MATLAB Parallel programming on GPUs, Cores and CPUs

Name: MATLAB Parallel programming on GPUs, Cores and CPUs
Rating: 4.1 (25 reviews)

With practical examples on every parallel programming concept including training NN and deep learning model

Created byHamdy egy

Last updated 1/2022

English

English [Auto],

What you'll learn

Run Deep learning models in parallel on GPUs
Learn the difference between cores, CPUs and GPUs
Learn the concept of multi-threading in MATLAB with examples
Learn the concept of multi-workers in MATLAB with examples
measuring the performance of each parallel computing code
Learn how to convert your code to parallel computing to increase the performance
Run MATLAB files and functions in the background
Using GPUs to execute and Run MATLAB functions (Excellent performance)

Course content

4 sections • 23 lectures • 3h 22m total length

The concept of cores, CPUs and GPUs (important)15:35
Explore the differences between cpu and gpu architectures, learn how multiple cores enable parallelism, and determine when to run tasks on cpu versus gpu to optimize Matlab applications.
The difference between the cores and the logical processors4:10
Explain how physical cores map to logical processors and how hyper-threading enables multiple threads per core, with operating systems reporting more logical processors than physical cores.

Practical examples on the multi-threading concept13:18
Explore MATLAB parallel computing with multithreading, showing eigenvalues and matrix multiplication on multiple cores while some operations use a single core, and learn to time code with tic/toc.
Practical examples on the multi-workers concept14:58
Explore the multi worker concept in MATLAB, compare parfor parallelism with serial for loops, and analyze execution time and overhead to gauge speedup.
How to control the number of cores that execute your code?(practical examples)16:31
Control the number of physical cores that execute your MATLAB code and see how the number of workers and iterations (n) affect execution time and CPU utilization in parfor.
4 mistakes when using parfor loop (be carful about them)4:11
Identify four common parfor pitfalls in MATLAB: avoid decimal increments; prevent indexing that depends on previous values; do not nest parfor loops; and never use break or continue inside parfor.
Nested parfor loops8:07
Learn how to handle nested parfor loops in MATLAB; you cannot nest parfor loops, and parallelize the outer loop to minimize overhead, as nested parfor rarely improves performance.
nested-loops: Calculating the amount of data sent to each core3:24
Explore how an outer loop minimizes data transferred from MATLAB to each core in a multi-worker setup, and compare serial inner loops with parallel execution across 24 workers.
Functions in parfor loop3:49
Discover how to use functions within parfor loops in MATLAB, including handling function references, function handles, and inner functions with eval, and learn to diagnose common errors.
Reduction variables (independence) in parfor loop6:12
Explore how loop variables behave in parfor, highlight reduction variables and independence mistakes, and distinguish temporary from reduction variables using a step-by-step example summing values from 1 to 10.
displaying values inside the parfor loop3:36
Explore how to display values inside a parfor loop in MATLAB, addressing workspace communication, variable scope, and the use of local variables to safely display updated array values during iterations.
Single instruction multiple data (SPMD)13:18
Explores single program multiple data (spmd) for running a function on different data sets across multiple workers. Highlights advantages over for loops and how per-worker variables are read for results.
Run MATLAB functions in background using parfeval11:50
Learn to run MATLAB functions in the background with parfeval and parfeval on all, manage outputs as objects, and control workers for parallel execution.
Run MATLAB files in background using batch jobs5:40
Explore running functions in the background with the Percival function and batch function in MATLAB, including scripts and MATLAB files, and learn to fetch results using load after completion.

Check if your GPU supports parallel computing with MATLAB5:01
Identify your Nvidia GPU architecture and install a graphics driver to enable MATLAB parallel computing with GPUs, then ensure your MATLAB release and the parallel computing toolbox support the device.
Read the GP information in MATLAB2:21
Learn to read GPU information in MATLAB with three functions: count GPUs, display a device table, and show all device details such as compute capability.
Running MATLAB functions on GPUs8:58
Prepare data in GPU memory, execute MATLAB functions on the GPU using GPU array, and compare eigenvalues and performance with CPU execution.
Important Notations to run functions on GPUs (Sharpen an image)4:52
Identify how to run functions on GPUs by placing inputs in GPU memory and using GPU arrays, so functions execute on the GPU or CPU depending on data location.
Running codes on multiple GPUs4:07
Learn to run MATLAB code on multiple GPUs by setting the GPU count in parallel preferences and choosing the number of workers; explore the for iteration and BMD methods.

Train the neural networks in parallel using GPUS in detail29:32
Explore data preparation, network creation and configuration, and training workflows in MATLAB, then compare CPU, multi-core, and GPU training using CUDA and the parallel computing toolbox.
Using multi-GPUs for training the neural networks4:36
Learn to train neural networks on multiple GPUs in MATLAB by configuring parallel workers, selecting GPU versus CPU execution, and comparing resource use and training speed.
Training deep learning model in parallel using GPUs11:39
Learn how to train deep learning models in MATLAB using CPUs, GPUs, and parallel pools, including selecting execution environments (auto, cpu, gpu, multi GPUs) and cloud options.
Train deep NN in the background in parallel6:30
learn to train deep neural networks in the background in parallel with MATLAB, using a background function to run on GPU and CPU without blocking.

Requirements

MATLAB basics

Description

This course helps students, researchers, and anyone using the MATLAB decrease the execution time they take to execute a program

All computers today and the laptops have multi-cores and GPUs. But not all users use the to run or execute the programs in parallel.

The purpose of the course is to fill this gap. Is to teach you with practical examples how to use all resources on your computer and also how to monitor them.

The course is divided into many sections:

The first is an introduction to the hardware of the CPUs, cores, and GPUs. It is better to understand the basic components of these items to be able to get the best utilization when you use them.
The second section is explaining two concepts. The multi-threading and the multi-workers. The first is a built-in mechanism to run some functions in parallel using many cores but we can't control the number of cores and the way that the functions execute. The second one (multi-workers) is used to run any function on multiple cores but here we can control the number of cores to optimize the program execution. Also, I explained some examples and measured the performance parameters to differentiate between the two concepts.
The third section is the GPU section. In the section, I explained how to run any function on the GPUs to make use of the hundred or thousands of cores that the GPUs have. There are some notations to get the best results and I explained all of these notations with examples.
Deep learning and neural networks: in this section, you will learn how to train any neural network in parallel on GPUs or multi-cores. And also how to run the training process in the background in order to be able to use MATLAB while it is running.

Who this course is for:

Students, researchers and engineers

MATLAB Parallel programming on GPUs, Cores and CPUs

What you'll learn

Explore related topics

Course content

Introduction2 lectures • 20min

Multi-threading and Multi-cores with practical examples12 lectures • 1hr 45min

GPU parallel programming in MATLAB5 lectures • 25min

Run Deep learning and Neural network GPUS (parallel)4 lectures • 52min

Requirements

Description

Who this course is for: