Module scalar_parallel 

Initiation-interval-one scalar arithmetic for Ed25519.

The group order is l = 2^252 + q, where q is only 125 bits. Reduction therefore folds high bits through multiplication by the fixed eight-limb radix-2^16 constant q. The multiplier below is fully spatial: each of 24 coefficients owns eight registered 16 x 16 products, and normalization advances two limbs per stage.

Structs

ConstQMulIn

Tagged input to ConstQMulPipe.

ConstQMulOut

Tagged output from ConstQMulPipe.

ConstQMulPipe

Seventeen-stage multiplication by the Ed25519 reduction constant q.

Radix16X17

Seventeen radix-2^16 input limbs, least significant first.

Radix16X25

Twenty-five radix-2^16 product limbs, least significant first.

Radix16X33

Exact 528-bit output of ScalarWideMulPipe.

ScalarMulAddIn

Tagged scalar multiply-add request.

ScalarMulAddOut

Tagged canonical scalar multiply-add response.

ScalarMulAddPipe

Fully pipelined a * b + c mod l engine.

ScalarReducePipe

Fifty-seven-stage, initiation-interval-one reduction modulo l.

ScalarReducePipeIn

Tagged 512-bit reduction request for ScalarReducePipe.

ScalarReducePipeOut

Tagged canonical scalar from ScalarReducePipe.

ScalarWideMulIn

Tagged request for ScalarWideMulPipe.

ScalarWideMulOut

Tagged response from ScalarWideMulPipe.

ScalarWideMulPipe

Twenty-two-stage, 512-DSP full scalar multiplier.

Functions

const_q_mul_kernel

State transition and hierarchy wiring for ConstQMulPipe.

scalar_muladd_kernel

State transition for ScalarMulAddPipe.

scalar_product_addend_kernel

Add a canonical scalar addend to a wide radix-2^16 product.

scalar_reduce_pipe_kernel

State transition for ScalarReducePipe.

scalar_wide_mul_kernel

State transition for ScalarWideMulPipe.