Fix the order of hosts in the ring (#2718)

Co-authored-by: Joseph Heck <j_heck@apple.com>

.circleci/config.yml

@@ -26,9 +26,9 @@ jobs:
           name: Install
           command: |
             xcodebuild -downloadComponent MetalToolchain
-            brew install python@3.9
+            brew install python@3.10
             brew install doxygen
-            python3.9 -m venv env
+            python3.10 -m venv env
             source env/bin/activate
             pip install --upgrade pip
             pip install --upgrade cmake
@@ -140,7 +140,7 @@ jobs:
       - run:
           name: Install Python package
           command: |
-            uv venv --python 3.9
+            uv venv --python 3.10
             uv pip install \
               nanobind==2.4.0 \
               cmake \
@@ -230,6 +230,9 @@ jobs:
             sudo mv ccache-4.11.3-linux-x86_64/ccache /usr/bin/ccache
             rm -rf ccache-4.11.3-linux-x86_64
             curl -LsSf https://astral.sh/uv/install.sh | sh
+      - run:
+          name: Set CCache size
+          command: ccache --max-size 1G
       - run:
           name: Install Python package
           command: |
@@ -260,7 +263,6 @@ jobs:
           command: |
             ccache --show-stats
             ccache --zero-stats
-            ccache --max-size 400MB
             ccache --cleanup
       - save_cache:
           key: cuda-<< parameters.image_date >>-{{ arch }}-{{ epoch }}
@@ -271,7 +273,7 @@ jobs:
     parameters:
       python_version:
         type: string
-        default: "3.9"
+        default: "3.10"
       xcode_version:
         type: string
         default: "26.0.0"
@@ -326,7 +328,7 @@ jobs:
             << parameters.build_env >> MLX_BUILD_STAGE=1 python -m build -w
       - when:
           condition:
-            equal: ["3.9", << parameters.python_version >>]
+            equal: ["3.10", << parameters.python_version >>]
           steps:
             - run:
                 name: Build common package
@@ -349,7 +351,7 @@ jobs:
     parameters:
       python_version:
         type: string
-        default: "3.9"
+        default: "3.10"
       build_env:
         type: string
         default: ""
@@ -385,7 +387,7 @@ jobs:
             bash python/scripts/repair_linux.sh
       - when:
           condition:
-            equal: ["3.9", << parameters.python_version >>]
+            equal: ["3.10", << parameters.python_version >>]
           steps:
             - run:
                 name: Build common package
@@ -482,7 +484,7 @@ workflows:
               ignore: /.*/
           matrix:
             parameters:
-              python_version: ["3.9", "3.10", "3.11", "3.12", "3.13"]
+              python_version: ["3.10", "3.11", "3.12", "3.13", "3.14"]
               macosx_deployment_target: ["13.5", "14.0", "15.0"]
               build_env: ["PYPI_RELEASE=1"]
               xcode_version: ["26.0.0"]
@@ -501,7 +503,7 @@ workflows:
               ignore: /.*/
           matrix:
             parameters:
-              python_version: ["3.9", "3.10", "3.11", "3.12", "3.13"]
+              python_version: ["3.10", "3.11", "3.12", "3.13", "3.14"]
               build_env: ["PYPI_RELEASE=1"]
       - build_cuda_release:
           filters:
@@ -544,13 +546,13 @@ workflows:
       - build_release:
           matrix:
             parameters:
-              python_version: ["3.9", "3.10", "3.11", "3.12", "3.13"]
+              python_version: ["3.10", "3.11", "3.12", "3.13", "3.14"]
               macosx_deployment_target: ["13.5", "14.0", "15.0"]
               xcode_version: ["26.0.0"]
       - build_linux_release:
           matrix:
             parameters:
-              python_version: ["3.9", "3.10", "3.11", "3.12", "3.13"]
+              python_version: ["3.10", "3.11", "3.12", "3.13", "3.14"]
       - build_cuda_release
 
   build_dev_release:
@@ -562,14 +564,14 @@ workflows:
       - build_release:
           matrix:
             parameters:
-              python_version: ["3.9", "3.10", "3.11", "3.12", "3.13"]
+              python_version: ["3.10", "3.11", "3.12", "3.13", "3.14"]
               macosx_deployment_target: ["13.5", "14.0", "15.0"]
               build_env: ["DEV_RELEASE=1"]
               xcode_version: ["26.0.0"]
       - build_linux_release:
           matrix:
             parameters:
-              python_version: ["3.9", "3.10", "3.11", "3.12", "3.13"]
+              python_version: ["3.10", "3.11", "3.12", "3.13", "3.14"]
               build_env: ["DEV_RELEASE=1"]
       - build_cuda_release:
           matrix:

.github/actions/build-cuda-release/action.yml

@@ -0,0 +1,24 @@
+name: 'Build CUDA wheel'
+description: 'Build CUDA wheel'
+
+inputs:
+  nvcc-location:
+    description: 'Location of nvcc compiler'
+    required: true
+
+runs:
+  using: "composite"
+  steps:
+    - name: Build package
+      shell: bash
+      env:
+        MLX_BUILD_STAGE: 2
+        CMAKE_ARGS: -DMLX_BUILD_CUDA=ON -DCMAKE_CUDA_COMPILER=${{ inputs.nvcc-location }}
+      run: |
+        pip install auditwheel build patchelf setuptools
+        python setup.py clean --all
+        python -m build -w
+
+        if [ -f "python/scripts/repair_cuda.sh" ]; then
+          bash python/scripts/repair_cuda.sh
+        fi

.github/actions/build-cuda/action.yml

@@ -0,0 +1,68 @@
+name: 'Build and Test with CUDA'
+description: 'Build and test MLX with CUDA'
+
+inputs:
+  build-type:
+    description: 'Build type (debug, release)'
+    required: false
+    default: 'debug'
+  run-tests:
+    description: 'Whether to run tests'
+    required: false
+    default: 'true'
+  nvcc-location:
+    description: 'Location of nvcc compiler'
+    required: true
+    default: '/usr/local/cuda-12.9/bin/nvcc'
+    # this value is dependent on the CUDA tools installed in the setup-linux workflow
+
+runs:
+  using: "composite"
+  steps:
+    - name: Install Python package
+      shell: bash
+      env:
+        DEBUG: 1
+        CMAKE_ARGS: -DMLX_BUILD_CUDA=ON -DCMAKE_COMPILE_WARNING_AS_ERROR=ON -DCMAKE_CUDA_COMPILER=${{ inputs.nvcc-location }}
+      run: pip install -e ".[dev]" -v
+
+    - name: Check if build actually worked
+      shell: bash
+      run: python -c "import mlx.core"
+
+    - name: Run Python tests - CPU
+      if: inputs.run-tests == 'true'
+      shell: bash
+      env:
+        LOW_MEMORY: 1
+        DEVICE: cpu
+      run: python -m unittest discover python/tests -v
+
+    - name: Run Python tests - GPU
+      if: inputs.run-tests == 'true'
+      shell: bash
+      env:
+        LOW_MEMORY: 1
+        DEVICE: gpu
+      run: python -m tests discover python/tests -v
+
+    - name: Build CPP only
+      if: inputs.build-type == 'debug'
+      shell: bash
+      run: |
+        cmake . -B build \
+          -DMLX_BUILD_CUDA=ON \
+          -DCMAKE_CUDA_COMPILER=${{ inputs.nvcc-location }} \
+          -DCMAKE_BUILD_TYPE=DEBUG
+        cmake --build build -j $(nproc)
+    
+    - name: Run CPP tests
+      if: ${{ inputs.build-type == 'debug' && inputs.run-tests == 'true' }}
+      shell: bash
+      run: ./build/tests/tests -sfe="*fft_tests.cpp,*linalg_tests.cpp"
+
+    - name: Build Python package
+      if: inputs.build-type == 'release'
+      uses: ./.github/actions/build-cuda-release
+      with:
+        nvcc-location: ${{ inputs.nvcc-location }}

.github/actions/build-docs/action.yml

@@ -0,0 +1,38 @@
+name: 'Build Documentation'
+description: 'Build documentation on a mac'
+
+runs:
+  using: "composite"
+  steps:
+    - name: Setup machine
+      uses: ./.github/actions/setup-macos
+
+    - name: Install dependencies
+      shell: sh
+      run: |
+        brew install doxygen
+        uv pip install --upgrade pip cmake
+        uv pip install -r docs/requirements.txt
+        uv pip install . -v
+  
+    - name: Build documentation
+      shell: bash
+      run: |
+        source .venv/bin/activate
+        cd docs
+        doxygen
+        make html O=-W
+    
+    - name: Create artifact tar
+      shell: sh
+      run: tar -cf artifact.tar --cd docs/build/html -L .
+
+    # Do it manually because upload-pages-artifact requires gtar
+    - name: Upload artifact
+      id: upload-artifact
+      uses: actions/upload-artifact@v5
+      with:
+        name: github-pages
+        path: artifact.tar
+        retention-days: 1
+        if-no-files-found: error

