skills.rest
openJiuwen-ai
openJiuwen-aiOFFICIAL
Explore all skills by this author →
WebsiteGitHub

ascend-moe-optimizer-auto-trace

Generate and validate TRACE_POINT instrumentation and Chrome trace workflows for Ascend MoE operators.

2.5k|422|Updated Mar 5, 2026
Search Tags:#profiling#instrumentation#ascend#moe tracing#chrome trace#point_map#compile hook
One-click install
npx skills add https://github.com/openJiuwen-ai/jiuwenswarm --skill ascend-moe-optimizer-auto-trace
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: ascend-moe-optimizer-auto-trace
Source: https://github.com/openJiuwen-ai/jiuwenswarm/tree/main/jiuwenswarm/resources/agent/workspace/skills/ascend-moe-optimizer-auto-trace
Command: npx skills add https://github.com/openJiuwen-ai/jiuwenswarm --skill ascend-moe-optimizer-auto-trace

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires torch, and includes scripts (resource) and references (resource) components.

What problem does it solve?

This Skill helps you add end-to-end TRACE_POINT instrumentation and MoE profiling capture for Ascend operators, so you can generate Chrome trace JSON that pinpoints performance bottlenecks across AIC/AIV execution paths.

Core Features & Use Cases

  • Full-chain instrumentation & profiling: inserts TRACE_POINT around real operator phase boundaries (not just shell entries), captures per-core profiling tensors, and decodes them into Chrome trace events.
  • Strict correctness guardrails (G1–G5): enforces trace preprocessor hook/point_map generation consistency, output arity and “profiling is last” ordering, compilation and example/UT synchronization + save steps, and absolute-path safety for spawned multiprocessing.
  • Toolchain deployment + compile hook integration: can deploy trace toolchain scripts and patch existing compile scripts with a preprocessor hook for the active build tree.
  • Validation workflow: runs validate_trace_points.py and check_compile_safety.py, then guides the required full compilation and profile/trace generation steps.

Quick Start

Ask for instrumentation and trace generation for an Ascend MoE operator (e.g., “Add TRACE_POINT + profiling and produce chrome_trace.json for this op”), and then follow the skill’s G1–G5 checklist, ending with running save_profiling_data after NPU synchronization and generating chrome_trace.json via trace_collector using the same build’s point_map.json.

Frequently Asked Questions about ascend-moe-optimizer-auto-trace

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I add TRACE_POINT instrumentation to Ascend MoE operators for profiling?▼

To add TRACE_POINT instrumentation to Ascend MoE operators, insert trace points around real operator phase boundaries for AIC/AIV paths, ensuring profiling tensors are captured per-core and decoded into Chrome trace events. This requires strict B/E pairing and point_map.json generation during compilation.

How does Chrome trace generation work for Ascend MoE operator profiling?▼

Chrome trace generation for Ascend MoE profiling works by capturing per-core profiling tensors from instrumented TRACE_POINT boundaries and decoding them into Chrome trace JSON events. It requires running trace_collector using the same build's point_map.json after device synchronization saves the profiling data.

Does Ascend MoE profiling require compilation hooks for point_map generation?▼

Yes, Ascend MoE profiling requires a trace preprocessor hook patched into existing compile scripts to generate point_map.json consistently. This ensures the trace_collector can accurately decode profiling tensors into Chrome trace events using the active build tree's mapping.

What is the best way to validate TRACE_POINT B/E pairing and compilation safety for Ascend operators?▼

The best way to validate TRACE_POINT pairing and compilation safety is by running validate_trace_points.py and check_compile_safety.py. These scripts enforce G1-G5 guardrails, verifying output arity, profiling-last ordering, and absolute-path safety before guiding the full compilation and trace generation steps.

Why does my Ascend MoE chrome_trace.json fail to generate after profiling?▼

Ascend MoE chrome_trace.json generation fails if profiling data is not saved after NPU synchronization or if point_map.json is missing. You must execute save_profiling_data after synchronization and run trace_collector using the exact point_map.json generated during the operator's compilation flow.

Can I use PyTorch to instrument Ascend MoE ops for Chrome trace profiling?▼

Yes, you can use PyTorch to instrument Ascend MoE ops for Chrome trace profiling, as the workflow relies on PyTorch dependencies. It inserts TRACE_POINT instrumentation across AIC/AIV execution paths and decodes captured profiling tensors into Chrome trace JSON format reliably.

Related Skills

Explore More Software Engineering Skills→
AscendAscend
31

ascend-model-migration

Migrate and train open-source models on Ascend NPU with distributed training.

Official
Advanced
openJiuwen-aiopenJiuwen-ai
2.5k

ascend-moe-optimizer-trace-analyzer

Convert Chrome/Perfetto trace.json into MoE phase timing tables and diagnosis reports.

Official
Advanced
rukikotoorukikotoo
6

stata

Execute Stata do-files via Bash and retrieve output logs.

Community
Advanced
AscendAscend
31

triton-operator-performance-eval

Diagnose Triton kernel performance bottlenecks on Ascend NPUs using msprof data.

Official
Advanced