This is a short post that explains how to write a high-performance matrix multiplication program on modern processors. In this tutorial I will use a single core of the Skylake-client CPU with AVX2, but the principles in this post also apply to other processors with different instruction sets (such as AVX512).
Matrix multiplication is a mathematical operation that defines the product of
| #ifndef _GNU_SOURCE | |
| #define _GNU_SOURCE | |
| #endif | |
| #include <execinfo.h> | |
| #include <stdio.h> | |
| #include <stdlib.h> | |
| #include <unistd.h> | |
| #define BT_BUF_SIZE 100 |
| #include <algorithm> | |
| #include <cassert> | |
| #include <cstdint> | |
| #include <cstdio> | |
| #include <cstring> | |
| #include <pthread.h> | |
| #include <unistd.h> | |
| namespace { | |
| /// A Block is a header to a managed memory buffer. Blocks are arranged as |
| #!/usr/bin/python | |
| from __future__ import print_function | |
| from bcc import BPF | |
| from bcc.utils import printb | |
| from time import sleep | |
| import sys | |
| import pdb | |
| if len(sys.argv) < 2: |
| FROM ubuntu:21.04 | |
| # We install some useful packages. | |
| RUN apt-get update -qq | |
| RUN DEBIAN_FRONTEND="noninteractive" apt-get -y install tzdata | |
| RUN apt-get install -y vim clang-format sudo python3 wget cmake g++ | |
| RUN apt-get install -y git clang linux-tools-generic ninja-build lldb zip curl | |
| RUN apt-get install -y firefox | |
| RUN apt-get install -y python3-pip | |
| RUN apt-get install -y libssl-dev |
| https://en.wikipedia.org/wiki/Continued_fraction | |
| from math import trunc, pi, sqrt | |
| def approximate(alpha, iters): | |
| r = [alpha] | |
| a = [] | |
| for i in range(iters): | |
| a.append(trunc(r[-1])) | |
| r.append(1/(r[-1] - a[-1])) |