viable/strict/1787680605: [inductor] Vendor and adapt QuACK symmetric GEMM (#194177)
- PyTorch: 49 events in the last 90 days
- PyTorch: 49th Release in the last 90 days
- Previous: earlier the same day · trunk/497054bc5632075da7cadbd6a8e14a49e8b534f7: Add XPU dispatch for linalg.polar (#188253)
What happened
Human commentary from the author: "and add architecture descriptions and examples to each PR to make it clear" AI-assisted change description: Summary A product such as X @ X.T is symmetric, so a conventional GEMM computes two equal output triangles. Vendor the existing QuACK symmetric GEMM from the pinned upstream revision and adapt it to PyTorch runtime interfaces. The kernel schedules one triangle and writes each…
Summary assembled by rule from the sources below