ywkang f0c5b9c33f gqa(adr): ADR-0065 Proposed -> Accepted ...

Flat-ops CompositeCmd + softmax_merge recipe is fully implemented and tested: structural restructure (P1-P4), dispatch-cost measurement (P5b/P6, 5.22x per-tile win), and data-mode numerics (D8 output handle/space + N1-N4 composite GEMM / recipe MATH / accumulate / opt2 parity). git mv EN -> docs/adr/, KO -> docs/adr-ko/ (verify_adr_lang_pairs OK).

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

2026-06-10 23:16:52 -07:00

..

ADR-0001-mem-physaddr-layout.md

ADR: translate adr-ko/ to Korean, fix ADR-0013 slug, refine Status check

2026-05-20 08:17:56 -07:00

ADR-0002-lat-routing-distance.md

ADR: translate adr-ko/ to Korean, fix ADR-0013 slug, refine Status check

2026-05-20 08:17:56 -07:00

ADR-0003-dev-target-system-hierarchy.md

ADR: translate adr-ko/ to Korean, fix ADR-0013 slug, refine Status check

2026-05-20 08:17:56 -07:00

ADR-0004-mem-memory-semantics-local-hbm.md

ADR: translate adr-ko/ to Korean, fix ADR-0013 slug, refine Status check

2026-05-20 08:17:56 -07:00

ADR-0005-dev-diagram-views-distance-layout.md

ADR: translate adr-ko/ to Korean, fix ADR-0013 slug, refine Status check

2026-05-20 08:17:56 -07:00

ADR-0006-dev-topology-compilation-distance-diagram.md

ADR: translate adr-ko/ to Korean, fix ADR-0013 slug, refine Status check

2026-05-20 08:17:56 -07:00

ADR-0007-api-runtime-api-boundaries.md

ADR: translate adr-ko/ to Korean, fix ADR-0013 slug, refine Status check

2026-05-20 08:17:56 -07:00

ADR-0008-api-tensor-deploy-and-allocation.md

ADR: translate adr-ko/ to Korean, fix ADR-0013 slug, refine Status check

2026-05-20 08:17:56 -07:00

ADR-0009-api-kernel-execution-messaging.md

ADR: translate adr-ko/ to Korean, fix ADR-0013 slug, refine Status check

2026-05-20 08:17:56 -07:00

ADR-0010-api-cli-surface-and-semantics.md

benches: package as kernbench.benches, add @bench registry + list subcommand

2026-05-20 14:42:10 -07:00

ADR-0011-mem-memory-addressing-simplification.md

ADR: translate adr-ko/ to Korean, fix ADR-0013 slug, refine Status check

2026-05-20 08:17:56 -07:00

ADR-0012-api-host-io-message-schema.md

ADR: translate adr-ko/ to Korean, fix ADR-0013 slug, refine Status check

2026-05-20 08:17:56 -07:00

ADR-0013-ver-verification-strategy.md

ADR: translate adr-ko/ to Korean, fix ADR-0013 slug, refine Status check

2026-05-20 08:17:56 -07:00

ADR-0014-dev-pe-pipeline-execution-model.md

ADR: translate adr-ko/ to Korean, fix ADR-0013 slug, refine Status check

2026-05-20 08:17:56 -07:00

ADR-0015-dev-component-port-wire-model.md

ADR: translate adr-ko/ to Korean, fix ADR-0013 slug, refine Status check

2026-05-20 08:17:56 -07:00

ADR-0016-dev-iochiplet-noc-and-memory-path.md

ADR: translate adr-ko/ to Korean, fix ADR-0013 slug, refine Status check

2026-05-20 08:17:56 -07:00

ADR-0017-dev-cube-noc-and-hbm-connectivity.md

ADR: translate adr-ko/ to Korean, fix ADR-0013 slug, refine Status check

2026-05-20 08:17:56 -07:00

ADR-0020-prog-data-execution-two-pass.md

ADR: bilingual structure — EN canonical in adr/, KO mirror in adr-ko/

2026-05-20 01:38:44 -07:00

ADR-0022-prog-program-id-2d-grid.md

ADR: translate adr-ko/ to Korean, fix ADR-0013 slug, refine Status check

2026-05-20 08:17:56 -07:00

ADR-0023-dev-ipcq-pe-collective.md

ADR: bilingual structure — EN canonical in adr/, KO mirror in adr-ko/

2026-05-20 01:38:44 -07:00

ADR-0024-par-sip-tp-launcher.md

adr: add ADR-0043/0044 (eval harnesses); reconcile ADR-0024/0032 for SIP w/h

2026-05-21 10:26:25 -07:00

ADR-0025-algo-ipcq-direction-addressing.md

ADR: bilingual structure — EN canonical in adr/, KO mirror in adr-ko/

2026-05-20 01:38:44 -07:00

ADR-0026-par-dppolicy-intra-device.md

policy(adr): ADR-0026 Revision 6 — document DPPolicy.cube_start

2026-06-08 14:18:27 -07:00

ADR-0027-par-megatron-tp.md

ADR: bilingual structure — EN canonical in adr/, KO mirror in adr-ko/

2026-05-20 01:38:44 -07:00

ADR-0032-algo-intercube-allreduce.md

