小编edo*_*ado的帖子

Haswell内存访问

我正在尝试使用AVX -AVX2指令集来查看连续阵列上的流媒体性能.所以我有下面的例子,我做基本的内存读取和存储.

#include <iostream>
#include <string.h>
#include <immintrin.h>
#include <chrono>
const uint64_t BENCHMARK_SIZE = 5000;

typedef struct alignas(32) data_t {
  double a[BENCHMARK_SIZE];
  double c[BENCHMARK_SIZE];
  alignas(32) double b[BENCHMARK_SIZE];
}
data;

int main() {
  data myData;
  memset(&myData, 0, sizeof(data_t));

  auto start = std::chrono::high_resolution_clock::now();

  for (auto i = 0; i < std::micro::den; i++) {
    for (uint64_t i = 0; i < BENCHMARK_SIZE; i += 1) {
      myData.b[i] = myData.a[i] + 1;
    }
  }
  auto end = std::chrono::high_resolution_clock::now();
  std::cout << (end - start).count() / …
Run Code Online (Sandbox Code Playgroud)

performance x86 cpu-architecture avx2 intel-pmu

19
推荐指数
1
解决办法
1841
查看次数

标签 统计

avx2 ×1

cpu-architecture ×1

intel-pmu ×1

performance ×1

x86 ×1