Appendix — Report → Wiki-Page Cross-Reference Map
This appendix is a provenance index over
neuronx_cc2.24.5133.0+58f8de22 wiki artifacts, not a binary-RE page. It is generated by scanning two corpora that live in this repository: the raw evidence reports underneuronx-cc/raw/(theD-*.txtdeep-dive notes) and the published wiki pages underneuronx-cc/wiki/src/. Every edge below is a realrgcitation found in a committed page — no mapping is hand-asserted. Because it is repo-derived, this page has no version-pinned addresses of its own; its accuracy is exactly the accuracy of the citations the authors wrote, and it goes stale the moment a page adds, drops, or renames aD-ID. Regenerate it (see §5) after any wiki edit that touches an evidence anchor.
Abstract
Each normative wiki page anchors its claims to one or more deep-dive reports — the D-<family><n>.txt notes under raw/ that record the binary-analysis evidence a page is built on (a disassembly trace, an enum dump, a struct layout, a verifier walk). A page cites a report by writing its D-ID inline in a BACKING: line or an evidence callout. This appendix inverts and totals those citations so a maintainer can answer two questions in O(1):
- Forward (§2): given a report
D-Xnn, which page(s) consume it? — so editing or correcting a report immediately shows every page that must be re-checked. - Reverse (§3): given a page, which reports back it? — so a reviewer can pull the exact evidence behind a page without re-grepping.
The index is also a completeness audit. It flags two failure modes that a provenance map exists to catch: a report cited by zero pages (an unconsumed evidence note — work captured but not yet written up), and a D-ID cited by a page that does not exist in raw/ (a dangling citation — a claim whose evidence is missing or mis-typed). The gap census is §4.
The map is the substance of this page; the prose is scaffolding. Its value is completeness and accuracy of the edges, not narrative. Every (report → page) row was produced by rg -l '\bD-ID\b' against wiki/src/, and every page slug named is a file that exists on disk.
| Report corpus | neuronx-cc/raw/D-*.txt — 506 files, 504 distinct D-IDs (D-P02/D-P02b and two D-Z06 variants share a stem) |
D-ID grammar | D-<FAMILY><NN> — family is 1–2 uppercase letters (A..Z, AA..AG), NN is a zero-padded ordinal; 33 families |
| Wiki corpus | neuronx-cc/wiki/src/**/*.md |
| Reports consumed | 235 / 504 distinct reports are cited by ≥ 1 page |
| Reports unconsumed | 269 / 504 — cited by zero pages (evidence captured, not yet written up) — §4.1 |
| Dangling citations | 0 — every D-ID cited by a page resolves to a real raw/ file — §4.2 |
| Pages citing ≥ 1 report | 170 distinct .md files |
| Method / confidence model | §0.3 Methodology & Confidence |
| Sibling provenance index | §14.4 Symbol / Offset Index |
0. How to read an edge
A single edge is a citation. When bir/dtype-tables.md contains the string D-D04, that is the page declaring "the dtype-table claims on this page are backed by the evidence in raw/D-D04-dtype-tables.txt." This appendix records that as the forward edge D-D04 → bir/dtype-tables.md and the reverse edge bir/dtype-tables.md ← D-D04.
Three properties of the citation graph a maintainer should know:
- Many-to-many. A report can back several pages (a struct-layout report like
D-E12accesspattern struct feeds four pages), and a page can rest on several reports (nki/bircodegen-collective.mdcites eleven). The map below lists every page for each report (forward) and every report for the highest-fan-in pages (reverse). - Word-boundary matching.
D-IDsare matched with\bD-[A-Z]+[0-9]+\b. This is deliberate: a naïverg 'D-SB2'would false-positive on the hardware termGPSIMD-SB2SBinwalrus/legality-dispatch.md. The\b…\banchoring rejects that substring, which is why the dangling count is exactly 0 rather than a spurious 1. Any future regeneration must keep the boundary anchors. - Stem vs file. Two reports were split across a base note and a continuation (
D-P02+D-P02b, plus twoD-Z06variants). They share oneD-IDstem, so the corpus is 506 files / 504 stems. Forward edges key on the stem.
QUIRK — a
D-IDappearing in a page is not proof the page is correct, only that it claims that report as evidence. This index audits linkage, not truth. Truth is the job of each page's own adversarial-verify section and of §0.3. A row here means "this page says report X backs it"; whether the page read X faithfully is out of scope.
1. Family map
Reports are grouped by the leading-letter family, which tracks the analysis theme the note belongs to. Knowing the family lets a maintainer guess a report's subject from its D-ID alone. The table gives each family's theme, total stem count, and how many of those stems are currently consumed by at least one page.
