# 0xE4B713EF AFC8: ATT array fast-clock image, derived and LAUNCHED (box1-two-v2 build worker, 2026-11-17)

Program 1: one Llama-3.2-3B drafter on 7 V80s (ATT TP4 x4 + MLPV x2; H200 verifies). Motto: never deny the idea; engineer toward the fastest flashable PDI.
Lever: `ATT_FASTCLK_20261117.md` (commit 6f16e0449). Base image: NFA8 0xE4B713CE (`COMBO3_NF_A8QOVL_20261117.md`, `AUDIT_NFA8_0x13CE_20261117.md`).
**Nothing was flashed, killed or removed.** The NFA8 build (buildx03 s3LAtp4CE) and the A8OV build (build00 s3LAtp4D3) were not touched; I only read from them. **NO FLASH before adjudication.**

## Calculations and hypothesis (everything that is not a measurement)
- The clk_wiz change follows from the measured VCO. LCM(66, 132, 286) = 1,716, and 2 x 1,716 = 3,432 MHz is the only multiple inside the MMCM VCO range. That is the same VCO the {66,264,286} request already resolves to, so the expected dividers are O1 52 / O2 26 / O3 12, at the same M 205.921875 / D 2. The {66,132} MLPV lineage already resolves to M 205.921875 / D 2 / O2 26 / O1 52 (`uni_pipeline_4c1v12_mlpv_dg30.sh:329`). The real dividers land in the routed XCI. The `ATTAFC-clkwiz-2to1` row reads them back (report-only).
- `DC_FAST_FREQHZ` seed = 131,999,881. Two sources point to it: the MLPV {66,132} BD-actual at the same M/D and O2, and the CE build's own fast BD-actual 263,999,763 (= 2 x 131,999,881.5). If the BD reports a different value, STEP4b learns it and retries once. That retry costs the BD stage only, and the CE build took it too.
- The per-tree gain projection stays in the lever report (up to -271 us/tree at d11). This build adds no new claim.

## 1. Image [M]
| field | value |
|---|---|
| build id | **0xE4B713EF** (label), probe `8'hEF`, reg100 short id **0xE4B711EF** |
| arm | AFC8 = NFA8 0xE4B713CE + `ATT_ARRAY_FAST_CLK_V 1'b1` |
| host / slot | **buildx02** / `s3LAtp4EF` (clone `/shared/build/s3LAtp4EF/aved_oneflip_p2s_head_asgard`) |
| stage | build host `/shared/build/lutg_d9_tp4stk_EF_stage`; box1 copy `/mnt/data/eagle4_scratch/att_fastclk_build_20261117/lutg_d9_tp4stk_EF_stage` |
| launch | 2026-11-17T21:22:21Z: `cd <repo> && TC3A_SLOT=s3LAtp4EF NFA8_MAXTHREADS=8 setsid nohup bash /shared/build/lutg_d9_tp4stk_EF_stage/launch_COMBO3.sh > .../afc8_pipeline.log 2>&1 < /dev/null &` |
| first state | STEP0 / STEP0.5 passed, including the CERT_TOP / CFG / GSAS / DCW checks that `gate_fail` on a mismatch. STEP1 PACKAGE was admitted at 21:22:22Z, and Vivado pid 3632834 was elaborating at 21:27Z with 0 ERROR lines. |
| pipeline logs (build host) | `<repo>/reports/results/UNI_DUALCLK_s3LAtp4EF_head_ASGARD_PIPELINE.log`, `GATE_VERDICTS_DUALCLK_s3LAtp4EF_head_ASGARD.txt` |

Host checks before launch (buildx02):
- `host_build_count_v3`: `0 173 251 1 builds=`, meaning 0 builds, 173 GB available, 251 GB total, and tunnel key 1. The only Vivados on the host are 2 srpol optdiff census jobs; neither is a build.
- SMBus smoke: `SMOKE_IS_LOCKED 0`. `/shared` has 100 GB free.
- AVED base: `build_design.tcl` is 0783d8aa, versus 4e3996d4 on buildx01/03. The only difference is the 60-min "QW LICENSE GATE" smbus wait in synth and impl. That wait printed `smbus checkable=1` on buildx02's own ATT build `s3LAtp4C1`, which reached a PDI. `create_design.tcl` 5c1958e1 and `create_bd_design_layout_b.tcl` 6d8071eb are equal on all three hosts.
- The pipeline's 24 `$REPO/...` paths have identical md5s on buildx02 and buildx03. 10 of them are legacy paths that are missing on both hosts.

