优化 OpenGL 计算着色器
问题描述 投票：0回答：0
一直在尝试制作自己的游戏，现在正在研究PBR Bloom Effect。我做了一个简单的布隆测试应用程序并试图让它变得更好。因此，我对此有几个问题。
这是我的代码。 注意：我更喜欢计算着色器而不是离屏渲染，以便更快地绽放。
#include "main.h"
#include "FileIO.h"
#include "stb_image.h"
#include "glm/gtc/type_ptr.hpp"
int main(void)
{
if (!InitGLFW())
return -1;
glEnable(GL_DEPTH_TEST);
//glPolygonMode(GL_FRONT_AND_BACK, GL_LINE);
glEnable(GL_CULL_FACE);
glEnable(GL_FRAMEBUFFER_SRGB);
Meshes::MeshProperties prop;
Meshes::loadMeshFromFile("resource\\mesh\\StandartCube.mesh", &prop);
glm::mat4 projM = glm::perspective(glm::radians(45.0f), (float)SCREEN_WIDTH / (float)SCREEN_HEIGHT, 0.1f, 25.0f);
glm::mat4 viewM;
unsigned int uniformBuffer;
glCreateBuffers(1, &uniformBuffer);
glNamedBufferData(uniformBuffer, 2 * sizeof(glm::mat4), nullptr, GL_DYNAMIC_DRAW);
glNamedBufferSubData(uniformBuffer, 0, sizeof(glm::mat4), &projM);
glBindBufferBase(GL_UNIFORM_BUFFER, 0, uniformBuffer);
float screenBox[20] =
{
-1.0f, -1.0f, 0.0f, 0.0f, 0.0f,
1.0f, -1.0f, 0.0f, 1.0f, 0.0f,
1.0f, 1.0f, 0.0f, 1.0f, 1.0f,
-1.0f, 1.0f, 0.0f, 0.0f, 1.0f,
};
unsigned int boxIndices[6] =
{
0,1,2,
2,3,0
};
unsigned int boxVertexArrayID, boxVertexBufferID, boxIndexBufferID;
glCreateVertexArrays(1, &boxVertexArrayID);
glCreateBuffers(1, &boxVertexBufferID);
glCreateBuffers(1, &boxIndexBufferID);
glNamedBufferData(boxVertexBufferID, sizeof(screenBox), screenBox, GL_STATIC_DRAW);
glNamedBufferData(boxIndexBufferID, sizeof(boxIndices), boxIndices, GL_STATIC_DRAW);
glEnableVertexArrayAttrib(boxVertexArrayID, 0);
glVertexArrayAttribBinding(boxVertexArrayID, 0, 0);
glVertexArrayAttribFormat(boxVertexArrayID, 0, 3, GL_FLOAT, GL_FALSE, 0);
glEnableVertexArrayAttrib(boxVertexArrayID, 1);
glVertexArrayAttribBinding(boxVertexArrayID, 1, 0);
glVertexArrayAttribFormat(boxVertexArrayID, 1, 2, GL_FLOAT, GL_FALSE, 12);
glVertexArrayVertexBuffer(boxVertexArrayID, 0, boxVertexBufferID, 0, 20);
glVertexArrayElementBuffer(boxVertexArrayID, boxIndexBufferID);
unsigned int vertexArrayID, vertexBufferID, indexBufferID;
glCreateVertexArrays(1, &vertexArrayID);
glCreateBuffers(1, &vertexBufferID);
glCreateBuffers(1, &indexBufferID);
glNamedBufferData(vertexBufferID, prop.getVertexSize(), prop.vertexData, GL_STATIC_DRAW);
glNamedBufferData(indexBufferID, prop.getIndexSize(), prop.indexData, GL_STATIC_DRAW);
glEnableVertexArrayAttrib(vertexArrayID, 0);
glVertexArrayAttribBinding(vertexArrayID, 0, 0);
glVertexArrayAttribFormat(vertexArrayID, 0, 3, GL_FLOAT, GL_FALSE, 0);
glEnableVertexArrayAttrib(vertexArrayID, 1);
glVertexArrayAttribBinding(vertexArrayID, 1, 0);
glVertexArrayAttribFormat(vertexArrayID, 1, 2, GL_FLOAT, GL_FALSE, 12);
glEnableVertexArrayAttrib(vertexArrayID, 2);
glVertexArrayAttribBinding(vertexArrayID, 2, 0);
glVertexArrayAttribFormat(vertexArrayID, 2, 3, GL_FLOAT, GL_FALSE, 20);
glVertexArrayVertexBuffer(vertexArrayID, 0, vertexBufferID, 0, 32);