| Family | Theme (from report titles) | Stems | Consumed | Unconsumed |
|---|---|---|---|---|
D-A* | front-end / driver / pipeline / job registry | 13 | 9 | 4 |
D-AA* | conv & resize device kernels | 9 | 3 | 6 |
D-AB* | distribution / SPMD / sharding | 9 | 5 | 4 |
D-AC* | custom-op packaging / GPSIMD memory model | 7 | 6 | 1 |
D-AD* | perf-sim / metric-store / profiling | 6 | 3 | 3 |
D-AE* | HBM / DRAM geometry | 4 | 1 | 3 |
D-AF* | CompileCommand flag catalog | 5 | 3 | 2 |
D-AG* | PWP / activation profile / error binding | 10 | 3 | 7 |
D-B* | HLO legalize / collective combiners / layout | 32 | 16 | 16 |
D-C* | op fusion / tensorizer / MLIR pipeline | 24 | 12 | 12 |
D-D* | enums / dtype / memory taxonomy | 14 | 13 | 1 |
D-E* | valid-engines / JSON / structs / legality | 22 | 12 | 10 |
D-F* | numeric simulators (matmul/act/MX/collective) | 18 | 6 | 12 |
D-G* | BirVerifier / barrier-check / per-op verify | 14 | 9 | 5 |
D-H* | NKI translate / allocator / local collectives | 40 | 14 | 26 |
D-I* | KLIR / BIR codegen dispatch | 28 | 9 | 19 |
D-J* | per-gen ISA encoders (CoreV2/V3/V4) | 36 | 12 | 24 |
D-K* | allocator drivers / loop-opt / cost stats | 20 | 7 | 13 |
D-L* | layout autotuner / reward loop | 10 | 2 | 8 |
D-M* | engine-info / PWP / per-engine ISA | 16 | 10 | 6 |
D-N* | tensor descriptors / indirect addressing | 12 | 2 | 10 |
D-O* | NKI production kernels (MoE/attn/norm) | 30 | 24 | 6 |
D-P* | NeuronCodegen / BIR-codegen / IR builder | 23 | 14 | 8 |
D-Q* | custom-op wire / GPSIMD builtins | 10 | 6 | 4 |
D-R* | Pelican / DAP debugger / BIR hierarchy | 10 | 3 | 7 |
D-S* | NEFF / BIR-JSON serialization | 13 | 5 | 8 |
D-T* | PWP bucket / control blob | 8 | 2 | 6 |
D-U* | Penguin pass roster / tensorizer / IR model | 8 | 3 | 5 |
D-V* | per-engine ISA encode (matmul/pool/MX) | 16 | 4 | 12 |
D-W* | NKI ISA / sema / type system / bindings | 12 | 7 | 5 |
D-X* | dtype promotion / cast / numeric results | 11 | 3 | 8 |
D-Y* | ISL dependence / Tonga simplifier | 8 | 3 | 5 |
D-Z* | dynamic DMA / dynamic-shape synthesis | 8 | 4 | 3 |
| Total | 33 families | 504 | 235 | 269 |
2. Forward map — report → consuming page(s)
The forward index, grouped by family. Each row is report → the page(s) that cite it. Only consumed reports appear here; the 269 unconsumed reports are listed in §4.1. Page slugs are relative to wiki/src/.
D-A* family
| Report | Title | Consuming page(s) |
|---|---|---|
D-A01 | hlo2penguin main | frontend/hlo2penguin-entry.md hlo-opt/hlo-ingestion-boundary.md hlo-opt/neuron-dialect-registry.md |
D-A03 | hlo opt pass registry | hlo-opt/allreduce-dynslice-rewrites.md hlo-opt/pass-registry.md hlo-opt/topk-legalize.md |
D-A05 | compile pipeline | frontend/opt-level-planes.md |
D-A06 | job registry | frontend/job-registry.md frontend/static-io-transpose.md |
D-A09 | client daemon ipc | frontend/client-daemon-ipc.md |
D-A10 | private tools | frontend/job-registry.md |
D-A11 | walrus driver cli | frontend/walrus-driver-cli.md walrus/local-collectives.md |
D-A12 | io transpose schema | frontend/static-io-transpose.md |
D-A13 | error catalog | appendix/error-message-catalog.md frontend/diagnostic-error-catalog.md nki/spmd-programming-model.md |
D-AA* family
| Report | Title | Consuming page(s) |
|---|---|---|
D-AA04 | conv device lowering selection | nki/conv-device-lowering.md |
D-AA05 | conv2d pbp kernels | nki/conv2d-pbp.md |
D-AA07 | resize kernels | nki/resize-kernels.md |
D-AB* family
| Report | Title | Consuming page(s) |
|---|---|---|
D-AB01 | spmd partitioner driver | distribution/distribution-strategy-seeding.md distribution/spmd-partitioner-driver.md |
D-AB05 | sharding algebra | distribution/sharding-algebra.md |
D-AB06 | shardy hlosharding bridge | distribution/distribution-strategy-seeding.md distribution/shardy-hlosharding-bridge.md |
D-AB07 | mesh replica group math | distribution/mesh-replica-group-math.md distribution/three-d-rank-model.md |
D-AB08 | lnc sharding constraint | distribution/distribution-strategy-seeding.md distribution/lnc-sharding-constraint.md frontend/flag-catalog.md |
D-AC* family
| Report | Title | Consuming page(s) |
|---|---|---|
D-AC01 | ucode lib packaging | custom-ops/findcustomopdata-staging.md |
D-AC02 | findcustomopdata staging | custom-ops/findcustomopdata-staging.md |
D-AC03 | custom op wire validators | custom-ops/customop-wire-validators.md |
D-AC05 | aten c10 dispatch surface | custom-ops/aten-c10-surface.md |