## 2. Build-id collision check [M]
The method reuses `scripts/4c1/qgptq2/qg_id_check.py` `sources()`: every non-QGPTQ2 c4_regmap int, the pre-Qwen c4_regmap, TRACKLIST, NOW.md, git log --all, every reports/docs *.md, tested_archive and the artifact file-name trees. I scanned each free low byte in the box1-two-v2 blocks 0x13B1-0x13FF that is not already in `box1two_reserved_ids_20261116.txt`.
- 0xE4B713**EF**: **0 label hits, 0 live reg100 hits**, and 9 retired-only reg100 hits (HQ/KG trees). It is the only clean byte left. Every other free byte had a label hit (DA, E1, E5, EE, FE, FF) or a live reg100 hit in c4_regmap or flash_records (D2, E3, E4, E6-EA, ED, EE).
- 13EF is inside the hq-ops-acked box1 block 0x13E0-0x13EF (`TRACKLIST_20260923.md:611`, `c4_regmap.py:1252`). It does not collide with the junior's ids (13A0, 13E5, 1380-138D) or with today's 0x1379 claim.
- `qg_id_check.py --free` printed nothing. It fails closed because the reserved-list md5 changed (pinned aa41ef6c, file had become 410d2d27 before my append). That is a Qwen-tool pin, not a collision.
- I appended 0xE4B713EF to the reserved list and wrote the claim and launch line into `4C1_DELIVERY_TRACKER.md` in the same turn. The c4_regmap IMAGE_ALT / EXPECT row goes in at adjudication, as for 0x13CE: `test_record_profile_identity.py` guards IMAGE_ALT.

## 3. Derivation [M]
Deriver: `scripts/derive_lutg_tp4_nfa8_fastclk.py` (24475912). Log: `att_fastclk_20261117/build_EF/derive_EF.out` (920089e6). Record: `DERIVATION_AFC8.txt` (d5c2ecc5).
1. It runs `derive_lutg_tp4_combo3nfa8.py` (d042b8cd, `NFA8_NF=1`) unmodified with 0xE4B713EF / EF. All of its own PROOF-REPRO checks re-pass. Evidence: **772 / 772 files equal** to the launched CE stage once the id, probe, slot and stage-name tokens are mapped (0 differ, 0 missing). The 6 NFA8 RTL files match.
2. PROOF-REPRO: `patch_att_fastclk.py` (7ca95bf3) run on the staged NFA8 files, with `--bridge` = the fork's `gemv_array_dualclk.sv` 0395872d, reproduces all **5 / 5** fork files byte-exact. The same bridge md5 is present in tp4mlpv_134d, mlpv_1320, reimpl1376 and g8own_v2 MLPV stages. Then the fork files are overlaid, with the originals kept as `*.pre_fastclk`. Full tree: **149 / 149** non-backup files under fork `src_full` equal the staged copies.
3. Strap: the staged config changes in exactly one line, `ATT_ARRAY_FAST_CLK_V 1'b0 -> 1'b1`. This is the F1_L100 sim arm's `-d ATT_ARRAY_FAST_CLK_V=1`. `ARRAY_FAST_CLK_V 1'b0`, `ARRAY_FAST_TP_V 1'b0` and `ATT_FASTCLK_RED_V 0` stay as residency. The top parameter ARRAY_FAST_CLK stays 0 (report section 5); the XCI row `ARRAY_FAST_CLK want 0` is unchanged.
4. I did not run a separate xsim of the staged copy. The staged RTL is byte-identical to the fork, and its one config difference is the define the F1 identity arm already used (13a16bea).
5. must_pass: `--stage` **PASS** (A2: 0 macro-ternary port widths, 2 parameter-ternary) and `--reddef` **PASS**. A4 (pkg-param coverage) does not apply because the strap is a header define compiled from the packaged header; packaging visibility is checked by the XCI `PARAM_VALUE.ATT_ARRAY_FAST_CLK` row. `test_record_profile_identity.py` was not run because no host file changed.

