diff --git a/.beads/interactions.jsonl b/.beads/interactions.jsonl
index 5ce16e88fd46dc4e7997091efb1f46d0d23e04fc..f62151e7ce6ef618958c464987dde25c03e9b394 100644
--- a/.beads/interactions.jsonl
+++ b/.beads/interactions.jsonl
@@ -33,3 +33,4 @@ {"id":"int-8dbcb031","kind":"field_change","created_at":"2026-05-23T18:38:56.610443Z","actor":"Eugene Blikh","issue_id":"tarantool-protobuf-m7u","extra":{"field":"status","new_value":"closed","old_value":"in_progress","reason":"Closed"}}
 {"id":"int-500abbb3","kind":"field_change","created_at":"2026-05-23T19:50:37.871661Z","actor":"Eugene Blikh","issue_id":"tarantool-protobuf-rc8","extra":{"field":"status","new_value":"closed","old_value":"in_progress","reason":"Closed"}}
 {"id":"int-0be8d1bc","kind":"field_change","created_at":"2026-05-23T19:56:04.251964Z","actor":"Eugene Blikh","issue_id":"tarantool-protobuf-ra6","extra":{"field":"status","new_value":"closed","old_value":"open","reason":"All 16 sub-tasks (3a-3l) plus rc8 strict-decode parity bug closed. Generic C runtime codec landed: descriptor plan compiler, encode/decode scalars + repeated/packed + sub-messages + maps + oneofs + 64-bit cdata + WKT hooks + unknown fields + proto2 + dispatch + strict-decode parity. Ready for 43t parity gate."}}
 {"id":"int-2a7d9cba","kind":"field_change","created_at":"2026-05-23T20:04:23.62044Z","actor":"Eugene Blikh","issue_id":"tarantool-protobuf-43t","extra":{"field":"status","new_value":"closed","old_value":"in_progress","reason":"Parity gate landed. Added test-c / test-all / bench-c Justfile recipes; bench.lua relabels runtime→c-runtime under PB_ENABLE_C=1 and refuses --baseline/--compare in that mode. Reuses existing 1043-test luatest matrix, Google conformance suite, and interop fixtures via the transitivity-through-reference strategy. c-generated column dropped from scope (c0i deferred per docs/c-accel.md); starwing column lives in bench/starwing_bench.lua separately."}}
+{"id":"int-a631a4f9","kind":"field_change","created_at":"2026-05-23T20:13:43.682922Z","actor":"Eugene Blikh","issue_id":"tarantool-protobuf-74c","extra":{"field":"status","new_value":"closed","old_value":"in_progress","reason":"Phase-2 EmmyLua annotations landed. Added ---@param/---@return on public surface of codec.lua (encode_message/decode_message/encode_field/compile_writers/compile_readers/merge_message), grpc.lua (loopback/multiplex/new_stream_pair/wrap_call/wrap_server_stream/wrap_server_view), json.lua (M.encode/M.decode with new pb.JsonEncodeOpts/pb.JsonDecodeOpts shapes), and wkt.lua (register/lookup/any_pack/any_unpack). Also corrected two pre-existing _types.lua signature lies: pb.register is (desc), not (full_name, desc); pb.any.pack is (desc, t, prefix?), not (t, type_url). just test 752/752, just test-c 1043/1043."}}
diff --git a/.beads/issues.jsonl b/.beads/issues.jsonl
index dbe804fef51d6e8d3b196b5784598dad7b59e131..09edf8d47753812f591739f8c40587a73add05eb 100644
--- a/.beads/issues.jsonl
+++ b/.beads/issues.jsonl
@@ -15,7 +15,7 @@ {"_type":"issue","id":"tarantool-protobuf-4kj","title":"Decoder: generated tag/length fast path for full-mode decode","description":"Inspection against starwing/lua-protobuf shows the 1KB+ Person decode gap is mostly repeated Lua-level tag/field dispatch, not a fatal JIT abort. On the 930B Person fixture, generated full decode is ~6.9-7.5 us/op, tag+length scan is ~2.1 us/op, skip_field scan is ~3.6 us/op, and an order-specialized decoder is ~4.3 us/op. Implement a full-mode generated fast path that decodes expected 1-byte tags and LEN prefixes inline at the call site, falling back to the generic decode_tag/skip path for unknown/out-of-order/multi-byte cases. Keep correctness for arbitrary field order, repeated occurrences, unknown fields, and proto2 semantics.","status":"closed","priority":1,"issue_type":"task","assignee":"Eugene Blikh","owner":"bigbes@gmail.com","created_at":"2026-05-17T16:02:41Z","created_by":"Eugene Blikh","updated_at":"2026-05-18T18:56:20Z","started_at":"2026-05-18T18:44:48Z","closed_at":"2026-05-18T18:56:20Z","close_reason":"Inlined wire.decode_tag's 1-byte fast path at every M.X_decode while-loop site, plus header localization of string.byte/bit.band/bit.rshift. Median-of-3 Person full decode: +7-15% across all sizes. Full encode flat to -2.6% (likely header upvalue layout). 745/745 tests, 37/37 JIT, 0 bridges. See bench/PERF_LOG.md entry. Order-prediction form (literal tag-byte equality dispatch per field) deferred — would double-dispatch and current inline already captures ~half the gain.","dependencies":[{"issue_id":"tarantool-protobuf-4kj","depends_on_id":"tarantool-protobuf-0an","type":"blocks","created_at":"2026-05-17T19:02:53Z","created_by":"Eugene Blikh","metadata":"{}"}],"dependency_count":1,"dependent_count":0,"comment_count":0}
