CUDA Programming

Constant Memory

Read-only memory optimized for broadcast accesses where all threads read the same value simultaneously. It is particularly efficient when all threads in a warp access the same address.

← Quay lại