nn1997729的博客

"C_Cpp.formatting": "clangFormat", "C_Cpp.intelliSenseEngineFallback": "Enabled", "clangd.detectExtensionConflicts": false, "editor.formatOnType": true, "editor.semanticTokenColorCustomizations": { "enabled": true, // enable for al.

一、基于线程的优化方法1.选取合适的gridDim和blockDimblockDim最好为32的整数倍：因为执行指令的基本单位为线程束，线程束内的所有线程统一执行广播下来的命令，而线程束的线程数量基本为32。当block被分到SM中去，其会被划分为多个线程束，若blockDim!=线程束内线程数整数倍，则会造成线程的浪费。2.减少存在分支的if因为线程束中的所有线程执行同一条的指令，若出现存在出现分支的if，即线程束内的部分线程符合if条件，部分符合else条件，因cuda编译不具备分支预测能力，则

栈和队列stack<int> st;st.push(x);st.pop();cout<<st.top()<<endl;//注意在确认栈不为空的情况下才可以查看topcout<<st.size()<<endl;if(!st.empty())queue<int> que;cout<<que.front()<<que.back()<<endl;string//栈赋值给string

#include <iostream>#include "opencv2/opencv.hpp"#include "opencv2/imgcodecs/legacy/constants_c.h"#include "opencv2/imgproc/types_c.h"using namespace std;using namespace cv;int main(){ cv::Mat img_encode; img_encode = imread("/home/cqy

序列化Engine// serialize the engine, then close everything downconst std::string engine_file{ "tensorrt_mnist.trt" };nvinfer1::IHostMemory* gieModelStream = mEngine->serialize(); // GIE modelfprintf(stdout, "allocate memory size: %d bytes\n", gieModelS

Tensor RT C++ 使用总体流程提供包含权重和网络结构的文件，如.onnx，相当于构建生产流水线所需的参数图纸Creating a TensorRT network definition from your model盖一个工厂（builder），放置生产流水线（engine）用来生产，流水线是加速网络，原料是输入，生产出结果Invoking the TensorRT builder to create an optimized runtime engine from the networ

#include <ctime>srand(unsigned(time(NULL)));int num = rand() % 100; //产生0~149共150个整数。float fnum = (rand() % 2000)/1000.0 - 1; //产生-1~1的浮点数。

命令行参数：打开摄像头检测pose：–camera 0处理指定文件夹里的所有图像：–image_dir C:\Users\chenQY\Desktop\pose\Openpose_test\x64\Release\1指定模型文件夹：–model_folder C:\Users\chenQY\Desktop\pose\Openpose_test\x64\Release\models指定模型：...

string text_time = exe_address + “\time.txt”; //得到耗时ofstream fout_time(text_time);fout_time << “running time = " << ((double)getTickCount() - start_time) / getTickFrequency() <<”...

double real_time = (double)getTickCount();...程序...cout << “cut time=” << ((double)getTickCount() - real_time) / getTickFrequency() << endl;

Mat Image1(High, Width, CV_64F)；//double类型Mat Image2(High, Width, CV_8UC1)；//uchar类型

vector x; //动态数组for(int i=0;i<10;i++){x.push_back(i);}cout<<x.size()<<endl; //输出结果为10int x[5]={1,2,3,4,5}; //静态数组cou...

int n=1;for(int i=0;i<10;i++) n++;int *a;a=new int[n];

int转charchar str[100];int a=1,b=2,c=3;sprintf(str,"a=%d,b=%d,c=%d",a,b,c);cout<<a<<" "<<b<<" "<<c<<endl;cout<<str<<endl;