developer tools
A GPU's Special Functions Are Rebuilt Bit for Bit
SFU-Sim reverse-engineers NVIDIA multifunction instructions and reproduces tested PTX paths across three GPU architectures without a lookup table.
Summary
SFU-Sim reverse-engineers NVIDIA multifunction instructions and reproduces tested PTX paths across three GPU architectures without a lookup table.
SFU-Sim models the exact numerical behavior of NVIDIA multifunction-unit instructions using black-box measurements from real GPUs. The work first reconstructs a bare instruction, then adds the surrounding operations observed in compiled PTX-to-SASS paths. For tested FP16 and FP32 sine and cosine paths on Ampere, Ada Lovelace and Blackwell GPUs, exhaustive comparison across the specified input domains produced zero bit mismatches. The open-source model computes outputs directly and is intended as a reference component for instruction simulators and hardware verification.
Why it matters
SFU-Sim reverse-engineers NVIDIA multifunction instructions and reproduces tested PTX paths across three GPU architectures without a lookup table.
Limits and context
No additional limitation was separately recorded.
Key claims
SFU-Sim reverse-engineers NVIDIA multifunction instructions and reproduces tested PTX paths across three GPU architectures without a lookup table.
Evidence: source-2026-11-12-015
Sources
- arXiv:2611.13879v1arXiv / Cornell University · primary research
Corrections
No corrections have been recorded for this story.