How a single SM actually runs work: warps of 32, cutaway of cores and memory, divergence tax, latency-hiding occupancy, and coalesced loads — instruments, not a catalog.

Streaming Multiprocessor deep dive: SIMT warps, CUDA and Tensor Cores, warp divergence, latency hiding, occupancy tradeoffs, and coalesced global memory on modern NVIDIA GPUs.

GPU Streaming Multiprocessor (SM)

A fine kernel can still profile poorly: warp divergence taxes throughput ~1/N; alone on a long scoreboard wait, SM issue util can hit 0%. Interactive demos.

Warps

Concepts Related to Warps

GPU Streaming Multiprocessor (SM)

GPU Pipeline Hazards: Scoreboard Stalls and Warp Divergence