| staged file | NFA8 (CE) | AFC8 (EF) |
|---|---|---|
| `top/eagle4_config.svh` | 89565763 | **38e7a756** (fork 6a58f888 + strap 1) |
| `top/eagle4_drafter.sv` | 92ab14a2 | **901c191a** |
| `top/gemv_stage_array_scaled_slr.sv` | 7f754cc3 | **640a4530** |
| `primitives/lutg_a8q.sv` | 68be50f9 | **31d6dfdd** |
| `top/gemv_array_dualclk.sv` | a44f3e9f | **0395872d** (record MLPV 0x1361 bridge) |
| `layers/premlp_normfuse.sv` | ddbc0795 | ddbc0795 |
| `uni_pipeline_lutg_d9.sh` | 5a4b168b | **ebaac8da** |
| `run_lutg_d9.sh` | 65211bd9 | **b5de36be** (CERT_TOP 901c191a / CFG 38e7a756 / GSAS 640a4530) |
| `MD5LIST_head_hxf_cons.txt` | 6ecf582c | **bb89a4ee** (392 lines; 392/392 OK on box1 and buildx02) |
| `check_v1_netlist_lutg.sh` | 5ff9db77 | **b5352856** |
| `constraints_sllfix/eagle4_drafter_dualclk_loc_pblock.xdc` | fe2a53bd | **7c1caf10** |
| `launch_COMBO3.sh` / `COMMON_MD5.txt` | | 9d89ad8f / ad941439 (COMMON OK) |
| `STAGE_SHA_0xE4B713EF.txt` | | 4c69f53d |

The full stage diff is in `att_fastclk_20261117/build_EF/NFA8_STAGE_TO_AFC8_STAGE.diff` (2cd39e2b).

## 4. Clock change [M]
- **Before**, measured on the running 0x13CE build (buildx03 `top_usr_clk_wiz_0.xci`): PRIM_IN 33.333, CLKFBOUT_MULT 205.921875, DIVCLK 2, so VCO = 3,431.997 MHz. Request {66,264,286}, O1 52.000 / O2 13.000 / O3 12.000. BD-actual ap_clk 65,999,940 and ap_clk_fast 263,999,763 (from that build's STEP4b line), HBM 285,999,740.
- **After**: request **{66,132,286}**, expected O1 52 / **O2 26** / O3 12 at the same VCO, so 66.0 and 286.0 stay bit-identical. 132.0 is exactly 2 x 66.0 off one MMCM.
- **Mechanism.** STEP2.6 still writes {66,264}, and the fork's `step29_bd_hbmclk.py` (unchanged) still makes it {66,264,286}. A new count-checked **report-only** sed then turns {66,264,286} into {66,132,286} (`uni_pipeline_lutg_d9.sh:2170-2177`, row `ATTAFC-clkwiz-request`). The CE build's final BD tcl has exactly 1 x {66,264,286} (line 266) and the live `clock_reset_clk_usr_1 -> eagle4_drafter_lb_0/ap_clk_fast` net (line 1302), so the sed's pre-count is 1.
- **Pipeline rows re-keyed:**
  - `DC_FAST_FREQHZ` seed 263999760 -> 131999881 (`:287`).
  - Sanity-synth `create_clock ap_clk_fast` changes from `DERATE_PERIOD/4` to `/2` (`:1215`).
  - `BD-clk300` now wants {66,132,286} x1 and {66,264,* x0 (`:2179`).
  - The clkwiz row wants O1 52 / O2 26 / O3 12 (`ATTAFC-clkwiz-2to1`, `:2857`).
  - `CERT_DCW_P` a44f3e9f -> 0395872d (`:701`, a STEP0.5 `gate_fail`; it passed).
- **XDC.** The existing in-build margin `margin_att_d1.xdc` already covers clk_usr_1 (setup 0.040 / hold 0.020, cancelled for FINAL_TRUE). On top of that, `g_lcache.g_lc_grp[0..2]` is added to `pblock_dc_grp[0..2]`, so each cache sits next to its lanes, as in the OOC `att_fc_pblocks.tcl` plan. `DCXDC_MD5` was re-pinned. No false path, multicycle or ASYNC_REG was added on the bridge crossings.