D-AC06 | percpu link merge topology | custom-ops/gpsimd-xtensa-layout.md |
D-AC07 | gpsimd memory model reconciliation | custom-ops/two-gpsimd-reconciliation.md |
D-AD* family
| Report | Title | Consuming page(s) |
|---|---|---|
D-AD01 | perf sim internals | walrus/perfsim-cost-model.md |
D-AD03 | metricstore | walrus/metricstore.md |
D-AD06 | nki profile ntff contract | nki/entrypoints.md |
D-AE* family
| Report | Title | Consuming page(s) |
|---|---|---|
D-AE03 | hbm dram geometry | arch/dram-hbm-geometry.md |
D-AF* family
| Report | Title | Consuming page(s) |
|---|---|---|
D-AF01 | compilecommand flags a–m | distribution/distribution-strategy-seeding.md |
D-AF02 | compilecommand flags n–z | frontend/opt-level-planes.md |
D-AF05 | options defaults config | frontend/subtool-argv.md |
D-AG* family
| Report | Title | Consuming page(s) |
|---|---|---|
D-AG07 | pwp bkt ctrl blob | activation/bkt-ctrl-blob.md activation/loadactfuncset.md |
D-AG08 | activation profile json | activation/activation-profile-json.md |
D-AG09 | errormessages binding | appendix/error-message-catalog.md frontend/diagnostic-error-catalog.md |
D-B* family
| Report | Title | Consuming page(s) |
|---|---|---|
D-B01 | argmax legalize | hlo-opt/argmax-argmin-legalize.md |
D-B02 | softmax legalize | hlo-opt/softmax-legalize.md |
D-B03 | topk legalize | hlo-opt/topk-legalize.md |
D-B04 | mx legalize | appendix/master-enum-dtype-table.md bir/dtype-tables.md |
D-B08 | allreduce combiner | hlo-opt/collective-combiners.md |
D-B09 | reducescatter combiner | hlo-opt/collective-combiners.md |
D-B10 | allgather combiner | hlo-opt/collective-combiners.md |
D-B13 | ccops decompose | frontend/diagnostic-error-catalog.md hlo-opt/ccops-decompose-legalize.md |
D-B14 | collectives to cc | hlo-opt/hlo-ingestion-boundary.md |
D-B15 | stream channel id | hlo-opt/collective-stream-channel-id.md nki/bircodegen-collective.md |
D-B17 | whileloop collective motion | hlo-opt/whileloop-collective-codemotion.md |
D-B23 | dedup permute | hlo-opt/allreduce-dynslice-rewrites.md |
D-B24 | looped einsum | hlo-opt/looped-einsum-collective-matmul.md |
D-B28 | boundary markers | hlo-opt/boundary-markers-layer-cut.md |
D-B29 | control deps | hlo-opt/hlo-ingestion-boundary.md |
D-B31 | layout passes | hlo-opt/layout-passes.md |
D-C* family
| Report | Title | Consuming page(s) |
|---|---|---|
D-C01 | fuse dotlogistic | hlo-opt/op-fusion-dot-elementwise.md |
D-C02 | fuse elementwise | hlo-opt/op-fusion-dot-elementwise.md |
D-C04 | fuse transcendental | hlo-opt/op-fusion-dot-elementwise.md |
D-C07 | controldep tuple | hlo-opt/controldep-tuple-flatten-mlir.md |
D-C09 | canon tensorizer 2 | hlo-opt/canonicalize-for-tensorizer.md nki/resize-kernels.md |
D-C10 | tensorizer legalize | hlo-opt/tensorizer-legalization.md |
D-C11 | canonicalize conv | hlo-opt/conv-canonicalization.md |
D-C12 | schedule fusion | hlo-opt/schedule-fusion-composite.md |
D-C14 | pyprinter driver | hlo-opt/mhlo-to-python-printer-driver.md |
D-C15 | pyprinter elementwise | hlo-opt/mhlo-to-python-printer-driver.md |
D-C21 | dialects g2s | hlo-opt/neuron-dialect-registry.md |
D-C22 | mlir pipeline order | hlo-opt/hlo2penguin-mlir-pipeline.md |
D-D* family
| Report | Title | Consuming page(s) |
|---|---|---|
D-D01 | instructiontype | appendix/master-enum-dtype-table.md bir/codegen-dve-rng-control.md walrus/dma-engine-binding.md walrus/dma-queues.md |
D-D02 | aluop modes | arch/pool-engine.md bir/aluop-modes.md bir/codegen-dve-rng-control.md |
D-D03 | engine arch | walrus/dma-engine-binding.md walrus/engine-lowering-set.md walrus/parserdumper.md |
D-D04 | dtype tables | appendix/master-enum-dtype-table.md bir/cross-language-wirekeys.md bir/dtype-tables.md bir/value-model.md isa/neuron-isa-tpb-capstone.md nki/isa-compute-intrinsics.md numerics/mx-matmul-legality.md |
D-D05 | memory taxonomy | bir/memory-location.md nki/isa-compute-intrinsics.md walrus/allocator-drivers.md walrus/parserdumper.md walrus/peephole-constprop-remat.md walrus/vnsplitter-shrink.md |
D-D07 | collective enums | bir/op-family-enums.md nki/bircodegen-collective.md walrus/local-collectives.md |
D-D08 | matmul branch enums | bir/codegen-dve-rng-control.md |
D-D09 | dma enums | appendix/master-enum-dtype-table.md bir/op-family-enums.md walrus/dma-engine-binding.md walrus/parserdumper.md |
D-D10 | rng sema enums | appendix/master-enum-dtype-table.md bir/op-family-enums.md |
