Acceleration method used:
- Memory locality: Input matrix reorganized
- Parallel (OpenMP with no more than 4 processing cores): Both gaussian filter and power law transformation.
- Look Up Table: Only in power law transformation.
Original source file is before_optimize.cpp
Accelerated source file is after_optimize.cpp
If you want more fun, checkout the playground branch!
All things you need to run these file is a up-to-date C++ compiler. If you are using g++ compiler and make tool, that you can use make file here to run the code.
Compile command of g++ for reference:
g++ -g -Wall -O0 -std=c++23 -march=native -fopenmp -o after_optimize after_optimize.cpp-
-std=c++23: Older version of C++ might be compatible but you should keep it as new as possible because lambda expression is only supported in modern C++. -
-march=native: If you want a portable executable, remove this option.