## 5. Report-only rows added or re-keyed [R]
| row | wants |
|---|---|
| `ATTAFC-staged` (STEP3) | staged `ATT_ARRAY_FAST_CLK_V 1'b1` x1 and 1'b0 x0, `ATT_FASTCLK_RED_V 0` x1, `ARRAY_FAST_CLK_V 1'b0` x1, bridge 0395872d, drafter 901c191a |
| `NFA8-staged` | lutg_a8q / array md5 re-pinned to 31d6dfdd / 640a4530 |
| XCI `PARAM_VALUE` | `ATT_ARRAY_FAST_CLK` 1, `ATT_FASTCLK_RED` 0 |
| `ATTAFC-SIBW-absent-bridge-present-netlist` | sibw_pump_e3 0, u_lane_s 0, gemv_dualclk_bridge >= 1, g_lcache >= 1 |
| `ATTAFC-netlist` | u_dc_bridge, u_a8q_s, g_lc_grp >= 1; `g_a8q.u_lead` and `g_a8q.u_a8q` 0 |
| `ATTAFC-fastclk-loaded` | routed clock_reset_clk_usr_1 endpoints >= 10,000 (replaces `LUTG-pumpclk-unloaded`, < 1000) |
| `check_v1_netlist_lutg.sh` SIBW-off | gemv_dualclk_bridge > 0 |

Known inherited FAIL rows on this lineage are also printed on 0x13CE: p2s_selfref selftest SKIP, 8-3848, PREMLPCHECK, OSPREYCHECK, enforce-manifest, portw-reg0-range and THREEC_A STEP3. On 0x13CE they account for the 12 FAIL lines at the pkg / STEP3 stage (pkg-census contributes 4 of them).

## 6. Watcher [M]
- `scripts/att_fastclk_20261117/watch_afc8.sh` (3cf06efc) runs on box1 as pid 369876 since 21:26:01Z. It polls every 5 min for up to 72 h, report-only. Its STATUS file is `/var/tmp/box1two_fastclk_20261117/watch_EF/STATUS`.
  - **pkg:** `must_pass_gate.sh --pkg --kind ATT --dual 1`, packaged `ATT_ARRAY_FAST_CLK` / `ARRAY_FAST_CLK` / NORMFUSE / A8Q_OVL / FREQ_HZ.
  - **opt dcp:** `scripts/att_fastclk_20261117/afc8_census.tcl` (4e7ab64e) in a separate Vivado (MemAvailable >= 60 G) under `/shared/scratch/afc8_census_EF_20261117`. It covers the bridge / u_a8q_s / per-group cache RAMB (want 4 each) / in-array quantiser 0; sample-FD clocks (lanes on clk_usr_1); clk_usr_0/1/2 periods (want 15.152 / 7.576 / 3.497 ns); B3 nets; and the NFA8 NF / A8Q_OVL rows on u_a8q_s.
  - **placed dcp:** `report_utilization -slr`, for SLR0 SLICE.
  - **routed:** WNS/WHS and the clk_usr_1 intra / inter rows.
  - **every tick:** the ATTAFC / BD-clk300 / DCCHECK gate rows.
- First watcher tick at 21:26:02Z: STEP1 PACKAGE admitted, 1 FAIL in GATE_VERDICTS (the inherited p2s_selfref selftest SKIP, same as 0x13CE).
- STEP1 pkg sanity synth (`dualclk_s3LAtp4EF_head_box1_pkg_vivado.log`): **Finished RTL Elaboration at 21:29Z, 0 ERROR lines** (cpu 6:21, peak 13.45 GB). The elaboration guards `g_att_afc_x` (`ERROR_ATT_ARRAY_FAST_CLK_REQUIRES_...`), `g_lutg_top_x` and the array's `LCACHE && !A8Q` guard therefore did not fire on the strap-1 configuration.
- Still to run on artifacts: F1 `--srpol` + `--srpol-optdiff` on the final routed dcp, plus B1/B2. No CSR moved, but this is the first build of the AFC8 lineage.

