https://github.com/gianmariaromano/pmc-translated-notes

The repository contains translated notes for the course "Programmazione di Sistemi Multicore" given by Professor De Sensi for the "Informatica" course at Sapienza Università di Roma.
https://github.com/gianmariaromano/pmc-translated-notes

cuda cuda-programming mpi multicore openmp parallel-computing parallel-programming pthreads

Last synced: 11 days ago
JSON representation

The repository contains translated notes for the course "Programmazione di Sistemi Multicore" given by Professor De Sensi for the "Informatica" course at Sapienza Università di Roma.

• Host: GitHub
• URL: https://github.com/gianmariaromano/pmc-translated-notes
• Owner: GianmariaRomano
• Created: 2025-09-23T19:50:34.000Z (8 months ago)
• Default Branch: main
• Last Pushed: 2025-10-01T13:46:11.000Z (8 months ago)
• Last Synced: 2025-10-01T14:25:09.643Z (8 months ago)
• Topics: cuda, cuda-programming, mpi, multicore, openmp, parallel-computing, parallel-programming, pthreads
• Language: C
• Homepage:
• Size: 913 KB
• Stars: 0
• Watchers: 0
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

          
# 📚 Programmazione di Sistemi Multicore – Translated Notes & Code

## ⭐ About this Repository

This repository contains translated lecture notes and codes for the **"Programmazione di Sistemi Multicore"** course, taught by **Professor De Sensi** as part of the **Bachelor of Science in Informatica** at **Sapienza Università di Roma**.

---

## 📍 Course Details

- **📅 Lecture Schedule:**

  - **Tuesdays:** 17:00 – 19:00 at *Aula Magna, Regina Elena*.

  - **Wednesdays:** 14:00 – 17:00 at *Aula 11, Via Scarpa, Engineering Department*.

- **🧪 Exam Structure:**

  - A project and an oral exam covering both the project and the theoretical topics discussed throughout the lectures.

    The evaluation will take the following factors into consideration:

    - **Project performance:** 10 points.

      - The *OpenMP + MPI* implementation is worth up to 5 points.

      - The *CUDA* implementation is worth up to 5 points.

      

      **‼ Since the code is tested on the cluster, make sure that it works on it as, otherwise, you will automatically fail the exam.**

    - **Report and analysis:** 6 points.

      - The *clarity* and *scaling analysis* are worth up to 4 points.

      - The *profiling techniques* are worth up to 2 points.

    - **Project defense:** 6 points.

      - The explanation of the *design choices* is worth up to 3 points.

      - Further *questions on (alternative) design choices* are worth up to 3 points.

    - **Oral exam:** 8 points.

    For more information, refer to the Moodle page of the course.

---

## 🚨 Executing Codes in Linux

The Professor uses Linux as the main terminal/compiler for the course.

If you are not familiar with it, here are some instructions for running a C file in a Linux environment:

1) `gcc [filename].c -o [filename]`: use this command to *compile and name the executable file (this also comes in handy to avoid ambiguities)*.

2) `./[filename]`: use this command to *run the executable file*.

3) `rm [filename]`: if needed, use this command to *delete the executable file*.

However, some APIs use different instructions for running these files:

- 📮 **OpenMPI**

  1) `mpicc -g -Wall -o [filename] [filename].c`: use this command to *compile the executable file*.

  2) `mpiexec -n [processes] ./[filename]`: use this command to *run the executable file* or, by adding `ddd`, to *debug the program*.

- 🐌 **PThreads**

  

  Use the `pthread_create()` function to *start the threads in the executable file*.

- 🧭 **OpenMP**

  1) ` gcc -g -Wall -fopenmp -o [filename] [filename].c`: use this command to *compile the executable file* in OpenMP.

  2) `./[filename] n`: use this command to *run the executable file* using *n threads*.

     **N.B.:** The scheduling order of the threads is non-deterministic.

- 🛰 **CUDA**

  1) `nvcc --arch=sm_20 [filename].cu --o [filename]`: use this command to *compile the executable file* using CC 2.0.

  2) `./[filename]`: use this command to *run the executable file*.

     **N.B.:** For formality, programs containing CUDA codes are characterized by the `.cu` extension.

     **N.B.:** If your laptop does not support CUDA GPUs, you can ask the Professor to give you access to a University cluster (refer to the form shown on the November 25 lecture) or run CUDA using Google Colab (refer to last year's materials).

---

## 🎬 Additional Information

- 📌 For official communications and course materials, please refer to the Moodle page of the course.

- 📩 For any questions or clarifications, feel free to contact me directly.

- 📈 The Professor will use a GitHub classroom to handle the exercises, which will be pushed and graded according to their correctness.

  **N.B.:** For this year, the grades of the exercises will **not** impact the final grade of the exam.

---

## 📖 Course Log

**23 September:** Introduction to the course and to **parallel computing**.

**24 September:** Introduction to the C programming language: **variables, operators, input/output activities, selective branches, iterative branches, arrays, strings, functions, pointers**.

~~**30 September:**~~ This lecture was cancelled.

**1 October:** Introduction to the C programming language: **memory allocation, structs, lists**.

**7 October:** Recap on the **von Neumann architecture** and introduction to **MPI**.

**8 October**: **Message matching** and **communication modes** in MPI.

~~**14 October:**~~ This lecture was cancelled.

~~**15 October:**~~ This lecture was cancelled.

**21 October:** **Parallel design strategies** (GPLS, GSLP) and **collective communication** in MPI.

**22 October:** Exercises on MPI and evaluation of parallel programming: **speed-up**, **efficiency** and **scalability**.

**28 October:** Estimating speed-up using **Amdahl's law** and **Gustafson's law**.

**29 October:** Exercises on MPI and recap of **PThreads**.

~~**4 November:**~~ This lecture was cancelled.

~~**5 November:**~~ This lecture was cancelled.

**11 November:** Introduction to the **OpenMP** API: **pragmas and clauses, variable scope and reductions**.

**12 November:** Exercises and **parallel for loops**.

~~**18 November:**~~ This lecture was cancelled.

~~**19 November:**~~ This lecture was cancelled.

**25 November:** Introduction to **CUDA**.

**26 November:** Recap on **thread scheduling in CUDA** and **memory hierarchies**.

***1 December:** Extra lecture on profiling for bottlenecks in high-performance computers*.

**2 December:** Recap on **OpenMP** and introduction to **synchronization constructs** and **data dependencies**.

**3 December:** Recap of **data dependencies** for parallel loops.

**9 December:** The main aspects of **caching** for **single-core** and **multicore architectures**.

**10 December:** The CUDA **memory hierarchy**.

**16 December:** Introduction to **virtual memory management** and handling the shared memory (**bank conflicts**) and the global memory (**coalescing**).

**17 December:** Performing **reductions in CUDA** and using MPI for **GPU data transfers**.

---