IAM and IEM muxes control code.

CRYSTALS-Kyber has been standardized by the National Institute of Standards and Technology (NIST) as a quantum-resistant algorithm in the post-quantum cryptography (PQC) competition. The bottleneck in performance of Kyber is the polynomial multiplication based on Number Theoretic transform (NTT). This work presents two parallel architectures adopting Multi-Path Delay Commutator (MDC) approach on target FPGA platform. Resource sharing technique is adopted to perform PWM operations using MDC NTT/INTT architecture. Moreover, we propose various optimizations at architectural level to minimize resource consumption such as FIFO-based memory units for buffering of input output coefficients, LUT-based modular multiplier and distributed-ROM memories for twiddle factor storage. The presented architectures are implemented on Xilinx Artix-7 XC7A100T-3 device using Vivado Design Suite 2022.2 and coded using Verilog HDL. Our BRAM and DSP-free designs achieve 68% improved area-time product with a comparable ATP for PWM operations. Additionally, the two-parallel MDC architecture outperforms state-of-the-art architectures, using 29% fewer resources.