 {"_type":"issue","id":"tarantool-protobuf-gcy","title":"Decoder: inline nested-message decode at the call site","description":"Today Person_decode calls Address_decode(slice) across a function boundary; the side trace into Address_decode may not stitch back (memory: luajit_side_trace_inlined_return). For field-typed messages in mode=full, emit the decode body inline at the call site. Expected: 20-40% on payloads with nested messages. Codegen growth is bounded by recursion depth — pick a depth limit and call out beyond it.","status":"closed","priority":1,"issue_type":"task","assignee":"Eugene Blikh","owner":"bigbes@gmail.com","created_at":"2026-05-17T15:47:09Z","created_by":"Eugene Blikh","updated_at":"2026-05-18T18:44:22Z","started_at":"2026-05-18T18:32:15Z","closed_at":"2026-05-18T18:44:22Z","close_reason":"Attempted but reverted. Inlining Address_decode body into Person_decode (singular non-recursive candidate) showed 745/745 tests passing and 37/37 JIT, but bench median-of-3 regressed Person 1KB/10KB/100KB decode by 5-8% and encode by 3-6%. Profile's '100% interpreter bail' claim turned out to be a vl trace-attribution artifact; LuaJIT was already inlining Address_decode into Person_decode's trace effectively. Larger root traces (22 stops) compiled less efficiently. See bench/PERF_LOG.md entry for full data and analysis. Real decode bottlenecks remain decode_string utf8 (6bb), decode_tag fast path (4kj), repeated append (cch).","labels":["codegen","decoder","perf"],"dependency_count":0,"dependent_count":0,"comment_count":0}
 {"_type":"issue","id":"tarantool-protobuf-h8v","title":"Encoder: codegen-time inline FFI writes (mode=full)","description":"Replace per-field wire.encode_* calls + ../table.concat chain with directly-emitted FFI byte writes at every field site in mode=full. Today every encode_len(s) is 'encode_varint(#s) .. s' — two allocations and a concat per string. Sidesteps the per-byte b:alloc(1) cliff that sank the earlier ibuf attempt (memory: tarantool_ibuf_perf). Expected: 1.5-2x encode throughput; alloc/op drops from ~130 KB at 100 KB Person to near zero. Biggest single perf move. Bench reference: bench/starwing_bench.lua + bench/COMPARISON.md show starwing C encoder at 2.3-3x ours across all sizes.","status":"closed","priority":1,"issue_type":"task","assignee":"Eugene Blikh","owner":"bigbes@gmail.com","created_at":"2026-05-17T15:47:04Z","created_by":"Eugene Blikh","updated_at":"2026-05-18T18:31:54Z","started_at":"2026-05-18T18:25:39Z","closed_at":"2026-05-18T18:31:54Z","close_reason":"Inlined 1-byte varint length prefix at every LEN emit site; +25-48% encode at 1KB+ Person. Full FFI-buffer rewrite deferred — this slice captured the dominant profile target without that level of disruption. See bench/PERF_LOG.md entry for details.","labels":["codegen","encoder","perf"],"dependency_count":0,"dependent_count":2,"comment_count":0}
 {"_type":"issue","id":"tarantool-protobuf-e4t","title":"LSP: plugin emits @class type stubs for generated messages","description":"Phase 3 of the LSP / LLM affordance work. Follows phase 1 (commit 8721548 — runtime annotations) and phase 2 (bd-74c — annotate codec/grpc/json/wkt).\n\nHighest-payoff phase for LLM grounding. Modify the protoc plugin (cmd/protoc-gen-tarantool/internal/gen) so each generated _pb.lua module emits ---@class blocks describing every message, plus typed @param/@return on the _encode/_decode/_descriptor surface. After this lands, anyone (human or LLM) who require('full.hello.hello_pb') gets full type info on Person, Address, etc., without reading the .proto.\n\nEmit shape (decision: inline at top of generated file; one file per .proto already includes all messages from that file):\n\n  ---@class hello.Person\n  ---@field name?           string\n  ---@field user_id?        ffi.cdata*           # uint64\n  ---@field age?            integer              # int32\n  ---@field weight_kg?      number               # double\n  ---@field emails?         string[]\n  ---@field address?        hello.Address\n  ---@field lucky_numbers?  integer[]\n  ---@field favorite_color? hello.Color          # enum alias\n  ---@field tags?           table\u003cstring,string\u003e # map\u003cstring,string\u003e\n  ---@field unknown_fields? string\n\n  ---@param tbl hello.Person\n  ---@return string\n  function M.Person_encode(tbl) ... end\n\n  ---@param bytes string\n  ---@return hello.Person\n  function M.Person_decode(bytes) ... end\n\nField-kind -\u003e Lua-type mapping:\n- scalar int32/uint32/sint32/fixed32/sfixed32/bool → integer/boolean\n- scalar int64/uint64/sint64/fixed64/sfixed64 → ffi.cdata* (LuaJIT cdata, per project convention)\n- scalar float/double → number\n- scalar string/bytes → string\n- enum → alias of integer (emit ---@alias hello.Color integer)\n- message → another @class reference\n- repeated T → T[]\n- map\u003cK,V\u003e → table\u003cK_lua, V_lua\u003e\n- oneof — all members are optional, generator should NOT emit a discriminator field; user calls view:which() / inspects which value is non-nil\n\nAll fields are optional (? suffix) since proto3 default-elision means absence-on-wire is indistinguishable from default value. Required fields in proto2 omit the ?.