.github/actions/build-linux/action.yml

@@ -0,0 +1,78 @@
+name: 'Build and Test on Linux'
+description: 'Build and test MLX on Linux'
+
+inputs:
+  build-type:
+    description: 'Build type'
+    required: false
+    default: 'debug'
+    type: choice
+    options:
+      - debug
+      - release
+  run-tests:
+    description: 'Whether to run tests'
+    required: false
+    default: 'true'
+    type: boolean
+
+runs:
+  using: "composite"
+  steps:
+    - name: Set DEBUG
+      shell: sh
+      if: inputs.build-type == 'debug'
+      run: echo "DEBUG=1" >> $GITHUB_ENV
+
+    - name: Install Python package
+      shell: sh
+      env:
+        CMAKE_ARGS: "-DCMAKE_COMPILE_WARNING_AS_ERROR=ON"
+      run: pip install -e ".[dev]" -v
+    
+    - name: Generate package stubs
+      shell: sh
+      run: |
+        pip install typing_extensions
+        python setup.py generate_stubs
+    
+    - name: Run Python tests
+      if: inputs.run-tests == 'true'
+      shell: bash
+      run: |
+        python -m unittest discover python/tests -v
+        mpirun --bind-to none --allow-run-as-root -host localhost:8 -np 8 python python/tests/mpi_test_distributed.py
+        mlx.launch --verbose -n 8 python/tests/ring_test_distributed.py -v 2> >(tee -a stderr.log >&2)
+        if grep -Fq '[WARN]' stderr.log ; then
+          grep -F '[WARN]' stderr.log
+          echo "Distributed ring test failed";
+          exit 1;
+        fi
+    
+    - name: Build CPP only
+      if: inputs.build-type == 'debug'
+      shell: bash
+      run: |
+        mkdir -p build && cd build
+        cmake .. -DMLX_BUILD_METAL=OFF -DCMAKE_BUILD_TYPE=DEBUG
+        make -j $(nproc)
+    
+    - name: Run CPP tests
+      if: ${{ inputs.build-type == 'debug' && inputs.run-tests == 'true' }}
+      shell: sh
+      run: ./build/tests/tests
+    
+    - name: Build Python package
+      if: inputs.build-type == 'release'
+      shell: bash
+      run: |
+        pip install auditwheel patchelf build
+        python setup.py clean --all
+        MLX_BUILD_STAGE=1 python -m build -w
+        if [ -f "python/scripts/repair_linux.sh" ]; then
+          bash python/scripts/repair_linux.sh
+        fi
+
+        python setup.py clean --all
+        MLX_BUILD_STAGE=2 python -m build -w
+        auditwheel repair dist/mlx_cpu*.whl --plat manylinux_2_35_x86_64

.github/actions/build-macos-release/action.yml

@@ -0,0 +1,22 @@
+name: 'Build macOS release'
+description: 'Build MLX releases macOS'
+
+inputs:
+  macos-target:
+    description: 'macOS build target'
+    required: false
+    default: '15.0'
+
+runs:
+  using: "composite"
+  steps:
+    - name: Build Python package(s)
+      shell: bash
+      env:
+        MACOSX_DEPLOYMENT_TARGET: ${{ inputs.macos-target }}
+      run: |
+        uv pip install build
+        uv run --no-project setup.py clean --all
+        MLX_BUILD_STAGE=1 uv run -m build -w
+        uv run --no-project setup.py clean --all
+        MLX_BUILD_STAGE=2 uv run -m build -w

.github/actions/build-macos/action.yml

@@ -0,0 +1,124 @@
+name: 'Build and Test on macOS'
+description: 'Build and test MLX on macOS'
+
+inputs:
+  build-type:
+    description: 'Build type (debug, release)'
+    required: false
+    default: 'debug'
+    type: choice
+    options:
+      - debug
+      - release
+  run-tests:
+    description: 'Whether to run tests'
+    required: false
+    default: 'true'
+  build-jit:
+    description: 'Whether to build with JIT'
+    required: false
+    default: 'true'
+
+runs:
+  using: "composite"
+  steps:
+    - name: Install dependencies
+      shell: sh
+      env:
+        DEBUG: 1
+        DEV_RELEASE: 1
+      run: |
+        uv pip install --upgrade pip cmake setuptools
+        uv pip install nanobind==2.4.0 \
+          numpy torch tensorflow unittest-xml-reporting
+        uv pip install -e . -v
+
+    - name: Generate package stubs
+      shell: bash
+      run: |
+        uv pip install typing_extensions
+        uv run --no-project setup.py generate_stubs
+
+    - name: Run Python tests
+      if: inputs.run-tests == 'true'
+      shell: bash
+      env:
+        LOW_MEMORY: 1
+      run: |
+        DEVICE=cpu uv run -m xmlrunner discover -v python/tests -o test-results/cpu
+        DEVICE=gpu METAL_DEVICE_WRAPPER_TYPE=1 METAL_DEBUG_ERROR_MODE=0 uv run -m xmlrunner discover -v python/tests -o test-results/gpu
+        mpirun --bind-to none -host localhost:8 -np 8 -x DYLD_LIBRARY_PATH=/opt/homebrew/lib/ python python/tests/mpi_test_distributed.py
+        mlx.launch --verbose -n 8 python/tests/ring_test_distributed.py -v 2> >(tee -a stderr.log >&2)
+        if $(grep "\[WARN\]" stderr.log); then echo "Distributed ring test failed"; exit 1; fi
+    
+    - name: Build example extension
+      if: inputs.run-tests == 'true'
+      shell: bash
+      run: |
+        cd examples/extensions
+        uv pip install -r requirements.txt
+        uv run --no-project setup.py build_ext --inplace
+        uv run --no-project test.py
+    
+    - name: Build CPP only
+      if: inputs.build-type == 'debug'
+      shell: bash
+      run: |
+        mkdir -p build
+        cd build
+        cmake ..
+        make -j $(sysctl -n hw.ncpu)
+    
+    - name: Run CPP tests
+      if: ${{ inputs.build-type == 'debug' && inputs.run-tests == 'true' }}
+      shell: bash
+      env:
+        DEVICE: gpu
+        METAL_DEVICE_WRAPPER_TYPE: 1
+        METAL_DEBUG_ERROR_MODE: 0
+      run: ./build/tests/tests
+    
+    - name: Build small binary with JIT
+      if: inputs.build-jit == 'true'
+      shell: bash
+      run: |
+        mkdir -p build
+        cd build
+        cmake .. -DCMAKE_BUILD_TYPE=MinSizeRel \
+          -DBUILD_SHARED_LIBS=ON \
+          -DMLX_BUILD_CPU=OFF \
+          -DMLX_BUILD_SAFETENSORS=OFF \
+          -DMLX_BUILD_GGUF=OFF \
+          -DMLX_METAL_JIT=ON
+        make -j $(sysctl -n hw.ncpu)
+    
+    - name: Run Python tests with JIT
+      if: ${{ inputs.build-jit == 'true' && inputs.run-tests == 'true' }}
+      shell: bash
+      env:
+        LOW_MEMORY: 1
+        DEVICE: gpu
+        METAL_DEVICE_WRAPPER_TYPE: 1
+        METAL_DEBUG_ERROR_MODE: 0
+      run: |
+        CMAKE_ARGS="-DMLX_METAL_JIT=ON" \
+          uv pip install -e . -v
+        uv run -m xmlrunner discover \
+            -v python/tests \
+            -o test-results/gpu_jit
+
+    - name: Build macOS 13 package
+      if: inputs.build-type == 'release'
+      uses: ./.github/actions/build-macos-release
+      with:
+        macos-target: 13.0
+    - name: Build macOS 14 package
+      if: inputs.build-type == 'release'
+      uses: ./.github/actions/build-macos-release
+      with:
+        macos-target: 14.0
+    - name: Build macOS 15 package
+      if: inputs.build-type == 'release'
+      uses: ./.github/actions/build-macos-release
+      with:
+        macos-target: 15.0

