Tutti Logo
0.1

Contents:

  • Tutti 入门指南 / Getting Started with Tutti
  • Architecture
  • Design Documents
    • Tutti 多加速器、多盘与 NVMe daemon 设计
    • Tutti Runtime 配置、资源与装配流程
    • Kernel Portability — the snvme module across Linux versions and GPU vendors
    • Backend SPI — the DataPath / Resolver / Binding contracts
  • Tutorials and Integration Guides
  • Build & SNVMe Testing Guide
  • Part 1 — Build
  • Part 2 — SNVMe Testing
  • Extending Tutti: Adding a Resolver + Binding + DataPath
  • GPU Vendor Porting Guide (Tutti cuda_like Framework)
  • Tutti local-NVMe hardware contract tests
  • tutti_daemon 启动与部署指南
Tutti
  • Design Documents
  • View page source

Design Documents

  • Tutti 多加速器、多盘与 NVMe daemon 设计
    • 1. 范围与相关文档
    • 2. 当前系统模型
    • 3. 身份模型
    • 4. 多 accelerator Runtime 模型
    • 5. daemon 配置模型
    • 6. daemon 架构和启动流程
    • 7. daemon 接口及调用方
    • 8. 多盘选择、queue reservation 与共享
    • 9. Runtime 与 DataPath 的集成边界
    • 10. 生命周期与失败处理
    • 11. 当前多 accelerator 示例
    • 12. 限制与实现漂移
    • 13. 边界结论
    • 14. 代码索引
  • Tutti Runtime 配置、资源与装配流程
    • 1. 范围
    • 2. 总体模型
    • 3. 模块职责总览
    • 4. Config parser 边界
    • 5. Config spec 边界
    • 6. TuttiRuntime 创建入口与环境校验
    • 7. Resource 边界
    • 8. Resolver factory 边界
    • 9. DataPath factory 边界
    • 10. 装配为 StorageRuntime
    • 11. 所有权与生命周期
    • 12. 两个 NVMe 示例的端到端数据流
    • 13. 边界结论
    • 14. 代码索引
  • Kernel Portability — the snvme module across Linux versions and GPU vendors
    • 1. Two maintained kernel trees
    • 2. The peer_memory backend layer (GPU P2P)
    • 3. Userspace ABI handshake
    • 4. Queue geometry — controller maximum
    • 5. Interrupt semantics under GPU-polled completions
    • 6. Build & test entry points
  • Backend SPI — the DataPath / Resolver / Binding contracts
    • 1. Design rules
    • 2. The three SPI surfaces
    • 3. Runtime routing & grouping
    • 4. Capability & capacity knobs
    • 5. What a backend gets for free
Previous Next

© Copyright 2024, xPU-IO.

Built with Sphinx using a theme provided by Read the Docs.