Alibaba

Guide MNN Metal backend development and optimization, from finding performance bottlenecks to changing GPU computation, combining operators, and adjusting execution scheduling. A task-based document index directs work toward relevant implementation rules, benchmarks, and correctness checks.

中文简介正在准备中。当前显示英语简介。

• Finding bottlenecks in Metal inference
• Adding or modifying Metal kernels
• Combining operators into one GPU dispatch

Run MNN checks and regression tests on a host computer or Android device, with stage summaries and failure logs. A separate workflow benchmarks LLM prompt processing and token generation on real iPhones and iPads, including branch comparisons.

中文简介正在准备中。当前显示英语简介。

• Run static checks and host regression tests
• Verify selected backends on an Android device
• Benchmark LLM speed and compare branches on iPhone or iPad

Find the right workflow for MNN CPU work, from investigating slow execution to implementing calculation routines and planning tests. Compare ARM, x86_64, and RISC-V backend structures before choosing the relevant guidance.

中文简介正在准备中。当前显示英语简介。

• Choosing where to start an MNN CPU performance investigation
• Finding guidance for CPU kernel implementation
• Comparing ARM, x86_64, and RISC-V backend structures

Add new operators to MNN through a staged workflow covering data definitions, output shapes, computation, and tests. It prioritizes combining existing operators, with CPU implementations and additional hardware backends used when needed.

中文简介正在准备中。当前显示英语简介。

• Adding a new operator to MNN
• Implementing output shape calculations
• Combining existing operators for new operations

Review lessons from difficult tasks, debugging, and failed attempts, then turn them into reusable guidance. Update relevant skill instructions or shared memory so future tasks can benefit from what was learned.

中文简介正在准备中。当前显示英语简介。

• Preserving lessons learned from debugging
• Improving skill instructions after repeated failed attempts
• Turning mistaken assumptions into reusable guidance

Improve MNN inference on OpenCL by measuring bottlenecks and refining GPU kernels, the code that performs calculations. The skill guides baseline measurements, correctness checks, Android device benchmarks, and integration of new OpenCL features with compatibility fallbacks.

中文简介正在准备中。当前显示英语简介。

• Find bottlenecks in a model's GPU inference
• Optimize a specific operator's GPU kernels
• Integrate an OpenCL extension with a fallback

Optimize and troubleshoot MNN's Hexagon DSP backend, which runs computation on a dedicated processor. The workflow uses device measurements to identify bottlenecks, compare v79 and v81 architectures, verify correctness, and investigate crashes or memory usage.

中文简介正在准备中。当前显示英语简介。

• Finding bottlenecks in Hexagon operators
• Checking changes on v79 and v81 devices
• Investigating cDSP crashes and device disconnects

Add support for a new language model in MNN, from downloading and analyzing it to mapping model fields, comparing intermediate results, and exporting files. Follow staged tests, with additional paths for image, audio, and unusual architectures.

中文简介正在准备中。当前显示英语简介。

• Adding a new text model to MNN
• Exporting a model from HuggingFace or ModelScope to MNN
• Checking intermediate outputs against the original model

Triage incorrect MNN outputs and regressions by matching symptoms to focused debugging guides. Use backend comparisons, minimal reproductions, and controlled experiments to investigate memory issues, quantization, concurrency, numeric limits, caches, and inconsistent results across runs.

中文简介正在准备中。当前显示英语简介。

• Investigating garbled or degraded MNN output
• Comparing results across execution backends
• Tracking regressions after optimization changes

Find why MNN CPU inference is slow by examining threads, scheduling, memory layout, and the processor instruction path actually used. Operator-level measurements and comparisons with other inference frameworks help identify which layer needs improvement across ARM, x86_64, and RISC-V.

中文简介正在准备中。当前显示英语简介。

• Investigate slow CPU operators or LLM inference
• Diagnose thread count and memory usage anomalies
• Locate the cause of a performance regression

Diagnose errors and incorrect inference results in MNN's QNN backend for Qualcomm HTP processors. Compare intermediate calculation data against CPU results to locate the first faulty operator, distinguish bugs from precision limits, and guide operator support and performance fixes.

中文简介正在准备中。当前显示英语简介。

• Investigate QNN runtime and model conversion errors
• Find the first operator producing incorrect results
• Distinguish implementation bugs from numerical precision limits

Use a defined database structure and business rules to query products, warehouses, stock levels, and stock movements. It supports inventory checks and SQL for identifying products whose combined warehouse stock has reached the reorder threshold.

中文简介正在准备中。当前显示英语简介。

• Checking stock quantities by product and warehouse
• Finding products at or below their reorder threshold
• Understanding records for stock movements

Use defined customer and order tables to build sales analysis queries with consistent business rules. Calculate revenue from completed orders, rank customers by spending, and identify high-value orders using the supplied thresholds and SQL example.

中文简介正在准备中。当前显示英语简介。

• Ranking the top ten customers by revenue over three months
• Calculating revenue from completed orders
• Finding orders with a total amount above 1,000

Code review behavioral standards. Reference during code review to ensure consistent quality checks from a QA perspective.

中文简介正在准备中。当前显示原文说明。

自监控指标、告警代码标准

中文简介正在准备中。当前显示原文说明。

This workflow, known as RIPER-5, is designed exclusively for handling complex software engineering tasks such as system design, architectural refactoring, bug diagnosis, performance optimization, and multi-component integration.

中文简介正在准备中。当前显示原文说明。

OMC agent catalog, available tools, team pipeline routing, commit protocol, and skills registry. Auto-loads when delegating to agents, using OMC tools, orchestrating teams, making commits, or invoking skills.

中文简介正在准备中。当前显示原文说明。

在进行 Code Review 时，使用这个技能对 LoongCollector 变更进行安全导向、架构一致性优先的深度代码评审。

中文简介正在准备中。当前显示原文说明。

LoongCollector project knowledge: architecture, terminology, codebase map, and coding standards (C++/Go).

中文简介正在准备中。当前显示原文说明。

按 Conventional Commits 规范编写提交信息。在创建 git commit 时使用。

中文简介正在准备中。当前显示原文说明。

LoongCollector E2E Testing Framework Development Guide. Use this rule when you need to understand E2E testing framework architecture, develop new test behavior functions, or extend testing capabilities. Contains detailed BDD testing framework usage, test behavior function development and registration workflows, and comprehensive reference of existing testing capabilities. Example: When adding custom test behaviors for new plugins, use this rule to understand how to write and register new test functions. Example: When developing E2E tests, use this rule to understand the e2e framework.

中文简介正在准备中。当前显示原文说明。

Whenever diagrams or mermaid is mentioned

中文简介正在准备中。当前显示原文说明。

Testing standards for LoongCollector: unit tests, e2e tests, benchmarks. Reference when writing or reviewing tests.

中文简介正在准备中。当前显示原文说明。

LoongCollector E2E 测试全流程指南：设计、编写、运行和调试。当需要编写新 E2E 测试、运行现有测试、或排查 E2E 测试失败时使用此 skill。

中文简介正在准备中。当前显示原文说明。