lance-graph-xlog: one resident hypercube as XLOG's execution context (probe, D-XLOG-NATIVE-7)

.claude/board/STATUS_BOARD.md

@@ -71,6 +71,7 @@ Plan: `.claude/plans/d-soc-fence-v1.md`. Plan only; nothing runs yet.
 | D-XLOG-NATIVE-4 | create `crates/lance-graph-xlog` (shape per Amendment C/D: Quack + mask-risc + DT64, no `xlog-*` or arrow dep, no plan type) | Shipped (#1907) | N1, N4, N5 met; N2 scalar oracle in-crate + both oracles in the probe; N3 typed refusals |
 | D-XLOG-NATIVE-6 | probe executes on `ndarray::simd` (`src/native.rs`): support via `eq_u32_to_mask` / `mask_scatter_or_u32` / `mask_and`, bag count via `masked_group_count_u32` ×2 + `checked_binary_i64`; DR `LocalMap` up to 256; register path kept as differential (plan Amendment B) | Shipped (#1884) | lab probe; layering fix is D-DYNBAG-0 |
 | D-XLOG-NATIVE-5 | bag count bound through the production bind store (`lance-graph-dt64` + `dt64::dr`), polyfill words by position; Epi5 reading of XLOG results pinned (plan Amendment A) | Shipped (#1884) | lab execution only |
+| D-XLOG-NATIVE-7 | resident hypercube context (`lance_graph_xlog::cube`): one intake for every relation, `Copy` view handles, terminals only run programs; XLOG plans executed on it in the XLOG fork (`xlog-cube`) | In PR (#1911; xlog#2): parity + 300-case differential; 12/230 integration rules admitted | recursion, derived inputs |
 
 ## D-LITHO — the photolithography contract (2026-10-30)
 

.claude/board/entries/2026-11-01-xlog-resident-hypercube-context.md

@@ -0,0 +1,66 @@
+# D-XLOG-NATIVE-7: one resident hypercube as XLOG's execution context (probe)
+
+**Date 2026-11-01 · ID D-XLOG-NATIVE-7 · Status PROBE · Plan `.claude/plans/d-xlog-native-v1.md` (Amendment F, pending beside E in #1910) · Crate `crates/lance-graph-xlog` (`cube`)**
+
+DECISION (direction, from the session): the experiment replaces the machine
+under XLOG's CUDA-facing API with one resident context, rather than speeding up
+operators one relation at a time. Observable results of admitted classical
+XLOG queries must match; the internal model is free.
+
+What `lance_graph_xlog::cube` does (VERIFIED-IN-CODE, `src/cube.rs`):
+
+- `Cube::new` looks up every id of every relation, both columns, once in one
+  tile's `LocalMap` under a `MapTable` frame. Nothing is looked up again.
+- `scan` / `bind` / `join` return `View`s: `Copy` handles into the cube's node
+  list. Making a view runs no program and builds no row.
+- Terminals: `support` (two `ScatterOrU32` + `mask_and`), `bag_count` (two
+  `GroupReduce Count` + checked product), `materialize_rows` (the one named
+  place tuples are built). Each re-checks the frame.
+- Any relation joins any other on either column from the same lanes: lanes
+  are stored per relation, not per join orientation.
+
+TEST-PINNED (`tests/cube.rs`, 6 tests):
+
+- parity with the per-call `support` / `bag_count` under five bound shapes;
+- all 36 (relation pair × column pair) joins over three relations match an
+  enumeration oracle, and `LocalMap::lookups()` is unchanged after intake;
+- 50 composed joins run 0 programs; `support` then runs exactly 2, builds 0 rows;
+- a bound outside the map short-circuits its side (1 program, not 2);
+- 300-case randomised differential (ids 0..12, either orientation, optional
+  bounds): bag count, support and rows; 150+ non-empty;
+- stale frame, foreign view, bound on the join column, non-join terminal refused.
+
+Disable runs, all red: bound ignored in programs; support on the wrong column;
+support skips the frame check; `materialize_rows` ignores the bound.
+
+XLOG side (MEASURED by reading `xlog-runtime` / `xlog-cuda`):
+
+- `CudaBuffer` and `CudaKernelProvider` are concrete structs, not traits.
+  `CudaBuffer` fields are `pub(crate)`; callers use 16 public methods.
+- `xlog-runtime` calls 79 of the provider's 397 public functions; ~25 are
+  telemetry/resets, ~30 are WCOJ/clique variants.
+- 12 `provider.device()` sites: 1 in the core dispatcher (join-index cache
+  key), 2 `synchronize`, 9 in `epistemic_workspace` / `resident` (outside the
+  13 RIR kinds).
+- `xlog-cuda/build.rs` panics without `nvcc` unless `XLOG_RUSTDOC_NO_CUDA=1`
+  (empty kernel data). With it, `cargo check -p xlog-runtime` finishes on a
+  host with no CUDA toolkit (MEASURED 2026-11-01, 3 warnings, 0 errors), so a
+  device-less build of the XLOG fork is reachable.
+
+Comparison with Amendment E (caller-level polyfill):
+
+- E keeps XLOG's API and moves its callers onto `lance-graph-xlog` one
+  operator at a time; every call still owns its inputs.
+- F gives the CUDA-facing handles a shared context: a `CudaBuffer` would be a
+  `View` plus the cube it belongs to, so a join between two XLOG buffers needs
+  no new lanes, no index (XLOG's `JoinIndexKey` cache becomes redundant) and no
+  device.
+- Neither is promoted. F is better for the 13 RIR kinds only if the dispatcher
+  can keep buffers as views between nodes; a node that downloads rows ends the
+  advantage at that node.
+
+**OPEN:** recursion (semi-naive delta over versioned masks), union and
+difference as views; more than one tile; n-ary relations; a `CudaBuffer::View` variant and a
+device-less provider in the XLOG fork; per-node cost vs the per-call path.
+Reasoning surfaces (NARS revision, thought layers, CausalEdge64) are not in this
+probe and need their own contracts.

.claude/board/entries/README.md

@@ -25,10 +25,11 @@ index row, (3) no duplicate entry id. Checks 1 and 2 are deliberately
 opposite directions; the stranding this convention prevents shows up in
 exactly one of them, never both.
 
-539 entries, 2026-08-06 .. 2026-11-01.
+540 entries, 2026-08-06 .. 2026-11-01.
 
 | date | entry id | finding | file |
 |---|---|---|---|
+| 2026-11-01 | `D-XLOG-NATIVE-7` |  | [2026-11-01-xlog-resident-hypercube-context.md](2026-11-01-xlog-resident-hypercube-context.md) |
 | 2026-11-01 | `D-DT64-7-5` |  | [2026-11-01-maptable-versioned-local-map.md](2026-11-01-maptable-versioned-local-map.md) |
 | 2026-11-01 | `D-XLOG-NATIVE-4` |  | [2026-11-01-lance-graph-xlog-crate.md](2026-11-01-lance-graph-xlog-crate.md) |
 | 2026-11-01 | `D-DYNBAG-0` |  | [2026-11-01-dynbag-0-xlog-through-quack-mask-risc.md](2026-11-01-dynbag-0-xlog-through-quack-mask-risc.md) |