Processing Elements and Matrix Compute Units
1.4.1Processing Element: MAC, Registers, and Local Control#
1.4.2Dot-Product Arrays, Outer-Product Arrays, and Reduction Trees#
1.4.3Accumulator Width and Partial-Sum Storage#
1.4.4Broadcast, Reduction, and Operand Distribution#
1.4.5PE-Level Pipelining and Significant-Bit Propagation#
1.4.6Multi-Precision Reuse, Zero Skipping, and Sparsity Metadata#