In tests, the MIT researchers used Halide to rewrite several common image-processing algorithms whose performance had already been optimized by seasoned programmers. The Halide versions were typically about one-third as long but offered significant performance gains — two-, three-, or even six-fold speedups. In one instance, the Halide program was actually longer than the original — but the speedup was 70-fold:
Jonathan Ragan-Kelley, a graduate student in the Department of Electrical Engineering and Computer Science, and Andrew Adams, a CSAIL postdoc, led the development of Halide, and they’ve released the code online. The SIGGRAPH 2012 paper is presented on-line too and its authors also have Adobe and Stanford University affiliations:
"Decoupling Algorithms from Schedules for Easy Optimization of Image Processing Pipelines", Jonathan Ragan-Kelley, Andrew Adams, Sylvain Paris, Marc Levoy, Saman Amarasinghe, Frédo Durand
 |
| The basic camera post-processing pipeline is a feedforward pipeline in which each stage either considers only nearby neighbors (denoise and demosaic), or is point-wise (color correct and tone curve). The best schedule computes the entire pipeline in small tiles in order to exploit producer-consumer locality. This introduces redundant computation in the overlapping tile boundaries, but the reduction in memory bandwidth more than makes up for it. |