.github/actions/setup-linux/action.yml

@@ -0,0 +1,83 @@
+name: 'Setup Linux Environment'
+description: 'Install dependencies for Linux builds'
+
+inputs:
+  runner-type:
+    description: 'Whether to set this up as a linux or CUDA runner'
+    required: false
+    default: 'linux'
+    type: choice
+    options:
+      - linux
+      - cuda
+  python-version:
+    description: 'Version of python to set up'
+    required: false
+    default: '3.10'
+
+runs:
+  using: "composite"
+  steps:
+    - name: Free disk space
+      shell: sh
+      if: inputs.runner-type == 'linux'
+      run: sudo rm -rf "$AGENT_TOOLSDIRECTORY"
+
+    - name: Install common dependencies
+      env:
+        TZ: Etc/UTC
+      shell: bash
+      run: |
+        sudo apt-get update
+        sudo apt-get install -y libblas-dev liblapack-dev liblapacke-dev tzdata zip
+        sudo apt autoremove -y
+
+    - uses: actions/setup-python@v6
+      with:
+        python-version: ${{ inputs.python-version }}
+        cache: 'pip'
+
+    - name: setup python venv
+      shell: bash
+      run: |
+        python -m venv .venv
+        source .venv/bin/activate
+        echo PATH=$PATH >> $GITHUB_ENV
+        pip install --upgrade pip cmake
+
+    - name: Install MPI
+      if: inputs.runner-type == 'linux'
+      shell: bash
+      run: sudo apt-get install -y openmpi-bin openmpi-common libopenmpi-dev
+
+    - name: Network CUDA installation from packages
+      id: install-cuda
+      if: inputs.runner-type == 'cuda'
+      env:
+        TZ: Etc/UTC
+      shell: bash ## Specific to Ubuntu 22.04 & Architecture x86_64
+      run: |
+        wget https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2204/x86_64/cuda-keyring_1.1-1_all.deb
+        sudo dpkg -i cuda-keyring_1.1-1_all.deb
+        sudo apt-get update
+        sudo apt-get install -y libcudnn9-dev-cuda-12 libnccl2 libnccl-dev cuda-toolkit-12-9
+      # Note: This installs CUDA 12.9, which is the latest supported by cuDNN 9.x and works with the NVidia 570 drivers
+      # cuda-toolkit by itself installs version 13 (+) and requires updated drives (580+), which require a reboot to function properly.
+      # Compatibility matrix: https://docs.nvidia.com/deeplearning/cudnn/backend/latest/reference/support-matrix.html
+      # This also drops `nvcc` into `/usr/local/cuda-12.9/bin/nvcc` - but it's *not* on the default PATH
+
+    - name: Package and Driver Report
+      if: inputs.runner-type == 'cuda'
+      shell: bash
+      run: |
+        sudo apt-get install -y ubuntu-drivers-common dkms
+        echo "NVIDIA Driver Packages Available:"
+        sudo ubuntu-drivers list --gpgpu
+        echo "NVIDIA Driver Version:"
+        cat /proc/driver/nvidia/version || echo "nvidia driver not found"
+        echo "Installed NVIDIA and CUDA packages:"
+        dpkg -l | egrep "cuda|nvidia" -i
+        echo "DKMS Status:"
+        dkms status || echo "dkms not found"
+        echo "NVIDIA-SMI Status:"
+        nvidia-smi || echo "nvidia-smi not found"

.github/actions/setup-macos/action.yml

@@ -0,0 +1,31 @@
+name: 'Setup macOS Environment'
+description: 'Install dependencies for macOS builds'
+
+inputs:
+  install-mpi:
+    description: 'Whether to install MPI'
+    required: false
+    default: 'true'
+    type: boolean
+  python-version:
+    description: 'Python version to use'
+    required: false
+    default: '3.10'
+
+runs:
+  using: "composite"
+  steps:
+    - name: Install Homebrew packages
+      shell: sh
+      if: inputs.install-mpi == 'true'
+      run: /opt/homebrew/bin/brew install openmpi
+    
+    - name: Verify MetalToolchain installed
+      shell: bash
+      run: xcodebuild -showComponent MetalToolchain
+    
+    - name: Setup uv
+      uses: astral-sh/setup-uv@v6
+      with:
+          python-version: ${{ inputs.python-version }}
+          activate-environment: true

.github/dependabot.yml

@@ -0,0 +1,6 @@
+version: 2
+updates:
+  - package-ecosystem: "github-actions"
+    directory: "/"
+    schedule:
+      interval: "weekly"

.github/workflows/documentation.yml

@@ -0,0 +1,28 @@
+name: Documentation
+
+on:
+  workflow_dispatch:
+
+permissions:
+  contents: read
+
+jobs:
+  build:
+    runs-on: [self-hosted, macos]
+    steps:
+      - uses: actions/checkout@v5
+      - uses: ./.github/actions/build-docs
+      
+  deploy:
+    needs: build
+    permissions:
+      pages: write
+      id-token: write
+    runs-on: ubuntu-latest
+    environment:
+      name: github-pages
+      url: ${{ steps.deployment.outputs.page_url }}
+    steps:
+      - name: Deploy to GitHub Pages
+        id: deployment
+        uses: actions/deploy-pages@v4

.github/workflows/nightly.yml

@@ -0,0 +1,93 @@
+name: Nightly Build
+
+on:
+  schedule:
+    - cron: 33 6 * * 1-5
+  workflow_dispatch:
+
+permissions:
+  contents: read
+
+jobs:
+  build_linux_release:
+    strategy:
+      fail-fast: false
+      matrix:
+        python_version: ["3.10", "3.14"]
+    runs-on: ubuntu-22.04
+    steps:
+      - uses: actions/checkout@v5
+      - uses: ./.github/actions/setup-linux
+      - uses: ./.github/actions/build-linux
+        with:
+          build-type: release
+          run-tests: false
+      - name: Upload mlx artifacts
+        uses: actions/upload-artifact@v5
+        with:
+          name: linux-wheels-${{ matrix.python_version }}
+          path: wheelhouse/mlx-*.whl
+          retention-days: 7
+      - name: Upload mlx-cpu artifacts
+        if: matrix.python_version == '3.10'
+        uses: actions/upload-artifact@v5
+        with:
+          name: mlx-cpu
+          path: wheelhouse/mlx_cpu-*.whl
+          retention-days: 7
+  
+  build_linux_with_tests:
+    strategy:
+      fail-fast: false
+      matrix:
+        python_version: ["3.10", "3.11", "3.12", "3.13", "3.14"]
+    runs-on: ubuntu-22.04
+    steps:
+      - uses: actions/checkout@v5
+      - uses: ./.github/actions/setup-linux
+        with:
+          python-version: ${{ matrix.python_version }}
+      - uses: ./.github/actions/build-linux
+
+  build_mac_release:
+    strategy:
+      matrix:
+        python-version: ["3.10", "3.13"]
+        # TODO: 3.14 had issues finding a compatible tensorflow
+    env:
+      MACOSX_DEPLOYMENT_TARGET: "15.0"
+    runs-on: [self-hosted, macos]
+    steps:
+      - uses: actions/checkout@v5
+      - uses: ./.github/actions/setup-macos
+        with:
+          python-version: ${{ matrix.python-version }}
+      - uses: ./.github/actions/build-macos
+
+  build_cuda_with_tests:
+    runs-on: gpu-t4-4-core
+    steps:
+      - uses: actions/checkout@v5
+      - uses: ./.github/actions/setup-linux
+        with:
+          runner-type: 'cuda'
+      - uses: ./.github/actions/build-cuda
+
+  build_cuda_release:
+    runs-on: ubuntu-22-large
+    steps:
+      - uses: actions/checkout@v5
+      - uses: ./.github/actions/setup-linux
+        with:
+          runner-type: 'cuda'
+      - name: Build Python package
+        uses: ./.github/actions/build-cuda-release
+        with:
+          nvcc-location: '/usr/local/cuda-12.9/bin/nvcc'
+      - name: Upload artifacts
+        uses: actions/upload-artifact@v5
+        with:
+          name: mlx-cuda
+          path: wheelhouse/mlx_cuda-*.whl
+          retention-days: 7
+

