feat: merge dcu branch features

.clang-format

+---
+BasedOnStyle: LLVM
+UseTab: Never
+IndentWidth: 2
+ColumnLimit: 80
+
+Language: Cpp
+Standard: c++17

.clang-tidy

+---
+InheritParentConfig: true
+ExtraArgs: []
+FormatStyle: file
+UseColor: true
+WarningsAsErrors: '*'
+# FIXME: Use `ExcludeHeaderFilterRegex` instead when all maintainers upgraded their `clang-tidy`
+HeaderFilterRegex: '^(?!.*(?:/|^)(3rdparty|tvm)/).*'
+# ExcludeHeaderFilterRegex: '^(3rdparty|tvm)/.*$'
+
+# NOTE: there must be no spaces before the '-', so put the comma last.
+Checks: >-
+  # 1. Retained categories: easier to find bugs/performance issues
+  clang-analyzer-*,
+  cppcoreguidelines-pro-type-static-cast-downcast,
+  cppcoreguidelines-pro-type-member-init,
+  cppcoreguidelines-pro-bounds-array-to-pointer-decay,
+  cppcoreguidelines-pro-bounds-pointer-arithmetic,
+  cppcoreguidelines-slicing,
+  cppcoreguidelines-narrowing-conversions,
+  performance-*,
+
+  # 2. Readability: only keep useful rules
+  readability-braces-around-statements,
+  readability-container-size-empty,
+  readability-delete-null-pointer,
+  readability-redundant-member-init,
+  readability-redundant-smartptr-get,
+  readability-redundant-string-cstr,
+
+  # 3. Disable all intrusive/style-breaking rules
+  -readability-identifier-length,
+  -readability-avoid-const-params-in-decls,
+  -readability-else-after-return,
+  -cppcoreguidelines-avoid-magic-numbers,
+  -modernize-use-trailing-return-type,
+  -modernize-use-nodiscard,
+  -modernize-use-auto,
+  -modernize-pass-by-value,
+  -modernize-return-braced-init-list,
+  -modernize-use-default-member-init,
+  -modernize-loop-convert,
+  -modernize-concat-nested-namespaces,
+  -llvm-include-order,
+  -bugprone-unused-return-value,
+  -clang-diagnostic-unused-result,
+  -cppcoreguidelines-special-member-functions,
+  -performance-noexcept-move-constructor,
+  -cppcoreguidelines-narrowing-conversions,
+  -clang-diagnostic-error,
+  -cppcoreguidelines-pro-type-member-init,
+  -clang-analyzer-optin.cplusplus.UninitializedObject,
+  -cppcoreguidelines-pro-type-static-cast-downcast,
+  -performance-unnecessary-value-param,
+  -performance-enum-size,
+  -cppcoreguidelines-pro-bounds-pointer-arithmetic,
+  -cppcoreguidelines-pro-bounds-array-to-pointer-decay,
+  -clang-analyzer-deadcode.DeadStores,
+  -clang-analyzer-optin.cplusplus.VirtualCall,
+  -clang-diagnostic-tautological-constant-compare,

.editorconfig

+# https://editorconfig.org/
+
+root = true
+
+[*]
+charset = utf-8
+end_of_line = lf
+indent_style = space
+indent_size = 4
+trim_trailing_whitespace = true
+insert_final_newline = true
+
+[*.{py,pyi}]
+indent_size = 4
+
+[*.{cpp,hpp,cxx,cc,c,h,cu,cuh}]
+indent_size = 2
+
+[{*.cmake,CMakeLists.txt}]
+indent_size = 2
+
+[*.{yaml,yml}]
+indent_size = 2
+
+[.clang-{format,tidy}]
+indent_size = 2
+
+[Makefile]
+indent_style = tab
+
+[*.sh]
+indent_size = 4
+
+[*.bat]
+indent_size = 4
+end_of_line = crlf
+
+[*.md]
+indent_size = 2
+x-soft-wrap-text = true
+
+[*.rst]
+indent_size = 4
+x-soft-wrap-text = true

.gitattributes

+* text eol=lf
+*.bat eol=crlf
+
+*.svg binary
+*.jpg binary
+*.jpeg binary
+*.png binary
+*.gif binary
+
+*.h linguist-language=C++

.github/ISSUE_TEMPLATE/bug-report.yml

+name: 🐛 Bug Report
+description: File an issue about a bug.
+title: "[BUG] "
+labels: [bug]
+assignees: []
+body:
+  - type: markdown
+    attributes:
+      value: >-
+        Please do your best to make the issue as easy to act on as possible,
+        and only submit here if there is clearly a problem with TileLang.
+
+  - type: checkboxes
+    id: steps
+    attributes:
+      label: Required prerequisites
+      description: Make sure you've completed the following steps before submitting your issue -- thank you!
+      options:
+        - label: I have read the documentation <https://tilelang.com>.
+          required: true
+        - label: >-
+            I have searched the [Issue Tracker](https://github.com/tile-ai/tilelang/issues)
+            that this hasn't already been reported. (comment there if it has.)
+          required: true
+
+  - type: input
+    id: version
+    attributes:
+      label: What version of TileLang are you using?
+      description: >-
+        Run command `python3 -c 'print(__import__("tilelang").__version__)'` in your shell
+        and paste the output here.
+      placeholder: E.g., 0.1.5
+    validations:
+      required: true
+
+  - type: textarea
+    id: system-info
+    attributes:
+      label: System information
+      description: |
+        Describe the characteristic of your environment:
+
+        - Describe how the library was installed (pip, conda, source, ...)
+        - Python version
+        - Versions of any other relevant libraries
+
+        ```python
+        import sys, tilelang, torch
+        print(sys.version, sys.platform)
+        print(tilelang.__version__)
+        print(torch.__version__)
+        ```
+
+        ```bash
+        python3 -m torch.utils.collect_env
+        ```
+    validations:
+      required: true
+
+  - type: textarea
+    id: description
+    attributes:
+      label: Problem description
+      description: >-
+        Provide a short description, state the expected behavior and what actually happens. Include
+        relevant information like what version of TileLang you are using, what system you are on, and
+        any useful commands / output.
+    validations:
+      required: true
+
+  - type: textarea
+    id: code
+    attributes:
+      label: Reproducible example code
+      description: >-
+        The code should be minimal, have minimal external dependencies, and isolate the functions
+        that cause breakage. Submit matched and complete snippets that can be easily run to diagnose
+        the issue.
+      value: |
+        The Python snippets:
+
+        ```python
+
+        ```
+    validations:
+      required: true
+
+  - type: textarea
+    id: traceback
+    attributes:
+      label: Traceback
+      description: Put the Python traceback information here.
+      placeholder: |
+        Traceback (most recent call last):
+          File ...
+      render: pytb
+
+  - type: textarea
+    id: expected
+    attributes:
+      label: Expected behavior
+      description: Provide a clear and concise description of what you expected to happen.
+
+  - type: textarea
+    id: additional-context
+    attributes:
+      label: Additional context
+      description: >-
+        Add any other context about the problem here. Screenshots may also be helpful.
+
+        If you know or suspect the reason for this bug, paste the code lines and suggest modifications.

