Command:
python -m unittest discover -s sim -vResult: ten tests passed.
Coverage:
- exhaustive
GF(2^8)serial-vs-polynomial multiplication: 65,536 pairs; - 100,000 deterministic random
GF(2^128)pairs; - multiplication identities and reduction boundary behavior;
- command-level XOR, MUL, SET, NONZERO, STATUS, CLEAR, and error behavior;
- exact no-stall cycle counts: XOR 33, MUL 161, SET 17, NONZERO 17, STATUS 5, CLEAR 1;
- 1,000 randomized cycle-model transactions with independent receive stalls and transmit backpressure;
- unknown-command error behavior.
The structural SystemVerilog smoke checker also reports balanced delimiters and block constructs for all RTL, testbench, and formal-harness files.
The environment lacked:
- Lean / Lake;
- Icarus Verilog or Verilator;
- Yosys / OpenLane;
- SymbiYosys / Boolector.
Package download and apt access were unavailable. Therefore the following must run in CI or a normal development machine:
cd lean && lake build
cd test && make sim
cd formal && sby -f gf8_mul.sbyThe project deliberately includes no proof placeholders or global axiom declarations. CI also checks for any
future sorry, admit, or axiom occurrence.
The checks below provide finite or model-scoped evidence only. They do not
connect the full LSC-1 SV controller to the frozen ISA and do not establish
RTL-to-netlist equivalence; see docs/PROOF_BOUNDARIES.md.
Three distinct descriptions are present:
- bit-streamed RTL transition;
- Python carry-less schoolbook product with polynomial long reduction;
- Lean
GF(2^8)bit-vector model checked bybv_decide.
The SymbiYosys harness contains a fourth, local schoolbook/reduction function for the eight-bit RTL specialization.