Parallel Programmings MIPT course

Quick start

Build

sudo apt install mpich libomp-dev
mkdir build && cd build
cmake .. && make -j

Launch

ctest --versbose

Configuration

CPU: Intel Core i5 11300H (4 cores, 8 threads)

Term 6

Lab1

Introduction to MPI

Admission

Task 1: Calculate pi using MPI

Results:

Calculated Pi: 3.14159
Actual Pi: 3.14159
Error: 1.11022e-13

Task 2: Measure time of communication between MPI nodes

Results:

Message size: 1 KB, Avg time: 2.62737 us, Bandwidth: 743.376 MB/s
Message size: 4 KB, Avg time: 3.94507 us, Bandwidth: 1980.32 MB/s
Message size: 16 KB, Avg time: 12.2503 us, Bandwidth: 2550.97 MB/s
Message size: 64 KB, Avg time: 33.9684 us, Bandwidth: 3679.89 MB/s
Message size: 256 KB, Avg time: 39.7788 us, Bandwidth: 12569.5 MB/s
Message size: 1024 KB, Avg time: 216.363 us, Bandwidth: 9243.74 MB/s

Task

Task: Solve transport equation

Results:

Lab2

Introduction to POSIX threads, synchronization primitives and processes communication

Admission

Task: Write custom parallel sort using threads and compare it with standard quick sort

Results:

Task

Task: Measure communication time between processes (through pipe) and between threads

Results:

Average process communication time is `109953` ns;
Average thread communication time is `32981` ns;

---

Term 7

Lab3

OpenMP Strassen matrix mul implementation

Results:

Without OpenMP:

With OpenMP: