The authors built an agentic compiler harness that asks an LLM to translate Triton kernels directly into NVIDIA’s PTX intermediate assembly.
LLM-generated PTX rivals autotuned Triton across three GPU generations
An agent translated Triton kernels directly into NVIDIA PTX, producing implementations ranging from 0.83x to 3.34x the performance of Triton’s conventional compiler.
Top University
François Costa · Charly Castes · Thomas Bourgeat · Azalia Mirhoseini
Stanford University · EPFL
Research Digest··2 min read
Costa et al.
Why this paper
From Stanford University and EPFL
In one line
An LLM agent compiles Triton kernels directly to PTX, reaching 0.83x-3.34x the performance of autotuned Triton across tested GPUs.
What we could check
- ·No code link found
- ·No weights link found
- ·No dataset link found
- ✓Compute or model size stated (hardware B200)
- ✓Limitations stated by the authors (2 noted)
- ✓Reports numbers on named benchmarks (4 benchmarks)
Observed from the paper text and links we have. Absence here means we did not find it, not that it does not exist.
§