.github/ISSUE_TEMPLATE/config.yml

+blank_issues_enabled: true

.github/ISSUE_TEMPLATE/feature-request.yml

+name: ✨ Feature Request
+description: Suggest an idea for this project.
+title: "[Feature Request] "
+labels: [enhancement]
+body:
+  - type: checkboxes
+    id: steps
+    attributes:
+      label: Required prerequisites
+      description: Make sure you've completed the following steps before submitting your issue -- thank you!
+      options:
+        - label: >-
+            I have searched the [Issue Tracker](https://github.com/tile-ai/tilelang/issues)
+            that this hasn't already been reported. (comment there if it has.)
+          required: true
+
+  - type: textarea
+    id: motivation
+    attributes:
+      label: Motivation
+      description: Outline the motivation for the proposal.
+      value: |
+        <!-- Please outline the motivation for the proposal.
+        Is your feature request related to a problem? E.g., "I'm always frustrated when [...]".
+        If this is related to another issue, please link here too. -->
+    validations:
+      required: true
+
+  - type: textarea
+    id: solution
+    attributes:
+      label: Solution
+      description: Provide a clear and concise description of what you want to happen.
+
+  - type: textarea
+    id: alternatives
+    attributes:
+      label: Alternatives
+      description: A clear and concise description of any alternative solutions or features you've considered.
+
+  - type: textarea
+    id: additional-context
+    attributes:
+      label: Additional context
+      description: Add any other context about the problem here. Screenshots may also be helpful.

.github/ISSUE_TEMPLATE/questions.yml

+name: 🤔 Questions / Help / Support
+description: Do you need support?
+title: "[Question] "
+labels: [question]
+body:
+  - type: checkboxes
+    id: steps
+    attributes:
+      label: Required prerequisites
+      description: Make sure you've completed the following steps before submitting your issue -- thank you!
+      options:
+        - label: I have read the documentation <https://tilelang.com>.
+          required: true
+        - label: >-
+            I have searched the [Issue Tracker](https://github.com/tile-ai/tilelang/issues)
+            that this hasn't already been reported. (comment there if it has.)
+          required: true
+
+  - type: textarea
+    id: questions
+    attributes:
+      label: Questions
+      description: Describe your questions with relevant resources such as snippets, links, images, etc.
+    validations:
+      required: true

.github/ISSUE_TEMPLATE/release-plan.yml

+name: "Release Plan"
+description: "Plan the next release"
+title: "[Release Plan] vX.Y.Z"
+labels:
+  - release-plan
+  - tracking
+assignees: []
+body:
+  - type: input
+    id: version
+    attributes:
+      label: "Version"
+      placeholder: "v0.2.0"
+    validations:
+      required: true
+
+  - type: input
+    id: milestone
+    attributes:
+      label: "Milestone"
+      description: "Link or name of the milestone for this release"
+      placeholder: "https://github.com/tile-ai/tilelang/milestone/XX"
+
+  - type: textarea
+    id: scope
+    attributes:
+      label: "Scope"
+      description: "Goals and non-goals (brief)"
+      placeholder: |
+        - Goals: ...
+        - Non-goals: ...
+
+  - type: textarea
+    id: tasks
+    attributes:
+      label: "Tasks"
+      description: "Task list; link issues/PRs"
+      value: |
+        - [ ] Features
+        - [ ] Fixes
+        - [ ] Docs
+        - [ ] API/Breaking changes
+        - [ ] Benchmarks
+        - [ ] Release notes
+
+  - type: checkboxes
+    id: readiness
+    attributes:
+      label: "Readiness"
+      options:
+        - label: "All planned issues closed or deferred"
+        - label: "Docs updated"
+        - label: "CI green; artifacts verified"
+        - label: "Release notes drafted"
+
+  - type: textarea
+    id: notes
+    attributes:
+      label: "Notes"
+      description: "Risks or communications (optional)"
+      placeholder: |
+        - Risk: ...
+        - Communication: ...

.github/dependabot.yml

+version: 2
+updates:
+  - package-ecosystem: "github-actions"
+    directory: "/"
+    schedule:
+      interval: "weekly"
+      day: "monday"
+      time: "12:00"
+      timezone: "Asia/Shanghai"
+    commit-message:
+      prefix: "[CI]"