D-D11 | activation enums | activation/act-function-catalog.md appendix/master-enum-dtype-table.md |
D-D12 | attr arg enums | bir/value-model.md formats/neff-feature-flags.md walrus/parserdumper.md |
D-D13 | remaining enums | appendix/master-enum-dtype-table.md bir/op-family-enums.md nki/neuroncodegen-control.md |
D-D14 | enum crosswalk | appendix/master-enum-dtype-table.md bir/op-family-enums.md |
D-E* family
| Report | Title | Consuming page(s) |
|---|---|---|
D-E02 | validengines activation | activation/act-function-catalog.md |
D-E03 | validengines dma sync | walrus/legality-dispatch.md |
D-E06 | json activation reduce | walrus/parserdumper.md |
D-E07 | json dma | bir/sim-elementwise-datamove.md |
D-E09 | json collective bn rng | nki/bircodegen-collective.md |
D-E12 | accesspattern struct | bir/value-model.md walrus/engine-lowering-set.md walrus/racecheck.md walrus/vnsplitter-shrink.md |
D-E13 | memorylocation struct | bir/memory-location.md walrus/memreserve-kernel-inline.md walrus/reorder-nonssa-presched.md walrus/sbuf-liveness-interference.md |
D-E14 | argument register struct | bir/codegen-dve-rng-control.md bir/codegen-sub-encoders.md bir/value-model.md |
D-E16 | loopaxis dmaqueue sync | walrus/dma-queues.md |
D-E19 | dependency sync model | bir/codegen-sub-encoders.md walrus/dma-queues.md walrus/engine-lowering-set.md walrus/reorder-nonssa-presched.md |
D-E21 | legality dispatch | numerics/mx-matmul-legality.md |
D-E22 | libbir verifiers | walrus/local-collectives.md |
D-F* family
| Report | Title | Consuming page(s) |
|---|---|---|
D-F02 | sim matmul | numerics/mx-matmul-legality.md numerics/mx-microscaling.md |
D-F03 | sim activation pwp | activation/pwp-model.md |
D-F04 | sim arithmetic | numerics/numeric-negative-results.md |
D-F08 | sim collective | bir/codegen-dve-rng-control.md distribution/three-d-rank-model.md nki/bircodegen-collective.md |
D-F13 | sim mx quantize | numerics/mx-matmul-legality.md numerics/mx-microscaling.md numerics/numeric-negative-results.md |
D-F18 | pwp sim | activation/pwp-model.md |
D-G* family
| Report | Title | Consuming page(s) |
|---|---|---|
D-G01 | birverifier structure | walrus/legality-dispatch.md |
D-G02 | verify compute | walrus/birverifier-per-op.md |
D-G03 | verify dma sync | walrus/dma-queues.md |
D-G04 | verify collective control | walrus/birverifier-per-op.md |
D-G05 | mx matmul verify | numerics/mx-matmul-legality.md walrus/birverifier-per-op.md |
D-G08 | barriercheck ranger | walrus/barriercheck.md |
D-G10 | parserdumper | walrus/parserdumper.md |
D-G11 | lnc verifier | walrus/barriercheck.md walrus/local-collectives.md |
D-G14 | perfsim | walrus/perfsim-cost-model.md |
D-H* family
| Report | Title | Consuming page(s) |
|---|---|---|
D-H01 | translate nki driver | walrus/memreserve-kernel-inline.md |
D-H03 | lnc splitter | walrus/bir-linker.md walrus/local-collectives.md |
D-H06 | dma legalize | walrus/local-collectives.md |
D-H10 | psum ac | numerics/mixed-precision-accumulation.md |
D-H13 | pre sched | bir/codegen-sub-encoders.md |
D-H16 | memreserve inline | nki/three-sink-kernel-model.md walrus/memreserve-kernel-inline.md |
D-H17 | allocator wiring | penguin/dge-level-dynamic-dma.md |
D-H21 | deps | bir/codegen-sub-encoders.md |
D-H23 | local collectives | nki/bircodegen-collective.md walrus/bir-linker.md walrus/local-collectives.md |
D-H24 | shared lifetimes | walrus/bir-linker.md walrus/shared-lifetimes.md |
D-H26 | perfsim wiring | walrus/perf-sim-wiring.md |
D-H35 | lower control | bir/codegen-dve-rng-control.md |
D-H38 | engine lowering | activation/set-cover.md |
D-H40 | late passes | walrus/vnc-cross-core-link.md |
D-I* family
| Report | Title | Consuming page(s) |
|---|---|---|
D-I01 | klir dispatch | bir/codegen-sub-encoders.md |
D-I09 | codegen collective | nki/bircodegen-collective.md |
D-I10 | codegen permute sendrecv | nki/bircodegen-collective.md |
D-I11 | codegen barrier rank | bir/codegen-dve-rng-control.md distribution/three-d-rank-model.md nki/bircodegen-collective.md |
D-I12 | codegen maxindex | nki/index-gather-broadcast.md |
D-I17 | codegen control register | bir/codegen-sub-encoders.md |
D-I21 | codegen operand | bir/codegen-sub-encoders.md |
D-I23 | codegen assembly | bir/codegen-sub-encoders.md |
D-I28 | nkiklir kernel node | nki/three-sink-kernel-model.md |
D-J* family
| Report | Title | Consuming page(s) |
|---|---|---|
D-J03 | cv2 pool reduce | isa/neuron-isa-tpb-capstone.md |
