Hello, if you have any need, please feel free to consult us, this is my wechat: wx91due

CS152: Computer Systems Architecture

Overview

Files Involved

Implementation guide

Try to use all ideas discussed in class, including AVX, cache-obliviousness and loop unrolling. Some ideas may not work well together. Achieve the best you can!

The sole function in the mult.cpp file, mult, takes in five arguments: a, b, c, matrix_size and thread_count. a and b are to be multiplied and stored in c. matrix_size defines the size of the matrix input, where there is matrix_size2 elements in the square matrix. thread_count define the number of threads to be spawned.

The mult function is currently implemented using a naive triple-loop method.

In order to build and run the project run make, and then ./matrix T M . T is the number of threads argument that will be passed to your mult function. N is the size of the matrix, where the number of elements in the matrix is N*N. The default value of N is 2048 unless specified.

Evaluation environment

Your code will be tested on various (but reasonable, desktop-class) machines, with different cache, processor and memory characteristics, using a different thread values. The matrix size will vary between 2048 to 16384.

All machines will be using Intel x86 processors with AVX2 and FMA support at least. Please let me know if you do not have access to such a machine.

The derby

Each submission will be given a code name for anonymity, and I will create charts and leaderboards to compare various implementations, as well as some implementations I will make with various levels of optimizations.

Grading will not be on a curve, so relax, and do your best