.github/workflows/ci.yml

+name: CI
+on:
+  pull_request:
+    types:
+      - labeled
+      - unlabeled
+      - opened
+      - synchronize
+      - reopened
+  # Allow to trigger the workflow manually
+  workflow_dispatch:
+
+permissions:
+  contents: read
+
+concurrency:
+  group: "${{ github.workflow }}-${{ github.ref }}"
+  cancel-in-progress: ${{ github.event_name == 'pull_request' }}
+
+env:
+  CLANG_TIDY_CMAKE_OPTIONS: "-DCMAKE_EXPORT_COMPILE_COMMANDS=ON" # to be updated
+  PYTHONDEVMODE: "1"
+  PYTHONUNBUFFERED: "1"
+  PYTHONPATH: "" # explicit cleanup
+  PIP_USER: "" # explicit cleanup
+  COLUMNS: "100"
+  FORCE_COLOR: "1"
+  CLICOLOR_FORCE: "1"
+  UV_INDEX_STRATEGY: "unsafe-best-match"
+  UV_HTTP_TIMEOUT: "600"
+  XDG_CACHE_HOME: "${{ github.workspace }}/.cache" # to be updated
+  PIP_CACHE_DIR: "${{ github.workspace }}/.cache/pip" # to be updated
+  UV_CACHE_DIR: "${{ github.workspace }}/.cache/uv" # to be updated
+  PRE_COMMIT_HOME: "${{ github.workspace }}/.cache/pip/.pre-commit" # to be updated
+
+jobs:
+  lint:
+    name: Quick Lint
+    runs-on: ubuntu-latest
+    timeout-minutes: 30
+    steps:
+      - name: Checkout repository
+        uses: actions/checkout@v6
+        with:
+          fetch-depth: 0
+          submodules: recursive
+
+      - name: Setup Python 3.8
+        id: setup-pylowest
+        uses: actions/setup-python@v6
+        with:
+          python-version: "3.8" # use lowest supported version for linting
+          update-environment: false
+
+      - name: Check AST with Python 3.8
+        run: |
+          "${{ steps.setup-pylowest.outputs.python-path }}" -m compileall -q -f tilelang
+
+      - name: Setup Python 3.9
+        uses: actions/setup-python@v6
+        with:
+          python-version: "3.9"
+          update-environment: true
+          cache: pip
+          cache-dependency-path: |
+            pyproject.toml
+            requirements*.txt
+            .pre-commit-config.yaml
+
+      - name: Pre-commit Lint
+        run: |
+          if ! pipx run pre-commit run --all-files --color=always --show-diff-on-failure; then
+            echo "::error::Pre-commit checks failed. Please run 'pre-commit install' and 'pre-commit run --all-files' locally to see the issues."
+            exit 1
+          fi
+
+  tests:
+    name: Test for Python ${{ matrix.python-version }} with ${{ matrix.runner.toolkit }} (on ${{ matrix.runner.name }})
+    if: |
+      github.repository_owner == 'tile-ai' &&
+      (github.event_name != 'pull_request' || !github.event.pull_request.draft)
+    needs: [lint]
+    runs-on: ${{ matrix.runner.tags }}
+    strategy:
+      matrix:
+        runner:
+          - tags: [self-hosted, nvidia]
+            name: self-hosted-nvidia
+            # Format: [Nightly-]CUDA-<major>.<minor>[.<patch>]. E.g., "CUDA-12.8" or "Nightly-CUDA-13.0".
+            # Use "Nightly-" prefix to use torch nightly builds.
+            toolkit: CUDA-12.8
+          - tags: [self-hosted, amd, gpu]
+            name: self-hosted-amd
+            # Format: [Nightly-]ROCm-<major>.<minor>[.<patch>]. E.g., "ROCm-6.4" or "Nightly-ROCm-7.0".
+            # Use "Nightly-" prefix to use torch nightly builds.
+            toolkit: Nightly-ROCm-7.1
+          - tags: [macos-latest]
+            name: macos-latest
+            toolkit: Metal # or Nightly-Metal
+        python-version:
+          - "3.12"
+      fail-fast: false
+    timeout-minutes: 120
+
+    steps:
+      - name: Checkout repository
+        uses: actions/checkout@v6
+        with:
+          fetch-depth: 0
+          submodules: recursive
+
+      - name: Set environment (self-hosted runners)
+        if: startsWith(matrix.runner.name, 'self-hosted')
+        run: |
+          # Hide sensitive data in logs for self-hosted runners
+          if [[ -n "${{ secrets.SECRET_PATH_PREFIXES }}" ]]; then
+            echo "::add-mask::${{ secrets.SECRET_PATH_PREFIXES }}"
+            # Colon separated list of secrets to mask
+            for secret in $(echo "${{ secrets.SECRET_PATH_PREFIXES }}" | tr ':' '\n'); do
+              echo "::add-mask::${secret}"
+            done
+          fi
+
+          # Use runner tool_cache as cache root for self-hosted runners to avoid internet connection
+          # issues and to share cache between jobs.
+          export XDG_CACHE_HOME="${{ runner.tool_cache }}/.ci-cache-${{ github.workflow }}"
+          echo "XDG_CACHE_HOME=${XDG_CACHE_HOME}" | tee -a "${GITHUB_ENV}"
+          echo "PIP_CACHE_DIR=${XDG_CACHE_HOME}/pip" | tee -a "${GITHUB_ENV}"
+          echo "UV_CACHE_DIR=${XDG_CACHE_HOME}/uv" | tee -a "${GITHUB_ENV}"
+          echo "PRE_COMMIT_HOME=${XDG_CACHE_HOME}/pip/.pre-commit" | tee -a "${GITHUB_ENV}"
+
+      # Do not use ccache on self-hosted runners, as it will download/upload caches which is slow.
+      # Self-hosted runners usually have more CPU power to compile without ccache.
+      - name: Setup ccache (GitHub-hosted runners)
+        id: setup-ccache
+        if: ${{ !startsWith(matrix.runner.name, 'self-hosted') }}
+        uses: hendrikmuhs/ccache-action@v1
+        with:
+          create-symlink: true
+          evict-old-files: "7d"
+          append-timestamp: false
+          key: ${{ runner.os }}-${{ runner.arch }}-${{ matrix.runner.toolkit }}-${{ hashFiles('**/*.cc') }}
+          restore-keys: |
+            ${{ runner.os }}-${{ runner.arch }}-${{ matrix.runner.toolkit }}-${{ hashFiles('**/*.cc') }}
+            ${{ runner.os }}-${{ runner.arch }}-${{ matrix.runner.toolkit }}
+            ${{ runner.os }}-${{ runner.arch }}
+
+      - name: Set environment (CUDA)
+        if: contains(matrix.runner.toolkit, 'CUDA')
+        run: |
+          TOOLKIT="${{ matrix.runner.toolkit }}"
+          CUDA_VERSION="${TOOLKIT##*-}"
+          CUDA_VERSION_MAJMIN="$(echo ${CUDA_VERSION} | cut -d '.' -f-2)"
+          CUDA_VERSION_MAJMIN_NODOT="${CUDA_VERSION_MAJMIN//./}"
+          if [[ "${TOOLKIT}" == "Nightly-"* ]]; then
+            # Use torch nightly builds
+            export PIP_EXTRA_INDEX_URL="https://download.pytorch.org/whl/nightly/cu${CUDA_VERSION_MAJMIN_NODOT}"
+          else
+            export PIP_EXTRA_INDEX_URL="https://download.pytorch.org/whl/cu${CUDA_VERSION_MAJMIN_NODOT}"
+          fi
+          export UV_INDEX="${PIP_EXTRA_INDEX_URL}"
+          export CLANG_TIDY_CMAKE_OPTIONS="${CLANG_TIDY_CMAKE_OPTIONS} -DUSE_CUDA=ON"
+
