Stories by matt_d

TorchLean: Verified Neural Networks in Lean

Deterministic Fully-Static Whole-Binary Translation Without Heuristics

Dynamic Persistent Tile Scheduling w/ Cluster Launch Control (CLC) on Blackwell

VibeServe: Can AI Agents Build Bespoke LLM Serving Systems?

CCL-Bench 1.0: A Trace-Based Benchmark for LLM Infrastructure

Microbenchmark-Driven Analytical Performance Modeling Across Modern GPUs

PyTorch DevLog

VDCores: Resource Decoupled Programming and Execution for Asynchronous GPU

Aurora: A Leverage-Aware Optimizer for Rectangular Matrices

The Two Abstractions of System Design: Hide or Reduce

Practical Formal Verification for MLIR Programs

Kerncap: Automated Kernel Extraction and Isolation for AMD GPUs

Capsules: Compile-time lock discipline in OxCaml

Data Race Freedom in OxCaml

cuda-oxide: a custom rustc backend for compiling GPU kernels in pure Rust

A case study with Aeneas and jxl-rs

Finite Functional Programming

CommFuse: Hiding Tail Latency via Communication Decomposition and Fusion

SPEC CPU: The Next Generation

Persistent Iterators with Value Semantics

1 2 3 4 5 6 7 8 9 10