adr: add ADR-0043/0044 (eval harnesses); reconcile ADR-0024/0032 for SIP w/h

2026-05-21 10:26:25 -07:00

ADR-0033-lat-latency-model-assumptions.md

ADR: translate adr-ko/ to Korean, fix ADR-0013 slug, refine Status check

2026-05-20 08:17:56 -07:00

ADR-0034-dev-hbm-controller-internal-design.md

ADR: translate adr-ko/ to Korean, fix ADR-0013 slug, refine Status check

2026-05-20 08:17:56 -07:00

ADR-0035-dev-m-cpu-and-m-cpu-dma-component-model.md

ADR: translate adr-ko/ to Korean, fix ADR-0013 slug, refine Status check

2026-05-20 08:17:56 -07:00

ADR-0036-dev-io-cpu-component-model.md

ADR: translate adr-ko/ to Korean, fix ADR-0013 slug, refine Status check

2026-05-20 08:17:56 -07:00

ADR-0037-dev-forwarding-component.md

ADR: translate adr-ko/ to Korean, fix ADR-0013 slug, refine Status check

2026-05-20 08:17:56 -07:00

ADR-0038-dev-pcie-ep-component-model.md

ADR: add 0038-0042 (pcie_ep, pe_mmu, pe_tcm, sram, tiling)

2026-05-20 14:43:03 -07:00

ADR-0039-dev-pe-mmu-component-model.md

ADR: add 0038-0042 (pcie_ep, pe_mmu, pe_tcm, sram, tiling)

2026-05-20 14:43:03 -07:00

ADR-0040-dev-pe-tcm-component-model.md

ADR: add 0038-0042 (pcie_ep, pe_mmu, pe_tcm, sram, tiling)

2026-05-20 14:43:03 -07:00

ADR-0041-dev-cube-sram-component-model.md

ADR: add 0038-0042 (pcie_ep, pe_mmu, pe_tcm, sram, tiling)

2026-05-20 14:43:03 -07:00

ADR-0042-prog-tile-plan-generators.md

ADR: add 0038-0042 (pcie_ep, pe_mmu, pe_tcm, sram, tiling)

2026-05-20 14:43:03 -07:00

ADR-0043-eval-allreduce-harness.md

eval: fold GEMM/allreduce harnesses into self-contained milestone benches

2026-05-22 15:19:52 -07:00

ADR-0044-eval-gemm-harness.md

eval: fold GEMM/allreduce harnesses into self-contained milestone benches

2026-05-22 15:19:52 -07:00

ADR-0045-prog-bench-module-contract.md

eval: fold GEMM/allreduce harnesses into self-contained milestone benches

2026-05-22 15:19:52 -07:00

ADR-0046-prog-tl-context-contract.md

adr: add ADR-0046-0049 — close G4 coverage gaps from /report

2026-05-22 10:25:04 -07:00

ADR-0047-par-ahbm-ccl-backend.md

adr: add ADR-0046-0049 — close G4 coverage gaps from /report

2026-05-22 10:25:04 -07:00

ADR-0048-mem-allocator-algorithms.md

adr: add ADR-0046-0049 — close G4 coverage gaps from /report

2026-05-22 10:25:04 -07:00

ADR-0049-ver-probe-subcommand.md

adr: add ADR-0046-0049 — close G4 coverage gaps from /report

2026-05-22 10:25:04 -07:00

ADR-0050-par-ccl-algorithm-module-contract.md

adr: add ADR-0050-0053 — close /report's second-pass G4 candidates

2026-05-22 10:52:42 -07:00

ADR-0051-lat-routing-helper-api.md

adr: add ADR-0050-0053 — close /report's second-pass G4 candidates

2026-05-22 10:52:42 -07:00

ADR-0052-dev-oplog-memory-store-schemas.md

adr: add ADR-0050-0053 — close /report's second-pass G4 candidates

2026-05-22 10:52:42 -07:00

ADR-0053-dev-topology-builder-algorithms.md

adr: add ADR-0050-0053 — close /report's second-pass G4 candidates

2026-05-22 10:52:42 -07:00

ADR-0054-eval-milestone-benches.md

eval: commit milestone bench output (track generated figures + results)

2026-05-22 15:37:27 -07:00

ADR-0060-algo-gqa-fused-attention-ahbm.md

gqa: tile-granular Ring KV (P3c) + rename to gqa_attention_* + ADR-0060/62/63/64 → Accepted

2026-06-10 16:20:00 -07:00

ADR-0062-prog-tl-async-load.md

gqa: tile-granular Ring KV (P3c) + rename to gqa_attention_* + ADR-0060/62/63/64 → Accepted

2026-06-10 16:20:00 -07:00

ADR-0063-prog-tl-scratch-scope.md

gqa: tile-granular Ring KV (P3c) + rename to gqa_attention_* + ADR-0060/62/63/64 → Accepted

2026-06-10 16:20:00 -07:00

ADR-0064-perf-cpu-issue-cost-model.md

gqa(adr-0065): P2 — softmax_merge recipe + TLContext prologue lowering

2026-06-10 19:48:29 -07:00

ADR-0065-prog-flat-ops-composite-softmax-merge.md

gqa(adr): ADR-0065 Proposed -> Accepted

2026-06-10 23:16:52 -07:00

INDEX.md

eval: fold GEMM/allreduce harnesses into self-contained milestone benches

2026-05-22 15:19:52 -07:00