Stories by matt_d

Inside vLLM: Anatomy of a High-Throughput LLM Inference System

simdjson Version 4.0.0 Released

Introducing BackendBench: how well LLMs and humans can write PyTorch backends

Rethinking Analytical Processing in the GPU Era

EUV: Lithography: History, Latest Results, Technology Roadmap [video]

Disaggregation: A New Architecture for Cloud Databases

GSoC 2025 – Byte Type: Supporting Raw Data Copies in the LLVM IR

The Future of Memory: Limits and Opportunities

Interposing on clone() system calls in-process, from Linux userspace

So you want to control flow in PyTorch 2

IRHash: Efficient Multi-Language Compiler Caching by IR-Level Hashing

Evolving the OCaml Programming Language (2025) [pdf]

Why ML Needs a New Programming Language – Chris Lattner – Signals and Threads

vLLM with torch.compile: Efficient LLM inference on PyTorch

DaCe AD: Unifying High-Performance Automatic Differentiation for ML and SciComp

Still Asking: How Good Are Query Optimizers, Really? [pdf]

Sphinx: A Succinct Perfect Hash Index for x86 [pdf]

DialEgg: Dialect-Agnostic MLIR Optimizer Using Equality Saturation with Egglog [video]

Dependent Types: Universes, or types of types

Translation Validation for LLVM's AArch64 Back End [pdf]

6
7
8
9
10
11
12
13
14
15