.github/workflows/pull_request.yml

@@ -1,20 +1,46 @@
-on:
-  pull_request:
-    branches:
-      - main
+name: Build and Test
+
+on: pull_request  
+
+permissions:
+  contents: read
 
 jobs:
   check_lint:
-    runs-on: ubuntu-latest
+    runs-on: ubuntu-22.04
     steps:
-      - uses: actions/checkout@v4
-      - uses: actions/setup-python@v4
+      - uses: actions/checkout@v5
+      - uses: ./.github/actions/setup-linux
+      - uses: pre-commit/action@v3.0.1
+
+  linux_build_and_test:
+    runs-on: ubuntu-22.04
+    steps:
+      - uses: actions/checkout@v5
+      - uses: ./.github/actions/setup-linux
+      - uses: ./.github/actions/build-linux
+
+  mac_build_and_test:
+    runs-on: [self-hosted, macos]
+    needs: check_lint
+    steps:
+      - uses: actions/checkout@v5
+      - uses: ./.github/actions/setup-macos
+      - uses: ./.github/actions/build-macos
+
+  cuda_build_and_test:
+    runs-on: gpu-t4-4-core
+    needs: check_lint
+    steps:
+      - uses: actions/checkout@v5
+      - uses: ./.github/actions/setup-linux
         with:
-          python-version: 3.8
-      - name: Install dependencies
-        run: |
-          python -m pip install --upgrade pip
-          pip install pre-commit black isort clang-format
-      - name: Run lint
-        run: |
-          pre-commit run --all-files
+          runner-type: 'cuda'
+      - uses: ./.github/actions/build-cuda
+
+  build_documentation:
+    runs-on: [self-hosted, macos]
+    needs: check_lint
+    steps:
+      - uses: actions/checkout@v5
+      - uses: ./.github/actions/build-docs

.github/workflows/release.yml

@@ -0,0 +1,188 @@
+name: PyPI Release
+
+on:
+  push:
+    tags:
+      - 'v*'
+  workflow_dispatch:
+
+permissions:
+  contents: read
+
+jobs:
+  build_documentation:
+    runs-on: [self-hosted, macos]
+    steps:
+      - uses: actions/checkout@v5
+      - uses: ./.github/actions/build-docs
+    
+  deploy_documentation:
+    needs: build_documentation
+    permissions:
+      pages: write
+      id-token: write
+    runs-on: ubuntu-latest
+    environment:
+      name: github-pages
+      url: ${{ steps.deployment.outputs.page_url }}
+    steps:
+      - name: Deploy to GitHub Pages
+        id: deployment
+        uses: actions/deploy-pages@v4
+
+  build_linux_release:
+    strategy:
+      matrix:
+        python_version: ["3.10", "3.11", "3.12", "3.13", "3.14"]
+    runs-on: ubuntu-22.04
+    env:
+      PYPI_RELEASE: 1
+    steps:
+      - uses: actions/checkout@v5
+      - uses: ./.github/actions/setup-linux
+        with:
+          python-version: ${{ matrix.python_version }}
+      - uses: ./.github/actions/build-linux
+        with:
+          build-type: release
+          run-tests: false
+      - name: Upload MLX artifacts
+        uses: actions/upload-artifact@v5
+        with:
+          name: linux-wheels-${{ matrix.python_version }}
+          path: wheelhouse/mlx-*.whl
+      - name: Upload CPU artifacts
+        if: matrix.python_version == '3.10'
+        uses: actions/upload-artifact@v5
+        with:
+          name: mlx-cpu
+          path: wheelhouse/mlx_cpu-*.whl
+  
+  build_mac_release:
+    strategy:
+      matrix:
+        python-version: ["3.10", "3.11", "3.12", "3.13"]
+        # TODO: 3.14 had issues finding a compatible tensorflow
+    runs-on: [self-hosted, macos]
+    env:
+      PYPI_RELEASE: 1
+    steps:
+      - uses: actions/checkout@v5
+      - uses: ./.github/actions/setup-macos
+        with:
+          python-version: ${{ matrix.python-version }}
+      - uses: ./.github/actions/build-macos
+        with:
+          build-type: release
+      - name: Upload MLX artifacts
+        uses: actions/upload-artifact@v5
+        with:
+          name: mac-wheels-${{ matrix.python-version }}
+          path: dist/mlx-*.whl
+      - name: Upload Metal artifacts
+        if: matrix.python-version == '3.10'
+        uses: actions/upload-artifact@v5
+        with:
+          name: mlx-metal
+          path: dist/mlx_metal-*.whl
+
+  build_cuda_release:
+    runs-on: ubuntu-22-large
+    env:
+      PYPI_RELEASE: 1
+    steps:
+      - uses: actions/checkout@v5
+      - uses: ./.github/actions/setup-linux
+        with:
+          runner-type: 'cuda'
+      - name: Build Python package
+        uses: ./.github/actions/build-cuda-release
+        with:
+          nvcc-location: '/usr/local/cuda-12.9/bin/nvcc'
+      - name: Upload artifacts
+        uses: actions/upload-artifact@v5
+        with:
+          name: mlx-cuda
+          path: wheelhouse/mlx_cuda-*.whl
+
+  pypi-publish:
+    name: Upload release to PyPI
+    runs-on: ubuntu-latest
+    needs: [build_linux_release, build_mac_release]
+    permissions:
+      id-token: write
+    environment:
+      name: pypi
+      url: https://pypi.org/p/mlx
+    steps:
+      - uses: actions/download-artifact@v6
+        with:
+          pattern: linux-wheels-*
+          merge-multiples: true
+          path: artifacts
+      - uses: actions/download-artifact@v6
+        with:
+          pattern: mac-wheels-*
+          merge-multiples: true
+          path: artifacts
+      - name: Display structure of downloaded files
+        run: ls -R artifacts
+      # - name: Publish package distributions to PyPI
+      #  uses: pypa/gh-action-pypi-publish@release/v1
+  
+  pypi-publish-cuda:
+    name: Upload CUDA release to PyPI
+    runs-on: ubuntu-latest
+    needs: build_cuda_release
+    permissions:
+      id-token: write
+    environment:
+      name: pypi
+      url: https://pypi.org/p/mlx-cuda
+    steps:
+      - uses: actions/download-artifact@v6
+        with:
+          name: mlx-cuda
+          path: artifacts
+      - name: Display structure of downloaded files
+        run: ls -R artifacts
+      # - name: Publish package distributions to PyPI
+      #  uses: pypa/gh-action-pypi-publish@release/v1
+
+  pypi-publish-cpu:
+    name: Upload CPU release to PyPI
+    runs-on: ubuntu-latest
+    needs: build_linux_release
+    permissions:
+      id-token: write
+    environment:
+      name: pypi
+      url: https://pypi.org/p/mlx-cpu
+    steps:
+      - uses: actions/download-artifact@v6
+        with:
+          name: mlx-cpu
+          path: artifacts
+      - name: Display structure of downloaded files
+        run: ls -R artifacts
+      # - name: Publish package distributions to PyPI
+      #  uses: pypa/gh-action-pypi-publish@release/v1
+
+  pypi-publish-metal:
+    name: Upload Metal release to PyPI
+    runs-on: ubuntu-latest
+    needs: build_mac_release
+    permissions:
+      id-token: write
+    environment:
+      name: pypi
+      url: https://pypi.org/p/mlx-metal
+    steps:
+      - uses: actions/download-artifact@v6
+        with:
+          name: mlx-metal
+          path: artifacts
+      - name: Display structure of downloaded files
+        run: ls -R artifacts
+      # - name: Publish package distributions to PyPI
+      #  uses: pypa/gh-action-pypi-publish@release/v1

.pre-commit-config.yaml

@@ -1,4 +1,10 @@
 repos:
+-   repo: https://github.com/pre-commit/pre-commit-hooks
+    rev: v6.0.0
+    hooks:
+    -   id: check-yaml
+    # -   id: end-of-file-fixer
+    # -   id: trailing-whitespace
 -   repo: https://github.com/pre-commit/mirrors-clang-format
     rev: v19.1.7
     hooks:

ACKNOWLEDGMENTS.md

@@ -19,7 +19,7 @@ MLX was developed with contributions from the following individuals:
 - Gleb Pobudzey: Added the `where` primitive, and groups in 1D and 2D convolutions.
 - Paul Paczuski: Improved stability of BCE loss calculation
 - Max-Heinrich Laves: Added `conv_transpose1d`, `conv_transpose2d`, and `conv_transpose3d` ops.
-- Gökdeniz Gülmez: Added the `Muon (MomentUm Orthogonalized by Newton-schulz)` optimizer.
+- Gökdeniz Gülmez: Added the `Muon (MomentUm Orthogonalized by Newton-schulz)` optimizer, and the `ReLU²` activation function.
 