D-J04 | cv2 tensorscalar | isa/tensorscalar-encoding.md |
D-J05 | cv2 tensortensor fill | isa/neuron-isa-tpb-capstone.md |
D-J06 | cv2 copy shuffle | isa/neuron-isa-tpb-capstone.md |
D-J09 | cv2 bn | isa/batchnorm-encoding.md |
D-J13 | cv2 rng activation | activation/loadactfuncset.md |
D-J14 | cv2 misc | arch/sp-engine.md custom-ops/customop-codegen.md |
D-J20 | cv4 quantizemx | numerics/numeric-negative-results.md |
D-J28 | codegenmode | walrus/codegen-driver.md |
D-J29 | codegen driver | walrus/codegen-driver.md walrus/legality-dispatch.md |
D-J32 | opcode master | isa/neuron-isa-tpb-capstone.md |
D-J34 | neff filewriter | formats/neff-json-sidecars.md |
D-K* family
| Report | Title | Consuming page(s) |
|---|---|---|
D-K01 | allocator driver | walrus/allocator-drivers.md walrus/pass-pipeline-optlevels.md |
D-K10 | psum allocator | arch/sbuf-psum-geometry.md |
D-K12 | full unroll | walrus/translate-nki-unroll.md |
D-K14 | loopopt transforms | penguin/backend-dependence-distance.md |
D-K15 | sb size legalize | walrus/sb-size-legalization.md |
D-K19 | optlevel7 pipeline | frontend/opt-level-planes.md |
D-K20 | cost stats | walrus/metricstore.md |
D-L* family
| Report | Title | Consuming page(s) |
|---|---|---|
D-L02 | layout autotuner | walrus/autotuner-orchestration.md |
D-L10 | reward loop | walrus/autotuner-orchestration.md |
D-M* family
| Report | Title | Consuming page(s) |
|---|---|---|
D-M04 | act info | activation/act-function-catalog.md |
D-M05 | pwp profiles | activation/activation-profile-json.md activation/pwp-model.md |
D-M06 | pwp bins | activation/bkt-ctrl-blob.md |
D-M07 | loadactfuncset | activation/loadactfuncset.md |
D-M08 | pe isa | front/worked-example-matmul.md isa/neuron-isa-tpb-capstone.md |
D-M10 | sp isa | isa/neuron-isa-tpb-capstone.md |
D-M11 | gpsimd isa | custom-ops/two-gpsimd-reconciliation.md walrus/local-collectives.md |
D-M12 | engineinfo | arch/dram-hbm-geometry.md arch/sbuf-psum-geometry.md |
D-M15 | vestigial gens | arch/vestigial-generations.md |
D-M16 | withln target | activation/withln-vs-trainium.md |
D-N* family
| Report | Title | Consuming page(s) |
|---|---|---|
D-N02 | tensor123 | isa/indirect-descriptors.md |
D-N05 | indirect | isa/indirect-descriptors.md |
D-O* family
| Report | Title | Consuming page(s) |
|---|---|---|
D-O01 | moe cte dispatch | nki/moe-cte-prefill.md |
D-O02 | bwmm shard block | nki/moe-cte-prefill.md |
D-O03 | bwmm shard i | nki/moe-cte-prefill.md |
D-O04 | bwmm shard block mx | nki/moe-cte-prefill.md |
D-O05 | bwmm shard i mx | nki/moe-cte-prefill.md |
D-O06 | moe block tkg | nki/moe-decode-tkg.md |
D-O08 | moe tkg mx proj | nki/moe-decode-tkg.md |
D-O09 | affinity masking | nki/moe-decode-tkg.md |
D-O10 | router topk | nki/router-topk.md |
D-O11 | rmsnorm | nki/normalization-kernels.md |
D-O12 | layernorm | nki/normalization-kernels.md |
D-O14 | attention tkg | nki/attention-tkg.md |
D-O15 | attention bwd | nki/attention-bwd.md |
D-O16 | gen mask | nki/attention-tkg.md |
D-O18 | mlp | nki/dense-mlp.md |
D-O22 | index broadcast | nki/index-gather-broadcast.md |
D-O23 | nki allocator tiling | nki/nkilib-infrastructure.md |
D-O24 | common types | nki/nkilib-infrastructure.md |
D-O25 | moe experimental | nki/experimental-kernels.md |
D-O26 | fg collectives | nki/bircodegen-collective.md nki/experimental-kernels.md |
D-O27 | conv loss | nki/experimental-kernels.md |
D-O28 | transformer block | nki/experimental-kernels.md |
D-O29 | mxfp load | nki/moe-cte-prefill.md nki/mxfp-weight-scale-load.md |
D-O30 | prod kernels | nki/internal-kernel-registry.md nki/production-kernel-inventory.md |
D-P* family
| Report | Title | Consuming page(s) |
|---|---|---|
D-P01 | neuroncodegen matmul | nki/neuroncodegen-forward-builder.md |
D-P02 | neuroncodegen tensorops (+D-P02b builder) | nki/nkicodegen-printer.md |
D-P04 | neuroncodegen collective | nki/bircodegen-collective.md nki/neuroncodegen-collectives.md |
D-P06 | neuroncodegen macro | nki/neuroncodegen-tiling.md |
D-P11 | bircodegen dma | nki/bircodegen-dma.md |
D-P12 | bircodegen collective | nki/bircodegen-collective.md |
D-P13 | bircodegen macro | nki/three-sink-kernel-model.md |
D-P14 | frontend bridge | nki/frontend-bridge-cache.md |
D-P15 | bircodegen ap | nki/bircodegen-ap.md |
D-P18 | internal registry | nki/internal-kernel-registry.md |
D-P19 | nki cache | nki/frontend-bridge-cache.md |
D-P20 | tracekernel | nki/trace-context.md |
