Kernel Performance Optimization on Graph and ML Kernels Aggregate 1.0 (Summary)
43.25SpeedupBest LLM
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| Best LLMEvaluation Approach=Best performing configuration per kernel2025.12 | 43.25 | 92 | 95 | 5.2 | — | |
| LLM AverageEvaluation Approach=Average across all kernels2025.12 | 6.81 | 85 | 82 | 15.6 | — | |
| Traditional AverageEvaluation Approach=Average of traditional compiler baselines2025.12 | 6.45 | 81 | 100 | 2.1 | — |