TY - GEN
T1 - High-level synthesis implementation of transform-exempted SATD architectures for low-power video coding
AU - Partanen, Tero
AU - Lemmetti, Ari
AU - Sjövall, Panu
AU - Vanne, Jarno
N1 - JUFOID=65438
Funding Information:
This work was supported in part by the European ECSEL project ADACORSA (under the grant agreement 876019).
Publisher Copyright:
© 2021 IEEE
PY - 2021
Y1 - 2021
N2 - This paper presents the first known high-level synthesis (HLS) implementation for the Sum of Absolute Transformed Differences (SATD) calculation. The proposed hardware architecture is designed for two SATD algorithms: a widespread Fast Walsh-Hadamard Transform (FWHT-SATD) and a recently introduced Transform Exempted scheme (TE-SATD). This 2-stage architecture is made up of two 1-D Walsh-Hadamard Transform (WHT) stages and a transpose buffer (TB) between them. The chosen HLS approach cuts down design time over contemporary design methods and thereby made it feasible to implement a set of dedicated FWHT-SATD and TE-SATD architectures for 4×4, 8×8, and 16×16 pixel blocks. All these six architectures were synthesized for 28 nm and 45 nm standard cell technologies, and their area and energy consumptions were analysed. TE-based implementations provide 6.0-8.3% total cell area savings and 6.9-12.7% better energy-efficiency than traditional FWHT approaches. Our proposal is the first to introduce TE-SATD architectures for up to 16×16 blocks and each of these tailored architectures was shown to provide better trade-off between silicon area and performance than their reference implementations.
AB - This paper presents the first known high-level synthesis (HLS) implementation for the Sum of Absolute Transformed Differences (SATD) calculation. The proposed hardware architecture is designed for two SATD algorithms: a widespread Fast Walsh-Hadamard Transform (FWHT-SATD) and a recently introduced Transform Exempted scheme (TE-SATD). This 2-stage architecture is made up of two 1-D Walsh-Hadamard Transform (WHT) stages and a transpose buffer (TB) between them. The chosen HLS approach cuts down design time over contemporary design methods and thereby made it feasible to implement a set of dedicated FWHT-SATD and TE-SATD architectures for 4×4, 8×8, and 16×16 pixel blocks. All these six architectures were synthesized for 28 nm and 45 nm standard cell technologies, and their area and energy consumptions were analysed. TE-based implementations provide 6.0-8.3% total cell area savings and 6.9-12.7% better energy-efficiency than traditional FWHT approaches. Our proposal is the first to introduce TE-SATD architectures for up to 16×16 blocks and each of these tailored architectures was shown to provide better trade-off between silicon area and performance than their reference implementations.
KW - Hadamard transform
KW - High-level synthesis (HLS)
KW - Low-power hardware design
KW - Sum of absolute transformed differences (SATD)
KW - Video coding
U2 - 10.1109/ISCAS51556.2021.9401399
DO - 10.1109/ISCAS51556.2021.9401399
M3 - Conference contribution
AN - SCOPUS:85108999723
SN - 9781728192017
T3 - IEEE International Symposium on Circuits and Systems proceedings
BT - 2021 IEEE International Symposium on Circuits and Systems, ISCAS 2021 - Proceedings
PB - IEEE
T2 - IEEE International Symposium on Circuits and Systems
Y2 - 22 May 2021 through 28 May 2021
ER -