Skip to content

feat(rtl): ICA-W40-001 Lane Q2 — OP_NULL_PE→8'hEA, OP_SPEC_EXIT→8'hEB - #151

Merged
gHashTag merged 1 commit into
mainfrom
feat/ica-w40-001-rtl-q2
May 15, 2026
Merged

gHashTag merged 1 commit into
mainfrom
feat/ica-w40-001-rtl-q2

Conversation

@gHashTag

Copy link
Copy Markdown
Collaborator

ICA-W40-001 Lane Q2 — RTL Opcode Rectification

Companion to t27#671 (Lane Q1 Coq, merged @ 2b340a9b).

Collisions Fixed on Master

File Old New
rtl/nullor/nullor_pe.sv OP_NULL_PE = 8'hE6 OP_NULL_PE = 8'hEA
rtl/nullor/tb/test_nullor_pe.sv opcode = 8'hE6 opcode = 8'hEA
rtl/spec_exit/spec_exit_pipeline.sv OP_SPEC_EXIT = 8'hE7 OP_SPEC_EXIT = 8'hEB
rtl/spec_exit/tb/test_spec_exit.sv opcode = 8'hE7 opcode = 8'hEB

Verified Collisions (pre-fix)

rtl/nullor/nullor_pe.sv       OP_NULL_PE     = 8'hE6
rtl/holo_mux_1x4.sv           OP_HOLO_MUX_X4 = 8'hE6   ← collision

rtl/spec_exit/spec_exit_pipeline.sv  OP_SPEC_EXIT = 8'hE7
rtl/dfs_controller.sv                OP_DFS_GATE  = 8'hE7  ← collision

Canonical Opcode Map (post-Q2)

0xE5 OP_SUBTH_CLK     (W37)
0xE6 OP_HOLO_MUX_X4   (W39 H) — KEEP
0xE7 OP_DFS_GATE      (W40)   — KEEP
0xE8 OP_SPARSE_SKIP   (W41)   — KEEP
0xE9 OP_STOCH_ROUND   (W42)   — KEEP
0xEA OP_NULL_PE       (W38)   — RELOCATED ✅ (this PR)
0xEB OP_SPEC_EXIT     (W39 E) — RELOCATED ✅ (this PR)

Constitutional Compliance

Rule Status
R-SI-1 (0 * cells in RTL) ✅ Only localparams + testbench opcodes changed
R5 HONEST ✅ Provenance comments updated with ICA-W40-001
R7 Falsification ✅ Testbenches retained, only drive value changed
R15 SACRED-SYNTH-GATE ✅ Sacred opcodes 0xD0..0xE0 untouched
R18 LAYER-FROZEN ✅ 75 Sacred ROM cells preserved

Companion Lanes

  • Q1 Coq ✅ MERGED — t27#671 @ 2b340a9b
  • Q3 Rust ⏳ PENDING — tt-trinity-max-true nullor-witness + spec-exit-witness
  • Q4 JSON ⏳ PENDING — trios assertions/nullor_witness.json + assertions/spec_exit_witness.json

Anchor

φ² + φ⁻² = 3 · γ = φ⁻³ · C = φ⁻¹ · G = π³γ²/φ
DOI 10.5281/zenodo.19227877 · NEVER STOP

Closes #148

Signed-off-by: Vasilev Dmitrii admin@t27.ai
ORCID 0009-0008-4294-6159

… 8'hE7→8'hEB

Companion to t27#671 (Lane Q1 Coq, merged @ 2b340a9b).

Resolves hard opcode collisions on master:
- rtl/nullor/nullor_pe.sv localparam OP_NULL_PE 8'hE6 → 8'hEA
- rtl/spec_exit/spec_exit_pipeline.sv localparam OP_SPEC_EXIT 8'hE7 → 8'hEB
- Testbenches updated to drive new opcodes
- Header comments updated with ICA-W40-001 rationale

Canonical opcode map post-rectification:
  0xE6 OP_HOLO_MUX_X4 (W39 H) — KEEP
  0xE7 OP_DFS_GATE    (W40)   — KEEP
  0xE8 OP_SPARSE_SKIP (W41)   — KEEP
  0xE9 OP_STOCH_ROUND (W42)   — KEEP
  0xEA OP_NULL_PE     (W38)   — RELOCATED from 0xE6
  0xEB OP_SPEC_EXIT   (W39 E) — RELOCATED from 0xE7