+          echo "USE_CUDA=ON" | tee -a "${GITHUB_ENV}"
+          echo "CUDA_VERSION=${CUDA_VERSION}" | tee -a "${GITHUB_ENV}"
+          echo "CUDA_VERSION_MAJMIN=${CUDA_VERSION_MAJMIN}" | tee -a "${GITHUB_ENV}"
+          echo "CUDA_VERSION_MAJMIN_NODOT=${CUDA_VERSION_MAJMIN_NODOT}" | tee -a "${GITHUB_ENV}"
+          echo "PIP_EXTRA_INDEX_URL=${PIP_EXTRA_INDEX_URL}" | tee -a "${GITHUB_ENV}"
+          echo "UV_INDEX=${UV_INDEX}" | tee -a "${GITHUB_ENV}"
+          echo "CLANG_TIDY_CMAKE_OPTIONS=${CLANG_TIDY_CMAKE_OPTIONS}" | tee -a "${GITHUB_ENV}"
+
+          if [[ ! -x "$(command -v nvcc)" ]]; then
+            export PATH="/usr/local/cuda/bin:${PATH}"
+            export LD_LIBRARY_PATH="/usr/local/cuda/lib64${LD_LIBRARY_PATH:+:${LD_LIBRARY_PATH}}"
+            echo "PATH=${PATH}" | tee -a "${GITHUB_ENV}"
+            echo "LD_LIBRARY_PATH=${LD_LIBRARY_PATH}" | tee -a "${GITHUB_ENV}"
+          fi
+          if [[ -x "$(command -v nvcc)" ]]; then
+            echo "\$ $(command -v nvcc) --version" && nvcc --version
+          else
+            echo "::warning::nvcc not found in PATH!"
+          fi
+
+      - name: Set environment (ROCm)
+        if: contains(matrix.runner.toolkit, 'ROCm')
+        run: |
+          TOOLKIT="${{ matrix.runner.toolkit }}"
+          ROCM_VERSION="${TOOLKIT##*-}"
+          ROCM_VERSION_MAJMIN="$(echo ${ROCM_VERSION} | cut -d '.' -f-2)"
+          ROCM_VERSION_MAJMIN_NODOT="${ROCM_VERSION_MAJMIN//./}"
+          if [[ "${TOOLKIT}" == "Nightly-"* ]]; then
+            # Use torch nightly builds
+            export PIP_EXTRA_INDEX_URL="https://download.pytorch.org/whl/nightly/rocm${ROCM_VERSION_MAJMIN}"
+          else
+            export PIP_EXTRA_INDEX_URL="https://download.pytorch.org/whl/rocm${ROCM_VERSION_MAJMIN}"
+          fi
+          export UV_INDEX="${PIP_EXTRA_INDEX_URL}"
+          export CLANG_TIDY_CMAKE_OPTIONS="${CLANG_TIDY_CMAKE_OPTIONS} -DUSE_ROCM=ON"
+
+          echo "USE_ROCM=ON" | tee -a "${GITHUB_ENV}"
+          echo "ROCM_VERSION=${ROCM_VERSION}" | tee -a "${GITHUB_ENV}"
+          echo "ROCM_VERSION_MAJMIN=${ROCM_VERSION_MAJMIN}" | tee -a "${GITHUB_ENV}"
+          echo "ROCM_VERSION_MAJMIN_NODOT=${ROCM_VERSION_MAJMIN_NODOT}" | tee -a "${GITHUB_ENV}"
+          echo "PIP_EXTRA_INDEX_URL=${PIP_EXTRA_INDEX_URL}" | tee -a "${GITHUB_ENV}"
+          echo "UV_INDEX=${UV_INDEX}" | tee -a "${GITHUB_ENV}"
+          echo "CLANG_TIDY_CMAKE_OPTIONS=${CLANG_TIDY_CMAKE_OPTIONS}" | tee -a "${GITHUB_ENV}"
+
+          if [[ ! -x "$(command -v hipcc)" ]]; then
+            export PATH="/opt/rocm/bin:${PATH}"
+            export LD_LIBRARY_PATH="/opt/rocm/lib${LD_LIBRARY_PATH:+:${LD_LIBRARY_PATH}}"
+            echo "PATH=${PATH}" | tee -a "${GITHUB_ENV}"
+            echo "LD_LIBRARY_PATH=${LD_LIBRARY_PATH}" | tee -a "${GITHUB_ENV}"
+          fi
+          if [[ -x "$(command -v hipcc)" ]]; then
+            echo "\$ $(command -v hipcc) --version" && hipcc --version
+          else
+            echo "::warning::hipcc not found in PATH!"
+          fi
+
+      - name: Set environment (Metal)
+        if: contains(matrix.runner.toolkit, 'Metal')
+        run: |
+          if [[ "${{ matrix.runner.toolkit }}" == "Nightly-"* ]]; then
+            # Use torch nightly builds
+            export PIP_EXTRA_INDEX_URL="https://download.pytorch.org/whl/nightly/cpu"
+            export UV_INDEX="${PIP_EXTRA_INDEX_URL}"
+            echo "PIP_EXTRA_INDEX_URL=${PIP_EXTRA_INDEX_URL}" | tee -a "${GITHUB_ENV}"
+            echo "UV_INDEX=${UV_INDEX}" | tee -a "${GITHUB_ENV}"
+          fi
+          export CLANG_TIDY_CMAKE_OPTIONS="${CLANG_TIDY_CMAKE_OPTIONS} -DUSE_METAL=ON"
+
+          echo "USE_METAL=ON" | tee -a "${GITHUB_ENV}"
+          echo "CLANG_TIDY_CMAKE_OPTIONS=${CLANG_TIDY_CMAKE_OPTIONS}" | tee -a "${GITHUB_ENV}"
+
+      - name: Setup Python and uv with caching
+        id: setup-uv
+        uses: astral-sh/setup-uv@v7
+        with:
+          python-version: ${{ matrix.python-version }}
+          activate-environment: true
+          # Do not use cache for self-hosted runners, as it will download/upload caches which is slow.
+          enable-cache: ${{ !startsWith(matrix.runner.name, 'self-hosted') }}
+          prune-cache: ${{ !startsWith(matrix.runner.name, 'self-hosted') }}
+          # Use runner tool_cache for self-hosted runners
+          cache-local-path: ${{ env.UV_CACHE_DIR }}
+          ignore-nothing-to-cache: true
+          # Extra cache key to upload/download caches on GitHub-hosted runners
+          cache-suffix: uv-${{ runner.os }}-${{ runner.arch }}-${{ matrix.python-version }}-${{ matrix.runner.name }}-${{ matrix.runner.toolkit }}
+          cache-dependency-glob: |
+            pyproject.toml
+            requirements*.txt
+            .pre-commit-config.yaml
+
+      - name: Setup venv
+        id: setup-venv
+        run: |
+          set -o pipefail
+
+          uv pip install --upgrade pip setuptools wheel
+          if [[ "${UV_INDEX}" == *"/nightly/"* ]]; then
+            uv pip install --prerelease=allow -v torch
+          fi
+          uv pip install -v -r requirements-test.txt
+          echo "import torch; print(f'torch: {torch.__version__}')" | uv run --no-project --script -
+          if [[ "${{ matrix.runner.toolkit }}" == *"CUDA"* ]]; then
+            uv pip install --no-build-isolation-package=flash-attn -v -r requirements-test-cuda.txt
+            echo "import flash_attn; print(f'flash_attn: {flash_attn.__version__}')" | uv run --no-project --script -
+          elif [[ "${{ matrix.runner.toolkit }}" == *"ROCm"* ]]; then
+            uv pip install -v -r requirements-test-rocm.txt
+          elif [[ "${{ matrix.runner.toolkit }}" == *"Metal"* ]]; then
+            uv pip install -v -r requirements-test-metal.txt
+          else
+            echo "::error::Unknown toolkit: ${{ matrix.runner.toolkit }}"
+            exit 1
+          fi
+          echo "::group::torch.utils.collect_env"
+          uv run --no-project -m -- torch.utils.collect_env
+          echo "::endgroup::"
+
+      - name: Clear uv cache for self-hosted runners (if setup failed)
+        if: >-
+          ${{
+            failure() &&
+            startsWith(matrix.runner.name, 'self-hosted') &&
+            (steps.setup-uv.conclusion == 'failure' || steps.setup-venv.conclusion == 'failure')
+          }}
+        run: |
+          echo "Clearing uv cache at ${UV_CACHE_DIR} due to failure."
