adjust ...

This commit is contained in:
kai
2025-04-26 13:22:56 +02:00
parent b28774b428
commit 5df59e9376
2 changed files with 92 additions and 18 deletions

View File

@@ -1,4 +1,3 @@
#include "matrix.h"
#include <chrono>
#include <cstdint>
#include <fstream>
@@ -8,19 +7,22 @@
struct bench_result {
double time;
double pi;
int threads;
};
double monteCarloPi(int64_t N) {
int localCount = 0;
std::mt19937 generator(std::random_device{}());
std::uniform_real_distribution<double> dist(0.0, 1.0);
#pragma omp parallel for reduction(+ : localCount)
for (int i = 0; i < N; ++i) {
double x = dist(generator);
double y = dist(generator);
if (x * x + y * y <= 1.0) {
localCount++;
#pragma omp parallel
{
std::mt19937 generator(std::random_device{}() + omp_get_thread_num());
std::uniform_real_distribution<double> dist(0.0, 1.0);
#pragma omp for reduction(+ : localCount)
for (int i = 0; i < N; ++i) {
double x = dist(generator);
double y = dist(generator);
if (x * x + y * y <= 1.0) {
localCount++;
}
}
}
@@ -32,32 +34,35 @@ bench_result benchmark(int numThreads, int64_t N, double function(int64_t)) {
// start time measurement
auto start = std::chrono::system_clock::now();
auto result = bench_result{0.0, 0.0};
auto result = bench_result{0.0, 0};
// perform benchmark
const int maxNumExec = 10;
for (int nExec = 0; nExec < maxNumExec; ++nExec) {
result.pi = function(N);
volatile double pi = function(N);
}
auto end = std::chrono::system_clock::now();
std::chrono::duration<double, std::milli> diff = end - start;
result.time = diff.count() / maxNumExec;
result.threads = numThreads;
return result;
}
int main() {
const int64_t max_n = 10000000;
const int numThreads = 12;
std::ofstream fout("multi_monte.csv", std::ios::app);
fout << "pi" << ",";
const int max_threads = 12;
std::ofstream fout("diff_theads.csv", std::ios::app);
fout << "threads" << ",";
fout << "max_n" << ",";
fout << "time" << "\n";
for (int64_t i = 1000; i <= max_n; i += 1000) {
const bench_result res = benchmark(numThreads, i, monteCarloPi);
for (int i = 1; i <= max_threads; i++) {
const bench_result res = benchmark(i, max_n, monteCarloPi);
fout << res.pi << ",";
fout << res.threads << ",";
fout << max_n << ",";
fout << res.time << "\n";
}
}

View File

@@ -0,0 +1,69 @@
#include "matrix.h"
#include <chrono>
#include <cstdint>
#include <fstream>
#include <iostream>
#include <omp.h>
#include <random>
struct bench_result {
double time;
int threads;
};
double monteCarloPi(int64_t N) {
int localCount = 0;
#pragma omp parallel
{
std::mt19937 generator(std::random_device{}() + omp_get_thread_num());
std::uniform_real_distribution<double> dist(0.0, 1.0);
#pragma omp for reduction(+ : localCount)
for (int i = 0; i < N; ++i) {
double x = dist(generator);
double y = dist(generator);
if (x * x + y * y <= 1.0) {
localCount++;
}
}
}
return 4.0 * localCount / N;
}
bench_result benchmark(int numThreads, int64_t N, double function(int64_t)) {
omp_set_num_threads(numThreads);
// start time measurement
auto start = std::chrono::system_clock::now();
auto result = bench_result{0.0, 0};
// perform benchmark
const int maxNumExec = 10;
for (int nExec = 0; nExec < maxNumExec; ++nExec) {
volatile double pi = function(N);
}
auto end = std::chrono::system_clock::now();
std::chrono::duration<double, std::milli> diff = end - start;
result.time = diff.count() / maxNumExec;
result.threads = numThreads;
return result;
}
int main() {
const int64_t max_n = 1000000;
const int max_threads = 12;
std::ofstream fout("dyn_n.csv", std::ios::app);
fout << "threads" << ",";
fout << "max_n" << ",";
fout << "time" << "\n";
for (int i = 1; i <= max_threads; i++) {
const bench_result res = benchmark(i, max_n * i, monteCarloPi);
fout << res.threads << ",";
fout << (max_n * i) << ",";
fout << res.time << "\n";
}
}