CUDA kernels

CUDA kernels live in their own monorepo, prime-kernels, checked out here as the git submodule deps/prime-kernels, alongside prime-rl's other submodules. That repo is the wheel root (setup.py, pyproject.toml) and prime_kernels/ inside it is the importable package: one folder per kernel, holding the kernel's Python surface and, for compiled kernels, its C++/CUDA sources under csrc/, all declared in the single manifest prime_kernels/kernels.toml. See deps/prime-kernels/README.md once the submodule is initialized.

Nothing about a kernel lives in prime-rl. prime-rl pins a prime-kernels commit for local source builds and a prime-kernels release for installs. prime-kernels builds and publishes its own wheels. prime-rl stays a pure-Python wheel; never add compiled extensions to it.

Living under deps/ means tool.ruff.extend-exclude = ["deps"] in pyproject.toml already covers it — prime-rl lints none of it.

Calling a kernel from prime-rl

Kernels are compiled for exact compute capabilities and may not be built at all, so always gate. Never import prime_kernels.<name> directly in training code:

import prime_kernels

if prime_kernels.is_available("flash_moe"):
    flash_moe = prime_kernels.load("flash_moe")

prime_kernels.status() maps every kernel to "available" or the reason it is not — log it once at startup rather than failing a run halfway through. unavailable_reason(name) is the same answer for one kernel (None when it is usable), which is what a test's skip guard wants; is_available is just that call compared to None.

flash_moe is a compiled fused MoE forward kernel (bf16 + mxfp8) on Blackwell tcgen05. Its trainer integration is dormant. mxfp8_moe is a Python-only registered kernel package for MXFP8 grouped GEMM and torch EP transport on SM100; it owns the MoE-specific torchao-derived orchestration and exports explicit BF16 boundaries instead of tensor-subclass interception.

What a kernel requires of its inputs — block sizes, alignments, shape constraints — belongs to prime-kernels, which exports it: flash_moe.BLOCK_M, flash_moe.MXFP8_SCALE_BLOCK, and flash_moe.unsupported_shape_reason(dim, hidden_dim, mxfp8=...). When the trainer integration is restored, call this once during setup so an unsupported model fails before training rather than mid-step. Never hardcode a 128 on this side: then every requirement change is a change in both repos.

Building locally

uv sync --extra kernels installs the prebuilt wheel (see "Pinning installs at the prebuilt wheels"); building from source is for changing kernels. It is manual by design — no uv sync may compile CUDA, so the extra resolves to release wheels, never to this source tree:

git submodule update --init deps/prime-kernels
uv pip install --no-build-isolation -e deps/prime-kernels

Requirements: nvcc on CUDA_HOME with the same CUDA major as torch (torch refuses to build extensions otherwise).

Kernels whose toolkit is unsuitable are skipped with a message and reported unavailable at runtime — the build still succeeds. PRIME_KERNELS=a,b builds a subset; PRIME_KERNELS_REQUIRE=1 turns any skip into an error.

Changing or adding a kernel

The work happens in the prime-kernels repo, not here. Inside deps/prime-kernels/:

1. Commit the sources under prime_kernels/<name>/csrc/.
2. Add a [<name>] table to prime_kernels/kernels.toml — sources, include-dirs, arch, cxx-std; paths are relative to the kernel folder. A vendored Python kernel uses python-only = true, may declare import checks in requires, and omits compiled extension fields.
3. Write prime_kernels/<name>/__init__.py — from . import _C, then per op a wrapper calling torch.ops.<ns>.<op> and a torch.library.register_fake. No torch.library.custom_op decorator: that defines a Python op, and TORCH_LIBRARY has already defined these C++ side — only the fake (meta) kernel is missing. A kernel used in training also needs torch.library.register_autograd, since a schema carries no backward. flash_moe is forward only and currently has no trainer integration.
4. Nothing else: setup.py and the runtime registry both read the manifest.

Rules the build assumes:

• The extension is always prime_kernels.<name>._C, so the C++ side defines PYBIND11_MODULE(_C, m) and registers ops with TORCH_LIBRARY*.
• arch matches the device exactly at runtime (10.0a runs only on sm_100a); no PTX is shipped to JIT from.
• Two packages registering the same torch.ops namespace collide. If a kernel's sources are also installed as a standalone package (e.g. prime_moe from prime-flash-moe, where flash_moe originally came from), uninstall it.
• Only the Python surface and the compiled _C ship in the wheel; csrc/ is build input.

Then land it in prime-rl as a submodule bump.

Bumping the pin

Kernel sources are pinned by the submodule commit, so picking up any kernel change — yours or someone else's — is a bump:

git -C deps/prime-kernels fetch origin
git -C deps/prime-kernels log --oneline HEAD..origin/main
git -C deps/prime-kernels checkout origin/main
git add deps/prime-kernels

Then, in order:

• Read the diff for host-side contract changes, not just kernel internals. A change to what the caller must pass (weight layout, scale packing, argument order) is silently wrong numbers, not a build error, and prime-rl's call sites have to absorb it.
• Rebuild and re-run the kernel repository's numerical coverage plus every PrimeRL runtime path that calls the changed kernel; the ABI is not checked for you.
• The bump alone ships nothing: installs resolve prime-kernels from a release wheel, so the new code only reaches users once a release rebuilds the wheels and the pin below moves.

Prebuilt wheels

The prime-kernels repository owns build_kernels.yaml. It builds prime-kernels, deep-ep, deep-gemm, and torchao for x86_64 and aarch64. It publishes the wheels as prime-kernels release assets.

Run that workflow in prime-kernels for each torch or CUDA bump:

gh workflow run build_kernels.yaml \
  --repo PrimeIntellect-ai/prime-kernels \
  -f release_tag=vX.Y.Z

The wheel version carries the ABI it was built against, e.g. prime_kernels-0.1.0+cu130torch2.13.0-cp312-cp312-linux_x86_64.whl. It imports only under that exact torch and CUDA ABI. prime-kernels CI controls the build dependency pins.

Pinning installs at the prebuilt wheels

uv sync --extra kernels installs prime-kernels from the wheels named in [tool.uv.sources]. deep-ep, deep-gemm, torchao, and vLLM use the same pattern. A sync must never compile these packages:

[tool.uv.sources]
prime-kernels = [
    { url = "https://github.com/PrimeIntellect-ai/prime-kernels/releases/download/vX.Y.Z/prime_kernels-0.1.0+cu130torch2.13.0-cp312-cp312-linux_x86_64.whl", marker = "platform_machine == 'x86_64'" },
    { url = "https://github.com/PrimeIntellect-ai/prime-kernels/releases/download/vX.Y.Z/prime_kernels-0.1.0+cu130torch2.13.0-cp312-cp312-linux_aarch64.whl", marker = "platform_machine == 'aarch64'" },
]

The wheels are prime-kernels release assets. Get the exact names before you move all pins:

gh release view vX.Y.Z \
  --repo PrimeIntellect-ai/prime-kernels \
  --json assets

Then run uv lock.

Move the pin when a kernel, torch, or CUDA pin changes. A stale pin ships stale kernels. A torch ABI mismatch fails at import, not at install.

Gotchas

• Keep the torch floor in prime-kernels aligned with prime-rl. The wheel imports only under the torch version it was compiled against.
• Never reintroduce prime-kernels as a path source: uv cannot read a source tree's metadata without building it, so every uv lock would then need nvcc. A release-asset URL has static metadata and does not.
• A fresh prime-rl clone without the submodule still installs and trains. Only a local source build needs it. prime-kernels CI sets PRIME_KERNELS_REQUIRE=1, so a release cannot omit kernels silently.