BEMAP (BEnchMarks for Automatic Parallelizer) is a benchmark used to measure performance for an automatic parallelizer.
All OpenCL code benchmarks covered in this project are done step-by-step alongside with hand-tunning.
Each tuning step executional time are measured in details with a comprehensive user interface and help option.
The exact implementation in native code (C++) is also provided in each project folder for reference.
By analyzing these benchmarks, one can analyze:
1. How to tune/optimize the auto-parallelizer's compiler
2. Analyze whether the appropriate optimization procedure is provided within the compiler
The latest package includes:
Gaussian Blur
Black Scholes
Linear Search
Grayscale
SIFT (a demo app included!)
Runlength
Back Projection
Monte Carlo
This project is partially funded by the Department of the New Energy and Industrial Technology Development Organization (NEDO).