(*closestResult),
toLightUnit);
if(result) {
isBlocked = true;
break;
}
}
if(isBlocked) {
continue;
}
// cast ray to each light checking if colliding with other
// spheres
for(const auto &light : lights) {
glm::vec3 toLight = light.pos - std::get<0>(*closestResult);
glm::vec3 toLightUnit = toLight / std::sqrt(
toLight.x * toLight.x
+ toLight.y * toLight.y
+ toLight.z * toLight.z);
bool isBlocked = false;
for(unsigned int idx = 0; idx < spheres.size(); ++idx) {
if(idx == std::get<2>(*closestResult)) {
continue;
}
auto result = spheres[idx].rayToSphere(
std::get<0>(*closestResult),
toLightUnit);
if(result) {
isBlocked = true;
break;
}
}
if(isBlocked) {
continue;
}
// at this point, it is known that no spheres blocks ray
// to light
// at this point, it is known that no spheres blocks ray
// to light
if(mutex) {
std::lock_guard<std::mutex> lock(*mutex);
light.applyLight(
std::get<0>(*closestResult),
image.getPixel(i, j));
} else {
light.applyLight(
std::get<0>(*closestResult),
image.getPixel(i, j));
}
}
}
};
if(threadCount <= 1) {
for(unsigned int j = 0; j < outputHeight; ++j) {
yIteration(j, nullptr);
}
} else {
std::vector<std::thread> threads;
std::mutex mutex;
unsigned int range = outputHeight / threadCount;
for(unsigned int threadIdx = 0; threadIdx < threadCount; ++threadIdx) {
unsigned int start = range * threadIdx;
unsigned int end = range * (threadIdx + 1);
if(threadIdx + 1 == threadCount) {
end = outputHeight;
}
threads.emplace_back(std::thread([&yIteration] (unsigned int start, unsigned int end, std::mutex *mutex) {
for(unsigned int y = start; y < end; ++y) {
yIteration(y, mutex);
}
},
start,
end,
&mutex));
}
for(std::thread &thread : threads) {
thread.join();
}
}
return image;