Louis Touzalin / at0m741

Systems notes at the edge of compilers, numerical kernels, and computational physics.

I write about low-level programming, C/C++ systems work, compiler tooling, Tensorium_lib, MLIR experiments, and numerical relativity workflows that connect tensor calculus with high-performance computation.

Read articles View projects GitHub

Current focus

Systems

C/C++ systems programming

Notes around memory layout, runtime behavior, low-level performance constraints, and code that stays close to the hardware boundary.

Compilers

MLIR, DSLs, tensor code generation

Experiments around compiler infrastructure for tensor calculus, lowering pipelines, vectorization, and backend design.

HPC

Tensorium_lib and numerical kernels

High-performance C++ tensor and linear algebra work with SIMD, OpenMP/MPI-oriented design, and computational physics constraints.

Physics

Numerical relativity from first principles

Notes on tensor calculus, general relativity, black hole simulations, elliptic systems, and spectral methods.

Latest articles

2026-06-19 · EN

Tensorium_lang: compiler-visible tensor calculus for numerical relativity

A web version of the Tensorium_lang paper: an experimental MLIR-based DSL for preserving tensor structure through compiler lowering and generated numerical kernels.

Tensorium_lang MLIR DSL numerical relativity compiler infrastructure

Project index

Active

Tensorium_lib

High-performance C++ tensor, linear algebra, and numerical relativity library with SIMD, OpenMP/MPI-oriented design, and compiler experimentation.

Research

Tensorium compiler / DSL

Experimental compiler infrastructure for tensor calculus, elliptic systems, spectral methods, and numerical relativity workflows.

Research

MLIR experiments

Experiments around MLIR dialects, lowering pipelines, vectorization, and compiler backends.

Research

HPC / SIMD numerical kernels

Experiments around cache-aware kernels, vectorization, OpenMP/MPI-oriented execution, and performance constraints in numerical code.