Paste a shader to see the patterns that slow it down on Apple GPUs, each with its fix. It runs in your browser; your shader never leaves this page.
The source alone shows patterns, not costs. Render passes, memory traffic and how long each pass takes need a capture of a frame: Ray Kiln Trace's frame analysis, coming to Ray Kiln Plus, reads one with Apple's own tools and finds loads, stores and memory nothing needs.
Vertex, fragment and kernel functions, structs, attributes and template types. Macros aren't expanded and templates are skipped; a part it can't read is marked, and the rest is still checked.