cachegrind

Out of curiosity I ran coded up several different versions of matrix Multiplication and ran cachegrind against it. In my results below, I was wondering which parts were L1,L2,L3 misses and references and what it all really means? Below is my code for the matrix multiplications also, in case anyone needs that. #define SLOWEST ==6933== Cachegrind, a cache and branch-prediction profiler ==6933== Copyright (C) 2002-2012, and GNU GPL'd, by Nicholas Nethercote et al. ==6933== Using Valgrind-3.8.1 and LibVEX; rerun with -h for copyright info ==6933== Command: ./a.out 500 ==6933== --6933-- warning: L3

Recently Herb Sutter gave a great talk on "Modern C++: What You Need to Know" . The main theme of this talk was efficiency and how data locality and accessing the memory matters. He has also explained how linear access of memory(array/vector) would be loved by CPU. He has taken one example from another classical reference "Game performance by Bob Nystrom" on this topic. After reading these articles, I got that there is two type of cache which impact the program performance: Data Cache Instruction Cache Cachegrind tool also measures both cache type instrumentation information of our program.

I intend to multiply 2 matrices using the cache-friendly method ( that would lead to less number of misses) I found out that this can be done with a cache friendly transpose function. But I am not able to find this algorithm. Can I know how to achieve this? The word you are looking for is thrashing . Searching for thrashing matrix multiplication in Google yields more results . A standard multiplication algorithm for c = a*b would look like void multiply(double[,] a, double[,] b, double[,] c) { for (int i = 0; i < n; i++) for (int j = 0; j < n; j++) for (int k = 0; k < n; k++) C[i, j] += a[i, k