\n\nBoth mode=full and mode=runtime emit the same _encode/_decode wrapper signatures, so the stubs apply uniformly.\n\nVerify by:\n1. Regenerating examples/expected/ via just gen\n2. Confirming the generated _pb.lua files load and pass tests (no behavior change)\n3. Opening examples/expected/full/hello/hello_pb.lua in an LSP-aware editor and checking that hover on hello.Person, M.Person_encode shows the @class + typed signature\n4. Sanity: require a generated module from spike code and confirm autocomplete on the result-table field names","acceptance_criteria":"Plugin emits ---@class blocks + typed _encode/_decode signatures into every generated _pb.lua; just gen + just test pass; LSP hover on require('full.hello.hello_pb').Person_encode shows ---@param tbl hello.Person ---@return string; ---@class hello.Person is reachable via hover on a decoded value","status":"open","priority":2,"issue_type":"feature","owner":"bigbes@gmail.com","created_at":"2026-05-23T09:30:19Z","created_by":"Eugene Blikh","updated_at":"2026-05-23T09:30:19Z","labels":["codegen","lsp"],"dependencies":[{"issue_id":"tarantool-protobuf-e4t","depends_on_id":"tarantool-protobuf-74c","type":"blocks","created_at":"2026-05-23T12:30:18Z","created_by":"Eugene Blikh","metadata":"{}"}],"dependency_count":1,"dependent_count":0,"comment_count":0}
-{"_type":"issue","id":"tarantool-protobuf-74c","title":"LSP: annotate codec/grpc/json/wkt with EmmyLua types","description":"Phase 2 of the LSP / LLM affordance work (phase 1 landed in commit 8721548 — .luarc.json, runtime/pb/_types.lua, init.lua + lazy.lua annotations).\n\nAdd ---@param / ---@return annotations to the remaining public-surface modules so editor hover and LLM context show types beyond just init.lua and the lazy views:\n\n- runtime/pb/codec.lua — encode(desc, t) / decode(desc, b); compile_writers / compile_readers (called from pb.finalize_message)\n- runtime/pb/grpc.lua — loopback / multiplex transport factories; the service client (factory(transport) -\u003e client_methods) and server (impl -\u003e {service, methods}) shapes\n- runtime/pb/json.lua — pb.json.encode(desc, t, opts) / pb.json.decode(desc, s, opts); opts shape (preserve_proto_field_names, emit_unpopulated, ...)\n- runtime/pb/wkt.lua — register(full_name, desc), lookup, any_pack/any_unpack, NULL sentinel\n\nFoundational @class declarations (pb.Descriptor, pb.Field, pb.GrpcTransport, pb.Json, pb.Wkt, pb.Module) already live in runtime/pb/_types.lua — extend as needed for opts shapes / map\u003cK,V\u003e generics.\n\nPure metadata change — no runtime behavior. Verify by running just test (must stay at 748 passing) and by opening a generated _pb.lua in an LSP-aware editor and confirming hover on require('pb').encode / .grpc.loopback / .json.encode shows the right signatures.","acceptance_criteria":"just test still green (748+ tests); hover in lua-language-server-aware editor shows typed signatures for pb.encode, pb.decode, grpc.loopback, json.encode, json.decode, wkt.register","status":"open","priority":2,"issue_type":"task","owner":"bigbes@gmail.com","created_at":"2026-05-23T09:29:55Z","created_by":"Eugene Blikh","updated_at":"2026-05-23T09:29:55Z","labels":["docs","lsp"],"dependency_count":0,"dependent_count":1,"comment_count":0}
+{"_type":"issue","id":"tarantool-protobuf-74c","title":"LSP: annotate codec/grpc/json/wkt with EmmyLua types","description":"Phase 2 of the LSP / LLM affordance work (phase 1 landed in commit 8721548 — .luarc.json, runtime/pb/_types.lua, init.lua + lazy.lua annotations).\n\nAdd ---@param / ---@return annotations to the remaining public-surface modules so editor hover and LLM context show types beyond just init.lua and the lazy views:\n\n- runtime/pb/codec.lua — encode(desc, t) / decode(desc, b); compile_writers / compile_readers (called from pb.finalize_message)\n- runtime/pb/grpc.lua — loopback / multiplex transport factories; the service client (factory(transport) -\u003e client_methods) and server (impl -\u003e {service, methods}) shapes\n- runtime/pb/json.lua — pb.json.encode(desc, t, opts) / pb.json.decode(desc, s, opts); opts shape (preserve_proto_field_names, emit_unpopulated, ...)\n- runtime/pb/wkt.lua — register(full_name, desc), lookup, any_pack/any_unpack, NULL sentinel\n\nFoundational @class declarations (pb.Descriptor, pb.Field, pb.GrpcTransport, pb.Json, pb.Wkt, pb.Module) already live in runtime/pb/_types.lua — extend as needed for opts shapes / map\u003cK,V\u003e generics.