D-P21 | decorators | nki/option-alloc-decorators.md |
D-P22 | irbuilder | nki/isa-compute-intrinsics.md nki/isa-reduce-dve-dma.md nki/tile-tensor-model.md penguin/ir-mlir-bir-mapping.md |
D-Q* family
| Report | Title | Consuming page(s) |
|---|---|---|
D-Q04 | builtin sort | custom-ops/bitonic-sort-topk.md custom-ops/gpsimd-xtensa-layout.md |
D-Q05 | customop abi | custom-ops/gpsimd-xtensa-layout.md |
D-Q06 | macro lowerers | custom-ops/customop-full-chain.md |
D-Q08 | customop wire | custom-ops/customop-codegen.md custom-ops/customop-full-chain.md custom-ops/customop-wire-layout.md isa/collective-customop-encoding.md nki/neuroncodegen-builtin-customop.md |
D-Q09 | customop codegen | custom-ops/customop-codegen.md |
D-Q10 | customop chain | custom-ops/customop-full-chain.md |
D-R* family
| Report | Title | Consuming page(s) |
|---|---|---|
D-R03 | pelican hierarchy | bir/pelican-hierarchy.md |
D-R08 | dap types | bir/birsim-dap-debugger.md |
D-R10 | bir structural hierarchy | bir/value-model.md |
D-S* family
| Report | Title | Consuming page(s) |
|---|---|---|
D-S01 | neff elf layout | formats/neff-container.md |
D-S02 | neff json | distribution/three-d-rank-model.md formats/neff-json-sidecars.md |
D-S03 | neff bom | formats/per-engine-bin.md |
D-S04 | neff features | formats/neff-feature-flags.md |
D-S06 | bir tojson | bir/json-writer.md |
D-T* family
| Report | Title | Consuming page(s) |
|---|---|---|
D-T04 | pwp bkt | activation/bkt-ctrl-blob.md |
D-T05 | pwp ctrl | activation/bkt-ctrl-blob.md activation/set-cover.md |
D-U* family
| Report | Title | Consuming page(s) |
|---|---|---|
D-U01 | penguin pass roster | frontend/opt-level-planes.md |
D-U02 | penguin tensorizer | penguin/affine-expr-algebra.md penguin/axis-loop-model.md penguin/tensor-op-family.md |
D-U08 | penguin ir model | penguin/affine-expr-algebra.md penguin/axis-loop-model.md penguin/tensor-op-family.md penguin/tonga-isl-simplifier.md |
D-V* family
| Report | Title | Consuming page(s) |
|---|---|---|
D-V01 | pe matmul corev2 encode | numerics/mixed-precision-accumulation.md |
D-V03 | mx matmul quantize corev4 encode | nki/isa-compute-intrinsics.md |
D-V05 | pool reduce encode | isa/dve-datamove-encoding.md |
D-V06 | isa enum ordinals | appendix/master-enum-dtype-table.md isa/sp-register-encoding.md isa/tensortensor-encoding.md |
D-W* family
| Report | Title | Consuming page(s) |
|---|---|---|
D-W02 | nki isa reduce dve dma | nki/isa-compute-intrinsics.md |
D-W04 | nki sema asserts | nki/sema-legality.md |
D-W05 | nki err catalog | appendix/error-message-catalog.md nki/sema-legality.md |
D-W09 | nki type system | nki/dtype-facade.md nki/type-metaclasses.md nki/type-system.md |
D-W10 | nki mask predicate | nki/mask-predicate-algebra.md |
D-W11 | nki tracing spmd | nki/nki-ctx-scopes.md nki/option-alloc-decorators.md |
D-W12 | framework binding surface | nki/entrypoints.md nki/option-alloc-decorators.md |
D-X* family
| Report | Title | Consuming page(s) |
|---|---|---|
D-X05 | dtype promotion lattice | numerics/dtype-catalog.md numerics/dtype-promotion-lattice.md |
D-X06 | cast to new dtype | numerics/dtype-catalog.md numerics/numeric-negative-results.md |
D-X11 | numeric negative results | hlo-opt/hlo-ingestion-boundary.md numerics/numeric-negative-results.md |
D-Y* family
| Report | Title | Consuming page(s) |
|---|---|---|
D-Y01 | penguin isl dependence | penguin/isl-dependence-graph.md penguin/tonga-isl-simplifier.md |
D-Y05 | tonga isl simplifier | penguin/tonga-isl-simplifier.md |
D-Y06 | affine isl pelican bridge | penguin/affine-expr-algebra.md |
D-Z* family
| Report | Title | Consuming page(s) |
|---|---|---|
D-Z02 | dge level dynamic dma | penguin/dge-level-dynamic-dma.md walrus/dmametrics-profiler.md |
D-Z03 | dynamic for loop | penguin/dynamic-for-loop.md |
D-Z06 | symbolic ap register (alu + lowering) | penguin/tonga-isl-simplifier.md |
D-Z07 | dynamic shape system synthesis | penguin/dynamic-shape-synthesis.md |
3. Reverse map — page → backing reports
The inverse of §2: for the highest-fan-in pages (those resting on the most distinct reports), the reports that back them. A page absent here either cites a single report — read it off the forward map — or cites none. The full reverse map for every page is reconstructible from §2 by transposing the edges; this section spotlights the pages a reviewer most wants to audit, because they integrate the most evidence.