 <a href="https://github.com/ml-explore/mlx/graphs/contributors">
   <img class="dark-light" src="https://contrib.rocks/image?repo=ml-explore/mlx&anon=0&columns=20&max=100&r=true" />

CMakeLists.txt

@@ -26,6 +26,7 @@ set(CMAKE_CXX_STANDARD 17)
 set(CMAKE_CXX_STANDARD_REQUIRED ON)
 set(CMAKE_POSITION_INDEPENDENT_CODE ON)
 set(CMAKE_INSTALL_MESSAGE NEVER)
+set(CMAKE_EXPORT_COMPILE_COMMANDS ON)
 
 # ----------------------------- Configuration -----------------------------
 option(MLX_BUILD_TESTS "Build tests for mlx" ON)
@@ -87,22 +88,26 @@ cmake_policy(SET CMP0135 NEW)
 
 add_library(mlx)
 
-if(MLX_BUILD_METAL)
-  set(METAL_LIB "-framework Metal")
-  set(FOUNDATION_LIB "-framework Foundation")
-  set(QUARTZ_LIB "-framework QuartzCore")
-endif()
+# Supress warnings: note: parameter passing for argument of type
+# ‘std::pair<float, float>’ when C++17 is enabled changed to match C++14 in GCC
+# 10.1
+target_compile_options(mlx PRIVATE -Wno-psabi)
 
 if(MLX_BUILD_CUDA)
   enable_language(CUDA)
 endif()
 
-if(MLX_BUILD_METAL AND NOT METAL_LIB)
-  message(STATUS "Metal not found. Unable to build GPU")
-  set(MLX_BUILD_METAL OFF)
-  set(MLX_METAL_DEBUG OFF)
-elseif(MLX_BUILD_METAL)
-  message(STATUS "Building METAL sources")
+if(MLX_BUILD_METAL)
+  find_library(METAL_LIB Metal)
+  find_library(FOUNDATION_LIB Foundation)
+  find_library(QUARTZ_LIB QuartzCore)
+  if(METAL_LIB)
+    message(STATUS "Metal found ${METAL_LIB}")
+  else()
+    message(
+      FATAL_ERROR
+        "Metal not found. Set MLX_BUILD_METAL=OFF to build without GPU")
+  endif()
 
   if(MLX_METAL_DEBUG)
     add_compile_definitions(MLX_METAL_DEBUG)
@@ -111,7 +116,8 @@ elseif(MLX_BUILD_METAL)
   # Throw an error if xcrun not found
   execute_process(
     COMMAND zsh "-c" "/usr/bin/xcrun -sdk macosx --show-sdk-version"
-    OUTPUT_VARIABLE MACOS_SDK_VERSION COMMAND_ERROR_IS_FATAL ANY)
+    OUTPUT_VARIABLE MACOS_SDK_VERSION
+    OUTPUT_STRIP_TRAILING_WHITESPACE COMMAND_ERROR_IS_FATAL ANY)
 
   if(${MACOS_SDK_VERSION} LESS 14.0)
     message(
@@ -173,7 +179,7 @@ if(MLX_BUILD_CPU)
     message(STATUS "Accelerate found ${ACCELERATE_LIBRARY}")
     set(MLX_BUILD_ACCELERATE ON)
   else()
-    message(STATUS "Accelerate or arm neon not found, using default backend.")
+    message(STATUS "Accelerate not found, using default backend.")
     set(MLX_BUILD_ACCELERATE OFF)
   endif()
 

README.md

@@ -2,7 +2,7 @@
 
 [**Quickstart**](#quickstart) | [**Installation**](#installation) |
 [**Documentation**](https://ml-explore.github.io/mlx/build/html/index.html) |
-[**Examples**](#examples) 
+[**Examples**](#examples)
 
 [![CircleCI](https://circleci.com/gh/ml-explore/mlx.svg?style=svg)](https://circleci.com/gh/ml-explore/mlx)
 
@@ -11,37 +11,37 @@ brought to you by Apple machine learning research.
 
 Some key features of MLX include:
 
- - **Familiar APIs**: MLX has a Python API that closely follows NumPy. MLX
+- **Familiar APIs**: MLX has a Python API that closely follows NumPy. MLX
    also has fully featured C++, [C](https://github.com/ml-explore/mlx-c), and
    [Swift](https://github.com/ml-explore/mlx-swift/) APIs, which closely mirror
    the Python API. MLX has higher-level packages like `mlx.nn` and
    `mlx.optimizers` with APIs that closely follow PyTorch to simplify building
    more complex models.
 
- - **Composable function transformations**: MLX supports composable function
-   transformations for automatic differentiation, automatic vectorization,
-   and computation graph optimization.
+- **Composable function transformations**: MLX supports composable function
+  transformations for automatic differentiation, automatic vectorization,
+  and computation graph optimization.
 
- - **Lazy computation**: Computations in MLX are lazy. Arrays are only
-   materialized when needed.
+- **Lazy computation**: Computations in MLX are lazy. Arrays are only
+  materialized when needed.
 
- - **Dynamic graph construction**: Computation graphs in MLX are constructed
-   dynamically. Changing the shapes of function arguments does not trigger
-   slow compilations, and debugging is simple and intuitive.
+- **Dynamic graph construction**: Computation graphs in MLX are constructed
+  dynamically. Changing the shapes of function arguments does not trigger
+  slow compilations, and debugging is simple and intuitive.
 
- - **Multi-device**: Operations can run on any of the supported devices
-   (currently the CPU and the GPU).
+- **Multi-device**: Operations can run on any of the supported devices
+  (currently the CPU and the GPU).
 
- - **Unified memory**: A notable difference from MLX and other frameworks
-   is the *unified memory model*. Arrays in MLX live in shared memory.
-   Operations on MLX arrays can be performed on any of the supported
-   device types without transferring data.
+- **Unified memory**: A notable difference from MLX and other frameworks
+  is the *unified memory model*. Arrays in MLX live in shared memory.
+  Operations on MLX arrays can be performed on any of the supported
+  device types without transferring data.
 
 MLX is designed by machine learning researchers for machine learning
 researchers. The framework is intended to be user-friendly, but still efficient
 to train and deploy models. The design of the framework itself is also
 conceptually simple. We intend to make it easy for researchers to extend and
-improve MLX with the goal of quickly exploring new ideas. 
+improve MLX with the goal of quickly exploring new ideas.
 
 The design of MLX is inspired by frameworks like
 [NumPy](https://numpy.org/doc/stable/index.html),
@@ -91,7 +91,7 @@ Checkout the
 [documentation](https://ml-explore.github.io/mlx/build/html/install.html#)
 for more information on building the C++ and Python APIs from source.
 
-## Contributing 
+## Contributing
 
 Check out the [contribution guidelines](https://github.com/ml-explore/mlx/tree/main/CONTRIBUTING.md) for more information
 on contributing to MLX. See the
@@ -110,7 +110,7 @@ Hannun, Jagrit Digani, Angelos Katharopoulos, and Ronan Collobert. If you find
 MLX useful in your research and wish to cite it, please use the following
 BibTex entry:
 
-```
+```text
 @software{mlx2023,
   author = {Awni Hannun and Jagrit Digani and Angelos Katharopoulos and Ronan Collobert},
   title = {{MLX}: Efficient and flexible machine learning on Apple silicon},

benchmarks/python/blas/bench_gemm.py

@@ -142,9 +142,7 @@ def bench_shape(B, M, N, K, np_dtype, transpose="nn"):
     t_b = (0, 1, 2) if transpose[1] == "n" else (0, 2, 1)
 
     c_mlx = a_mx.transpose(t_a) @ b_mx.transpose(t_b)
-    c_npy = a_np.transpose(t_a).astype(np.float32) @ b_np.transpose(t_b).astype(
-        np.float32
-    )
+    c_npy = a_np.transpose(t_a).astype(np_dtype) @ b_np.transpose(t_b).astype(np_dtype)
 
     atol = 1e-5 if np_dtype == np.float32 else 1e-4
 
@@ -163,7 +161,7 @@ def get_gflop_count(B, M, N, K):
 if __name__ == "__main__":
     parser = argparse.ArgumentParser(description="Run gemm benchmarks")
 