\n\nPure metadata change — no runtime behavior. Verify by running just test (must stay at 748 passing) and by opening a generated _pb.lua in an LSP-aware editor and confirming hover on require('pb').encode / .grpc.loopback / .json.encode shows the right signatures.","acceptance_criteria":"just test still green (748+ tests); hover in lua-language-server-aware editor shows typed signatures for pb.encode, pb.decode, grpc.loopback, json.encode, json.decode, wkt.register","status":"closed","priority":2,"issue_type":"task","assignee":"Eugene Blikh","owner":"bigbes@gmail.com","created_at":"2026-05-23T09:29:55Z","created_by":"Eugene Blikh","updated_at":"2026-05-23T20:13:44Z","started_at":"2026-05-23T20:09:08Z","closed_at":"2026-05-23T20:13:44Z","close_reason":"Phase-2 EmmyLua annotations landed. Added ---@param/---@return on public surface of codec.lua (encode_message/decode_message/encode_field/compile_writers/compile_readers/merge_message), grpc.lua (loopback/multiplex/new_stream_pair/wrap_call/wrap_server_stream/wrap_server_view), json.lua (M.encode/M.decode with new pb.JsonEncodeOpts/pb.JsonDecodeOpts shapes), and wkt.lua (register/lookup/any_pack/any_unpack). Also corrected two pre-existing _types.lua signature lies: pb.register is (desc), not (full_name, desc); pb.any.pack is (desc, t, prefix?), not (t, type_url). just test 752/752, just test-c 1043/1043.","labels":["docs","lsp"],"dependency_count":0,"dependent_count":1,"comment_count":0}
 {"_type":"issue","id":"tarantool-protobuf-rmf","title":"ra6 3k: WKT override-hook passthrough","description":"If a plan's descriptor has desc.encode and desc.decode set (the WKT pattern from runtime/pb/wkt.lua), the C runtime must call those Lua functions instead of walking fields. Plan compiler stores luaL_ref to those functions; encode/decode entry points check first. This is also the extension point for future per-message codegen C (c0i) — same hook. Depends on 3a only (just adds a check at entry to the encode/decode loop). Acceptance: hello.Event (uses Timestamp, Duration, Any, FieldMask, Wrappers, Struct) round-trips with PB_ENABLE_C=1 producing byte-equal output to mode=full; the WKT module is unmodified.","status":"closed","priority":2,"issue_type":"task","assignee":"Eugene Blikh","owner":"bigbes@gmail.com","created_at":"2026-05-18T20:21:13Z","created_by":"Eugene Blikh","updated_at":"2026-05-23T18:22:01Z","started_at":"2026-05-23T18:17:19Z","closed_at":"2026-05-23T18:22:01Z","close_reason":"Closed","dependencies":[{"issue_id":"tarantool-protobuf-rmf","depends_on_id":"tarantool-protobuf-mq7","type":"blocks","created_at":"2026-05-18T23:21:55Z","created_by":"Eugene Blikh","metadata":"{}"}],"dependency_count":1,"dependent_count":1,"comment_count":0}
 {"_type":"issue","id":"tarantool-protobuf-m7u","title":"ra6 3i: proto2 — required, defaults, groups, extensions","description":"Proto2 semantic differences from proto3 that the C runtime must honor: (1) Required fields: decode-time enforcement when explicitly opted in (or default to skipping per existing pure-Lua behavior — match exactly); encode-time enforcement same. (2) Field defaults: present in the plan; emit them when missing on encode and substitute on decode where proto2 semantics require. (3) Groups (SGROUP/EGROUP wire format): encode/decode the framed group syntax. (4) Extensions: walk extension ranges, use plan-cached extension descriptor pointers, store in result._extensions. (5) Closed enum semantics for proto2. Depends on 3b + 3c + 3d + 3e + 3g + 3j (it's the join point — proto2 touches scalars, sub-messages, repeated, oneofs, unknown-fields). Acceptance: proto2_basic.BenchPayload and test_messages_proto2 round-trip byte-equal to mode=full at all bench sizes; conformance proto2 suite passes under PB_ENABLE_C=1.","status":"closed","priority":2,"issue_type":"task","assignee":"Eugene Blikh","owner":"bigbes@gmail.com","created_at":"2026-05-18T20:21:03Z","created_by":"Eugene Blikh","updated_at":"2026-05-23T18:38:57Z","started_at":"2026-05-23T18:26:23Z","closed_at":"2026-05-23T18:38:57Z","close_reason":"Closed","dependencies":[{"issue_id":"tarantool-protobuf-m7u","depends_on_id":"tarantool-protobuf-hwe","type":"blocks","created_at":"2026-05-18T23:22:21Z","created_by":"Eugene Blikh","metadata":"{}"},{"issue_id":"tarantool-protobuf-m7u","depends_on_id":"tarantool-protobuf-jc9","type":"blocks","created_at":"2026-05-18T23:22:22Z","created_by":"Eugene Blikh","metadata":"{}"},{"issue_id":"tarantool-protobuf-m7u","depends_on_id":"tarantool-protobuf-mz6","type":"blocks","created_at":"2026-05-18T23:22:20Z","created_by":"Eugene Blikh","metadata":"{}"},{"issue_id":"tarantool-protobuf-m7u","depends_on_id":"tarantool-protobuf-w3u","type":"blocks","created_at":"2026-05-18T23:22:22Z","created_by":"Eugene Blikh","metadata":"{}"},{"issue_id":"tarantool-protobuf-m7u","depends_on_id":"tarantool-protobuf-wyp","type":"blocks","created_at":"2026-05-18T23:22:23Z","created_by":"Eugene Blikh","metadata":"{}"},{"issue_id":"tarantool-protobuf-m7u","depends_on_id":"tarantool-protobuf-y1n","type":"blocks","created_at":"2026-05-18T23:22:20Z","created_by":"Eugene Blikh","metadata":"{}"}],"dependency_count":6,"dependent_count":1,"comment_count":0}
