feat(dec): Decaps D1 - byteDecode_d + Decompress -> u'/v'

K-PKE.Decrypt step 1 (FIPS 203 Alg 15): decode+decompress the ciphertext.

- comp_decomp_sync instance made mode-selectable: Encaps C1/C2 still compress
  (mode 0), Decaps ST_DEC_DECOMP decompresses (mode 1) with d=du/dv.
- New ST_DEC_DECOMP state with an inline byteDecode_d walker (reverse of the
  C1/C2 bit-packer): walks c_in_bram bytes, accumulates LSB-first into a bit
  buffer, extracts d-bit symbols, feeds comp_decomp, writes each decompressed
  coeff (mod q) to a bank.
    c1 = K polys, d=du -> u'[i] in bank_se rel slot i (0..K-1)
    c2 = 1 poly,  d=dv -> v'   in bank_t rel slot DEC_VSLOT=2 (avoids UPSUM=1)
- dbg_slot_i widened 4->6 bits so the TB can read v' (abs slot 26 at K=4).
- bse/bt write muxes gain the DECOMP writeback paths (dec_u_we / dec_v_we).

Verification: examples/dump_decaps.rs (ml-kem-r worktree) emits per-stage
golden (u'/v'/s_hat/u_hat/w/m') into vectors/decgold/. TB verify_d1 reads back
u'[i] and v' and compares all 256 coeffs each.

Bring-up note: dbg coeff readback latency is bank(1)+dbg_coeff_r(1); the TB's
rdcoeff initially waited 2 cyc and saw data shifted by one index -> fixed to 3.

Verified: dec D1 K=2/3/4 all cases PASS; KeyGen + Encaps unregressed.
This commit is contained in:
2026-06-29 17:32:03 +08:00
parent e46d2258d9
commit ecc00d6dd5
111 changed files with 25841 additions and 19 deletions

View File

@@ -62,7 +62,7 @@
## 实现阶段(逐阶段 dbg/KAT 对拍)
- **D0 — 脚手架 + dk/c 载入 + 解析** ✅:op_i 加宽 2-bit(00 KG/01 Enc/10 Dec),ST_DEC_LOAD(D0 暂直接→DONE)。dk 流入按 region 路由:dk_pke→dkp_bram、ek_pke→ek_bram、h→hek_r、z→z_r;ct→c_in_bram(独立于 ct_bram)。dbg 验证 h/z/ek_pke/dk_pke。**踩坑1:载入路由用 k_r 但 k_r 在 start_i 才锁存 → 预载期 region 边界全 0,路由全错。改用 LIVE k_i 边界(dkp_bytes_ld 等)。踩坑2:旧 KG/Enc TB 未接新端口(dk_in_*/c_in_*/dbg_*)→ X 漂入 write mux,KeyGen/Encaps 超时回归。补 tie-off 0。** runner = `./run_tb.sh dec [K] [CASE]`。K=2/3/4 D0 全过,KG/Enc 回归通过。
- **D1 — byteDecode_d + Decompress → u'/v'**:新解包器 + comp_decomp mode=1。dbg 对 u'/v'(ml-kem-r golden,新 dump_decaps)
- **D1 — byteDecode_d + Decompress → u'/v'** ✅:复用 comp_decomp(改 mode 可选:Encaps C1/C2 compress mode=0,Decaps DECOMP mode=1)。ST_DEC_DECOMP 内联 byteDecode 走子机:逐字节读 c_in_bram,LSB-first 累进 bit buffer,凑够 d 位抽符号→comp_decomp 解压→写 bank。c1(K 多项式,d=du)→u'[i] bank_se rel i;c2(1 多项式,d=dv)→v' bank_t rel DEC_VSLOT=2(避开 UPSUM=1)。dbg_slot_i 加宽 4→6 bit(K=4 v' 在 slot 26)。dump_decaps.rs(ml-kem-r 工作树)产 u'/v'/s_hat/u_hat/w/m' golden 到 vectors/decgold/。**踩坑:dbg coeff 读回延迟 = bank(1)+dbg_coeff_r(1),TB rdcoeff 等 2 拍少一拍 → 数据整体错位一格;改 3 拍修正。** K=2/3/4 全过,KG/Enc 回归通过
- **D2 — s_hat 解码 + u_hat = NTT(u')**:TDEC 复用(dk_pke 源)、ntt_core fwd。dbg 对 s_hat / u_hat。
- **D3 — w = v' INTT(Σ s∘u_hat)**:V 机器 SUB 变体。dbg 对 w。
- **D4 — m' = byteEncode₁(Compress₁(w))**:打包器 d=1。dbg 对 m'(== KAT 解密的 m')。