Infrastructure & Type Systems
GPU-native type systems, unified memory architectures, compiler synthesis, and declarative UI without runtime.
III: A GPU-Native Type System for Knowledge Graph Inference
Knowledge graph inference systems face a fundamental architectural tension: the semantic richness required for meaningful entity relationships conflicts with the computational demands of real-time traversal. Current approaches—SPARQL engines, Cypher interpreters, and various graph databases—execute on CPUs, accepting the inherent serialization overhead as unavoidable. This paper presents III (pronounced "triple-I"), a domain-specific language and type system designed for GPU-native knowledge graph operations.
From Swift to C: Architectural Patterns for Declarative UI Frameworks Without Runtime Dependencies
The declarative paradigm for user interface construction, exemplified by Apple's SwiftUI framework, represents a significant advance in how programmers express graphical layouts. However, these modern frameworks carry substantial runtime dependencies: garbage collection, dynamic dispatch, and platform-specific language runtimes. This paper examines the Chasm framework, an experimental effort to express SwiftUI's declarative component model in pure C, eliminating runtime dependencies while preserving the essential character of the API.
Unified Memory as Development Cache: Exploiting NVIDIA Grace-Hopper Architecture for 10-30x Faster File Operations
Software development workflows suffer from a fundamental information-theoretic bottleneck: the latency of storage I/O destroys developer cognitive flow. Traditional SSD-based file operations introduce 2-50ms latency per access, causing working memory decay and context-switching overhead that compounds across thousands of daily operations. We present a system architecture that exploits NVIDIA Grace-Hopper's unified memory to eliminate this bottleneck, achieving validated speedups of 7.5x to 33.5x across common development operations.
C Compiler Architecture: Synthesized Reference
Modern C compilers have evolved from simple translators into sophisticated optimization engines. The key architectural shift has been toward **modular, retargetable designs** with multiple intermediate representations enabling optimization at appropriate abstraction levels. **Key Insight**: No single compiler dominates performance across all workloads.