| Page | # reports | Backing reports |
|---|---|---|
nki/bircodegen-collective.md | 11 | D-B15 D-D07 D-E09 D-F08 D-H23 D-I09 D-I10 D-I11 D-O26 D-P04 D-P12 |
appendix/master-enum-dtype-table.md | 9 | D-B04 D-D01 D-D04 D-D09 D-D10 D-D11 D-D13 D-D14 D-V06 |
bir/codegen-sub-encoders.md | 8 | D-E14 D-E19 D-H13 D-H21 D-I01 D-I17 D-I21 D-I23 |
walrus/local-collectives.md | 8 | D-A11 D-D07 D-E22 D-G11 D-H03 D-H06 D-H23 D-M11 |
bir/codegen-dve-rng-control.md | 7 | D-D01 D-D02 D-D08 D-E14 D-F08 D-H35 D-I11 |
isa/neuron-isa-tpb-capstone.md | 7 | D-D04 D-J03 D-J05 D-J06 D-J32 D-M08 D-M10 |
walrus/parserdumper.md | 6 | D-D03 D-D05 D-D09 D-D12 D-E06 D-G10 |
nki/moe-cte-prefill.md | 6 | D-O01 D-O02 D-O03 D-O04 D-O05 D-O29 |
bir/value-model.md | 5 | D-D04 D-D12 D-E12 D-E14 D-R10 |
numerics/numeric-negative-results.md | 5 | D-F04 D-F13 D-J20 D-X06 D-X11 |
numerics/mx-matmul-legality.md | 5 | D-D04 D-E21 D-F02 D-F13 D-G05 |
bir/op-family-enums.md | 5 | D-D07 D-D09 D-D10 D-D13 D-D14 |
nki/isa-compute-intrinsics.md | 5 | D-D04 D-D05 D-P22 D-V03 D-W02 |
distribution/distribution-strategy-seeding.md | 4 | D-AB01 D-AB06 D-AB08 D-AF01 |
walrus/dma-queues.md | 4 | D-D01 D-E16 D-E19 D-G03 |
penguin/tonga-isl-simplifier.md | 4 | D-U08 D-Y01 D-Y05 D-Z06 |
4. Gaps — unconsumed reports and dangling citations
A provenance index is only as useful as the gaps it surfaces. Two are actionable.
4.1 Unconsumed reports (269)
These reports exist in raw/ but are cited by zero wiki pages. They represent evidence that was captured during analysis but is not yet written up in a normative page — the to-do backlog of the wiki, not an error. A maintainer planning new pages should mine this list first. Listed by family (the consumed ones are in §2):
D-A* (4): D-A02, D-A04, D-A07, D-A08
D-AA* (6): D-AA01, D-AA02, D-AA03, D-AA06, D-AA08, D-AA09
D-AB* (4): D-AB02, D-AB03, D-AB04, D-AB09
D-AC* (1): D-AC04
D-AD* (3): D-AD02, D-AD04, D-AD05
D-AE* (3): D-AE01, D-AE02, D-AE04
D-AF* (2): D-AF03, D-AF04
D-AG* (7): D-AG01, D-AG02, D-AG03, D-AG04, D-AG05, D-AG06, D-AG10
D-B* (16): D-B05, D-B06, D-B07, D-B11, D-B12, D-B16, D-B18, D-B19, D-B20, D-B21, D-B22, D-B25, D-B26, D-B27, D-B30, D-B32
D-C* (12): D-C03, D-C05, D-C06, D-C08, D-C13, D-C16, D-C17, D-C18, D-C19, D-C20, D-C23, D-C24
D-D* (1): D-D06
D-E* (10): D-E01, D-E04, D-E05, D-E08, D-E10, D-E11, D-E15, D-E17, D-E18, D-E20
D-F* (12): D-F01, D-F05, D-F06, D-F07, D-F09, D-F10, D-F11, D-F12, D-F14, D-F15, D-F16, D-F17
D-G* (5): D-G06, D-G07, D-G09, D-G12, D-G13
D-H* (26): D-H02, D-H04, D-H05, D-H07, D-H08, D-H09, D-H11, D-H12, D-H14, D-H15, D-H18, D-H19, D-H20, D-H22, D-H25, D-H27, D-H28, D-H29, D-H30, D-H31, D-H32, D-H33, D-H34, D-H36, D-H37, D-H39
D-I* (19): D-I02, D-I03, D-I04, D-I05, D-I06, D-I07, D-I08, D-I13, D-I14, D-I15, D-I16, D-I18, D-I19, D-I20, D-I22, D-I24, D-I25, D-I26, D-I27
D-J* (24): D-J01, D-J02, D-J07, D-J08, D-J10, D-J11, D-J12, D-J15, D-J16, D-J17, D-J18, D-J19, D-J21, D-J22, D-J23, D-J24, D-J25, D-J26, D-J27, D-J30, D-J31, D-J33, D-J35, D-J36
D-K* (13): D-K02, D-K03, D-K04, D-K05, D-K06, D-K07, D-K08, D-K09, D-K11, D-K13, D-K16, D-K17, D-K18
D-L* (8): D-L01, D-L03, D-L04, D-L05, D-L06, D-L07, D-L08, D-L09
D-M* (6): D-M01, D-M02, D-M03, D-M09, D-M13, D-M14
D-N* (10): D-N01, D-N03, D-N04, D-N06, D-N07, D-N08, D-N09, D-N10, D-N11, D-N12
D-O* (6): D-O07, D-O13, D-O17, D-O19, D-O20, D-O21
D-P* (8): D-P03, D-P05, D-P07, D-P08, D-P09, D-P10, D-P16, D-P17
D-Q* (4): D-Q01, D-Q02, D-Q03, D-Q07
D-R* (7): D-R01, D-R02, D-R04, D-R05, D-R06, D-R07, D-R09
D-S* (8): D-S05, D-S07, D-S08, D-S09, D-S10, D-S11, D-S12, D-S13
D-T* (6): D-T01, D-T02, D-T03, D-T06, D-T07, D-T08
D-U* (5): D-U03, D-U04, D-U05, D-U06, D-U07
D-V* (12): D-V02, D-V04, D-V07, D-V08, D-V09, D-V10, D-V11, D-V12, D-V13, D-V14, D-V15, D-V16
D-W* (5): D-W01, D-W03, D-W06, D-W07, D-W08
D-X* (8): D-X01, D-X02, D-X03, D-X04, D-X07, D-X08, D-X09, D-X10
D-Y* (5): D-Y02, D-Y03, D-Y04, D-Y07, D-Y08
D-Z* (3): D-Z01, D-Z04, D-Z05
4.2 Dangling citations (0)
No page cites a D-ID that fails to resolve to a raw/ file. Every evidence anchor in the wiki points at a report that exists.
