mlkem-sync

fallensigh/mlkem-sync

Fork 0

Commit Graph

Author	SHA1	Message	Date
FallenSigh	3a53993754	refactor(kg): make ML-KEM K a runtime input k_i instead of a parameter mlkem_top now sizes storage for KMAX=4 (worst case) and selects the active ML-KEM parameter set at start_i via the k_i input. All K-derived quantities (eta1, slot bases, ek/dk byte counts, H(ek) block count, FSM bounds) are computed at runtime from the captured k_r. Verified byte-exact against NIST KAT for all three parameter sets: K=2 (512) cases 0-4, K=3 (768) cases 0-2, K=4 (1024) cases 0-2 -> 11/11 PASS (ek==pk, dk==sk).	2026-06-28 03:24:58 +08:00
FallenSigh	b2bf798454	feat(mlkem_top): parameterize K in {2,3,4} (ML-KEM 512/768/1024) Generalize KeyGen from K=2-hardcoded to compile-time parameter K: - eta1 derived (3 for K=2, else 2); slot layout SLOT_S/E/T = KK+{0,K,2K}, NUM_SLOTS = KK+3K; SAW=5 slot-addr width. - A-stage: explicit a_i/a_j row-major counters (slot = iK+j) instead of K=2 bit-tricks. C/N stages: parametric slot bases, 2K polys. - M-stage: m_i/m_j widened to 3-bit (must reach K=4); slots iK+j etc. - E-stage: 2K polys, e_is_dk split, rho offset 384*K. - H(ek): H_NBLK=ceil((EK_BYTES+1)/136), H_LAST padding generalized; h_blk 4-bit. Byte mems sized EK_BYTES/DK_BYTES. - Widen dbg_byte_idx_i to [10:0] (ek up to 1568B for K=4). Parametric TB (tb_mlkem_kg_katK, KP generic + CASE plusarg). Verified byte-exact vs NIST KAT: K=2 (512): cases 0..4 ek 800B / dk 1632B K=3 (768): cases 0..2 ek 1184B / dk 2400B (~36k cyc) K=4 (1024): cases 0..2 ek 1568B / dk 3168B (~54k cyc) run_tb.sh top runs all three parameter sets.	2026-06-28 02:59:58 +08:00

Author

SHA1

Message

Date

FallenSigh

3a53993754

refactor(kg): make ML-KEM K a runtime input k_i instead of a parameter

mlkem_top now sizes storage for KMAX=4 (worst case) and selects the
active ML-KEM parameter set at start_i via the k_i input. All K-derived
quantities (eta1, slot bases, ek/dk byte counts, H(ek) block count, FSM
bounds) are computed at runtime from the captured k_r.

Verified byte-exact against NIST KAT for all three parameter sets:
  K=2 (512)  cases 0-4, K=3 (768) cases 0-2, K=4 (1024) cases 0-2
  -> 11/11 PASS (ek==pk, dk==sk).

2026-06-28 03:24:58 +08:00

FallenSigh

b2bf798454

feat(mlkem_top): parameterize K in {2,3,4} (ML-KEM 512/768/1024)

Generalize KeyGen from K=2-hardcoded to compile-time parameter K:
- eta1 derived (3 for K=2, else 2); slot layout SLOT_S/E/T = K*K+{0,K,2K},
  NUM_SLOTS = K*K+3K; SAW=5 slot-addr width.
- A-stage: explicit a_i/a_j row-major counters (slot = i*K+j) instead of
  K=2 bit-tricks. C/N stages: parametric slot bases, 2K polys.
- M-stage: m_i/m_j widened to 3-bit (must reach K=4); slots i*K+j etc.
- E-stage: 2K polys, e_is_dk split, rho offset 384*K.
- H(ek): H_NBLK=ceil((EK_BYTES+1)/136), H_LAST padding generalized;
  h_blk 4-bit. Byte mems sized EK_BYTES/DK_BYTES.
- Widen dbg_byte_idx_i to [10:0] (ek up to 1568B for K=4).

Parametric TB (tb_mlkem_kg_katK, KP generic + CASE plusarg). Verified
byte-exact vs NIST KAT:
  K=2 (512):  cases 0..4  ek 800B  / dk 1632B
  K=3 (768):  cases 0..2  ek 1184B / dk 2400B  (~36k cyc)
  K=4 (1024): cases 0..2  ek 1568B / dk 3168B  (~54k cyc)
run_tb.sh top runs all three parameter sets.

2026-06-28 02:59:58 +08:00

2 Commits