tandem-mojo

Mojo implementation of Tandem8x32 in one file, tandem.mojo. It produces the stream of the specification bit for bit, with SIMD fills on CPUs and MAX fills on GPUs.

• API: the generator, draws, fills, bounded integers, normals, exponentials and GPU fills.
• Design: how the fills, bounded integers, normals and exponentials work.
• Tests: what the suite checks, where the fixtures come from, and what CI runs.
• Performance: Apple M4 and A100 figures.

Install

pixi install          # Mojo 1.1 and MAX 26.6 from the Modular conda channel

The whole port is one file, tandem.mojo. The GPU fills need the max package. On an NVIDIA driver older than 580, set MODULAR_NVPTX_COMPILER_PATH to a CUDA 12.8 ptxas. On macOS, the Metal toolchain needs a full Xcode install.

AI assistance

This port was written with the help of large language models under human direction. The design and the specification are human work, as is much of the Julia implementation. The code is tested bit for bit against every vector of the specification and against long stream dumps from the Julia implementation, and every value must match. The output does not depend on who or what wrote the code.