Faster than the Speed of BRAM: In-Memory Computing for Next Generation FPGAs
MC 2 -BRAM is a Memory-Centric Compute BRAM architecture for FPGAs that integrates arithmetic units and dataflow control directly within on-chip block memory, enabling Processor-in-Memory execution for machine learning workloads and achieves the lowest inter-PIM accumulation latency among all proposed custom PIM tile d...