glVertexArrayElementBuffer(vertexArrayID, indexBufferID);
// --------------------------------------------
unsigned int FBOCAPTURE, RBO, tex, finalTexture;
glCreateTextures(GL_TEXTURE_2D, 1, &tex);
glTextureParameteri(tex, GL_TEXTURE_MIN_FILTER, GL_LINEAR);
glTextureParameteri(tex, GL_TEXTURE_MAG_FILTER, GL_LINEAR);
glTextureParameteri(tex, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE);
glTextureParameteri(tex, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE);
glTextureStorage2D(tex, 1, GL_R11F_G11F_B10F, FBO_SIZE, FBO_SIZE);
glBindTextureUnit(0, tex);
glCreateTextures(GL_TEXTURE_2D, 1, &finalTexture);
glTextureParameteri(finalTexture, GL_TEXTURE_MIN_FILTER, GL_LINEAR);
glTextureParameteri(finalTexture, GL_TEXTURE_MAG_FILTER, GL_LINEAR);
glTextureParameteri(finalTexture, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE);
glTextureParameteri(finalTexture, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE);
glTextureStorage2D(finalTexture, 1, GL_R11F_G11F_B10F, FBO_SIZE, FBO_SIZE);
glBindTextureUnit(1, finalTexture);
glCreateRenderbuffers(1, &RBO);
glNamedRenderbufferStorage(RBO, GL_DEPTH24_STENCIL8, FBO_SIZE, FBO_SIZE);
glCreateFramebuffers(1, &FBOCAPTURE);
glNamedFramebufferTexture(FBOCAPTURE, GL_COLOR_ATTACHMENT0, tex, 0);
glNamedFramebufferTexture(FBOCAPTURE, GL_COLOR_ATTACHMENT1, finalTexture, 0);
unsigned int drawBuffers[2] = {GL_COLOR_ATTACHMENT0, GL_COLOR_ATTACHMENT1};
glNamedFramebufferDrawBuffers(FBOCAPTURE, 2, drawBuffers);
glNamedFramebufferRenderbuffer(FBOCAPTURE, GL_DEPTH_STENCIL_ATTACHMENT, GL_RENDERBUFFER, RBO);
unsigned int bloomTextures[5];
glCreateTextures(GL_TEXTURE_2D, 5, bloomTextures);
unsigned int size[2] = { FBO_SIZE, FBO_SIZE};
for (unsigned int i = 0; i < 5; i++)
{
size[0] /= 2;
size[1] /= 2;
glTextureParameteri(bloomTextures[i], GL_TEXTURE_MIN_FILTER, GL_LINEAR);
glTextureParameteri(bloomTextures[i], GL_TEXTURE_MAG_FILTER, GL_LINEAR);
glTextureParameteri(bloomTextures[i], GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE);
glTextureParameteri(bloomTextures[i], GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE);
glTextureStorage2D(bloomTextures[i], 1, GL_R11F_G11F_B10F, size[0], size[1]);
}
unsigned int cubeProgram = glCreateProgram();
glAttachShader(cubeProgram, Shaders::loadShaderfromFile("resource\\cubeVertex.shader", GL_VERTEX_SHADER));
glAttachShader(cubeProgram, Shaders::loadShaderfromFile("resource\\cubeFragment.shader", GL_FRAGMENT_SHADER));
glLinkProgram(cubeProgram);
glValidateProgram(cubeProgram);
unsigned int downSampler = glCreateProgram();
glAttachShader(downSampler, Shaders::loadShaderfromFile("resource\\downSampler.shader", GL_COMPUTE_SHADER));
glLinkProgram(downSampler);
unsigned int upSampler = glCreateProgram();
glAttachShader(upSampler, Shaders::loadShaderfromFile("resource\\upSampler.shader", GL_COMPUTE_SHADER));
glLinkProgram(upSampler);
