Zaneham/Booth

Health v2 · maintenance only

Adoption not part of the score

1731 stars · 92 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

Booth is an open-source compiler that takes CUDA C, HIP, or Triton kernel source and emits binaries for AMD RDNA 2/3/4 GPUs, NVIDIA PTX, Tenstorrent Metalium, RV32IM, or plain x86-64 CPUs without requiring LLVM. It also supports Fortran `do concurrent` kernels via LFortran and OCaml kernels type-checked by ocamlc.

Use cases

• compile triton kernels to run on a cpu without a gpu
• compile cuda c to amd rdna binaries without nvcc
• run gpu kernels on machines with no gpu
• compile hip code for nvidia gpus
• write gpu kernels in ocaml with type checking
• compile fortran do concurrent kernels for gpus
• target tenstorrent hardware from cuda or triton source

When to choose

• you need to run CUDA/Triton kernels on non-NVIDIA hardware like AMD RDNA or Tenstorrent
• you want to execute GPU kernels on CPU-only machines
• you want a dependency-free compiler binary for Linux, macOS, or Windows
• you want to write kernels in OCaml or Fortran with upfront type checking

When to avoid

• you need full production parity with nvcc or ROCm feature sets and vendor support
• you rely on the latest CUDA libraries, cuDNN, or ecosystem tooling
• you need guaranteed performance parity with vendor compilers for your specific kernels

Facets

cli-tool · maturity active

compiler gpu-computing machine-learning developer-tools compilers gpu-computing machine-learning developer-tools programming-languages windows cli cpp cuda triton hip mlir amd-rdna nvidia-ptx tenstorrent fortran ocaml risc-v x86-64 kernel-compilation linux macos gpu

Member repositories

For agents

markdown · JSON · MCP: product_card(name="Zaneham/Booth")