Best Practice Webinar for Multicore Programming
With the wider adoption of multicore architectures, programmers are faced with a number of additional design challenges that come with the development of parallel programs or the parallelization of existing programs.
- How to optimize performance through parallelization and how to identify potential bottle necks
- How to test and verify parallel programs
- How to optimize your applications for specific architectures
- How to reuse your parallel code on different target platforms
- How to greatly speed up your multicore development flow
The webinar will show you the most common techniques to parallelize your code, avoid conflicts in parallel programs and use code transformations to expose the inherent parallelism inside of your applications.
You will also learn more about our solution – the automated parallelization in emmtrix Parallel Studio.
Tuesday, May 8, 10am (GMT+2), duration 90 minutes. Language German.
Tuesday, May 22, 10am (GMT+2), duration 90 minutes. Language English.
Contents of the Webinar
Utilizing multicore processors to cope with customer demands
– Increased performance
– Faster response time
– Lower power consumption
Elements for successful performance assessment
– Profiling on hardware
– Software-based simulation
– Static analysis of source code
Common pitfalls in parallel programs
– Race conditions
– Synchronization between the cores
– Failure to resolve dependencies
Basic parallelization techniques
– Task level parallelization
– Task level pipelining
– Code transformations
– Automating tedious tasks
– Problems with automation
Overview of emmtrix Parallel Studio
– The automated workflow
– Interaction with the GUI
– Live demonstration
Examples from data processing and control flow applications