SparseDitto uses LLM agents guided by structural matrix features and target-GPU measurements to generate custom CUDA kernels for SpMV, SpMM, and SpGEMM, beating cuSPARSE by 2.68x to 2.79x on average.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.DC 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
SparseDitto: Customizing GPU Kernels for Different Sparsity Patterns with LLM-Based Agentic System
SparseDitto uses LLM agents guided by structural matrix features and target-GPU measurements to generate custom CUDA kernels for SpMV, SpMM, and SpGEMM, beating cuSPARSE by 2.68x to 2.79x on average.