warpfront
- 9 skills
- 0 followers
- 9 hours ago last updated
- ▌ Astrea · warpfront bundleUse for hipfire quant calibration, imatrix-driven experiments, KLD/PPL quality evaluation, k-map/format selection, MQ/HFQ/HFP/MFP tradeoff work, ParoQuant-style weight transform planning, and KV policy planning. Use when deciding whether a calibrated model candidate should be promoted, rejected, packaged, or sent through Atlas for AR/DFlash perf validation.
- ▌ Hipfire Diag · warpfront bundleRun and interpret hipfire GPU diagnostics for ROCm/HIP bring-up, missing kernels, test_kernels failures, inference smoke failures, and install/runtime environment problems. Use when a user asks to diagnose hipfire, check GPU readiness, run baseline tests, or explain diagnostic output.
- ▌ Hipfire Tester · warpfront bundleGuide a tester through hipfire bring-up, serve smoke, claim-scoped harnesses, and benchmark reporting on AMD RDNA/CDNA GPUs. Use when the user wants a route-oriented test plan or an upstream-ready tester report. Retired batteries are not current evidence — use [`docs/VALIDATION.md`](../../../docs/VALIDATION.md).
- ▌ Hipfire Autoheal · warpfront bundleTriage and repair hipfire runtime failures such as daemon hangs, stale serve.pid, port 11435 conflicts, ROCm include-path problems, missing precompiled kernels, VRAM OOM, kernel JIT failures, and multi-turn recall regressions. Use after diagnostics identify a likely runtime issue or when the user asks to fix a broken hipfire serve/run flow.
- ▌ Hipfire Arch Port · warpfront bundlePort hipfire compute kernels to a new RDNA / CDNA architecture (gfx1201/gfx1200/gfx94x/gfx1150/etc.). Use when adding support for a new GPU arch, fixing arch-specific kernel codegen failures (e.g. "Cannot select intrinsic %llvm.amdgcn.wmma..."), or refactoring rdna-compute arch-conditional dispatch. Captures the workflow for WMMA/MFMA ports, ArchCaps routing, channel/speed validation selection, contributor onboarding, and known correctness traps. Triggers on phrases like "port to gfx12", "9070 XT support", "R9700 support", "WMMA gfx12", "Cannot select intrinsic wmma", "amdgcn.wmma", "new arch port", "cross-arch kernel".
- ▌ Rebase Onto Modular · warpfrontUse when porting a hipfire feature/fix branch authored against pre-0.1.20 master onto post-modular master. Walks through the engine→hipfire-runtime + per-arch-crate split mechanically, then surfaces semantic conflicts that need human judgment.
- ▌ Hipfire Kernel Atlas · warpfront bundleUse Kernel Atlas to collect phase-aware hipfire measurements and render ISA Fit View visualizations for AMD GPU kernels, quant formats, and architectures. Use when a user asks how MQ/HFQ/HFP/Q8 quants occupy hardware, asks for an ASCII ISA visualization, wants to compare gfx1010/gfx1030/gfx11/gfx12 kernel fit, or wants an agent-readable "left on table" summary from Atlas rows. Atlas output is measurement evidence only — never runtime or promotion proof.
- ▌ Hipfire Kernel Tuning · warpfront bundleOptimize hipfire HIP/compute kernels — pick one tuning lever (multi-row, K-tile depth, prefetch, wave-size port, WMMA/MFMA, fused projections, ISA flags) and validate it with profile → ISA → fresh-process measurement. Use when a hot kernel is identified, you want a real perf win, and you must not regress adjacent archs or promote unverified deltas. Codifies the methodology from this repo's perf history (wave64 CDNA3 port 4105035, nontemporal-load revert 34eb024, gfx12 WMMA PR
- ▌ Redline Retained Replay · warpfrontDiscovery hook for retained AQL/PM4 (Redline) work. Use when admitting a model to retained replay, changing recorder/tape/PM4/queue policy, model reset or fallback, kernels on a retained route, or making a Redline-attributed bench claim. Routes to docs/REDLINE.md and docs/VALIDATION.md — does not fork procedure.