 {"_type":"issue","id":"tarantool-protobuf-asz","title":"ra6 3h: maps (entry-as-pseudo-message)","description":"map\u003cK,V\u003e fields are wire-encoded as repeated messages with synthetic Entry { key=1; value=2 } shape. Compile a map plan: outer field is K_REPEATED_MESSAGE with a synthesized Entry sub-plan. Encode: walk the Lua map via pairs(), emit each entry (this is the documented JIT exception — map fields are allowed pairs() in the hot path per CLAUDE.md). Decode: each entry yields a key + value, lua_settable into the result map. Depends on 3d (uses sub-message machinery). Acceptance: Person.ages_by_nickname / nickname_by_age / addresses_by_label round-trip; multi-key map fixtures behave correctly even though encode byte order differs (the existing map_test pattern).","status":"closed","priority":2,"issue_type":"task","assignee":"Eugene Blikh","owner":"bigbes@gmail.com","created_at":"2026-05-18T20:20:57Z","created_by":"Eugene Blikh","updated_at":"2026-05-23T17:07:06Z","started_at":"2026-05-23T16:45:29Z","closed_at":"2026-05-23T17:07:06Z","close_reason":"ra6 3h: maps (entry-as-pseudo-message) — encode walks the user map with lua_next (CLAUDE.md JIT exception), builds each entry payload in a stack-backed sub-buffer with synthetic tag(1,key) + tag(2,value), proto3-elides defaults independently for key and value, and emits outer tag + len + payload into the parent buffer. Map\u003c,message\u003e resolves the value sub-plan once and recurses through encode_body. Decode reads the entry payload bounded, dispatches inner id=1/id=2 (skipping anything else per spec), and lua_rawsets into a lazy-created result map table; missing key or value falls back to the proto3 zero. Reuses the existing list_stack_idx[] slot for the lazy map cache (a field is either repeated or map, never both). 12 new tests cover round-trip for ages_by_nickname (string→int32), nickname_by_age (int32→string), and addresses_by_label (string→message), plus default-elision, multi-key correctness, empty maps, and unknown-inner-id tolerance. Full suite 964/964 with PB_ENABLE_C=1.","dependencies":[{"issue_id":"tarantool-protobuf-asz","depends_on_id":"tarantool-protobuf-hwe","type":"blocks","created_at":"2026-05-18T23:22:19Z","created_by":"Eugene Blikh","metadata":"{}"}],"dependency_count":1,"dependent_count":1,"comment_count":0}
@@ -32,7 +32,7 @@ {"_type":"issue","id":"tarantool-protobuf-kot","title":"Codegen: localize wire.* upvalues at top of each generated message function","description":"examples/expected/full/hello/hello_pb.lua has 106 wire.* accesses — each is a hash-table lookup on the upvalue. LuaJIT hoists this when traces stay hot, but every nested-message boundary breaks the trace (see gcy notes), forcing re-lookup in the side trace / interpreter. Fix: codegen-emit at the top of each Type_encode and Type_decode function the locals it actually uses — only for the wire.* entries referenced in that function body — so the in-body calls are direct LJ_FUNCC dispatches with no table lookup. One-line codegen change; broad impact whenever traces stitch poorly. Pairs naturally with h8v and 4kj (both move more work into the same generated function bodies, magnifying the per-call lookup cost). Expected: 3-8% across the board; larger when traces break.","notes":"Codegen-side change in the Go plugin (protoc-gen-tarantool). Pre-compute the set of wire.* symbols used in each function (e.g., {encode_varint, decode_string, decode_tag}) and emit 'local encode_varint = wire.encode_varint' style preambles. Keep the existing 'local wire = pb.wire' so non-emitted entries still work. Verify with bench/jit_trace.lua that we don't increase trace size past LuaJIT inline budget on large messages (test_messages_proto3 is 4212 lines).","status":"open","priority":2,"issue_type":"task","owner":"bigbes@gmail.com","created_at":"2026-05-18T16:51:23Z","created_by":"Eugene Blikh","updated_at":"2026-05-18T16:51:23Z","dependency_count":0,"dependent_count":0,"comment_count":0}
 {"_type":"issue","id":"tarantool-protobuf-lkz","title":"Encoder: single-pass two-phase (size + emit) into single buffer","description":"Eliminate the per-call out table and the per-field intermediate string allocations by emitting once into a pre-sized buffer. Two viable shapes (per drm notes):\n\n1. Two-pass: walk fields once to compute exact size, allocate one buffer, walk again to emit. Mirrors vtproto's Size()+MarshalTo() pattern — vtproto's 1-alloc encode is what gives it 13-20x throughput over us at 1KB-10KB Person.\n2. Single-pass with backpatched length varints: write tag + 1-byte placeholder for the length, recurse, fill in (memmove if final length \u003e= 128). Skips the size pass (~3 us at 1KB).