菜单
Generate and analyze T1/T2/R roofline reports for PyTorch OOB workloads comparing Intel XPU and NVIDIA CUDA. Use when working with eager profiling artifacts, per-model reports, fleet summaries, graph consistency, or XPU-vs-CUDA software efficiency analysis.
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
Analyze UT (unit test) results for a torch-xpu-ops PR. Use when asked to check test results, analyze CI failures, or evaluate test coverage for a PR. Produces a structured report of new failures, failure relevance, and new test coverage.
Rules for automatically determining which disable_* labels to apply to a PR based on the file paths changed. Used by the auto-label workflow.
Extract GPU ISA from oneDNN ngen-JIT kernels. This is the ONLY codegen path that bypasses the standard SYCL/SPIR-V/IGC stack. oneDNN uses its own native code generator (ngen) that directly emits GPU ISA bytes — no SPIR-V, no IGC, no zebin ELF, no .debug_line. Use when extracting ASM from oneDNN kernels (gemm_kernel, gen_conv_kernel), matmul, linear, conv, or SDPA-graph ops dispatched via mkldnn.
Extract GPU ISA from AOT-compiled SYCL binaries using NEO driver runtime dump. Use when the kernel is a SYCL symbol (_ZTS...) from an AOT binary such as libtorch_xpu.so, SYCL-TLA FMHA, or any DPC++ binary compiled with -fsycl-targets=spir64_gen -Xs "-device <gpu>".
Extract GPU ISA from JIT-compiled SYCL kernels. The binary only contains SPIR-V; IGC compiles it to native ISA at first launch. The zebin only exists in memory at runtime, so IGC_ShaderDumpEnable=1 is required to capture it. Use when extracting ASM from SYCL JIT kernels, torch-xpu-ops built with TORCH_XPU_ARCH_LIST=none, or standalone DPC++ without AOT flags.
Extract GPU ISA from Triton kernels on XPU. Triton compiles through the same IGC backend as SYCL (Triton IR → SPIR-V → IGC → zebin). Extraction is identical to sycl-jit: IGC_ShaderDumpEnable=1 captures the zebin at runtime. Use when extracting ASM from torch.compile fusions (triton_per_fused, triton_poi, triton_red), standalone @triton.jit kernels, or Inductor-generated XPU kernels.
| name | oob-perf-analysis |
| description | Generate and analyze T1/T2/R roofline reports for PyTorch OOB workloads comparing Intel XPU and NVIDIA CUDA. Use when working with eager profiling artifacts, per-model reports, fleet summaries, graph consistency, or XPU-vs-CUDA software efficiency analysis. |
Analyzes OOB eager-mode performance using T1/T2/R roofline methodology to compare XPU and CUDA software efficiency.
Core outputs (all under agent_space_xpu/, git-ignored):
agent_space_xpu/reports/<session>/models/agent_space_xpu/reports/<session>/summary_eager_inference.mdagent_space_xpu/reports/<session>/graph_consistency_eager_inference.mdagent_space_xpu/reports/<session>/insights_summary.md| File | Purpose |
|---|---|
| methodology.md | T1/T2/R definitions, formulas, classification thresholds |
| inputs.md | Input file formats, completeness rules, output paths |
| per-model-report.md | Per-model analysis steps and 5-section report structure |
| fleet-summary.md | Fleet aggregation steps and 7-section report structure |
| graph-consistency.md | Graph consistency analysis |
| insights.md | Developer-facing insights summary |
| troubleshooting.md | Diagnosing abnormal R, trace issues, and data problems |
User has raw artifacts and wants model-level T1/T2/R analysis.
Follow: methodology.md → inputs.md → per-model-report.md
User wants fleet-wide comparison, model scorecards, op ranking, or projection-quality aggregation.
Follow: fleet-summary.md (references graph-consistency.md for Section 7)
User wants to compare CUDA vs XPU computational graphs.
Follow: graph-consistency.md
Per-model reports and fleet summary exist; user wants a concise developer-facing summary.
Follow: insights.md
User asks why R is abnormal, why traces disagree, or why results look suspicious.
Follow: troubleshooting.md
methodology.md before computing any metric.inputs.md before accessing any artifact.If a calling workflow explicitly requires a skill marker, append this exact literal final line: Custom skills applied: oob-perf-analysis.