Compiler Optimization on tensorflow
0.125I_O3CFAST
Evaluation results
| Method | Links | |
|---|---|---|
| CFASTSearch Strategy=Traditional Search Heuristic2026.01 | 0.125 | |
| GRACESearch Strategy=Traditional Search Heuristic2026.01 | 0.1225 | |
| ECCOBackbone=Qwen2.5-7B, Sampling Strategy=Best-of-322026.01 | 0.1172 | |
| CompTunerSearch Strategy=Traditional Search Heuristic2026.01 | 0.1083 | |
| PDCATSearch Strategy=Traditional Search Heuristic2026.01 | 0.1033 | |
| RIOSearch Strategy=Traditional Search Heuristic2026.01 | 0.0988 | |
| OpenTunerSearch Strategy=Traditional Search Heuristic2026.01 | 0.0971 | |
| GASearch Strategy=Traditional Search Heuristic2026.01 | 0.0958 | |
| TPESearch Strategy=Traditional Search Heuristic2026.01 | 0.0907 | |
| Qwen3-CoderSearch Strategy=Direct LLM Prompting, Sampling Strategy=Best-of-322026.01 | 0.0439 | |
| GPT5-chatSearch Strategy=Direct LLM Prompting, Sampling Strategy=Best-of-322026.01 | 0.0351 | |
| Kimi-K2Search Strategy=Direct LLM Prompting, Sampling Strategy=Best-of-322026.01 | 0.0331 | |
| DeepSeek-V3.2Search Strategy=Direct LLM Prompting, Sampling Strategy=Best-of-322026.01 | 0.0301 |