Max Pooling 3D on CUDA-LLM kernels task suite
5.62Time (s)OptiML
Evaluation results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| OptiMLOptimization=G+X, Hardware=A100 80GB2026.02 | 5.62 | 1.08 | 30.71 | 29.18 | 37.55 | 494 | 1,863 | 801 | |
| HPC-Coder-V2 + OptiML-XOptimization=OptiML-X, Hardware=A100 80GB2026.02 | 5.88 | 1.04 | 29.02 | 30.41 | 37.11 | 508 | 1,891 | 809 | |
| StarCoder2 + OptiML-XOptimization=OptiML-X, Hardware=A100 80GB2026.02 | 6.02 | 1.01 | 27.61 | 31.18 | 36.02 | 532 | 1,978 | 831 | |
| GPT 5.1 + OptiML-XOptimization=OptiML-X, Hardware=A100 80GB2026.02 | 6.09 | — | 28.14 | 31.47 | 36.72 | 524 | 1,944 | 824 | |
| Qwen2.5-Coder + OptiML-XOptimization=OptiML-X, Hardware=A100 80GB2026.02 | 6.38 | 0.95 | 26.88 | 32.12 | 35.31 | 548 | 2,032 | 843 |