+          uv cache clean
+
+      - name: Enable core dump generation (Linux / GitHub-hosted runners)
+        if: ${{ runner.os == 'Linux' && !startsWith(matrix.runner.name, 'self-hosted') }}
+        run: |
+          sudo sysctl -w kernel.core_pattern="core.${{ matrix.python-version }}.${{ matrix.runner.toolkit }}.%P"
+          sudo sysctl -w kernel.core_uses_pid=0
+          sudo sysctl -w fs.suid_dumpable=1
+          sysctl kernel.core_pattern kernel.core_uses_pid fs.suid_dumpable
+
+      - name: Enable core dump generation (macOS / GitHub-hosted runners)
+        if: ${{ runner.os == 'macOS' && !startsWith(matrix.runner.name, 'self-hosted') }}
+        run: |
+          sudo sysctl -w kern.corefile="core.${{ matrix.python-version }}.${{ matrix.runner.toolkit }}.%P"
+          sudo sysctl -w kern.coredump=1
+          sudo sysctl -w kern.sugid_coredump=1
+          sysctl kern.corefile kern.coredump kern.sugid_coredump
+
+      - name: Install project (wheel form)
+        run: |
+          uv pip install -v .
+
+      - name: Run clang-tidy
+        id: clang-tidy
+        if: runner.os == 'Linux'
+        run: |
+          echo "\$ $(command -v clang-tidy) --version" && clang-tidy --version
+
+          # Download run-clang-tidy script
+          RCT_URL=https://raw.githubusercontent.com/llvm/llvm-project/refs/heads/release/21.x/clang-tools-extra/clang-tidy/tool/run-clang-tidy.py
+          echo "Downloading run-clang-tidy script from ${RCT_URL}"
+          echo "import urllib.request; url = '${RCT_URL}'.rstrip('/'); urllib.request.urlretrieve(url, url.split('/')[-1])" | uv run --no-project --script -
+          RUN_CLANG_TIDY=(uv run --no-project --script -- run-clang-tidy.py)
+
+          if [[ -x "$(command -v clang-apply-replacements)" ]]; then
+            echo "Using clang-apply-replacements from $(command -v clang-apply-replacements)"
+            RUN_CLANG_TIDY+=(-fix -clang-apply-replacements-binary="$(command -v clang-apply-replacements)")
+          else
+            echo "::warning::clang-apply-replacements not found in PATH, automatic fixing disabled."
+          fi
+
+          # Run cmake to create the build directory with compile_commands.json
+          cmake -S . -B cmake-build --fresh ${CLANG_TIDY_CMAKE_OPTIONS}  # no quotes here
+          echo "::group::compile_commands.json"
+          ls -alh cmake-build/compile_commands.json
+          uv run --no-project -m -- json.tool --no-ensure-ascii cmake-build/compile_commands.json
+          echo "::endgroup::"
+
+          CXX_FILES=$(find src -type f -iname "*.[ch]pp" -o -iname "*.cc" -o -iname "*.c" -o -iname "*.h")
+          rc=0
+          echo "::group::run-clang-tidy"
+          "${RUN_CLANG_TIDY[@]}" -clang-tidy-binary="$(command -v clang-tidy)" \
+            -exclude-header-filter='^(3rdparty|tvm)/.*$' \
+            -p="cmake-build" ${CXX_FILES} || rc="$?"
+          echo "::endgroup::"
+          rm -rf cmake-build run-clang-tidy.py
+          if (( rc != 0 )); then
+            echo "::error::clang-tidy found issues (exit code: ${rc}). Please run 'clang-tidy --fix' locally to fix them."
+            git diff --color=always || true
+            exit "${rc}"
+          fi
+
+      - name: Run examples with Python ${{ matrix.python-version }} (${{ matrix.runner.toolkit }})
+        if: contains(matrix.runner.toolkit, 'CUDA')
+        run: |
+          cd testing
+          PYTEST=(
+            uv run --no-project -m --
+            pytest --verbose --color=yes --durations=0 --showlocals --cache-clear
+          )
+          "${PYTEST[@]}" --maxfail=3 --numprocesses=4 \
+            ../examples
+
+      # NVIDIA CUDA tests
+      - name: Run CUDA tests with Python ${{ matrix.python-version }} (${{ matrix.runner.toolkit }})
+        id: cuda-tests
+        if: contains(matrix.runner.toolkit, 'CUDA')
+        run: |
+          cd testing
+          PYTEST=(
+            uv run --no-project -m --
+            pytest --verbose --color=yes --durations=0 --showlocals --cache-clear
+          )
+          "${PYTEST[@]}" --maxfail=3 --numprocesses=4 \
+            --ignore=./python/jit/test_tilelang_jit_cutedsl.py \
+            ./python
+
+      # CuTeDSL JIT tests require GEMM v1 (must be set before importing tilelang).
+      # Run them in a dedicated step to avoid changing the default GEMM selection
+      # (and to keep the rest of the CUDA tests on GEMM v2).
+      - name: Run CuTeDSL JIT tests (GEMM v1) with Python ${{ matrix.python-version }} (${{ matrix.runner.toolkit }})
+        id: cutedsl-tests
+        if: contains(matrix.runner.toolkit, 'CUDA')
+        env:
+          TILELANG_USE_GEMM_V1: "1"
+        run: |
+          cd testing
+          PYTEST=(
+            uv run --no-project -m --
+            pytest --verbose --color=yes --durations=0 --showlocals --cache-clear
+          )
+          # Avoid xdist contention on a single GPU by running this file in one worker.
+          "${PYTEST[@]}" --maxfail=3 --numprocesses=1 \
+            ./python/jit/test_tilelang_jit_cutedsl.py
+
+      # AMD ROCm tests
+      - name: Run ROCm tests with Python ${{ matrix.python-version }} (${{ matrix.runner.toolkit }})
+        id: rocm-tests
+        if: contains(matrix.runner.toolkit, 'ROCm')
+        run: |
+          cd testing
+          PYTEST=(
+            uv run --no-project -m --
+            pytest --verbose --color=yes --durations=0 --showlocals --cache-clear
+          )
+          "${PYTEST[@]}" --maxfail=3 --numprocesses=4 \
+            ./python/amd
+
+      # Apple Metal tests
+      - name: Run Metal tests with Python ${{ matrix.python-version }} (${{ matrix.runner.toolkit }})
+        id: metal-tests
+        if: contains(matrix.runner.toolkit, 'Metal')
+        run: |
+          cd testing
+          PYTEST=(
+            uv run --no-project -m --
+            pytest --verbose --color=yes --durations=0 --showlocals --cache-clear
+          )
+          "${PYTEST[@]}" --maxfail=3 --numprocesses=4 \
+            -k metal \
+            ./python
+
+      - name: List generated files
+        if: ${{ !cancelled() }}
+        run: |
+          find . -type f -name '*.py[co]' -delete
+          find . -depth -type d -name "__pycache__" -exec rm -r "{}" +
+          if git status --ignored --porcelain | grep -qvE '/$'; then
+            ls -alh $(git status --ignored --porcelain | grep -vE '/$' | grep -oE '\S+$')
+          fi

