Direct kernels
Broader topicGPU workflows
Prepare resident simulations, manage transfers and checkpoints, or use direct kernels.
Follow the science
Direct kernels
Control individual CUDA process calls and explicit transfers.
Topics
Direct kernels expose the boundary between host state, device buffers, and individual aerosol processes. Each call needs compatible configuration and capacity; a successful launch alone does not establish physical agreement.
A question to investigate
How does a direct condensation or coagulation step compare with an independent reference at matched initial conditions?