unsigned int finalProgram = glCreateProgram();
glAttachShader(finalProgram, Shaders::loadShaderfromFile("resource\\samplerVertex.shader", GL_VERTEX_SHADER));
glAttachShader(finalProgram, Shaders::loadShaderfromFile("resource\\finalFragment.shader", GL_FRAGMENT_SHADER));
glLinkProgram(finalProgram);
glValidateProgram(finalProgram);
glUseProgram(finalProgram);
glUniform1i(glGetUniformLocation(finalProgram, "texBloom"), 0);
glUniform1i(glGetUniformLocation(finalProgram, "texHDR"), 1);
glUniform1f(glGetUniformLocation(finalProgram, "exposure"), 1.0f);
glUniform1f(glGetUniformLocation(finalProgram, "bloomStrength"), 0.4f);
while (GLenum error = glGetError())
{
std::cout << error << std::endl;
}
while (!glfwWindowShouldClose(window))
{
//Input process
currentFrame = (float)glfwGetTime();
deltaFrame = currentFrame - lastFrame;
lastFrame = currentFrame;
KeyboardInputs(2.0f);
viewM = glm::lookAt(cameraPos, cameraPos + cameraFront, cameraUp);
glNamedBufferSubData(uniformBuffer, sizeof(glm::mat4), sizeof(glm::mat4), &viewM);
//Render
glViewport(0, 0, FBO_SIZE, FBO_SIZE);
glBindFramebuffer(GL_FRAMEBUFFER, FBOCAPTURE);
glClear(GL_COLOR_BUFFER_BIT | GL_DEPTH_BUFFER_BIT);
glUseProgram(cubeProgram);
glBindVertexArray(vertexArrayID);
glDrawElements(GL_TRIANGLES, 36, GL_UNSIGNED_INT, nullptr);
glBindFramebuffer(GL_FRAMEBUFFER, NULL);
//DownSampling
glUseProgram(downSampler);
glBindImageTexture(0, tex, 0, GL_FALSE, 0, GL_READ_WRITE, GL_R11F_G11F_B10F);
glBindImageTexture(1, bloomTextures[0], 0, GL_FALSE, 0, GL_READ_WRITE, GL_R11F_G11F_B10F);
glDispatchCompute(FBO_SIZE / 16, FBO_SIZE / 8, 1);
glMemoryBarrier(GL_SHADER_IMAGE_ACCESS_BARRIER_BIT);
glBindImageTexture(0, bloomTextures[0], 0, GL_FALSE, 0, GL_READ_WRITE, GL_R11F_G11F_B10F);
glBindImageTexture(1, bloomTextures[1], 0, GL_FALSE, 0, GL_READ_WRITE, GL_R11F_G11F_B10F);
glDispatchCompute(FBO_SIZE / 32, FBO_SIZE / 16, 1);
glMemoryBarrier(GL_SHADER_IMAGE_ACCESS_BARRIER_BIT);
glBindImageTexture(0, bloomTextures[1], 0, GL_FALSE, 0, GL_READ_WRITE, GL_R11F_G11F_B10F);
glBindImageTexture(1, bloomTextures[2], 0, GL_FALSE, 0, GL_READ_WRITE, GL_R11F_G11F_B10F);
glDispatchCompute(FBO_SIZE / 64, FBO_SIZE / 32, 1);
glMemoryBarrier(GL_SHADER_IMAGE_ACCESS_BARRIER_BIT);
glBindImageTexture(0, bloomTextures[2], 0, GL_FALSE, 0, GL_READ_WRITE, GL_R11F_G11F_B10F);
glBindImageTexture(1, bloomTextures[3], 0, GL_FALSE, 0, GL_READ_WRITE, GL_R11F_G11F_B10F);
glDispatchCompute(FBO_SIZE / 128, FBO_SIZE / 64, 1);
glMemoryBarrier(GL_SHADER_IMAGE_ACCESS_BARRIER_BIT);
glBindImageTexture(0, bloomTextures[3], 0, GL_FALSE, 0, GL_READ_WRITE, GL_R11F_G11F_B10F);
glBindImageTexture(1, bloomTextures[4], 0, GL_FALSE, 0, GL_READ_WRITE, GL_R11F_G11F_B10F);
glDispatchCompute(FBO_SIZE / 256, FBO_SIZE / 128, 1);
glMemoryBarrier(GL_SHADER_IMAGE_ACCESS_BARRIER_BIT);
//UpSampling
glUseProgram(upSampler);
glBindImageTexture(0, bloomTextures[4], 0, GL_FALSE, 0, GL_READ_WRITE, GL_R11F_G11F_B10F);
