How to make MATLAB code execute faster
조회 수: 101 (최근 30일)
이전 댓글 표시
Hi,
How to make MATLAB code execute faster?
댓글 수: 2
채택된 답변
Andreas Goser
2014년 2월 2일
Identify bottlenecks by using the Profiler tool within MATLAB.
Reflect on you really need that size of data and if the goal of your computation / project can maybe achieved by using less data.
Work with your IT department and MathWorks on finding ways for computing your stuff on a cluster.
추가 답변 (3개)
Jan
2014년 2월 2일
Use a clean and efficient code. E.g.:
- pre-allocation
- move repeated calculations out of the loops
- prefer efficient operations, e.g. a multiplication is faster than a division, sqrt() is faster than ^0.5
- Process the data in column direction, because it is faster to access elements, which are stored side by side in the memory
- Vectorized code is often, but not always, faster in Matlab. This depends mainly on the´size of the required temporary arrays
- Use an efficient structure to store your data. If you have to search in a list of strings, which is changed rarely only, using a sorted list can reduce the processing time massively, because this allows for a binary search
- For large data sets, the RAM can be the limiting factor. Then a deeply nested struct might be better to keep the overview during programming, but one large cell can be faster for processing.
- Never optimize prematurely. If you start programming with optimizing the code too much, the time required for debugging might explode. It is not useful to reduce the processing time by some seconds, if the debug time is some month. So start with creating a clean and correct code, take into account a efficient data structure, and finally optimize the bottlenecks only.
댓글 수: 0
참고 항목
Community Treasure Hunt
Find the treasures in MATLAB Central and discover how the community can help you!
Start Hunting!