2 Commits

Author	SHA1	Message	Date
modelbeast	df28940d81	docs: root-cause the sdpa finding — head_dim=56 misses the fast path ... Shape-spy on a real 2048 forward + exact-shape sweeps on M1/M3 Ultra: fused==unfused at hdim 56 (silent fallback) on both machines, so use_sdpa buys nothing and its transpose choreography costs ~4.5x on M1. Padding to hdim 64 engages the fast kernel: 2.2x (M3) / 2.7x (M1) on the dominant global-attention shape despite 14% extra FLOPs.	2026-07-16 23:05:13 +10:00
modelbeast	a9f6b75ef2	docs: M-series fleet ablation (M3 Ultra / M1 Ultra / M4 Pro 24GB) ... Ran scripts/bench_optimizations.py unmodified across three machines + an engine-level tiled×compile extension with ground-truth alpha scoring. Headlines: sdpa = 4.5x regression on M1 at 2048 (bf16 innocent — isolated per-toggle on both Ultras); stage_gc harmful on Ultra-class; tiled 768/64 + compile = best 2048 config on both Ultras; 24GB Macs swap-bound at 2048 full-frame (tiled required).	2026-07-16 22:48:06 +10:00