Link measured Forward+ follow-up from baseline study

docs/studies/22-p3-lighting-benchmark-2026-09-24.md

@@ -46,4 +46,4 @@ With shadows on, four point lights request 24 faces but only 12 fit atomically i
 
 `gpu_ms` includes the renderer's synchronous framebuffer capture copy, while `cpu_ms` includes submission and wait. Neither is an interactive game frame rate. The gate deliberately uses GPU raster timestamps; the post-occlusion raster pass is included when active. The fixed geometry and light arrangement, GPU clock state, and one driver limit generalization to other scenes and devices. The retained raw data allow this conclusion to be recomputed without trusting the prose.
 
-The next P3 checkpoint must implement depth-free 16 × 16 tiled Forward+ with a full-light fallback for overflowing tiles, compare rendered RGB against the current full-scan path, and remeasure tile-build **plus** raster time on this same workload. The simple path remains available as a correctness reference until that result is recorded.
+The [Forward+ follow-up](23-p3-forward-plus-2026-09-24.md) implements depth-free 16 × 16 tiles with a full-light fallback for overflowing tiles, compares rendered RGB against the full-scan path, and records tile-build **plus** raster time. Its dense-scene result keeps the simple forward path as the default; localized lights are measured separately.