.github/workflows/dist.yml

+name: Dist
+on:
+  workflow_dispatch:
+  schedule:
+    # gemini said this is 6:00 china time
+    - cron: "0 22 * * *"
+  pull_request:
+    types:
+      - opened
+      - synchronize
+      - reopened
+      - ready_for_review
+    paths:
+      - setup.py
+      - setup.cfg
+      - pyproject.toml
+      - MANIFEST.in
+      - CMakeLists.txt
+      - version_provider.py
+      - .github/workflows/dist.yml
+  release:
+    types:
+      - published
+
+permissions:
+  contents: read
+
+concurrency:
+  group: "${{ github.workflow }}-${{ github.ref }}"
+  cancel-in-progress: true
+
+env:
+  PYTHONDEVMODE: "1"
+  PYTHONUNBUFFERED: "1"
+  COLUMNS: "100"
+  FORCE_COLOR: "1"
+  CLICOLOR_FORCE: "1"
+
+jobs:
+  build-sdist:
+    name: Build SDist
+    if: |
+      github.repository_owner == 'tile-ai' &&
+      (github.event_name != 'pull_request' || !github.event.pull_request.draft)
+    runs-on: macos-latest
+    timeout-minutes: 30
+    env:
+      # `NO_VERSION_LABEL=ON` disables embedding the toolchain / git commit hash in version metadata.
+      # Otherwise, the version of the SDist has a git hash suffix (e.g., 0.1.0+gitabcdef12),
+      # but the package built from the SDist has no way to get the git hash (it is not a git repo),
+      # leading to inconsistent versions between SDist and built packages (+gitabcdef12 vs. +gitunknown).
+      NO_VERSION_LABEL: 'ON'
+
+    steps:
+      - name: Checkout repository
+        uses: actions/checkout@v6
+        with:
+          fetch-depth: 1
+          submodules: recursive
+
+      - name: Setup Python and uv with caching
+        id: setup-uv
+        uses: astral-sh/setup-uv@v7
+        with:
+          python-version: "3.12"
+          activate-environment: true
+
+      - name: Build SDist
+        run: |
+          uv run --no-project --with=build -m -- build --sdist --outdir=dist
+
+      - name: Setup ccache
+        uses: hendrikmuhs/ccache-action@v1
+        with:
+          create-symlink: true
+          evict-old-files: "7d"
+          append-timestamp: false
+          key: sdist-${{ runner.os }}-${{ runner.arch }}-${{ hashFiles('**/*.cc') }}
+          restore-keys: |
+            sdist-${{ runner.os }}-${{ runner.arch }}-${{ hashFiles('**/*.cc') }}
+            sdist-${{ runner.os }}-${{ runner.arch }}
+            ${{ runner.os }}-${{ runner.arch }}
+
+      - name: Test SDist buildable
+        run: |
+          TEMP_DIR="$(mktemp -d -t tilelang-sdist-test)"
+          cp -r dist "${TEMP_DIR}/dist"
+          cd "${TEMP_DIR}"
+          uv pip install -v dist/*.tar.gz
+          python3 -c "import tilelang; print(tilelang.__version__)"
+
+      - name: Upload SDist
+        # Not PR to save artifact storage, as SDist is only needed for releases.
+        if: github.event_name != 'pull_request' || contains(github.event.pull_request.title, '[Release]')
+        uses: actions/upload-artifact@v6
+        with:
+          name: sdist
+          path: dist/*.tar.gz
+          if-no-files-found: error
+
+  build-wheels:
+    name: Build wheels for Python ${{ matrix.python-version }} on ${{ matrix.target.runner }} with ${{ matrix.target.toolkit }}
+    if: |
+      github.repository_owner == 'tile-ai' &&
+      (github.event_name != 'pull_request' || !github.event.pull_request.draft)
+    strategy:
+      matrix:
+        target:
+          - { runner: ubuntu-latest, toolkit: "CUDA-12.8" }
+          - { runner: ubuntu-24.04-arm, toolkit: "CUDA-12.8" }
+          - { runner: macos-latest, toolkit: "Metal" }
+        python-version:
+          # Wheels are built with Python 3.8 Limited API, they should work with all Python >= 3.8.
+          # Only build wheels against Python 3.8 Limited API to save CI resources.
+          - "3.9"
+      fail-fast: false
+    timeout-minutes: 120
+    runs-on: ${{ matrix.target.runner }}
+    env:
+      NO_VERSION_LABEL: ${{ github.event_name == 'release' && 'OFF' || 'ON' }}
+
+    steps:
+      - name: Checkout repository
+        uses: actions/checkout@v6
+        with:
+          fetch-depth: 1
+          submodules: recursive
+
+      - name: Setup ccache
+        uses: hendrikmuhs/ccache-action@v1
+        with:
+          create-symlink: true
+          evict-old-files: "7d"
+          append-timestamp: false
+          key: wheel-${{ runner.os }}-${{ runner.arch }}-${{ matrix.target.toolkit }}-${{ hashFiles('**/*.cc') }}
+          restore-keys: |
+            wheel-${{ runner.os }}-${{ runner.arch }}-${{ matrix.target.toolkit }}-${{ hashFiles('**/*.cc') }}
+            wheel-${{ runner.os }}-${{ runner.arch }}-${{ matrix.target.toolkit }}
+            wheel-${{ runner.os }}-${{ runner.arch }}
+            ${{ runner.os }}-${{ runner.arch }}-${{ matrix.target.toolkit }}
+            ${{ runner.os }}-${{ runner.arch }}
+
+      - name: Set CIBW_BUILD
+        run: |
+          PYTHON_VERSION="${{ matrix.python-version }}"
+          PYTHON_VERSION_MAJMIN="$(echo "${PYTHON_VERSION}" | cut -d '.' -f-2)"
+          PYTHON_VERSION_MAJMIN_NODOT="${PYTHON_VERSION_MAJMIN//./}"
+          echo "CIBW_BUILD=cp${PYTHON_VERSION_MAJMIN_NODOT}-*" | tee -a "${GITHUB_ENV}"
+
+          if [[ "${{ matrix.target.toolkit }}" == *"CUDA"* ]]; then
+            CUDA_VERSION="${{ matrix.target.toolkit }}"
+            CUDA_VERSION="${CUDA_VERSION#CUDA-}"
+            echo "CUDA_VERSION=${CUDA_VERSION}" | tee -a "${GITHUB_ENV}"
+          fi
+
+          if [[ "${{ runner.os }}" == "Linux" ]]; then
+            HOST_CCACHE_DIR="$(ccache --get-config cache_dir)"
+            echo "CIBW_BEFORE_BUILD_LINUX=yum install -y ccache && ccache -o cache_dir=/host${HOST_CCACHE_DIR}" | tee -a "${GITHUB_ENV}"
+          fi
+
+      - name: Build wheels
+        uses: pypa/cibuildwheel@v3.3
+        with:
+          package-dir: .
+          output-dir: wheelhouse
+          config-file: "{package}/pyproject.toml"
+
+      - name: Upload wheels
+        # Not PR to save artifact storage, as wheels are only needed for releases.
+        if: github.event_name != 'pull_request' || contains(github.event.pull_request.title, '[Release]')
+        uses: actions/upload-artifact@v6
+        with:
+          name: wheels-${{ matrix.python-version }}-${{ runner.os }}-${{ runner.arch }}-${{ matrix.target.toolkit }}
+          path: wheelhouse/*.whl
+          if-no-files-found: error
+
+  list-artifacts:
+    name: List artifacts
+    # Not PR to save artifact storage, as artifacts are only needed for releases.
+    if: github.event_name != 'pull_request' || contains(github.event.pull_request.title, '[Release]')
+    runs-on: ubuntu-latest
+    needs: [build-sdist, build-wheels]
+    timeout-minutes: 15
+    steps:
+      - name: Download built SDist
+        uses: actions/download-artifact@v7
+        with:
+          # unpacks default artifact into dist/
+          # if `name: artifact` is omitted, the action will create extra parent dir
+          name: sdist
+          path: dist
+
+      - name: Download built wheels
+        uses: actions/download-artifact@v7
+        with:
+          pattern: wheels-*
+          path: dist
+          merge-multiple: true
+
+      - name: List distributions
+        run: ls -lh dist/*
+
+      - name: Upload artifacts
+        uses: actions/upload-artifact@v6
+        with:
+          name: artifacts
+          path: dist/*
+          if-no-files-found: error