\n\nImplementation can live in codegen (mode=full) — emit a _encode_to_buf per message that takes (data, ibuf, offset) and returns new_offset, then a public Person_encode that wraps it with the buffer allocation. Constraint: per-field closure count must stay \u003c= current pb.encode writers count, else the per-field dispatch cost re-emerges (drm: two prior ibuf prototypes both regressed by ~2x because of dispatch).\n\nRelates to h8v (codegen FFI direct writes) which is the per-field building block this work composes.","notes":"drm notes have the deeper analysis including why M6 ibuf path naive byte-write fails (20x interpreter dispatch overhead). The codegen approach sidesteps that by inlining all writes at codegen time so there's no runtime closure dispatch.","status":"open","priority":2,"issue_type":"task","owner":"bigbes@gmail.com","created_at":"2026-05-18T16:49:16Z","created_by":"Eugene Blikh","updated_at":"2026-05-18T16:49:16Z","dependencies":[{"issue_id":"tarantool-protobuf-lkz","depends_on_id":"tarantool-protobuf-h8v","type":"blocks","created_at":"2026-05-18T19:49:21Z","created_by":"Eugene Blikh","metadata":"{}"}],"dependency_count":1,"dependent_count":0,"comment_count":0}
 {"_type":"issue","id":"tarantool-protobuf-2sn","title":"Decoder: emit table.new(N, 0) for repeated-field lists","description":"Decode result table rehashes as nested fields populate. u39 covers table.new(0, N) for the message table itself, but per-field list tables (result.emails = {}, result.lucky_numbers = {}, ...) are also bare {} and re-hash as they grow. Worth a follow-up: emit table.new(N, 0) for repeated lists where the encoded count is recoverable from a single quick scan (count tag occurrences for non-packed, or read the LEN prefix and divide by per-element size for packed). Quick scan cost vs allocation savings is the tradeoff to measure — on the 1KB Person fixture, 26 emails + 5 packed lucky_numbers means 2 re-hash cycles per repeated list. Pairs with u39 to fully close the rehash-on-grow alloc pattern.","notes":"See bench/COMPARISON.md and bd show drm. For packed scalars where per-element size is fixed (fixed32/fixed64/float/double), count = payload_len / elem_size — O(1). For varint-packed and non-packed repeated, count requires a scan; might still be net-positive on payloads with \u003e8 elements but needs measurement.","status":"open","priority":2,"issue_type":"task","owner":"bigbes@gmail.com","created_at":"2026-05-18T16:49:15Z","created_by":"Eugene Blikh","updated_at":"2026-05-18T16:49:15Z","dependencies":[{"issue_id":"tarantool-protobuf-2sn","depends_on_id":"tarantool-protobuf-u39","type":"blocks","created_at":"2026-05-18T19:49:20Z","created_by":"Eugene Blikh","metadata":"{}"}],"dependency_count":1,"dependent_count":0,"comment_count":0}
 {"_type":"issue","id":"tarantool-protobuf-2nl","title":"Repo: unstealth Beads tracking files","description":"Make the repo-local Beads state visible to Git by removing the local .git/info/exclude rule for .beads/. Keep .beads/.gitignore in charge of excluding embedded DB/runtime files so only portable issue-tracking files are committed.","status":"closed","priority":2,"issue_type":"task","owner":"bigbes@gmail.com","created_at":"2026-05-17T18:39:50Z","created_by":"Eugene Blikh","updated_at":"2026-05-17T18:42:26Z","closed_at":"2026-05-17T18:42:26Z","close_reason":"Removed the local .git/info/exclude rule that hid .beads/ from Git. Portable .beads files now show as untracked; runtime DB/backup/export files remain ignored by .beads/.gitignore.","dependency_count":0,"dependent_count":0,"comment_count":0}
-{"_type":"issue","id":"tarantool-protobuf-43t","title":"C accel: conformance and parity gate for C paths","description":"Extend tests/benches so C runtime and generated-C modes must pass the same luatest, interop, JSON/text, proto2, lazy passthrough where applicable, and Google conformance suites as Lua modes. Add bench output columns for lua-full, lua-runtime, c-runtime, c-generated, starwing. C acceleration must remain byte-equal with current encoders and preserve unknown fields/extensions/groups.","notes":"Architecture finalized in docs/c-accel.md. Parity strategy: REUSE existing test suites — no separate Lua-vs-C diff harness. (1) just test (luatest, 639 tests already parameterized over full/runtime modes) runs once with PB_ENABLE_C unset and once with PB_ENABLE_C=1. Both must pass. (2) just conformance (Google proto3 + proto2) runs both modes. (3) Interop fixtures (test/interop/fixtures/*.bin) cover both modes via the existing parity.full_vs_runtime tests; extend the matrix to also cover c-runtime. Logic: every assert is against a reference output (golden bytes, txtpb, conformance result). If Lua passes and C passes, both equal the reference, so Lua == C by transitivity. No new test infrastructure required. bench/bench.lua gains a c-runtime column.","status":"in_progress","priority":2,"issue_type":"task","assignee":"Eugene Blikh","owner":"bigbes@gmail.com","created_at":"2026-05-17T16:32:53Z","created_by":"Eugene Blikh","updated_at":"2026-05-23T19:56:10Z","started_at":"2026-05-23T19:56:10Z","dependencies":[{"issue_id":"tarantool-protobuf-43t","depends_on_id":"tarantool-protobuf-ra6","type":"blocks","created_at":"2026-05-17T19:33:44Z","created_by":"Eugene Blikh","metadata":"{}"},{"issue_id":"tarantool-protobuf-43t","depends_on_id":"tarantool-protobuf-wky","type":"blocks","created_at":"2026-05-17T19:33:44Z","created_by":"Eugene Blikh","metadata":"{}"}],"dependency_count":2,"dependent_count":0,"comment_count":0}
