Email
Like
Liked ×
Save
Private Content
Embed
Loading embed code…
We have emailed the verification/download link to "".
Login to your email and click the link to download the file directly.

To request the link at a different email address, update it here. Close
Validation messages. Success message. Fail message.

Check your bulk/spam folders if you can't find our mail.
Loading

+ Follow

Parallel Development in VS10

by Valdis Iljuconoks

296 views

Accessibility

View text version

Upload Details

Uploaded via SlideShare as Microsoft PowerPoint

Usage Rights

File a copyright complaint

1 Embed 2

http://www.linkedin.com

Statistics

Likes: 0
Downloads: 13
Comments: 0
Embed Views: 2
Views on SlideShare: 294
Total Views: 296

.NET has had support for parallel programming since version 1.0, now referred to as classicthreading, but it was hard to use and made you think too much about managing the parallel aspects ofyour program, which detracts from focusing on what needs to be done.
Parallel execution doesn’t come for free. There are overhead costs associated with setting up andmanaging parallel programming features. If you have only a small amount of work to perform, theoverhead can outweigh the performance benefit.
Don’t just assume that a parallel solution will give you better performanceand move on.
However the thread pool isn’t so great at letting you know when work has been completed orcancelling running threads. The thread pool also doesn’t have any information about the context inwhich the work is created.
Dataparallelisim – makeanimage,splitintomanysmallpieciesandperformthesameoperationoneachofthem.Taskparalellisim – takeallsetoftasksandperformthemallatonce.
Chunksize – toobig/toosmallChunklayout – memorycache (L2)
Ordering– Pipilines, FuturesandDynamicTasks (recursion)
Race conditions arise in software when separate processes or threads of execution depend on some shared state
A deadlock is a situation in which two or more competing actions are waiting for the other to finish, and thus neither ever does
Thread starvation can be caused by creating too many threads
Optimizing code for different machine environments
If your application is multithreaded is it running in parallel?Operating system allocates time with theCPU to each thread and then rapidly switches between them (known as time slicing)
So how does the task scheduler work?1. When tasks are created, they are added to a global task queue.2. The thread pool will create a number of “worker” threads. The exact number that are createddepends on a number of factors such as the number of cores on the machine, current workload, type of work load, and so on. The thread pool utilizes a hill-climbing algorithm thatdynamically adjusts the thread pool to use the optimum number of threads. For example, if thethread pool detects that many threads have an I/O bottleneck, it will create additional threadsto complete the work more quickly. The thread pool contains a background thread that checksevery 0.5 seconds to see whether any work has been completed. If no work has been done (andthere is more work to do), a new thread will be created to perform this work.3. Each worker thread picks up tasks from the global queue and moves it onto its local queue forexecution.4. Each worker thread processes the tasks on its queue.5. If a thread finishes all the work in its local queue, it steals work from other queues to ensure thatwork is processed as quickly as possible. Note that tasks will steal work from the end of the othertask’s queues to minimize the chance that the task has started operating with the work already.
The .NET Framework 4 ThreadPool also features a work-stealing algorithm to help make sure that no threads are sitting idle while others still have work in their queues.Stealing work from other thread’s local queue tail.

Parallel Development in VS10 Presentation Transcript

Powered by MVP
Independent Experts. Real World Answers.
10. februārī, Rīgā
Parallel Computing withVisual Studio 2010
Valdis Iljuconoks
Microsoft C# MVP, Latvia
Whywecareaboutthis?
Performance
Avoidmultithreading
Parallelprogrammingishard
It’sforexperts
Parallelism Will Increase Your
Application's Complexity
The keyis
measurement
Comparison Matrix (Amdahl'slaw)
Sojust usethreads ?!
Crapcode runningin parallel
is justParallelizedcrap code
Where do I start?
TasksvsData
ControlFlow
ControlandDataFlow
DataParallelism
Chunksize
Chunklayout
Datadependencies
TaskParallelism
Taskcount
Work per task
Dependencies
ControlandDataFlow
TaskConstraints
ExternalConstraints
Orderings
Availability - Framework 4
Some Functionality in 3.5
Reactive Extensions (Rx)
Parallel Extensions CTP
ParallelLibraryContent
Task Parallel Library (TPL) and Concurrency and Coordination Runtime (CCR)
Parallel LINQ (PLINQ)
New debugging and profiling tools
Coordination data structures
Parallel Pattern Library(PPL) C++ only
ParallelLibraryArchitecture
Some of the issues
Some of the issues
Race conditions
Some of the issues
Race conditions
Deadlocks
Some of the issues
Race conditions
Deadlocks
Thread starvation
Some of the issues
Race conditions
Deadlocks
Thread starvation
Difficult to code and debug
Some of the issues
Race conditions
Deadlocks
Thread starvation
Difficult to code and debug
Environmental
Important Concepts
Concurrent != Parallel
Singlecore machine?
Parallelization features
Canslow down
Getting started?
Task Factory
demo
Parallel.For()Parallel.ForEach()Parallel.Invoke()
Cancelling Task
demo
Avoid
Console.WriteLine()
Debugging Enhancements
Parallel Task Window
Parallel Stack Window
demo
Deadlocks
demo
Parallelism Will Increase Your
Application's Complexity
Work Stealing
PLINQ
demo
SuccessKeys:
Frameworksandlibraries
Tools (VS2010)
Guidance
Maslow’sHammer
Questions ?
Future Considerations
Dormant GPUs
Distribute work across
multiple machines?
Async CTP
Valdis Iljuconoks
Software Architect
Microsoft C# MVP, Latvia
valdis.iljuconoks@dotnet.lv
http://dotnet.lv/blogs/vi
Powered by MVP
Independent Experts. Real World Answers.
10. februārī, Rīgā

Parallel Development in VS10

by Valdis Iljuconoks

Accessibility

Categories

Upload Details

Usage Rights

1 Embed 2

Statistics

Parallel Development in VS10 Presentation Transcript