-    dtypes = ("float32", "float16")
+    dtypes = ("float32", "float16", "complex64")
     transposes = ("nn", "nt", "tn")
     shapes = (
         (16, 234, 768, 3072),
@@ -187,7 +185,7 @@ if __name__ == "__main__":
                 diff = gflops_mx / gflops_pt - 1.0
 
                 print(
-                    f"{B:3d}, {M:4d}, {N:4d}, {K:4d}, {dtype}, {transpose}, {gflops_pt:05.3f}, {gflops_mx:05.3f}, {100. * diff:+5.2f}%"
+                    f"{B:3d}, {M:4d}, {N:4d}, {K:4d}, {dtype}, {transpose}, {gflops_pt:05.3f}, {gflops_mx:05.3f}, {100.0 * diff:+5.2f}%"
                 )
                 if gflops_pt >= 2.0 * gflops_mx:
                     print("ATTENTION ^^^^^^^")

benchmarks/python/blas/bench_gemv.py

@@ -196,7 +196,7 @@ def bench_with_out_len(ax, out_vec_len, in_vector_lens, dtype, transpose):
 
 
 for transpose in (False, True):
-    for dtype in ("float32", "float16"):
+    for dtype in ("float32", "float16", "complex64"):
         fig, axs = plt.subplots(
             len(in_vec_sizes), 2, figsize=(8.5, 11), layout="constrained"
         )
@@ -215,7 +215,7 @@ for transpose in (False, True):
         fig.suptitle(f"{device_name}: {dtype} {op_name}")
         fig.savefig(
             os.path.join(
-                results_dir, f'{device_name.replace(" ", "_")}_{dtype}_{op_name}.pdf'
+                results_dir, f"{device_name.replace(' ', '_')}_{dtype}_{op_name}.pdf"
             )
         )
         plt.close(fig)

docs/src/install.rst

@@ -16,7 +16,7 @@ silicon computer is
 To install from PyPI your system must meet the following requirements:
 
 - Using an M series chip (Apple silicon)
-- Using a native Python >= 3.9
+- Using a native Python >= 3.10
 - macOS >= 13.5
 
 .. note::
@@ -39,7 +39,7 @@ requirements:
 - Nvidia driver >= 550.54.14
 - CUDA toolkit >= 12.0
 - Linux distribution with glibc >= 2.35
-- Python >= 3.9
+- Python >= 3.10
 
 
 CPU-only (Linux)
@@ -55,7 +55,7 @@ To install the CPU-only package from PyPi your system must meet the following
 requirements:
 
 - Linux distribution with glibc >= 2.35
-- Python >= 3.9
+- Python >= 3.10
 
 
 Troubleshooting

docs/src/python/nn/functions.rst

@@ -27,6 +27,7 @@ simple functions.
    mish
    prelu
    relu
+   relu2
    relu6
    selu
    sigmoid

docs/src/python/nn/layers.rst

@@ -50,6 +50,7 @@ Layers
    QuantizedLinear
    RMSNorm
    ReLU
+   ReLU2
    ReLU6
    RNN
    RoPE

docs/src/python/ops.rst

@@ -112,6 +112,7 @@ Operations
    max
    maximum
    mean
+   median
    meshgrid
    min
    minimum

docs/src/usage/compile.rst

@@ -130,8 +130,8 @@ Now make an array, and benchmark both functions:
 .. code-block:: python
 
   x = mx.random.uniform(shape=(32, 1000, 4096))
-  timeit(nn.gelu, x)
-  timeit(mx.compile(nn.gelu), x)
+  timeit(gelu, x)
+  timeit(mx.compile(gelu), x)
 
 On an M1 Max the times are 15.5 and 3.1 milliseconds. The compiled ``gelu`` is
 five times faster.

docs/src/usage/distributed.rst

@@ -184,7 +184,7 @@ almost identical to the example above:
 
     def step(model, x, y):
         loss, grads = loss_grad_fn(model, x, y)
-        grads = mlx.nn.average_gradients(grads) # <---- This line was added
+        grads = mx.nn.average_gradients(grads)  # <---- This line was added
         optimizer.update(model, grads)
         return loss
 

docs/src/usage/export.rst

@@ -164,11 +164,11 @@ to export a function which can be used for inputs with variable shapes:
 
 .. code-block:: python
 
-  mx.export_function("fun.mlxfn", mx.abs, mx.array(0.0), shapeless=True)
+  mx.export_function("fun.mlxfn", mx.abs, mx.array([0.0]), shapeless=True)
   imported_abs = mx.import_function("fun.mlxfn")
 
   # Ok
-  out, = imported_abs(mx.array(-1.0))
+  out, = imported_abs(mx.array([-1.0]))
 
   # Also ok
   out, = imported_abs(mx.array([-1.0, -2.0]))

docs/src/usage/indexing.rst

@@ -107,8 +107,20 @@ same array:
   >>> a
   array([1, 2, 0], dtype=int32)
 
+Note that unlike NumPy, slicing an array creates a copy, not a view. So
+mutating it does not mutate the original array:
 
-Note, unlike NumPy, updates to the same location are nondeterministic:
+.. code-block:: shell
+
+  >>> a = mx.array([1, 2, 3])
+  >>> b = a[:]
+  >>> b[2] = 0
+  >>> b
+  array([1, 2, 0], dtype=int32)
+  >>> a
+  array([1, 2, 3], dtype=int32)
+
+Also unlike NumPy, updates to the same location are nondeterministic:
 
 .. code-block:: shell
 

mlx/array.cpp

@@ -241,8 +241,8 @@ array::ArrayDesc::ArrayDesc(
     std::vector<array> inputs)
     : shape(std::move(shape)),
       dtype(dtype),
-      status(Status::unscheduled),
       primitive(std::move(primitive)),
+      status(Status::unscheduled),
       inputs(std::move(inputs)) {
   init();
 }

mlx/backend/common/matmul.h

@@ -13,7 +13,7 @@ inline std::tuple<Shape, Strides, Strides> collapse_batches(
     const array& a,
     const array& b) {
   if (a.ndim() == 2) {
-    return {{1}, {0}, {0}};
+    return {Shape{1}, Strides{0}, Strides{0}};
   }
 
   Shape A_bshape{a.shape().begin(), a.shape().end() - 2};
@@ -38,7 +38,7 @@ inline std::tuple<Shape, Strides, Strides> collapse_batches(
 inline std::tuple<Shape, Strides, Strides, Strides>
 collapse_batches(const array& a, const array& b, const array& c) {
   if (a.ndim() == 2) {
-    return {{1}, {0}, {0}, {0}};
+    return {Shape{1}, Strides{0}, Strides{0}, Strides{0}};
   }
 
   Shape A_bshape{a.shape().begin(), a.shape().end() - 2};

mlx/backend/common/ternary.h

@@ -11,6 +11,8 @@ namespace mlx::core {
 enum class TernaryOpType {
   ScalarScalarScalar,
   VectorVectorVector,
+  VectorVectorScalar,
+  VectorScalarVector,
   General,
 };
 
@@ -25,6 +27,14 @@ get_ternary_op_type(const array& a, const array& b, const array& c) {
       (a.flags().col_contiguous && b.flags().col_contiguous &&
        c.flags().col_contiguous)) {
     topt = TernaryOpType::VectorVectorVector;
+  } else if (
+      b.data_size() == 1 && a.flags().row_contiguous &&
+      c.flags().row_contiguous) {
+    topt = TernaryOpType::VectorScalarVector;
+  } else if (
+      c.data_size() == 1 && a.flags().row_contiguous &&
+      b.flags().row_contiguous) {
+    topt = TernaryOpType::VectorVectorScalar;
   } else {
     topt = TernaryOpType::General;
   }
@@ -59,6 +69,8 @@ inline void set_ternary_op_output_data(
             b.flags());
       }
       break;
+    case TernaryOpType::VectorVectorScalar:
+    case TernaryOpType::VectorScalarVector:
     case TernaryOpType::General:
       // Try to donate an input which is row_contiguous
       if (!((a.flags().row_contiguous && maybe_donate(a)) ||

mlx/backend/cpu/conv.cpp

@@ -996,131 +996,6 @@ void explicit_gemm_conv_1D_cpu(
   encoder.add_temporaries(std::move(temps));
 }
 
