Cache Memory Energy Efficiency Exploration for the HEVC Motion Estimation

• Anderson Martins UFPEL
• Wagner Penny UFPEL
• Matheus Weber UFPEL
• Daniel Palomino UFPEL
• Júlio Mattos UFPEL
• Marcelo Porto UFPEL
• Luciano Agostini UFPEL
• Bruno Zatt UFPEL

Resumo

The intense demand for video applications from mobile devices brings a challenge to the hardware design, especially in the energy consumption. This work presents a design space exploration to define energy-efficient cache memory configurations for the ME process considering different video sequences and HEVC encoder configurations. We focus on the Motion Estimation (ME) process, known as the most processing and memory intensive task within the video encoder. Our evaluation considered the widely used TZ Search algorithm, 23 video sequences with distinct resolutions, and four QPs under 32 different cache configurations. A cache simulator was developed and CACTI was used to obtain time and energy results. Thus, it was possible to identify optimal cache configurations for each scenario. Considering the optimal cache configuration for each scenario, the use of cache may lead to bandwidth savings of up to 97.37%, reaching a reduction of 25.48 GB/s to 548.53MB/s in one case. The energy reduction reaches up to 93.95%, which corresponds, an energy reduction from 5.02 mJ to 0.30 mJ, when comparing different cache configurations.

Palavras-chave: Encoding, Cache memory, Bandwidth, Memory management, Video sequences, Motion estimation, Algorithm design and analysis, energy saving, HEVC, memory bandwidth reduction, motion estimation