Stories by matt_d

Using Task Graph Caching to Accelerate TVM Code Generation

Google's Training Supercomputers from TPU v2 to Ironwood: Five Generations

The Return of Rigorous Full-System Timing Simulation

Language integrated LLMs as an OCaml function

Using OxCaml to implement type-safe reference counting between OCaml and Python

Scalable GPU Acceleration of Scalar Functions in Analytical Databases

Compiling Strassen-Like Matrix Multiplication Algorithms to Fast CUDA Kernels

Programming Language Design and Implementation (PLDI) 2026 Live Streams

Puzzling Success of Overparameterization: Lottery Tickets or Escape Dimensions?

One More Type in the Tiny Type Theory

A Galois Field Arithmetic Primer

An O(x)Caml book that runs

Type Theory Forall #62 – Dependent Haskell – Vladislav Zavialov [video]

Trip report: June 2026 ISO C++ standards meeting (Brno, Czechia)

UnpredictaBench: A Benchmark for Evaluating Distributional Randomness in LLMs

Linear Algebra Kernels for the Age of Research

Latent learning: episodic memory complements parametric learning

NEURA: A Unified and Retargetable Compilation Framework for CGRAs

System Call Stack Alignment

Making FlashAttention-4 faster for inference

1 2 3 4 5 6 7 8 9 10