https://github.com/sandialabs/math-bytecode

A C++ library to compile C-like code at runtime and produce GPU-interpretable bytecode
https://github.com/sandialabs/math-bytecode

scr-2689 snl-science-libs

Last synced: 3 months ago
JSON representation

A C++ library to compile C-like code at runtime and produce GPU-interpretable bytecode

• Host: GitHub
• URL: https://github.com/sandialabs/math-bytecode
• Owner: sandialabs
• License: mit
• Created: 2021-10-01T15:50:19.000Z (over 3 years ago)
• Default Branch: main
• Last Pushed: 2022-11-02T18:06:03.000Z (over 2 years ago)
• Last Synced: 2025-01-11T15:32:44.649Z (4 months ago)
• Topics: scr-2689, snl-science-libs
• Language: C++
• Homepage:
• Size: 83 KB
• Stars: 0
• Watchers: 5
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
Math Bytecode

=============

`math-bytecode` is a C++17 library that uses `ParseGen` to implement a compiler for a subset of the C language.

The language that `math-bytecode` supports is basically the body of a C function, allowing only variables of

type `double` and operations on them.

The body of the function is provided as a C++ `std::string` and is then compiled into bytecode stored in

an object of type `math_bytecode::compiled_function`.

The bytecode can then be executed by an object of type `math_bytecode::executable_function`

as if it were a regular C++ function call.

The executable function object is designed to be used in the interior of "hot loops" inside

physics simulation codes, and `math-bytecode` is an ideal way to accept arbitrary mathematical

functions from users for things like initial and boundary conditions.

`math-bytecode` uses `P3A` to be portably performant across HPC hardware like Intel CPUs,

NVIDIA GPUs, and AMD GPUs.

At Sandia, Math Bytecode is SCR# 2689.0