.github/workflows/pr-perfbench-bot.yml

+name: Performance Benchmark Bot
+
+on:
+  issue_comment:
+    types:
+      - created
+
+permissions:
+  contents: read
+
+concurrency:
+  group: "${{ github.workflow }}-${{ github.ref }}"
+  cancel-in-progress: true # always cancel in-progress
+
+env:
+  PYTHONDEVMODE: "1"
+  PYTHONUNBUFFERED: "1"
+  PYTHONPATH: "" # explicit cleanup
+  PIP_USER: "" # explicit cleanup
+  COLUMNS: "100"
+  FORCE_COLOR: "1"
+  CLICOLOR_FORCE: "1"
+  XDG_CACHE_HOME: "${{ github.workspace }}/.cache" # to be updated
+  PIP_CACHE_DIR: "${{ github.workspace }}/.cache/pip" # to be updated
+
+jobs:
+  perfbench:
+    name: Benchmark between PR and main
+    if: |
+      github.repository_owner == 'tile-ai' &&
+      github.event.issue.pull_request &&
+      (contains(github.event.comment.body, '/performance-report') || contains(github.event.comment.body, '/perf'))
+    runs-on: [self-hosted, nvidia]
+    steps:
+      - name: Checkout repository
+        uses: actions/checkout@v6
+        with:
+          ref: refs/pull/${{ github.event.issue.number }}/merge
+          fetch-depth: 0
+          submodules: recursive
+
+      - name: Setup Python
+        uses: actions/setup-python@v6
+        with:
+          python-version: "3.12"
+          update-environment: true
+          cache: pip
+          cache-dependency-path: |
+            pyproject.toml
+            requirements*.txt
+
+      - name: Install merged version
+        run: |
+          python -m venv tll
+          source tll/bin/activate
+          pip install -r requirements-test.txt
+          pip install .
+
+      - name: Install original version
+        run: |
+          echo "Check files to be deleted!"
+          git clean -dxf -e tll/
+          echo "Delete files completed!"
+          git checkout main
+          python -m venv tl
+          source tl/bin/activate
+          pip install -r requirements-test.txt
+          pip install .
+
+      - name: Run performance test
+        id: perfbench
+        run: |
+          source tl/bin/activate
+          python maint/scripts/ci_performance.py
+
+      - name: Post test results as PR comment
+        uses: actions/github-script@v8
+        with:
+          github-token: ${{ secrets.GITHUB_TOKEN }}
+          script: |
+            github.rest.issues.createComment({
+              owner: context.repo.owner,
+              repo: context.repo.repo,
+              issue_number: context.issue.number,
+              body: '📊 ​**Performance Test Results** (triggered by @' + context.payload.comment.user.login + '):\n\n' +
+                'Run listed here: ${{ github.server_url }}/${{ github.repository }}/actions/runs/${{ github.run_id }}\n\n' +
+                "${{ steps.perfbench.outputs.stdout }}"
+            })

.github/workflows/pr-reminder-bot.yml

+name: PR Reminder Bot
+
+on:
+  pull_request_target:
+    types:
+      - opened
+
+jobs:
+  remind:
+    runs-on: ubuntu-latest
+    if: github.repository_owner == 'tile-ai'
+    steps:
+      - name: Remind
+        uses: actions/github-script@v8
+        with:
+          github-token: ${{ secrets.GITHUB_TOKEN }}
+          script: |
+            github.rest.issues.createComment({
+              owner: context.repo.owner,
+              repo: context.repo.repo,
+              issue_number: context.issue.number,
+              body: '👋 Hi! Thank you for contributing to the **TileLang** project.\n\n' +
+                'Please remember to run `pre-commit run --all-files` in the root directory of the project ' +
+                'to ensure your changes are properly linted and formatted. ' +
+                'This will help ensure your contribution passes the format check.\n\n' +
+                'We appreciate you taking this step! ' +
+                'Our team will review your contribution, and we look forward to your awesome work! 🚀'
+            })

