MATLAB Performance Tuning and Acceleration Overview
This intermediate-level MATLAB Performance Tuning and Acceleration training course teaches attendees how to improve the run-time performance (speed) of their MATLAB programs. Students learn how to leverage tools for profiling and measuring performance. In addition, students learn a wide variety of speedup techniques, including loop optimization, data caching and chunking, I/O optimization, mathematical/physical identities, storage type modifications, compiled (binary) code, vectorization, parallelization, graphics, memory-related techniques, and more.
Note: This course can be condensed to a half-day with fewer examples and less hands-on practice. We strongly recommend the full-day version if possible.
Location and Pricing
This course is taught as a private, live online class for teams of 3 or more. All our courses are hands-on, instructor-led, and tailored to fit your group’s goals and needs. Most Accelebrate classes can be flexibly scheduled for your group, including delivery in half-day segments across a week or set of weeks. To receive a customized proposal and price quote for online corporate training, please contact us.
In addition, some courses are available as live, instructor-led training from one of our partners.
Objectives
- Understand tradeoffs in performance and cost-effectiveness of MATLAB code
- Learn how to measure and profile MATLAB’s run-time performance
- Discover top performance hotspots/bottlenecks in run-time code
- Acquire soft-skills such as knowing when to optimize and when not to bother
- Discover multiple possible ways of improving MATLAB run-time speed
- Understand memory’s effects on performance, and how to use this information
- Write MATLAB code that is highly performant and responsive
- Learn how to improve perceived speedup when actual speedup is not possible
Prerequisites
Attendees should have taken Accelebrate's From MATLAB Scripts to Complete Programs course or have equivalent knowledge. Students should already be comfortable using the MATLAB environment and have at least basic MATLAB programming experience.
Outline
Expand All | Collapse All
Introduction
Profiling MATLAB Performance
- When to profile and when not to bother
- When should we stop optimizing the code?
- Profiling techniques
- Real-time profiling limitations
- MATLAB’s JIT and its effect on profiling
- Trade-offs: performance, maintainability, robustness, development cost, etc.
- Vertical vs. horizontal scalability
Standard Programming Techniques
- Loop optimizations
- Caching data
- Smart checks bypass
- Exception handling and performance
- Sizing data sets
- Inlining code
- Externally-connected systems
- Perceived vs. actual performance
- Using mathematical identities
MATLAB-Specific Techniques
- Using different storage types
- Object-orient MATLAB and performance
- Using internal helper functions
- Strings and dates/times
- MATLAB’s Startup Accelerator
Using Binary Code
- Mex
- MATLAB Compiler vs. MATLAB Coder
- 3rd-party libraries
I/O Speedup Techniques
- XLS/CSV read/write
- Binary vs. text format
- Reducing disk access
- Buffered, consolidated and chunked I/O
Vectorization and Parallelization Techniques
- Vertical vs. horizontal scaling
- Parallelization mechanisms in MATLAB
- Vectorization
- Explicit parallelization
- Amdahl’s Law
- Using the GPU
Graphics and GUI Techniques
- Initial graphs creation
- Updating graphs in real-time
- GUI preparation
- GUI responsiveness
- Feedback for long-duration tasks
- Dynamic updates/refresh
- Asynchronous updates/refresh
- Avoiding common pitfalls
Memory-Related Techniques
- Why memory affects performance
- Profiling memory usage in MATLAB
- MATLAB’s memory storage
- Optimizing loop ordering
- Pre-allocation of data
- Minimizing run-time memory allocations
- In-place data manipulations
- Using global and persistent variables
Conclusion
Training Materials
All MATLAB training students will receive comprehensive courseware.
Software Requirements
- Any Windows, Linux, or macOS operating system
- A recent version of MATLAB