glBindImageTexture(1, bloomTextures[3], 0, GL_FALSE, 0, GL_READ_WRITE, GL_R11F_G11F_B10F);
glDispatchCompute(FBO_SIZE / 128, FBO_SIZE / 64, 1);
glMemoryBarrier(GL_SHADER_IMAGE_ACCESS_BARRIER_BIT);
glBindImageTexture(0, bloomTextures[3], 0, GL_FALSE, 0, GL_READ_WRITE, GL_R11F_G11F_B10F);
glBindImageTexture(1, bloomTextures[2], 0, GL_FALSE, 0, GL_READ_WRITE, GL_R11F_G11F_B10F);
glDispatchCompute(FBO_SIZE / 64, FBO_SIZE / 32, 1);
glMemoryBarrier(GL_SHADER_IMAGE_ACCESS_BARRIER_BIT);
glBindImageTexture(0, bloomTextures[2], 0, GL_FALSE, 0, GL_READ_WRITE, GL_R11F_G11F_B10F);
glBindImageTexture(1, bloomTextures[1], 0, GL_FALSE, 0, GL_READ_WRITE, GL_R11F_G11F_B10F);
glDispatchCompute(FBO_SIZE / 32, FBO_SIZE / 16, 1);
glMemoryBarrier(GL_SHADER_IMAGE_ACCESS_BARRIER_BIT);
glBindImageTexture(0, bloomTextures[1], 0, GL_FALSE, 0, GL_READ_WRITE, GL_R11F_G11F_B10F);
glBindImageTexture(1, bloomTextures[0], 0, GL_FALSE, 0, GL_READ_WRITE, GL_R11F_G11F_B10F);
glDispatchCompute(FBO_SIZE / 16, FBO_SIZE / 8, 1);
glMemoryBarrier(GL_SHADER_IMAGE_ACCESS_BARRIER_BIT);
glBindImageTexture(0, bloomTextures[0], 0, GL_FALSE, 0, GL_READ_WRITE, GL_R11F_G11F_B10F);
glBindImageTexture(1, tex, 0, GL_FALSE, 0, GL_READ_WRITE, GL_R11F_G11F_B10F);
glDispatchCompute(FBO_SIZE / 8, FBO_SIZE / 4, 1);
glMemoryBarrier(GL_SHADER_IMAGE_ACCESS_BARRIER_BIT);
glViewport(0, 0, SCREEN_WIDTH, SCREEN_HEIGHT);
glClear(GL_COLOR_BUFFER_BIT | GL_DEPTH_BUFFER_BIT);
glUseProgram(finalProgram);
glBindVertexArray(boxVertexArrayID);
glDrawElements(GL_TRIANGLES, 6, GL_UNSIGNED_INT, nullptr);
glfwSwapBuffers(window);
glfwPollEvents();
}
glfwTerminate();
return 0;
}
void key_callback(GLFWwindow* window, int key, int scancode, int action, int mods)
{
if (action == GLFW_PRESS)
{
switch (key)
{
case GLFW_KEY_ESCAPE:
glfwSetWindowShouldClose(window, true);
break;
}
}
}
我的问题是，
- 无论是否使用glMemoryBarrier，为什么我都无法使用async compute flight？
- PS Warp Can't Launch部分有一个峰值，请问有什么办法可以避免，还是正常的？
- 在我的计算着色器中，我使用 (Usage1)x = 4, y = 4, z = 1 次调用。我应该使用 (Usage2)x = 8, y = 4, z = 1 来使用所有的扭曲吗？
- 而且我认为情况很奇怪。 当我尝试使用像 Usage1 这样的调用时，我的 GPU 使用率有所下降（与 Usage2 相比）但是我的 GPU 内存时钟频率在 800Mhz-6000Mhz 之间不断下降和上升？我不知道为什么？
最新问题
- 分类瀑布图未正确堆叠
- 无法为 SVG 组件设置 className 属性
- 如果报告表中存在报告，如何选择批准日期？
- 我们可以在VBA中使用条件编译在函数内声明变量吗？
- Python 脚本不会重命名文件夹中最后一个 .xlsx 文件
- CameraX - 设置 LENS_FOCUS_DISTANCE 在 Samsung S22 Ultra 上不起作用
- 使用 Excel VBA 抓取 Investing.com
- 将元框添加到管理 WooCommerce 产品（如果它具有特定产品标签）
- 如何使用Python 3.11在Raspberry Pi 5上安装SMB库
- 将 gtable 转换为 R ggplot2 中的 ggplot
- 通过 Snaplogic Rest API 连接到 HubSpot
- laravel 11 应用程序在自定义VerifyCsrfToken 中间件上失败
- _hyperscript 中的滚动事件监听器
- Alembic 看不到 SQLModel 模型并创建空迁移
- 检查产品（帖子）是否具有特定标签 - 如何进行？
- couchbase 7.2 升级后数据库调用 java 代码的测试覆盖率
- 如何在 Pyspark 中使用复杂数据类型
- 忽略ggplot2中的箱小提琴图中的异常值
- 请求权限提示未向用户显示 android
- 如何使用yubikey在GCP中启用MFA到超级管理员帐户
© www.soinside.com 2019 - 2024. All rights reserved.