-void explicit_gemm_conv_2D_cpu(
-    const array& in,
-    const array& wt,
-    array out,
-    const std::vector<int>& padding_lo,
-    const std::vector<int>& padding_hi,
-    const std::vector<int>& wt_strides,
-    const std::vector<int>& wt_dilation,
-    Stream stream) {
-  const int N = in.shape(0); // Batch size, should be the same as out.shape(0)
-  const int iH = in.shape(1); // Input spatial dim
-  const int iW = in.shape(2); // Input spatial dim
-  const int oH = out.shape(1); // Output spatial dim
-  const int oW = out.shape(2); // Output spatial dim
-  const int O = wt.shape(0); // Out channels
-  const int C = wt.shape(3); // In channels
-  const int wH = wt.shape(1); // Weight spatial dim
-  const int wW = wt.shape(2); // Weight spatial dim
-
-  auto conv_dtype = out.dtype();
-  auto& encoder = cpu::get_command_encoder(stream);
-
-  // Pad input
-  Shape padded_shape = {
-      N,
-      iH + padding_lo[0] + padding_hi[0],
-      iW + padding_lo[1] + padding_hi[1],
-      C};
-  array in_padded(padded_shape, conv_dtype, nullptr, {});
-
-  // Fill with zeros
-  std::vector<array> temps;
-  temps.push_back(array(0, conv_dtype));
-  copy_cpu(temps.back(), in_padded, CopyType::Scalar, stream);
-
-  // Pick input slice from padded
-  size_t data_offset = padding_lo[0] * in_padded.strides()[1] +
-      padding_lo[1] * in_padded.strides()[2];
-  array in_padded_slice(in.shape(), in_padded.dtype(), nullptr, {});
-  in_padded_slice.copy_shared_buffer(
-      in_padded,
-      in_padded.strides(),
-      in_padded.flags(),
-      in_padded_slice.size(),
-      data_offset);
-  temps.push_back(in_padded_slice);
-
-  // Copy input values into the slice
-  copy_cpu_inplace(in, in_padded_slice, CopyType::GeneralGeneral, stream);
-
-  // Make strided view
-  Shape strided_shape = {N, oH, oW, wH, wW, C};
-
-  Strides strided_strides = {
-      in_padded.strides()[0],
-      in_padded.strides()[1] * wt_strides[0],
-      in_padded.strides()[2] * wt_strides[1],
-      in_padded.strides()[1],
-      in_padded.strides()[2],
-      in_padded.strides()[3]};
-  auto flags = in_padded.flags();
-
-  array in_strided_view(strided_shape, in_padded.dtype(), nullptr, {});
-  in_strided_view.copy_shared_buffer(
-      in_padded, strided_strides, flags, in_strided_view.size(), 0);
-
-  // Materialize strided view
-  Shape strided_reshape = {N * oH * oW, wH * wW * C};
-  array in_strided(strided_reshape, in_strided_view.dtype(), nullptr, {});
-  copy_cpu(in_strided_view, in_strided, CopyType::General, stream);
-  temps.push_back(in_strided);
-
-  // Check wt dtype and prepare
-  auto gemm_wt = wt;
-  auto gemm_out = out;
-
-  if (wt.dtype() != float32 || !wt.flags().row_contiguous) {
-    auto ctype =
-        wt.flags().row_contiguous ? CopyType::Vector : CopyType::General;
-    gemm_wt = array(wt.shape(), float32, nullptr, {});
-    copy_cpu(wt, gemm_wt, ctype, stream);
-    temps.push_back(gemm_wt);
-  }
-
-  if (out.dtype() != float32) {
-    gemm_out = array(out.shape(), float32, nullptr, {});
-    gemm_out.set_data(allocator::malloc(gemm_out.nbytes()));
-    temps.push_back(gemm_out);
-  }
-
-  encoder.set_input_array(in_strided);
-  encoder.set_input_array(gemm_wt);
-  encoder.set_output_array(gemm_out);
-
-  encoder.dispatch([in_strided_ptr = in_strided.data<float>(),
-                    gemm_wt_ptr = gemm_wt.data<float>(),
-                    gemm_out_ptr = gemm_out.data<float>(),
-                    strided_reshape = std::move(strided_reshape),
-                    O]() {
-    // Perform gemm
-    cblas_sgemm(
-        CblasRowMajor,
-        CblasNoTrans, // no trans A
-        CblasTrans, // transB
-        strided_reshape[0], // M
-        O, // N
-        strided_reshape[1], // K
-        1.0f, // alpha
-        in_strided_ptr,
-        strided_reshape[1], // lda
-        gemm_wt_ptr,
-        strided_reshape[1], // ldb
-        0.0f, // beta
-        gemm_out_ptr,
-        O // ldc
-    );
-  });
-
-  // Copy results if needed
-  if (out.dtype() != float32) {
-    copy_cpu_inplace(gemm_out, out, CopyType::Vector, stream);
-  }
-  encoder.add_temporaries(std::move(temps));
-}
-
 void explicit_gemm_conv_ND_cpu(
     const array& in,
     const array& wt,

mlx/backend/cpu/eig.cpp

@@ -46,7 +46,6 @@ void eig_impl(
     int info;
     {
       T work;
-      int iwork;
       geev<T>(
           &jobl,
           &jobr,

mlx/backend/cpu/gemms/bnns.cpp

@@ -1,5 +1,4 @@
 // Copyright © 2023-2024 Apple Inc.
-
 #include <Accelerate/Accelerate.h>
 
 #include "mlx/array.h"
@@ -49,9 +48,15 @@ void matmul_bnns(
   size_t K = a_shape[ndim - 1];
 
   BNNSDataType bnns_dtype = to_bnns_dtype<T>();
-
 #pragma GCC diagnostic push
 #pragma GCC diagnostic ignored "-Wdeprecated-declarations"
+  if (beta != 1.0 && beta != 0.0) {
+    // scale the output
+    for (auto i = 0; i < batch_size * M * N; ++i) {
+      out[i] *= beta;
+    }
+    beta = 1.0;
+  }
   const BNNSLayerParametersBroadcastMatMul gemm_params{
       /* float alpha = */ alpha,
       /* float beta = */ beta,

mlx/backend/cpu/gemms/cblas.cpp

@@ -88,4 +88,47 @@ void matmul<double>(
   }
 }
 
+template <>
+void matmul<complex64_t>(
+    const complex64_t* a,
+    const complex64_t* b,
+    complex64_t* out,
+    bool a_transposed,
+    bool b_transposed,
+    size_t lda,
+    size_t ldb,
+    size_t ldc,
+    float alpha,
+    float beta,
+    size_t batch_size,
+    const Shape& a_shape,
+    const Strides& a_strides,
+    const Shape& b_shape,
+    const Strides& b_strides) {
+  auto ndim = a_shape.size();
+  size_t M = a_shape[ndim - 2];
+  size_t N = b_shape[ndim - 1];
+  size_t K = a_shape[ndim - 1];
+  auto calpha = static_cast<complex64_t>(alpha);
+  auto cbeta = static_cast<complex64_t>(beta);
+
+  for (int i = 0; i < batch_size; ++i) {
+    cblas_cgemm(
+        CblasRowMajor,
+        a_transposed ? CblasTrans : CblasNoTrans, // transA
+        b_transposed ? CblasTrans : CblasNoTrans, // transB
+        M,
+        N,
+        K,
+        &calpha,
+        a + elem_to_loc(M * K * i, a_shape, a_strides),
+        lda,
+        b + elem_to_loc(K * N * i, b_shape, b_strides),
+        ldb,
+        &cbeta,
+        out + M * N * i,
+        ldc);
+  }
+}
+
 } // namespace mlx::core

mlx/backend/cpu/masked_mm.cpp

@@ -215,18 +215,18 @@ void BlockMaskedMM::eval_cpu(const std::vector<array>& inputs, array& out) {
 
   encoder.set_input_array(a);
   encoder.set_input_array(b);
-  const void* a_mask_ptr;
-  const void* b_mask_ptr;
-  const void* out_mask_ptr;
+  const void* a_mask_ptr = nullptr;
+  const void* b_mask_ptr = nullptr;
+  const void* out_mask_ptr = nullptr;
   Shape a_mask_shape;
   Shape b_mask_shape;
   Shape out_mask_shape;
   Strides a_mask_strides;
   Strides b_mask_strides;
   Strides out_mask_strides;
-  bool a_mask_bool;
-  bool b_mask_bool;
-  bool out_mask_bool;
+  bool a_mask_bool = false;
+  bool b_mask_bool = false;
+  bool out_mask_bool = false;
   if (has_op_mask) {
     auto& a_mask = inputs[inputs.size() - 2];
     auto& b_mask = inputs[inputs.size() - 1];
@@ -423,7 +423,6 @@ void GatherMM::eval_cpu(const std::vector<array>& inputs, array& out) {
   auto& rhs_indices = inputs[3];
 
   auto batch_shape = get_batch_dims(out.shape());
-  int batch_ndim = batch_shape.size();
 
   auto batch_shape_A = get_batch_dims(a.shape());
   auto batch_strides_A = get_batch_dims(a.strides());

mlx/backend/cpu/matmul.cpp

@@ -91,7 +91,6 @@ void matmul_general(
   auto [b_transposed, ldb, b] = check_transpose(b_pre);
   size_t M = a.shape(-2);
   size_t N = b.shape(-1);
-  size_t K = a.shape(-1);
   if (M == 0 || N == 0) {
     return;
   }
@@ -108,6 +107,9 @@ void matmul_general(
   } else if (out.dtype() == float64) {
     matmul_dispatch<double>(
         a, b, out, a_transposed, b_transposed, lda, ldb, alpha, beta, stream);
+  } else if (out.dtype() == complex64) {
+    matmul_dispatch<complex64_t>(
+        a, b, out, a_transposed, b_transposed, lda, ldb, alpha, beta, stream);
   } else {
     throw std::runtime_error("[Matmul::eval_cpu] Invalid type.");
   }
@@ -128,10 +130,6 @@ void Matmul::eval_cpu(const std::vector<array>& inputs, array& out) {
 }
 
 void AddMM::eval_cpu(const std::vector<array>& inputs, array& out) {
-  if (out.dtype() != float32) {
-    throw std::runtime_error(
-        "[AddMM::eval_cpu] Currently only supports float32.");
-  }
   if (out.size() == 0) {
     out.set_data(allocator::malloc(out.nbytes()));
     return;

mlx/backend/cpu/quantized.cpp

@@ -1,8 +1,11 @@
 // Copyright © 2023 Apple Inc.
 
