graph-processing

Input-aware CUDA auto-scheduler for CSR SpMM, SDDMM, and sparse attention in graph neural networks.

Exploring computational scaling of spiking neural network simulations across hardware architectures

CUDA/PyTorch operator for fused neighborhood sampling and mean aggregation, with reproducible KSEM 2026 experiments.

Official CUDA implementation of "Streaming Right Multiplication over Grammar-Compressed Matrices: A Memory-Bounded GPU Engine for Genotype and Graph Data" (ALENEX 2027). Enables memory-bounded matrix-vector products and graph algorithms on massive dataset via grammar compression.

A Python wrapper over the GraphGen system