Expand description
Constant-pattern variable-base scalar multiplication.
VariableBaseMul computes [scalar]P for a decoded affine point. Every
one of the 256 scalar bits executes one complete Edwards doubling and one
complete mixed addition; a final multiplexer selects the bit result. The
operation count and memory access pattern therefore do not reveal scalar
bits.
Four radix-2^51 multipliers execute each formula round together. Twenty
independent contexts hide their 17-cycle latency, allowing one four-product
round to issue each clock. A lane consumes 1,024 issue clocks per scalar;
the strict verifier uses two lanes for a sustained request interval near 512
clocks.
Structs§
- Variable
Base In - Tagged variable-base multiplication request.
- Variable
Base Mul - Twenty-context, four-multiplier variable-point scalar engine.
- Variable
Base Out - Result and flow control from
VariableBaseMul.
Functions§
- variable_
base_ kernel - State transition for
VariableBaseMul.