Cache Associativity 

constexpr int N = 64;

for (auto j = 0u; j < 1024 / N; ++j) {
  for (auto i = 0u; i < N; ++i) {
    sum += arr1[i] + arr2[i];
  }
}

constexpr int N = 128;

for (auto j = 0u; j < 1024 / N; ++j) {
  for (auto i = 0u; i < N; ++i) {
    sum += arr1[i] + arr2[i];
  }
}

* The benchmark is run under Apple Macbook Air M2.

* For the full benchmark code, please refer here.

* For illustration purpose only, see FAQ for more details.