+{"_type":"issue","id":"tarantool-protobuf-43t","title":"C accel: conformance and parity gate for C paths","description":"Extend tests/benches so C runtime and generated-C modes must pass the same luatest, interop, JSON/text, proto2, lazy passthrough where applicable, and Google conformance suites as Lua modes. Add bench output columns for lua-full, lua-runtime, c-runtime, c-generated, starwing. C acceleration must remain byte-equal with current encoders and preserve unknown fields/extensions/groups.","notes":"Architecture finalized in docs/c-accel.md. Parity strategy: REUSE existing test suites — no separate Lua-vs-C diff harness. (1) just test (luatest, 639 tests already parameterized over full/runtime modes) runs once with PB_ENABLE_C unset and once with PB_ENABLE_C=1. Both must pass. (2) just conformance (Google proto3 + proto2) runs both modes. (3) Interop fixtures (test/interop/fixtures/*.bin) cover both modes via the existing parity.full_vs_runtime tests; extend the matrix to also cover c-runtime. Logic: every assert is against a reference output (golden bytes, txtpb, conformance result). If Lua passes and C passes, both equal the reference, so Lua == C by transitivity. No new test infrastructure required. bench/bench.lua gains a c-runtime column.","status":"closed","priority":2,"issue_type":"task","assignee":"Eugene Blikh","owner":"bigbes@gmail.com","created_at":"2026-05-17T16:32:53Z","created_by":"Eugene Blikh","updated_at":"2026-05-23T20:04:24Z","started_at":"2026-05-23T19:56:10Z","closed_at":"2026-05-23T20:04:24Z","close_reason":"Parity gate landed. Added test-c / test-all / bench-c Justfile recipes; bench.lua relabels runtime→c-runtime under PB_ENABLE_C=1 and refuses --baseline/--compare in that mode. Reuses existing 1043-test luatest matrix, Google conformance suite, and interop fixtures via the transitivity-through-reference strategy. c-generated column dropped from scope (c0i deferred per docs/c-accel.md); starwing column lives in bench/starwing_bench.lua separately.","dependencies":[{"issue_id":"tarantool-protobuf-43t","depends_on_id":"tarantool-protobuf-ra6","type":"blocks","created_at":"2026-05-17T19:33:44Z","created_by":"Eugene Blikh","metadata":"{}"},{"issue_id":"tarantool-protobuf-43t","depends_on_id":"tarantool-protobuf-wky","type":"blocks","created_at":"2026-05-17T19:33:44Z","created_by":"Eugene Blikh","metadata":"{}"}],"dependency_count":2,"dependent_count":0,"comment_count":0}
 {"_type":"issue","id":"tarantool-protobuf-z7x","title":"C accel: C table and string handling strategy","description":"Design and implement the low-level C helpers for Lua table access and string/buffer handling. Cache field-name strings/registry refs in descriptor plans, pre-size result tables where possible, build output strings with luaL_Buffer or exact-size allocation, avoid per-field Lua stack churn, and handle 64-bit cdata consistently with the current Lua API.","notes":"Architecture finalized in docs/c-accel.md. Strategy decisions validated by bench/c_accel/ spike: (1) Field-name strings cached as luaL_ref slots at plan-compile time; each decode does lua_rawgeti from the cached ref instead of re-interning. (2) Output buffer: 4KB stack-backed cap with malloc promotion on overflow (pattern in bench/c_accel/person_codec.c — validates against tiny-msg overhead and big-msg correctness). (3) Result tables pre-sized via lua_createtable(0, n_fields) from descriptor stats. (4) Per-field stack indices cached for repeated/packed array fields for the duration of decode_message — naive lazy-getfield was 2x slower at 100KB (spike Phase B). (5) 64-bit cdata via luaT_pushuint64/luaT_pushint64 (Tarantool extensions in module.h). (6) Sub-messages use separate buf with stack-allocated header; backpatching avoided (the bench showed sub-buf approach is fast enough).","status":"closed","priority":2,"issue_type":"task","assignee":"Eugene Blikh","owner":"bigbes@gmail.com","created_at":"2026-05-17T16:32:53Z","created_by":"Eugene Blikh","updated_at":"2026-05-19T04:24:12Z","started_at":"2026-05-19T04:10:45Z","closed_at":"2026-05-19T04:24:12Z","close_reason":"C-side strategy formalized in docs/specs/c_accel_strategy.md. Spec covers: pb_plan / pb_plan_field struct layout, field-name luaL_ref caching (spike-validated), 4 KB stack-backed pb_buf with malloc promotion, lua_createtable(0, n_fields) result pre-sizing, cached per-field stack indices for repeated/packed (spike Phase B: 2x speedup at 100 KB), 64-bit cdata via luaT_pushint64/checkint64, sub-buffer (chosen) vs backpatching (deferred) for sub-messages, two-pass packed encode, map iteration via lua_next, oneof grouping, unknown field passthrough, deferred micro-opts (manual varint inlining, SIMD scan), C99/-O2 build environment, deferred PB_C_TRACE/PB_C_CHECK_PARITY env knobs. ra6/mq7 references this spec for low-level decisions.","dependency_count":0,"dependent_count":3,"comment_count":0}
