HOME ABOUT MORE
Tasks &
SIMPLIFIED
PARALLELISM
INTRODUCTION HOME ABOUT MORE
Simplified Parallelism is a way of making
computers work better. It involves doing lots of
things at the same time. This is done by
splitting big jobs into smaller parts that can be
done at the same time. This makes the
computer work better.
HOME ABOUT MORE
1 EXMAPLE
Imagine a car factory where there is an assembly line. Instead of one person
building a car from start to finish, the process is divided into several stages:
Stage 1: One person puts in the chassis.
Stage 2: Another person installs the engine.
Stage 3: A third person assembles the doors.
Stage 4: Finally, another person paints the car.
These stages are done in parallel for different cars at the same time, allowing
multiple cars to be assembled simultaneously, speeding up the overall process.
HOME ABOUT MORE
2 EXMAPLE
When editing an image or applying filters,
simplified parallelism allows you to split the
image into several parts (e.g. blocks of pixels)
and process each part in parallel, making editing
faster.
BENEFITS OF SIMPLIFIED HOME ABOUT MORE
PARALLELISM
Speed: Completes tasks in less time by
distributing work across multiple processing
cores.
Efficiency: Optimises the use of hardware
resources, avoiding waste.
Scalability: Facilitates the handling of larger and
more complex tasks by increasing hardware
resources.
DATA HOME ABOUT MORE
PARALLELISM
Data Parallelism & TPL
Data Parallelism: Concurrently performs the same
operation on different elements of a collection or array.
Source is partitioned for multi-threaded processing.
Task Parallel Library (TPL): Simplifies parallel operations
using [Link] and [Link].
Automatically handles threads, locks, and workload
balancing.
Features & Customization
Task Scheduler: Automatically partitions and balances
workloads.
Advanced Controls:
ParallelOptions for concurrency control.
ParallelLoopState and CancellationToken for state
management and cancellation.
HOME ABOUT MORE
PARALLEL PROGRAMMING IN C#
Pros and Cons of Parallel Programming Parallel Programming in C# Using [Link] for Improvement
Advantages: Scenario: Execute heavy Parallel Execution: Tasks
✅ Faster Execution: computations using both run simultaneously
Enables multiple actions to sequential and parallel (hardware-dependent).
run simultaneously, approaches. Result: Total time is less
reducing operation time. Sequential Example: than the sum of task
✅ Efficient Resource Actions run one after times.
Utilization: Optimizes another.
hardware usage when Total Time: Sum of C#
parallelism is well- individual task times. [Link](
implemented. () =>
Disadvantages: C# HeavyComputation("A"),
❌ Complexity: Harder to HeavyComputation("A"); () =>
write, read, and debug. HeavyComputation("B"); HeavyComputation("B"),
❌ Hardware Dependency: // ... other tasks // ... other tasks
Behavior may vary based );
on hardware
configurations.
HOME ABOUT MORE
PARALLEL PROGRAMMING IN C#
Key Notes on [Link]
Execution Time: Reduces
overall time, but
parallelism isn’t
guaranteed.
Order: Execution order is
not preserved.
Other Parallel Methods in TPL
[Link]: For-loop with
parallel iterations.
[Link]: Parallel
iteration over collections.
[Link]:
Added in .NET 6 for
asynchronous execution.
HOME ABOUT MORE
COMPARISON OF
SEQUENTIAL VS
PARALLEL
EXECUTION
PARALLEL
EXECUTION
HEAVYCOMPUTATIONASYNC(STRIN
G NAME) METHOD
SIMULATES INTENSIVE TASKS,
JUST LIKE THE ORIGINAL METHOD,
BUT IS DESIGNED AS AN
ASYNCHRONOUS METHOD (ASYNC
TASK<INT>).
ASYNCHRONOUS
SYNCHRONIZATION: USING ASYNC
ALLOWS THIS METHOD TO BE
EXECUTED CONCURRENTLY WITH
OTHER ASYNCHRONOUS METHODS.
PARALLEL
EXECUTION MAIN METHOD
PARALLEL TASKS: THE
HEAVYCOMPUTATIONASYNC TASKS FOR "A", "B",
"C", "D", AND "E" ARE STARTED SIMULTANEOUSLY
AND GROUPED INTO AN ARRAY (TASKS).
[Link]: COORDINATES THE EXECUTION
OF ALL TASKS IN PARALLEL AND WAITS FOR ALL
TASKS TO FINISH BEFORE CONTINUING.
MEASURING TOTAL TIME: BY MEASURING TIME
WITH STOPWATCH, WE CAPTURE THE DURATION
OF THE ENTIRE PROCESS (ALL TASKS RUNNING
IN PARALLEL).
RESULTS
THIS SHOWS HOW THE DIFFERENCE BETWEEN
SEQUENTIAL PROCESSES IS MUCH LONGER THAN
PARALLEL PROCESSES.
HOME ABOUT MORE
RUNTIME 6000
Tiempo de Ejecución (ms)
COMPARISON:
5000
SEQUENTIAL VS
4000
PARALLEL
3000
2000
1000
0
Secuencial paralelo
Time (s) Secuencial (%) Paralelo (%)
100 CPU USAGE COMPARISON
(SECUENTIAL VS
80 PARALLEL)
60
40
20
0
1 2 3 4 5
HOME ABOUT MORE
THANK YOU
[Link]