.github/workflows/publish-docs.yml

+name: Documentation
+
+on:
+  pull_request_target:
+    types:
+      - closed
+  workflow_dispatch:
+
+permissions:
+  contents: write
+
+jobs:
+  docs:
+    name: Build and Publish Docs
+    if: |
+      github.repository_owner == 'tile-ai' &&
+      (
+        (
+          github.event_name == 'pull_request_target' &&
+          github.event.pull_request.merged == true &&
+          github.event.pull_request.base.ref == 'main'
+        ) ||
+        github.event_name == 'workflow_dispatch'
+      )
+    runs-on: [self-hosted, nvidia]
+    steps:
+      - name: Checkout repository
+        uses: actions/checkout@v6
+        with:
+          fetch-depth: 0
+          submodules: recursive
+
+      - name: Setup Python
+        uses: actions/setup-python@v6
+        with:
+          python-version: "3.10"
+
+      - name: Build docs
+        run: |
+          bash -ex maint/scripts/build_docs.sh
+
+      - name: Push built docs to another repo
+        run: |
+          # Hide sensitive info in logs
+          echo "::add-mask::${{ secrets.TARGET_TOKEN }}"
+          echo "::add-mask::${{ secrets.TARGET_REPO }}"
+          TARGET_REPO_URL="https://github.com/${{ secrets.TARGET_REPO }}.git"
+
+          git clone "${TARGET_REPO_URL}" -b main target_repo
+          cd target_repo
+          git config --local user.name "github-actions[bot]"
+          git config --local user.email "github-actions[bot]@users.noreply.github.com"
+          find . -mindepth 1 -maxdepth 1 ! -name ".github" ! -name "." ! -name ".git" -exec rm -rf {} +
+          cp -r ../docs/_build/html/* ./
+          git add .
+          if [[ -n "$(git status --porcelain)" ]]; then
+            # If there are changes, commit and push
+            git commit -m "Update docs"
+            git push "https://github-actions[bot]:${{ secrets.TARGET_TOKEN }}@${TARGET_REPO_URL##*://}" main
+          else
+            echo "No changes detected, skipping commit and push."
+          fi

.gitignore

+# Compiled Object files
+*.slo
+*.lo
+*.o
+*.so
+*.obj
+*.pyc
+
+# Precompiled Headers
+*.gch
+*.pch
+
+# emacs
+*~
+
+# vim
+*.swp
+*.swo
+
+debug/
+build/
+*dist/
+dist*/
+wheelhouse/
+__pycache__
+nnfusion.tar.gz
+
+# makeenv and test intermediate files
+tmp/
+
+.env
+.envrc
+.venv
+env/
+venv/
+ENV/
+env.bak/
+venv.bak/
+.vscode/
+.vs/
+
+# VisualGDB files
+VisualGDB/
+toolchain.cmake
+
+# docbuild artifacts
+doc/sphinx/build/*
+doc/doxygen/*.xml
+doc/doxygen/*.html
+doc/doxygen/man/*
+doc/doxygen/latex/*
+doc/doxygen/xml/*
+doc/doxygen/html/*
+
+# git merge
+*.orig
+\#*
+\.#*
+
+# idea
+.idea/*
+
+# python egg
+*.egg-info
+
+# Macos
+**/.DS_Store
+
+nnfusion_rt/
+models/frozenmodels/
+
+# log
+*.log
+
+# pkl
+*.pkl_*
+
+# .pytest_cache
+.pytest_cache
+
+# .hypothesis
+.hypothesis
+
+# .ruff_cache
+.ruff_cache
+
+# exclude debug testing folder
+!testing/python/debug
+
+# ignore lib with develop mode
+tilelang/lib
+
+# cython
+tilelang/jit/adapter/cython/.cycache
+
+# cache directory for clangd
+.cache/
+
+# claude
+**/.claude
+
+# CMake
+cmake-build/
+cmake-build-*/
+
+# Git version for sdist
+.git_commit.txt
+
+# pre-commit cache
+.pre-commit-cache/*
+
+# host checks logs
+maint/host_checks/logs/*
+
+# ncu
+*.ncu-rep
+
+# csv
+*.csv
+
+# clang-tidy
+/run-clang-tidy.py

.gitmodules

+[submodule "3rdparty/cutlass"]
+	path = 3rdparty/cutlass
+	url = https://github.com/NVIDIA/cutlass
+[submodule "3rdparty/tvm"]
+	path = 3rdparty/tvm
+	url = https://github.com/TileLang/tvm
+[submodule "3rdparty/composable_kernel"]
+	path = 3rdparty/composable_kernel
+	url = https://github.com/ROCm/composable_kernel

.pre-commit-config.yaml

+# See https://pre-commit.com for more information
+# See https://pre-commit.com/hooks.html for more hooks
+ci:
+  autofix_prs: false
+  autofix_commit_msg: "[Lint]: [pre-commit.ci] auto fixes [...]"
+  autoupdate_commit_msg: "[CI] [pre-commit.ci] autoupdate"
+  autoupdate_schedule: monthly
+default_stages: [pre-commit, pre-push, manual]
+exclude: '^(build|3rdparty)/.*$'  # exclude build and 3rdparty directories
+repos:
+  - repo: https://github.com/pre-commit/pre-commit-hooks
+    rev: v6.0.0
+    hooks:
+      - id: check-symlinks
+      - id: destroyed-symlinks
+      # FIXME: enable these hooks
+      # - id: trailing-whitespace
+      # - id: end-of-file-fixer
+      - id: check-added-large-files
+      - id: check-merge-conflict
+        fail_fast: true
+      # FIXME: enable these hooks
+      # - id: check-executables-have-shebangs
+      # - id: check-shebang-scripts-are-executable
+      - id: detect-private-key
+      - id: check-yaml
+      - id: check-toml
+      - id: check-ast
+        fail_fast: true
+      - id: debug-statements
+      - id: file-contents-sorter
+        args: [--ignore-case]
+        files: ^docs/spelling_wordlist\.txt$
+  - repo: https://github.com/pre-commit/mirrors-clang-format
+    rev: v21.1.7  # sync with requirements-lint.txt
+    hooks:
+      - id: clang-format
+        types_or: [c++, c]
+  - repo: https://github.com/astral-sh/ruff-pre-commit
+    rev: v0.14.9  # sync with requirements-lint.txt
+    hooks:
+      - id: ruff-check
+        args: [--fix, --exit-non-zero-on-fix]
+      - id: ruff-format
+        args: [--exit-non-zero-on-format]
+  - repo: https://github.com/codespell-project/codespell
+    rev: v2.4.1  # sync with requirements-lint.txt
+    hooks:
+      - id: codespell
+        additional_dependencies: [".[toml]"]
+        exclude: |
+          (?x)(
+            ^.+\.(cpp|hpp|cxx|cc|c|h|cu|cuh)$|
+            ^.+\.svg$|
+            ^.*\brequirements\b.*\.txt$
+          )

3rdparty/.gitignore

+clang*
+
+llvm*

composable_kernel @ b38bb492

+Subproject commit b38bb492a1a55b5abb0c345962143c0f9c482cfb