chore: one CONTRIBUTING and one LICENSE for the repository

AGENTS.md

@@ -67,8 +67,8 @@ all CI: example compilation, additional example lint scopes, book builds, and
 CodeQL have separate jobs. Read the recipe and relevant workflow when choosing
 validation for a change.
 
-For Tile and shared-host changes, follow `cutile-rs/CONTRIBUTING.md` and
-`.github/workflows/cutile-rs.yml`. The CPU and assembler suite in
+For Tile and shared-host changes, follow the cutile-rs section of
+`CONTRIBUTING.md` and `.github/workflows/cutile-rs.yml`. The CPU and assembler suite in
 `cutile-rs/scripts/run_cpu_tests.sh` runs with `cutile-rs/` as its working
 directory and includes the CPU-safe shared-host tests. CPU-only does not mean
 toolkit-free: some suites require CUDA headers and the Tile IR assembler. GPU tests have a separate

CONTRIBUTING.md

@@ -111,14 +111,35 @@ Your `Signed-off-by` name and email must match your Git configuration
 Pull requests that do not meet the requirements below or lack proper DCO
 sign-off will not be merged.
 
+### Branch and pull request naming
+
+Prefix branches and pull request titles with the kind of change. Branches use
+`/` as the separator and titles use `:`, lowercase: `fix/kernel-launch-sync`
+and `fix: resolve kernel launch sync race`.
+
+| Prefix | Use for |
+| --- | --- |
+| `feat` | New features |
+| `fix` | Bug fixes |
+| `docs` | Documentation changes |
+| `refactor` | Code changes with no behavior change |
+| `perf` | Performance improvements with no functional change |
+| `test` | Test additions or changes |
+| `build` | Build system and dependency changes |
+| `ci` | CI changes |
+| `chore` | Maintenance tasks |
+
+A scope in parentheses is welcome when a change is confined to one component,
+for example `docs(cutile-rs): ...`.
+
 ## Code Requirements
 
 ### Toolchains
 
 The shared host crates at the repository root use the root stable pin.
 Each product has its own workspace and toolchain under `cuda-oxide/` or
-`cutile-rs/`. For Tile development, see [cutile-rs/CONTRIBUTING.md](cutile-rs/CONTRIBUTING.md).
-The setup and validation commands below cover cuda-oxide.
+`cutile-rs/`. The setup commands below cover cuda-oxide; the Tile checks have
+their own section further down.
 
 cuda-oxide requires the Rust nightly toolchain with `rustc_private` support.
 See the [cuda-oxide README](cuda-oxide/README.md) for setup instructions.
@@ -150,6 +171,28 @@ codegen backend, a Python virtualenv, or GitHub's own infrastructure. The
 step when a workflow changes; the commands below are the ones worth knowing by
 hand even so.
 
+### Running the cutile-rs checks
+
+cutile-rs builds on the stable toolchain pinned in `cutile-rs/rust-toolchain.toml`
+and needs a CUDA 13.2 or newer toolkit with the Tile IR assembler, `tileiras`;
+the [cutile-rs book](https://nvidia.github.io/cuda-rust/cutile/) covers setup.
+Its test scripts locate the Tile workspace from their own path and reach the
+shared host crates at the repository root on their own, so they can be run
+from anywhere:
+
+```bash
+cutile-rs/scripts/run_cpu_tests.sh   # CPU and assembler tests; needs the toolkit, not a GPU
+cutile-rs/scripts/run_gpu_tests.sh   # GPU integration tests
+cutile-rs/scripts/run_all.sh         # both of the above, then the examples and benchmarks
+```
+
+Run `cutile-rs/scripts/run_all.sh` before marking a Tile pull request ready
+for review.
+CI runs the same suites in `.github/workflows/cutile-rs.yml`. Because the Tile
+lanes use NVIDIA self-hosted runners, that workflow triggers on the
+`pull-request/<n>` branches that copy-pr-bot mirrors rather than on the pull
+request itself.
+
 ### Formatting and Style
 
 - Run `(cd cuda-oxide && cargo oxide fmt)` before submitting. All code must be

README.md

@@ -81,14 +81,8 @@ See [CONTRIBUTING.md](CONTRIBUTING.md) for commit sign-off requirements, pull re
 ## License
 
 CUDA Rust is licensed under the [Apache License 2.0](./LICENSE). The root
-`LICENSE` covers the whole repository. Each component also includes a copy with
-its sources and built artifacts.
-
-| Component | License | License file |
-| --- | --- | --- |
-| `cuda-oxide` | Apache-2.0 | root [`LICENSE`](./LICENSE) |
-| `cutile-rs` | Apache-2.0 | [`cutile-rs/LICENSE`](./cutile-rs/LICENSE) |
-| `cuda-bindings`, `cuda-core`, `cuda-core-derive`, `cuda-async` | Apache-2.0 | root [`LICENSE`](./LICENSE) |
+`LICENSE` covers the whole repository: cuda-oxide, cutile-rs, and the shared
+host crates. Every published crate declares `license = "Apache-2.0"`.
 
 Third-party attributions are listed in
 [`cuda-oxide/THIRD_PARTY_NOTICES`](./cuda-oxide/THIRD_PARTY_NOTICES).

cuda-oxide/README.md

@@ -375,6 +375,6 @@ cuda-oxide is one of several Rust + GPU efforts under active development. Projec
 
 ## License
 
-cuda-oxide is licensed under the Apache License, Version 2.0: [LICENSE](LICENSE).
+cuda-oxide is licensed under the Apache License, Version 2.0: [LICENSE](../LICENSE).
 Third-party components retain the licenses stated in their files; see
 [dependency-licenses.csv](dependency-licenses.csv) for the tracked license inventory.

cuda-oxide/THIRD_PARTY_NOTICES

@@ -4,7 +4,8 @@ THIRD PARTY NOTICES
 This file attributes third-party open source software whose source code is
 embedded in cuda-oxide.
 
-cuda-oxide's own code is licensed under Apache-2.0 (see LICENSE). Third-party
+cuda-oxide's own code is licensed under Apache-2.0 (see the LICENSE at the
+repository root). Third-party
 crates that cuda-oxide merely depends on through Cargo are not redistributed
 here and are therefore not listed below; their licenses are recorded in
 `dependency-licenses.csv` and enforced by `deny.toml`.
@@ -29,7 +30,7 @@ The upstream license texts are retained in place at
 carries its own identical copies at
 `crates/fuzzer/rustlantis/rangemap/LICENSE-APACHE` and
 `crates/fuzzer/rustlantis/rangemap/LICENSE-MIT`. The Apache-2.0 text is
-identical to the one in this directory (LICENSE). The MIT text is
+identical to the repository's root LICENSE. The MIT text is
 reproduced here:
 
     Permission is hereby granted, free of charge, to any