 {"_type":"issue","id":"tarantool-protobuf-47e","title":"C accel: preserve pure-Lua fallback and public API compatibility","description":"Define compatibility boundaries for C acceleration. Existing generated Lua modules and runtime pb APIs must keep working without a compiler or C module. New generated-C mode should be opt-in initially. Errors, 64-bit cdata behavior, WKT shapes, unknown fields, extension representation, lazy decode API, and text/JSON integration must remain compatible.","notes":"Architecture finalized in docs/c-accel.md. Concrete contract: (1) PB_ENABLE_C=1 env var is the only activation switch, default off. (2) require('pb') returns the same Lua surface in both modes — no API change. (3) 64-bit ints stay LuaJIT cdata (int64_t/uint64_t) in both modes. (4) WKT shapes unchanged. (5) Unknown fields round-trip identically. (6) Errors land as the same Lua error types. (7) Pure-Lua install must work without a C compiler — rockspec opt-builds the C module, install never fails on a host without cc. (8) When C module fails to load or PB_ENABLE_C is unset, runtime/pb/init.lua transparently uses the pure-Lua path.","status":"closed","priority":2,"issue_type":"task","assignee":"Eugene Blikh","owner":"bigbes@gmail.com","created_at":"2026-05-17T16:32:52Z","created_by":"Eugene Blikh","updated_at":"2026-05-19T04:23:56Z","started_at":"2026-05-19T04:10:44Z","closed_at":"2026-05-19T04:23:56Z","close_reason":"Compatibility contract formalized in docs/specs/c_accel_compat.md. Spec covers: PB_ENABLE_C=1 single activation switch, silent fallback, full public surface preservation (pb.encode/decode/decode_lazy/parse/from_pb/json/text/grpc), generated module wrappers, 64-bit cdata invariant, WKT shapes, unknown fields, extensions, error compatibility, install paths (pure-Lua and C-enabled), ABI versioning, conformance via existing-suite re-run, explicit out-of-scope list. Lazy view stays Lua-only per spike rationale. Downstream tasks (mq7, ra6, 43t, c0i) reference this spec for compat boundaries.","dependency_count":0,"dependent_count":3,"comment_count":0}
 {"_type":"issue","id":"tarantool-protobuf-wky","title":"C accel: build and packaging support for C modules","description":"Add build-system and packaging support for C acceleration. Needs Justfile targets, rockspec support for compiled modules, platform naming, local dev build, CI matrix integration, and fallback when the C module is absent. Generated C backend must not make pure-Lua install impossible unless explicitly selected.","notes":"Architecture finalized in docs/c-accel.md. Concrete deliverables: (1) Justfile recipe 'just build-c' that builds the C runtime module (mirror bench/c_accel/Makefile auto-detection of TT_INC). (2) Rockspec optionally builds the C module — install on host without cc must succeed and produce a pure-Lua install. (3) runtime/pb/init.lua does pcall(require, 'pb.c_runtime') only when os.getenv('PB_ENABLE_C') == '1'. (4) Sourcehut CI build manifest gains one extra job that runs the full test+conformance suite with PB_ENABLE_C=1 set — single .build.yml, both modes covered on every push to master. (5) CI also runs one variant without the C module compiled to confirm pure-Lua install works.","status":"closed","priority":2,"issue_type":"task","assignee":"Eugene Blikh","owner":"bigbes@gmail.com","created_at":"2026-05-17T16:32:52Z","created_by":"Eugene Blikh","updated_at":"2026-05-19T04:24:13Z","started_at":"2026-05-19T04:10:45Z","closed_at":"2026-05-19T04:24:13Z","close_reason":"Build and packaging scaffolding landed: docs/specs/c_accel_build_packaging.md spec, runtime/pb/init.lua PB_ENABLE_C=1 opt-in pcall hook (exposed as pb.c_runtime field), Justfile build-c/clean-c recipes (stub erroring cleanly until bd-ra6 lands runtime/pb/c/), Justfile lua_cpath constant + LUA_CPATH on test/test-one, .gitignore for compiled .so/.dylib, sourcehut .builds/{pure-lua,c-enabled}.yml manifests (ubuntu/noble, parallel jobs via separate files since sourcehut has no matrix), .sourcehut/conformance.yml for manual conformance triggers. Rockspec changes deferred to first C-source landing (described in spec). All 745 tests pass with PB_ENABLE_C unset and PB_ENABLE_C=1 (silent fallback verified).","dependencies":[{"issue_id":"tarantool-protobuf-wky","depends_on_id":"tarantool-protobuf-pf6","type":"blocks","created_at":"2026-05-17T19:33:11Z","created_by":"Eugene Blikh","metadata":"{}"}],"dependency_count":1,"dependent_count":1,"comment_count":0}