Episode 4.5 - Parallel Loops, Private and Shared Variables, Scheduling

Table of Contents:
00:14 - Simple example of OpenMP for-loop
00:17 - Loop-Centric Parallelism
00:23 - Iterations distribution
00:56 - #pragma omp parallel for
01:04 - Work items distribution
01:17 - Iterations and number of threads
01:32 - OpenMP is not over-subscribing the system
01:41 - Memory model in OpenMP
01:48 - Shared variables
02:04 - Private variables
02:15 - Clauses: private() and shared()
02:39 - Implicit declaration of private variable
02:50 - Code example
03:24 - OpenMP scheduling
03:38 - Static, dynamic, and guided scheduling
04:07 - In the next episode: Fork-Join model

[@travel.tales.official]

Loved the animation. Would love more such contents.

[@marcinziajkowski3870]

3:35 For dynamic blocks on left, there is double 9 iteration. Is that mistake or during dynamic schedule could be occurred something like that?

[@VadimKarpusenko]

It is a mistake - iteration should be executed only once in any scheduling schema. and WOW! for noticing this mistake!!!

[@drdopesec]

Nice observation.

[@jewelossom6737]

how can i use a break statement in such a parallel loop? tried it but getting errors

[@aaronstaight4586]

is parelize the same thing?

[@VadimKarpusenko]

Can you elaborate?

[@dron_kroft]

найти бы на русском...

[@VadimKarpusenko]

+Andrey Novokshonov Есть желание попробовать привлечь сообщесто habrahabr.ru для перевода. Только в прошлый раз попытка была очень неудачной. Если будет время на следующей неделе - сделаю. Если нет, но придется отложить до ноября - много выездных трейнингов и SeperComputing 2015 там же.