## 7. Files
| file | md5 |
|---|---|
| `scripts/derive_lutg_tp4_nfa8_fastclk.py` | 24475912 |
| `scripts/att_fastclk_20261117/watch_afc8.sh`, `afc8_census.tcl` | 3cf06efc, 4e7ab64e |
| `att_fastclk_20261117/build_EF/{STAGE_SHA_0xE4B713EF.txt, DERIVATION_AFC8.txt, derive_EF.out, NFA8_STAGE_TO_AFC8_STAGE.diff}` | 4c69f53d, d5c2ecc5, 920089e6, 2cd39e2b |
| `scripts/4c1/box1two_reserved_ids_20261116.txt` (+0xE4B713EF), `4C1_DELIVERY_TRACKER.md` (+ launch line) | |

## Housekeeping
- Nothing was removed or moved. New dirs: box1 `/mnt/data/eagle4_scratch/att_fastclk_build_20261117/` (the stage plus `.repro` and `.repro_fastclk` scratch, about 62 MB; `/mnt/data` has 1.6 GB free), and buildx02 `/shared/build/lutg_d9_tp4stk_EF_stage` and `/shared/build/s3LAtp4EF`.
- A box1-side ssh client from the launch call is still attached to the remote `bash -c` wrapper (pid 3631590). That client is my own process. The pipeline itself is `setsid` + `nohup` (pid 3631592) and does not depend on it.

## 8. Closure-recipe sibling 0xE4B7137F AFC8X (LAUNCHED buildx01 2026-11-17T21:46:05Z) [M]
Ordered by box1-two-v2 to cut closure risk on EF. The OOC margin was +0.118 / +0.010.

**Stage**
- The RTL and the MD5LIST are identical to EF: MD5LIST bb89a4ee, config 38e7a756, drafter 901c191a, bridge 0395872d, locality XDC 7c1caf10.
- It was produced by re-running `derive_lutg_tp4_nfa8_fastclk.py` with id 0x137F / 7F. All proofs re-passed: 772/772 vs CE, PROOF-REPRO 5/5, 149/149 fork files.
- EF vs 7F after mapping the id tokens: only 2 comment lines differ (the id inside a comment). Then `scripts/derive_afc8_recipe_sibling.py` (e5044cec) changed the recipe only.
- Stage: buildx01 `/shared/build/lutg_d9_tp4stk_7F_stage`; box1 copy `/mnt/data/eagle4_scratch/att_fastclk_build_20261117/lutg_d9_tp4stk_7F_stage`.
- Checks: MD5LIST 392/392 and COMMON_MD5 OK on host; must_pass `--stage` PASS.

**Recipe diff vs EF** (`att_fastclk_20261117/build_7F/EF_TO_7F_RECIPE.diff`, 5abec301)
| knob | EF 0x13EF | AFC8X 0x137F | why |
|---|---|---|---|
| place_design | `-subdirective {Floorplan.ForceSpreading.high GPlace.ForceSpreading.high GPlace.ReduceCongestion.high DPlace.ReducePinDensity.high}` | **`-directive Explore`** + the same subdirectives | **Lineage evidence:** BC-ETO re-placed 0x13BC's own opt dcp with exactly this delta and reached FINAL_TRUE **+0.050 / +0.011**, vs the 0x13BC pipeline's +0.040 / +0.010 (`G6_REISSUE_BC_ADJUDICATION_20261111.md:40-42`). The 0x13BC loss was congestion, so the spreading set stays. `ExtraTimingOpt` is not a Versal 2025.1 place directive (4-2308 revert). **Project mode:** DIRECTIVE Explore and SUBDIRECTIVE are accepted together (buildx01 `/shared/scratch/afc8x_dirtest_20261117/t.log`, rc 0). |
| STEP3 "place DIRECTIVE set" check | `gate_fail` | report-only row `AFC8X-place-directive` (wants `DIRECTIVE Explore` x1) | |
| in-build setup margin, clock_reset_clk_usr_1 (132 MHz) | 0.040 | **0.080** | The fast clock is the thin domain. `cancel_hook_d1.tcl` (unchanged) zeroes it before the final report and write_device_image, so FINAL_TRUE is unmargined. All other clocks stay at 0.040 / 0.020. Margin XDC 3c5ebe84, md5 re-pinned. |
| route / phys_opt / threads / floorplan | AggressiveExplore / AggressiveExplore / 8 / sllfix XDC | unchanged | |