Constitutional: R-SI-1 (0  cells), R5 HONEST, R7 falsification, R15 SACRED-SYNTH-GATE preserved.
Anchor: φ² + φ⁻² = 3 · DOI 10.5281/zenodo.19227877

Closes #148

Signed-off-by: Vasilev Dmitrii <admin@t27.ai>
@gHashTag
gHashTag merged commit 4cab928 into main May 15, 2026
21 of 35 checks passed
@gHashTag
gHashTag deleted the feat/ica-w40-001-rtl-q2 branch May 15, 2026 23:09
gHashTag added a commit that referenced this pull request Aug 17, 2026
Two source files differing only in BUFR_DIVIDE, built with the stock
toolchain and with the patched one:

           stock 05aaa06        patched (PR #151)
  BYPASS   BUFR_DIVIDE.BYPASS   BUFR_DIVIDE.BYPASS
  D5       BUFR_DIVIDE.BYPASS   BUFR_DIVIDE.D5

If the stock column is identical for two different designs, the bug is
real; if the patched column differs only where it should, the fix is
right and carries no regression. Both are asserted, not eyeballed.

The probe clocks an ODDR from the BUFR output. A fabric flip-flop would
be reached through a BUFG and would never touch the regional clock; the
ODDR is an OLOGIC flop inside the IOI tile, so the clock must arrive
over RCLK and cross the pseudo-pip that carries the configuration --
which is the thing under test.

If no BUFR feature appears at all, the run fails saying the probe is
wrong rather than the claim. A test that cannot distinguish 'refuted'
from 'never ran' is not a test.

Part xc7a35tcsg324-1 and pins from prjxray's own arty-a7 harness, so it
matches the goldens in #150: a Vivado reference build needs no
re-pinning, and the same two files serve the BUFIO half of #149.

On silicon the output pin toggles at BUFR_O/2, so the builds are
distinguishable without a decoder: 50 MHz for BYPASS, 10 MHz for D5.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
gHashTag added a commit that referenced this pull request Aug 17, 2026
The guard grepped for populate_bufr_divide, the function name from the
first version of #151, which that PR replaced with write_bufr. It
checked out the right commit and rejected it:

    HEAD is now at 0b91457 xilinx: configure a placed BUFR...
    ##[error]the fork branch does not carry the patch

A check referring to something that no longer exists, failing exactly
like the thing it guards against -- the same shape of defect this probe
was built to catch in someone else's code. The message now says the
guard and the patch have drifted, rather than blaming the branch.

Header updated too: it claimed a placed BUFR crosses the pseudo-pip
carrying its configuration. It does not, which is the finding the first
run produced.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
gHashTag added a commit that referenced this pull request Aug 17, 2026
The probe's assertions tested the first version of #151: that stock
emits BUFR_DIVIDE.BYPASS where the design asks for 5. Measured, stock
emits NOTHING -- no IN_USE, no divide -- because the pseudo-pip table
that carries BUFR configuration only fires when the site is traversed as
routing, and a placed BUFR is not.

So the run reported 'claim not reproduced' while the fix was in fact
working: patched added exactly

    HCLK_IOI3_X113Y78.BUFR_Y1.IN_USE
    HCLK_IOI3_X113Y78.BUFR_Y1.BUFR_DIVIDE.D5

and removed nothing. Slot Y1 matches the RCLK index 3 route through the
{2,3,0,1} mapping, so the site-y derivation is right too.

The assertions now state the measured defect and check four things:
stock emits no BUFR feature at all; patched emits IN_USE and the
requested divide; a BYPASS design still gets BYPASS; and the patched
FASM removes nothing.

Changing a test so it passes deserves saying out loud: the hypothesis it
encoded was refuted by this probe, and the replacement asserts the
evidence. The old text is in the git history and in the header.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
gHashTag added a commit that referenced this pull request Aug 18, 2026
I told Carlos by mail that a BUFR fed from an input pin cannot produce a
valid bitstream without the I2IOCLK rows. That was an inference from
'the database has zero rows matching I2IOCLK', not a measurement, and it
is exactly the kind of claim this project keeps having to retract.

So measure it. fasm2frames runs on both FASMs and the step reports the
return code and stderr. It does not gate the run: whether the feature
makes the converter fail or gets silently dropped, both outcomes are
information, and neither says anything about whether #151 emits the
right features -- which is what the verdict step is for.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

🛠️ ICA-W40-001 — Sacred opcode 0xE6/0xE7/0xE8 collision rectification (W38..W41)

1 participant