We present PULPO, a floating-point baseband-processing accelerator for massive multi-user multiple-input multiple-output (MU-MIMO) basestations (BSs). PULPO accelerates matrix-vector products, not only with a matrix but also with its Hermitian, as well as affine transforms and nonlinear projections used in iterative algorithms that outclass traditional linear methods in various applications. PULPO is integrated in a system-on-chip (SoC) with a tight integration to the system's data memory, facilitating data exchange and co-operation with 8 RISC-V cores. The fabricated accelerator achieves comparable efficiency as recently-proposed fixed-point baseband processors, while eliminating the burdens associated with fixed-point design, thus simplifying massive MU-MIMO BS development.
Castaieda O., Benini L., Studer C. (2022). A 283 pJ/b 240 Mb/s Floating-Point Baseband Accelerator for Massive MU-MIMO in 22FDX [10.1109/ESSCIRC55480.2022.9911311].
A 283 pJ/b 240 Mb/s Floating-Point Baseband Accelerator for Massive MU-MIMO in 22FDX
Benini L.;
2022
Abstract
We present PULPO, a floating-point baseband-processing accelerator for massive multi-user multiple-input multiple-output (MU-MIMO) basestations (BSs). PULPO accelerates matrix-vector products, not only with a matrix but also with its Hermitian, as well as affine transforms and nonlinear projections used in iterative algorithms that outclass traditional linear methods in various applications. PULPO is integrated in a system-on-chip (SoC) with a tight integration to the system's data memory, facilitating data exchange and co-operation with 8 RISC-V cores. The fabricated accelerator achieves comparable efficiency as recently-proposed fixed-point baseband processors, while eliminating the burdens associated with fixed-point design, thus simplifying massive MU-MIMO BS development.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
