菜单
Answer questions about orkid's threading, concurrency, operation queues (OPQ), LockedResource, thread pools, futures, and task graphs. Use when the user asks about async, threading, mutexes, concurrent queues, or parallelism in orkid.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Orchestrate multi-agent engine work on orkid — decomposing milestone specs into lanes, assigning machines/branches, briefing implementer and gate-runner subagents, reviewing, adjudicating gates, and integrating. Use when coordinating subagent-based implementation of a spec/milestone rather than implementing directly.
How to author procedural content in orkid using HyperSyn (Reality Synthesizer) — a Python-as-DSL system for composing particle systems, 2D/3D procedural textures, mesh/SDF operations, and (eventually) audio/sequencer graphs into compact runtime experiences. Use when the user asks about HyperSyn, ptex2d, ptex3d, hypermesh, dataflow DSL, FXV2 codegen from dataflow, subprocess graph validation, or the hypergraph composition layer.
Operate THIS machine as an obtnet fleet node — launching the node daemon correctly, what the controller may do to this machine (jobs, builds into the staging, checkout gitsync/sync), node-side etiquette and troubleshooting. Use when setting up, running, or debugging the node side of the fleet (the controller-side skill is `obtnet`).
Drive the obtnet fleet (distributed build/run/sync across machines) via obt.net.py — node routing, async jobs, artifact fetch, git-free tree sync. Use whenever work should run on another machine or files must move between machines without git.
Answer questions about orkid's OBT build system, ork.build.py, CMake structure, module layout, ork.python environment, staging directories, pyext builds, and environment variables. Use when the user asks about building, compiling, CMake, or project structure.
Answer questions about orkid's ComputeInterface (CI), compute shader dispatch, indirect dispatch, SSBO binding for compute, memory barriers, image/sampler binding, dispatch phases. Use when the user asks about compute shaders, GPU compute, or dispatch.
| name | orkcore-threading |
| description | Answer questions about orkid's threading, concurrency, operation queues (OPQ), LockedResource, thread pools, futures, and task graphs. Use when the user asks about async, threading, mutexes, concurrent queues, or parallelism in orkid. |
| user-invocable | false |
When answering questions about threading or concurrency in orkid, consult the source files listed below. All paths are relative to the orkid repo root.
| Component | Header | Implementation |
|---|---|---|
| Operation Queue (OPQ) | ork.core/inc/ork/kernel/opq.h | ork.core/src/kernel/opq.cpp |
| Thread | ork.core/inc/ork/kernel/thread.h | ork.core/src/kernel/thread.cpp |
| Mutex / LockedResource | ork.core/inc/ork/kernel/mutex.h | (header-only) |
| Concurrent Queues | ork.core/inc/ork/kernel/concurrent_queue.h | (header-only templates) |
| Ring Buffer (MPMC) | ork.core/inc/ork/kernel/ringbuffer.hpp | (header-only template) |
| Thread Pool | ork.core/inc/ork/kernel/thread_pool.h | ork.core/src/kernel/thread_pool.cpp |
| Future | ork.core/inc/ork/kernel/future.hpp | ork.core/src/kernel/future.cpp |
| Task Graph | ork.core/inc/ork/kernel/taskgraph.h | ork.core/src/kernel/taskgraph.cpp |
| Semaphore | ork.core/inc/ork/kernel/semaphore.h | ork.core/src/kernel/semaphore.cpp |
| Atomic | ork.core/inc/ork/kernel/atomic.h | (header-only wrapper) |
OperationsQueue — dynamic thread pool with enqueue/drain/syncopq::concurrentQueue(), opq::mainSerialQueue(), opq::updateSerialQueue(), opq::ioQueue()ConcurrencyGroup — subdivide OPQ with max-inflight limits; MakeSerial() for serial executionCompletionGroup — batch operations and join() to wait for allLOW_LATENCY (5us), BALANCED (25us), IO (2500us)LockForRead(), LockForWrite(), UnLock()atomicOp(lambda) — execute lambda under lockatomicCopy(), atomicExchange() — convenience methodsSPSCQueue<T,N> — lock-free single-producer single-consumer, fixed capacityMpMcBoundedQueue<T,N> — multi-producer multi-consumer via MpMcRingBuf (Vyukov algorithm)push/pop) and non-blocking (try_push/try_pop)TaskGraph — DAG of phases with async executionTaskPhase — collection of tasks sharing an executorTaskExecutor::createOPQParallel() — parallel via OPQTaskExecutor::createSerial() — sequential executionLockedResource<VarMap>signal(result), waitForSignal(), isSignaled()svar160_t (160-bit variant)opq.h for the API, opq.cpp for thread management detailsLockedResource or atomicOp patternsconcurrentQueue() for CPU work, ioQueue() for I/O, serial queues for ordered work