DITRON: Distributed Multi-level Tiling Compiler for Parallel Tensor Programs
2026-05-06T08:52:04Z•4780efa510dd76de110a1913745714e6e083871c22c1aba0ead3197262547281
AI-governanceAMDDITRONGPUHoare-logicLLM-inferenceNVIDIAPact-choreographyRocqagentic-systemscertified-puritycompilercryptographic-attestationcyber-physical-systemsdecidabilitydefinitional-interpreterdistributed-tensor-compilerformal-verificationgovernance-operatorinteraction-treesinteractive-theorem-provermedical-device-safetyneural-network-verificationperformancesemantic-transparency
What happened
Collection of programming-languages and systems research focused on high-performance distributed tensor compilation, formal verification, and governance for AI workflows. Key contributions: DITRON, a hierarchical tile-level compiler that matches or exceeds expert-tuned CUDA libs and delivers significant GPU-hour savings and inference/training speedups across NVIDIA and AMD; Pact, a choreographic language that models agent self-interest via game-theoretic mapping for multi-party protocols; advances in Hoare logics and definitional interpreters (including dynamic frames, behavioral subtyping, &/
Why it matters
A reviewed impact interpretation has not been published for this record.
Evidence and limitations
- Source ID
- arxiv_cs_pl
- Record identifier
- 4780efa510dd76de110a1913745714e6e083871c22c1aba0ead3197262547281
- Enrichment time
- 2026-05-06T08:52:04Z
- AI-assisted enrichment
- Yes
This record may overlap with other records. Its enrichment can be incomplete or wrong, and machine assistance was used. Validate consequential decisions against the linked source and your own environment.