Interactive simulators that I host here in case others find them useful. Each one is a single web page that animates a piece of hardware or an algorithm step by step.
Simulators
CUDA Tiled GEMM Visualizer
An animated walk through a tiled GEMM (general matrix multiply) CUDA kernel computing C = A × B for 16×16 matrices. The output is split into 2×2 tiles, and each tile is computed by one thread block (a group of GPU threads that can share fast on-chip memory), giving 64 blocks of 4 threads. Click any block of C to watch it run through 8 phases, one per 2-wide slice of the shared dimension. The simulator was inspired by Zachary Huang's video Give Me 30 min, I'll Make CUDA Click Forever, whose load, synchronize, compute pattern it animates, and its visual style is borrowed from Addy Osmani's X post showing how browsers work in a 40-second animation drawn frame by frame in JavaScript (which he AI generated with Opus 5.5 as did I).
single web page, no install · tiled or naive kernel with configurable grid and tile size · play, pause, step, or jump to the end at 0.25× to 3× speed · kernel source with the active line highlighted