trunk/c96c0d945cc30cd4317ca02911049a75e43972b7: flex_gemm: separate layout recognition from reduction analysis (#198421)
- PyTorch 近 90 天出现 1672 次
- PyTorch 近 90 天第 1653 次发版
- 上一次:同一天稍早 · ciflow/torchtitan/199089
发生了什么
AI-assisted summary of the author-requested refactor-only base: Move grouped-view recognition and main-output contraction matching into gemm_epilogue_layout.py . Keep reduction dataflow and output ownership in gemm_epilogue_analysis.py ; codegen and the existing layout tests import recognition from its canonical owner. This is a behavior-preserving refactor. Existing recognition bodies, supported expressions, guards…
摘要按规则整理自下方来源原文