Application-Specific Cache and Prefetching for HEVC CABAC Decoding

Philipp Habermann, Chi Ching Chi, Mauricio Alvarez-Mesa, Ben Juurlink
2017 IEEE Multimedia  
Context-based Adaptive Binary Arithmetic Coding (CABAC) is the entropy coding module in the HEVC/H.265 video coding standard. As in its predecessor, H.264/AVC, CABAC is a well-known throughput bottleneck due to its strong data dependencies. Besides other optimizations, the replacement of the context model memory by a smaller cache has been proposed for hardware decoders, resulting in an improved clock frequency. However, the effect of potential cache misses has not been properly evaluated. This
more » ... work fills the gap by performing an extensive evaluation of different cache configurations. Furthermore, it demonstrates that application-specific context model prefetching can effectively reduce the miss rate and increase the overall performance. The best results are achieved with two cache lines consisting of four or eight context models. The 2 × 8 cache allows a performance improvement of 13.2 percent to 16.7 percent compared to a non-cached decoder due to a 17 percent higher clock frequency and highly effective prefetching. The proposed HEVC/H.265 CABAC decoder allows the decoding of high-quality Full HD videos in real-time using few hardware resources on a low-power FPGA. High Efficiency Video Coding (HEVC/H.265; here, HEVC) is the most recent video coding standard developed by the Joint Collaborative Team on Video Coding (JCT-VC). 1 HEVC allows video compression with the same perceptive quality as its predecessor H.264/AVC 2 (here, H.264) while requiring only half the bitrate. Due to its sequential nature, Context-based Adaptive Binary Arithmetic Coding (CABAC) 3,4 was a throughput bottleneck in H.264 decoding. This is still the case in HEVC. High-throughput requirements must be fulfilled to achieve real-time decoding of high-quality, high-resolution videos. This is a challenging task, especially on mobile devices and other power-and performance-constrained systems, making energy-efficient hardware solutions necessary.
doi:10.1109/mmul.2017.12 fatcat:aoiwpo3bgvc77pik4i5gbqfgcq