Mohamed A. Bamakhrama mohamed

High-Performance Matrix Multiplication

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).

Intro

Matrix multiplication is a mathematical operation that defines the product of

Using the DECA board with LiteX

Follow LiteX installation steps outlined here
Install Intel Quartus Prime (Lite Edition) from here
Build and run the example shipped with LiteX:

$ cd litex-boards/litex_boards/targets

Setting up RISC-V toolchain and simulator

This guide explains the steps needed to build the compiler and simulator toolchain for RISC-V ISA. It assumes RV32GC instruction extension.

Tool links:

GNU GCC Compiler: https://github.com/riscv/riscv-gnu-toolchain
LLVM Compiler: https://github.com/llvm/llvm-project.git
Proxy Kernel (to map syscalls from binary to OS): https://github.com/riscv-software-src/riscv-pk

Compiling baremetal C program for Arm Cortex-M0+

Install Arm GNU GCC compiler. Download the tar.xv archive from: https://developer.arm.com/downloads/-/arm-gnu-toolchain-downloads Define the environmet variable $ARM_GCC_ROOT to point to the location where you extract this archive. This guide was tested with arm-gnu-toolchain-11.3.rel1-x86_64-arm-none-eabi
Update your $PATH and $LD_LIBRARY_PATH as follows:
```
setenv PATH $ARM_GCC_ROOT/bin:$PATH
```

setenv LD_LIBRARY_PATH $ARM_GCC_ROOT/lib:$LD_LIBRARY_PATH

	#!/usr/bin/env sh

	set -euv

	# Updated on April 2019 to reflect changes in GN
	# See:
	# https://github.com/ninja-build/ninja
	# https://gn.googlesource.com/gn/

	# We need a recent git

	#!/usr/bin/env bash
	# journal.sh
	# ==========
	#
	# One daily text file to rule them all.
	#
	# Copyright: 2022 Tyler Cipriani <[email protected]
	# License: GPLv3

	set -euo pipefail

	import pyaudio
	import matplotlib.pyplot as plt
	import numpy
	import wave


	FORMAT = pyaudio.paInt16 # We use 16bit format per sample
	CHANNELS = 1
	RATE = 44100
	CHUNK = 1024 # 1024bytes of data red from a buffer

	"""
	Tarjan's strongly connected components algorithm + Topological sort
	Taken from: http://www.logarithmic.net/pfh/blog/01208083168
	Public domain. Use it as you will
	"""


	def strongly_connected_components(graph):
	"""
	Tarjan's Algorithm (named for its discoverer, Robert Tarjan) is a graph

	# Run with make SIM=verilator

	TOPLEVEL_LANG = verilog
	VERILOG_SOURCES = $(shell pwd)/skid_buffer.sv
	TOPLEVEL = skid_buffer
	MODULE = skid_buffer_tb
	EXTRA_ARGS += --coverage --trace-fst --trace-structs

	include $(shell cocotb-config --makefiles)/Makefile.sim