**Pipeline file md5s:** pipeline 3eb67f1b, run ffe6be54.

**Host buildx01 at launch**
- Memory: 251 GB total, 137 GB available, 61 GB free (`free -g`).
- Builds: `host_build_count_v3` reports 0. The 3 Vivados on the host are srpol optdiff census jobs, not builds.
- `/shared` has 107 GB free; SMBus smoke `IS_LOCKED 0`; tunnel key 1.
- AVED base is 4e3996d4 / 5c1958e1 / 6d8071eb, the same as buildx03 (the 0x13CE host). The 24 repo paths the pipeline uses are equal to buildx02's.
- STEP1 PACKAGE was admitted at 21:46:07Z with Vivado alive.

**Build id**
- 0xE4B7137F: 0 label hits, 0 live reg100 hits (reg100 0xE4B7117F has 7 retired-only hits).
- 0x13B1-0x13FF had no clean byte left after EF, so this id is the top of box1-two-v2's other reserved block 0x1360-0x137F.
- **Caveat:** the 11-12 tracker split (`4C1_DELIVERY_TRACKER.md:1664`, `c4_regmap.py:162`) names 0x1370-0x137F as box1-qwen-junior's range. 137F was unused. It is now in the reserved list, the tracker and this report, so `qg_id_check` will see it as a label hit.
- Other unused candidates were 0x1370, 0x137A, 0x137B and 0x137E (same range).

**Watcher**
- `scripts/att_fastclk_20261117/watch_afc8x.sh` runs on box1 as pid 416492 since 21:46:36Z. STATUS file: `/var/tmp/box1two_fastclk_20261117/watch_7F/STATUS`.
- It is `watch_afc8.sh` plus: the AFC8X / kvc2-margin gate rows; the impl `runme.log` place command; the 4-2302 Explore / 4-2301 / 4-2308 counts; and `^KVC2_MARGIN_NONVACUITY`.
- First tick: STEP1 admitted, 1 FAIL (the inherited selftest SKIP).

## 9. Verdict on the two EF-only FAIL rows (2026-11-18 ~00:35Z): both are check misfires, not defects in the image [M]
0x13EF has 13 FAIL rows in GATE_VERDICTS, and so does 0x137F. 11 of them are inherited and appear identically on 0x13CE. The two below are new on EF / 7F, and **both are mis-keyed checks**. Neither build was touched, and no fixed sibling is needed for these two rows.

**(1) `ATTAFC-staged`: drafter b82eaad0, but the row wants 901c191a.** This is my row, and it was mis-authored.
- I pinned the md5 of the **fork** drafter. The row reads the **staged-IP** copy, and the pipeline stamps one line into that copy before STEP3 (PROBE_OVERRIDE).
- EF: `diff <stage fork> <iprepo src>/eagle4_drafter.sv` gives exactly one changed line, `2408c2408`: `PROBE_REV = 8'h15` becomes `8'hEF;  // ONEFLIP-P2S head (pipeline-staged; fork source stays 8'h15)`. Reverting that one line in the staged file gives **901c191a**, the fork md5, byte for byte.
- 7F: the staged drafter is 8db74034. Reverting the `8'h7F` stamp also gives **901c191a**.
- 0x13CE gets the same stamp: one changed line (`2398c2398`), staged 091e8d23 vs fork 92ab14a2. Its NFA8-staged row never checked the drafter, so it stayed silent.
- Every other term of the row PASSES on both images: `ATT_ARRAY_FAST_CLK_V 1'b1` x1 (1'b0 x0), `ATT_FASTCLK_RED_V 0` x1, `ARRAY_FAST_CLK_V 1'b0` x1, bridge 0395872d.
- The staged IP still carries the fork content. EF iprepo `src/eagle4_drafter.sv` has:
  - `:1245` `localparam bit ATT_AFC = `EAGLE4_ATT_ARRAY_FAST_CLK;`
  - `:7687` `... u_a8q_s (`
  - `:7706` `.FAST_TP(DC_FTP), .FAST_RST_SYNC(ATT_AFC),`
  - `:7709` `) u_dc_bridge (`
  - `:7865` `.RST_PIPE(AFC ? 2 : 0), .LCACHE(AFC), ...`
- In `src/eagle4_config.svh`, `:457` reads `ATT_ARRAY_FAST_CLK_V 1'b1`.
- Staged md5s on **both** EF and 7F: array 640a4530, bridge 0395872d, lutg_a8q 31d6dfdd, premlp_normfuse ddbc0795 (all equal to the fork / PROOF-REPRO outputs), and staged config 288ff72d.

