#hardware implementation

The former is more time consuming, but you are already writing something meant to be implemented in hardware, might as well make it applicable in real life.

It has been weeks since I started studying the CRYSTALS-Kyber key encapsulation scheme. Honestly, the more I look at it the more I realize I might have just scraped the surface. Every time I thought I understand it enough and moved on to the next step (math equations to software implementation to hardware implementation), I realized I did not understand it that well and went back to the specifications.

Good news: I have a better understanding on the hardware schemes.

Bad news: An implementation that is heavily optimized in time might not be possible due to resource constraints (FPGA board).

I might have neglected to mention that I am implementing the scheme on FPGA, and I think I will have to go for encryption module first. Since the resource will have to be reused, I guess finishing an encapsulation module is completing half of the decapsulation module.

Before going back to the specs and pseudo-code, I forgot the IND-CPA encryption scheme requires hash functions to generate the noise vectors. It is not a cryptographic implementation if I don't include hash functions.