Geometric
Evolutionary algorithms for optimised CUDA kernels.
Posts
Sol-Shock: Hacking your way to the Speed of Light on Solexecbench
Dense Matmuls are faster when the inputs are sparse
Kernel Swapping Affects Model Performance
Evolving GPU Kernels with LLMs
Diversity Is All You Need (To Converge): Why Evolutionary Algorithms Need Diversity Management
AST Edits: The Code Editing Format Nobody Uses
Are you using the right baseline for your kernel benchmarks? An analysis of
torch.compile
modes
Open-sourcing kernels on the HuggingFace Hub