**(2) `STEP3: staged gemv_stage_array_scaled_slr.sv lacks the g_lutg / g_a8q arms`.** This is an inherited text check that the fast-clock patch is designed to break.
- The check is `uni_pipeline_lutg_d9.sh:2321`: `grep -q 'if (LUTG) begin : g_lutg' && grep -q 'if (A8Q) begin : g_a8q'`.
- `patch_att_fastclk.py` renames the in-array quantiser guard to `A8Q_INT = A8Q && !LCACHE` and adds the external-quantiser arm. In the staged file, `:2437` reads `if (A8Q_INT) begin : g_a8q` and `:2478` reads `end else if (A8Q_EXT) begin : g_a8q_ext`.
- The `g_lutg` half matches (`:1728` `if (LUTG) begin : g_lutg`); only the literal `if (A8Q)` no longer exists. 0x13CE's array has it once.
- With LCACHE the in-array quantiser is meant to be ABSENT; the quantiser runs at 66 MHz in front of the bridge as `g_dc.g_qs.u_a8q_s`. That is the design, proven by `F1_L100` identity (ATT_FASTCLK_20261117.md section 4).
- `gate_fail` in this pipeline records the row and returns 0 (`:322-326`), so neither build stopped. EF is in STEP4 build_all since the 22:28:32Z retry; 7F since 23:23:24Z.

**Why both builds took one STEP4b retry.** The BD reported ap_clk_hbm 285,999,744 against the seed 285,999,740. ap_clk 65,999,940 and the fast clock 131,999,881 matched their seeds.
- The resolved clk_wiz on EF (`top_usr_clk_wiz_0.xci`) is PRIM_IN 33.333, M 205.921875, D 2, **O1 52 / O2 26 / O3 12**, request {66,132,286}. That is the predicted solution.
- 0x13CE has the same M / D / O1 / O3, with only O2 13. So clk_out1 and clk_out3 are physically identical to 0x13CE; the 4 Hz difference is only in the BD's FREQ_HZ annotation.

**Deciding evidence still to come: the `afc8_census.tcl` rows on the opt dcp.**
- The census will report bridge and slow-quantiser primitives, LCACHE RAMB per group, in-array `g_a8q` primitives (want 0), sample-FD clocks (lanes on clock_reset_clk_usr_1), and the clk_usr_0/1/2 periods.
- Landmark for timing, measured on 0x13CE (buildx03): STEP4 retry 19:53:11Z, synth_1 dcp 23:02:43Z, and impl in the opt post-hooks at 00:35Z.
- [P] Applying the same intervals: EF opt dcp ≈ **03:10Z**, 7F ≈ **04:05Z** (2026-11-18).
- The watchers launch the census automatically once place starts with ≥ 60 GB free.

**Fix for future siblings (not applied to the running builds):**
- Mask the PROBE_REV line before comparing the drafter md5, or grep the five fast-clock signatures above.
- Re-key `:2321` to `if (A8Q_INT) begin : g_a8q` + `begin : g_a8q_ext`.
- At adjudication these two rows are cleared by this section, together with the census.