CORRECTION — the one apparent dangle is a regex artifact, not a real one. A loose scan (
rg 'D-SB2'without word boundaries) reportsD-SB2as cited-but-missing. It is not a citation: it is the substringSB2inside the hardware termGPSIMD-SB2SBonwalrus/legality-dispatch.md(a SBUF-to-SBUF GPSIMD op,bir::InstGPSIMDSB2SB). Anchoring the match with\bD-[A-Z]+[0-9]+\brejects it and the dangling count is 0. This is the canonical reason the regeneration recipe in §5 mandates word boundaries — drop them and this index manufactures a phantom gap.
5. How this page is regenerated
This appendix is a pure function of the repo state at generation time. To rebuild it after a wiki edit, re-run the same scans the edges above were produced from. The recipe is the page's reproducibility contract.
## step 1 — corpus of distinct D-IDs (stems) in raw/ -> 504
ls neuronx-cc/raw/ | rg '^D-' | rg -o '^D-[A-Z]+[0-9]+' | sort -u
## step 2 — distinct D-IDs cited anywhere in the wiki (WORD-BOUNDED) -> 236
rg --no-filename --only-matching -g '*.md' '\bD-[A-Z]+[0-9]+\b' \
neuronx-cc/wiki/src/ | sort -u
## step 3 — forward edge for one report: which page(s) cite D-Xnn
rg -l -g '*.md' '\bD-Xnn\b' neuronx-cc/wiki/src/
## step 4 — consumed / unconsumed / dangling split (comm against the two sets)
comm -12 raw_dids cited_dids # consumed -> 235
comm -23 raw_dids cited_dids # unconsumed -> 269
comm -13 raw_dids cited_dids # dangling -> 0 (with \b...\b anchors)
Two invariants any regeneration must preserve, or it will produce wrong findings:
- Always anchor with
\b...\b. Without word boundaries, hardware tokens likeGPSIMD-SB2SBmasquerade as citations and inflate the dangling count (§4.2). - Key forward edges on the stem, not the filename.
D-P02/D-P02band the twoD-Z06notes share a stem; counting files (506) instead of stems (504) double-counts them.
6. Adversarial verification
Five forward edges and the three census totals, re-checked by hand against the repo. Every check passed.
| Check | Method | Result |
|---|---|---|
D-D04 → 7 pages | rg -l '\bD-D04\b' wiki/src/ | matches the 7 slugs in §2 D-D*; raw/D-D04-dtype-tables.txt exists — PASS |
D-A13 → error-message-catalog | rg -c '\bD-A13\b' …/error-message-catalog.md | 8 hits — PASS |
D-Q08 → 5 pages | rg -l '\bD-Q08\b' wiki/src/ | matches the 5 custom-op/isa/nki slugs; raw/D-Q08-customop-wire.txt exists — PASS |
D-AC07 → two-gpsimd-reconciliation | rg -c … + ls raw/ | 3 hits; raw/D-AC07-…-reconciliation.txt exists — PASS |
D-AE03 → dram-hbm-geometry | rg -c … + ls raw/ | 1 hit; raw/D-AE03-hbm-dram-geometry.txt exists — PASS |
| Corpus total | ls raw/ | rg -c '^D-' | 506 files / 504 stems — PASS |
| Consumed + unconsumed | 235 + 269 | = 504 (every stem is exactly consumed xor unconsumed) — PASS |
| Dangling | comm -13 with \b...\b | 0 — PASS |
See also
- §0.3 Methodology & Confidence Model — what a
CONFIRMED/INFERREDtag means and how reports earn one. - §14.4 Symbol / Offset Index — the sibling provenance index, mapping binary symbols and addresses to the pages that document them.
- §14.1 Master Opcode Reference, §14.2 Master Enum & Dtype Table, §14.3 Error-Message Catalog — the consolidated lookup appendices whose own backing reports appear in §2.