SIMD-instructions/parallism

Parallelism standards for accelerating performance on calculations for detection of positive DNA selection

Table of contents

• General Info
• Features
• Prerequisites
• Setup
• How to run
• Acknowledgements

General Information

Streaming SIMD Extensions (SSE), MPI and Pthreads for parallelization the calculation of a simplified form of OMEGA statistical, to detect positive selection in DNA sequences.

Repeated for a set of N DNA sites, extract statistical OMEGA outputs

$$num = \frac{L+R}{\left(\frac{m*(m-1.0)}{2.0}\right) + \left(\frac{n*(n-1.0)}{2.0}\right)}$$
$$den= \frac{C-L-R}{m*n}$$
$$\omega= \frac{num}{den+0.01}$$

The above calculations are repeated iteratively for a set of N DNA sites, $N$ is a variable for which $N ≥ 1$ . All input data is arranged in arrays of length N (user variable).

Interested for

$$max_{\omega}$$	$$min_{\omega}$$	$$avg_{\omega}$$

BONUS: Implementation of a different memory layout for better performance with SSE commands.

Features

• Reference
• SSE instructions
• Parallel standards (pthreads/MPI) + SSE

The implementation was guided by the reference file Benchmarked on Intel(R) Core(TM) i7-1065G7 @ 1.30GHz 1.50 GHz with 8GB DDR3 memory.

Prerequisites

SIMD
MPI

How to run

1. GCC Installation

$ gcc --version
$ sudo apt install gcc
$ sudo apt install lam-runtime

Supposing you can navigate the proper folder to compile the desired source code from the terminal

2. MPI Installation Download the source code MPI release

$ tar -xzf mpich-X-X-X.tar.gz
$ cd mpich-X-X-X

where X-X-X mpi version

$ ./configure
$ make; sudo make install

3. Check MPI version

$ mpiexec --version

4. LAM installation

sudo apt-get install lam-runtime

Reference

gcc -o reference reference.c
./reference 'N'

SSE

gcc -o jam jam.c
./jam 'N'
gcc -o unroll unroll.c
./unroll 'N'

Add -msse4.2 library

gcc -o SSE SSE.c -msse4.2
./SSE 'N'

Pthreads

gcc -pthread SSE_pthreads.c -o SSE_pthreads -msse4.2
./SSE_pthreads 'N' 'N_threads'

MPI

mpicc -pthread SSE_Mpthreads.c -o SSE_Mpthreads -msse4.2 -lm
lamboot -v host
mpiexec -n 'N_proc' ./SSE_Mpthreads 'N' 'N_threads'

Bonus

gcc -o bonus bonus.c -msse4.2
./bonus 'N'

Setup

Script variables initialized as:

• N = 10000000.
• Threads = [2 4]
• Processors = [2 4].

Acknowledgements

• This project was created for the requirements of the lesson Architecture of Parallel and Distributed Computers