{
  "id": "okf_xinetzone_specweave_bundles_chaos_apache_53cbc38e",
  "kind": "okf",
  "category": "dados",
  "name": "Apache TVM 深度学习编译器",
  "tagline": "Apache TVM 四层栈架构知识包，涵盖 FFI 基础设施、TIR 张量 IR、Relax 图级 IR、Runtime 执行引擎、MetaSchedule 自动调度及 LLM 推理支持",
  "body": "# Apache TVM 深度学习编译器\n\nApache TVM 是一个开源的深度学习编译器栈，通过四层架构（FFI 基础设施 → TIR 张量 IR → Relax 图级 IR → Runtime 执行引擎）实现深度学习模型的跨硬件高性能编译与部署。本知识包基于 TVM 源码（版本 0.26.dev0）整理，包含 22 篇概念文档、实践示例和完整的事实参考。\n\n## 概念文档\n\n### 第一批：基础架构\n\n- [00 架构总览](concepts/00-overview.md) — TVM 四层栈架构与编译流水线\n- [01 FFI 基础设施](concepts/01-ffi-foundation.md) — 跨语言函数调用、C ABI 与类型系统\n- [02 Object 对象系统](concepts/02-object-system.md) — 引用计数、容器系统与反射\n- [03 Pass 基础设施](concepts/03-pass-infrastructure.md) — PassContext 与 IRModule 变换框架\n- [04 Target 与代码生成](concepts/04-target-codegen.md) — 多后端描述与 LLVM/C 代码生成\n\n### 第二批：TIR 与调度\n\n- [05 TIRx 中间表示](concepts/05-tirx-ir.md) — 表达式/语句节点、SBlock、PrimFunc\n- [06 Buffer/Var/IterVar 核心类型](concepts/06-buffer-var-itervar.md) — 内存布局、变量系统与迭代空间\n- [07 SBlock 声明式调度](concepts/07-sblock-schedule.md) — Schedule 类、RV 体系与 Trace\n- [08 调度原语](concepts/08-schedule-primitives.md) — 循环变换、computeAt、缓存、张量化\n- [09 MetaSchedule 自动调度](concepts/09-meta-schedule.md) — 搜索策略、CostModel 与 dlight GPU 调度\n- [10 Arith 整数分析器](concepts/10-arith-analyze",
  "url": "https://github.com/xinetzone/SpecWeave/blob/HEAD/bundles/chaos/apache-tvm/index.md",
  "status": "live",
  "origin": "okf_github",
  "origin_id": "xinetzone/SpecWeave:bundles/chaos/apache-tvm/index.md",
  "install": "https://raw.githubusercontent.com/xinetzone/SpecWeave/HEAD/bundles/chaos/apache-tvm/index.md",
  "source": "https://github.com/xinetzone/SpecWeave",
  "transporte": "",
  "ns": "xinetzone",
  "versao": "0.2",
  "oficial_status": "",
  "repo_host": "github.com",
  "topico": "ai",
  "linguagem": "Python",
  "licenca": "Apache-2.0",
  "repo_topics": " agent-specification agentspec ai-agents ai-collaboration coding-standards development-framework mermaid multi-agent-systems specification workflow-automation ",
  "repo_slug": "xinetzone/specweave",
  "stars": 1,
  "forks": 1,
  "prs_abertos": 0,
  "pushed_at": "2026-08-30T02:46:24Z",
  "repo_estado": "ativo",
  "likes": 0,
  "comments": 0,
  "visits": 0,
  "readme_bytes": 8676,
  "readme_api": "https://agentalog.com/api/listings/okf_xinetzone_specweave_bundles_chaos_apache_53cbc38e/readme",
  "readme_fonte": "repo",
  "created_at": "2026-09-09 12:07:44",
  "updated_at": "2026-09-09 12:07:44",
  "mine": false,
  "api": "https://agentalog.com/api/listings/okf_xinetzone_specweave_bundles_chaos_apache_53cbc38e",
  "go": "https://agentalog.com/api/go/okf_xinetzone_specweave_bundles_chaos_apache_53cbc38e",
  "comments_api": "https://agentalog.com/api/listings/okf_xinetzone_specweave_bundles_chaos_apache_53cbc38e/comments",
  "readme_sha": "681c7794528c23d5a73c2c2f70a7cc42f2fc034f64f508cbc1faf561e0a7da81",
  "readme_em": "2026-09-14 00:36:18",
  "payload": {
    "name": "xinetzone/SpecWeave:bundles/chaos/apache-tvm/index.md",
    "title": "Apache TVM 深度学习编译器",
    "version": "0.2",
    "okf_bundle_dir": "bundles/chaos/apache-tvm",
    "okf_index_path": "bundles/chaos/apache-tvm/index.md",
    "okf_conceitos": 34,
    "okf_ref": "HEAD",
    "okf_type": "Index"
  },
  "volatil_em": "2026-09-13 23:02:16"
}