From 1ada7c2af04a705c66c688b1ff45c412acb5a766 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ingmar=20L=C3=BCbbe?= Date: Sun, 25 Oct 2026 16:39:14 +0100 Subject: [PATCH] Issue #2936: gpu-pipeline.md says what is measured about the remaining kernel time Co-Authored-By: Claude Opus 5.5 Claude-Session: https://claude.ai/code/session_018QWU3BMJMENFqFqxDHzV3b --- docs/gpu-pipeline.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/docs/gpu-pipeline.md b/docs/gpu-pipeline.md index ecf52603f..a6d78a4e1 100644 --- a/docs/gpu-pipeline.md +++ b/docs/gpu-pipeline.md @@ -306,7 +306,7 @@ On Windows / DX12 the kernels cost more: FXC compiles each pipeline single-threa `create_compute_pipeline` (issue #471). Built one after another, the 27 took ~2.2 s on an Intel Core Ultra 7 165U iGPU (`app.gpu` 2.4 s median, up to 7 s on the first run after an install), so `ctx::build_kernels` shares them out to worker threads (issue #2936): ~1.05 s there (`app.gpu` 1.2 s), -now bounded by the largest kernel (`finish.wgsl`). wgpu's error scopes are per thread, so every worker +most likely the largest kernel (`finish.wgsl`) alone (not timed per kernel). wgpu's error scopes are per thread, so every worker opens its own; a refused kernel still means no GPU. The `gpu:` info log line gives the kernel count and build time.