+#include "mlx/backend/common/unary.h"
 #include "mlx/backend/cpu/copy.h"
 #include "mlx/backend/cpu/encoder.h"
 #include "mlx/backend/cpu/simd/simd.h"
+#include "mlx/backend/cpu/unary.h"
+#include "mlx/backend/cpu/unary_ops.h"
 #include "mlx/fast_primitives.h"
 #include "mlx/primitives.h"
 #include "mlx/utils.h"
@@ -445,7 +448,6 @@ void mxfp4_qmm(
     int K) {
   constexpr int group_size = 32;
   constexpr int pack_factor = get_pack_factor(4, 8);
-  constexpr int bytes_per_pack = get_bytes_per_pack(4);
   constexpr int packs_in_group = group_size / pack_factor;
 
   for (int m = 0; m < M; m++) {
@@ -487,7 +489,6 @@ void mxfp4_qmm_t(
     int K) {
   constexpr int group_size = 32;
   constexpr int pack_factor = get_pack_factor(4, 8);
-  constexpr int bytes_per_pack = get_bytes_per_pack(4);
   constexpr int packs_in_group = group_size / pack_factor;
 
   for (int m = 0; m < M; m++) {
@@ -1104,4 +1105,44 @@ void fast::Quantize::eval_cpu(
   });
 }
 
+void fast::ConvertFP8::eval_cpu(
+    const std::vector<array>& inputs,
+    std::vector<array>& outputs) {
+  auto& in = inputs[0];
+  auto& out = outputs[0];
+  set_unary_output_data(in, out);
+  auto& encoder = cpu::get_command_encoder(stream());
+  encoder.set_input_array(in);
+  encoder.set_output_array(out);
+  encoder.dispatch([in = array::unsafe_weak_copy(in),
+                    out = array::unsafe_weak_copy(out),
+                    to_fp8 = to_fp8_]() mutable {
+    if (to_fp8) {
+      switch (in.dtype()) {
+        case float16:
+          unary_op<float16_t, uint8_t>(in, out, detail::ToFP8());
+          break;
+        case bfloat16:
+          unary_op<bfloat16_t, uint8_t>(in, out, detail::ToFP8());
+          break;
+        default:
+          unary_op<float, uint8_t>(in, out, detail::ToFP8());
+          break;
+      }
+    } else {
+      switch (out.dtype()) {
+        case float16:
+          unary_op<uint8_t, float16_t>(in, out, detail::FromFP8());
+          break;
+        case bfloat16:
+          unary_op<uint8_t, bfloat16_t>(in, out, detail::FromFP8());
+          break;
+        default:
+          unary_op<uint8_t, float>(in, out, detail::FromFP8());
+          break;
+      }
+    }
+  });
+}
+
 } // namespace mlx::core

mlx/backend/cpu/simd/accelerate_simd.h

@@ -1,5 +1,6 @@
 #pragma once
 
+#include <arm_neon.h>
 #include <simd/math.h>
 #include <simd/vector.h>
 
@@ -9,7 +10,7 @@
 
 #include "mlx/backend/cpu/simd/base_simd.h"
 
-// There seems to be a bug in sims/base.h
+// There seems to be a bug in simd/base_simd.h
 // __XROS_2_0 is not defined, the expression evaluates
 // to true instead of false setting the SIMD library
 // higher than it should be even on macOS < 15
@@ -200,6 +201,15 @@ SIMD_DEFAULT_COMPARISONS(<=)
 SIMD_DEFAULT_COMPARISONS(==)
 SIMD_DEFAULT_COMPARISONS(!=)
 
+template <typename T, int N>
+Simd<T, N> clz(Simd<T, N> x) {
+  auto a = *(uint32x4_t*)(&x);
+  auto b = *((uint32x4_t*)(&x) + 1);
+  a = vclzq_u32(a);
+  b = vclzq_u32(b);
+  return asd::make_uint8(a, b);
+}
+
 template <typename T, int N>
 Simd<T, N> atan2(Simd<T, N> a, Simd<T, N> b) {
   return asd::atan2(a.value, b.value);

mlx/backend/cpu/simd/base_simd.h

@@ -171,6 +171,11 @@ DEFAULT_BINARY(&)
 DEFAULT_BINARY(&&)
 DEFAULT_BINARY(||)
 
+template <typename T>
+Simd<T, 1> clz(Simd<T, 1> x_) {
+  return __builtin_clz(x_.value);
+}
+
 template <typename T>
 Simd<T, 1> remainder(Simd<T, 1> a_, Simd<T, 1> b_) {
   T a = a_.value;

mlx/backend/cpu/sort.cpp

@@ -15,6 +15,18 @@ namespace mlx::core {
 
 namespace {
 
+// NaN-aware comparator that places NaNs at the end
+template <typename T>
+bool nan_aware_less(T a, T b) {
+  if constexpr (std::is_floating_point_v<T> || std::is_same_v<T, complex64_t>) {
+    if (std::isnan(a))
+      return false;
+    if (std::isnan(b))
+      return true;
+  }
+  return a < b;
+}
+
 template <typename T>
 struct StridedIterator {
   using iterator_category = std::random_access_iterator_tag;
@@ -27,7 +39,7 @@ struct StridedIterator {
   StridedIterator() = default;
 
   explicit StridedIterator(T* ptr, int64_t stride, difference_type offset = 0)
-      : ptr_(ptr + offset * stride), stride_(stride) {}
+      : stride_(stride), ptr_(ptr + offset * stride) {}
 
   explicit StridedIterator(array& arr, int axis, difference_type offset = 0)
       : StridedIterator(arr.data<T>(), arr.strides()[axis], offset) {}
@@ -130,7 +142,7 @@ void sort(array& out, int axis) {
     StridedIterator st(data_ptr, axis_stride, 0);
     StridedIterator ed(data_ptr, axis_stride, axis_size);
 
-    std::stable_sort(st, ed);
+    std::stable_sort(st, ed, nan_aware_less<T>);
     src_it.step();
   }
 }
@@ -184,6 +196,15 @@ void argsort(const array& in, array& out, int axis) {
     std::stable_sort(st, ed, [data_ptr, in_stride](IdxT a, IdxT b) {
       auto v1 = data_ptr[a * in_stride];
       auto v2 = data_ptr[b * in_stride];
+
+      // Handle NaNs (place them at the end)
+      if (std::is_floating_point<T>::value) {
+        if (std::isnan(v1))
+          return false;
+        if (std::isnan(v2))
+          return true;
+      }
+
       return v1 < v2 || (v1 == v2 && a < b);
     });
   }
@@ -219,7 +240,7 @@ void partition(array& out, int axis, int kth) {
     StridedIterator md(data_ptr, axis_stride, kth);
     StridedIterator ed(data_ptr, axis_stride, axis_size);
 
-    std::nth_element(st, md, ed);
+    std::nth_element(st, md, ed, nan_aware_less<T>);
   }
 }
 
@@ -276,6 +297,15 @@ void argpartition(const array& in, array& out, int axis, int kth) {
     std::nth_element(st, md, ed, [data_ptr, in_stride](IdxT a, IdxT b) {
       auto v1 = data_ptr[a * in_stride];
       auto v2 = data_ptr[b * in_stride];
+
+      // Handle NaNs (place them at the end)
+      if (std::is_floating_point<T>::value) {
+        if (std::isnan(v1))
+          return false;
+        if (std::isnan(v2))
+          return true;
+      }
+
       return v1 < v2 || (v1 == v2 && a < b);
     });
   }

mlx/backend/cpu/svd.cpp

@@ -83,8 +83,6 @@ void svd_impl(
 
     auto jobz = (u_ptr) ? "A" : "N";
 
-    // Will contain the number of singular values after the call has returned.
-    int ns = 0;
     T workspace_dimension = 0;
 
     // Will contain the indices of eigenvectors that failed to converge (not

mlx/backend/cpu/unary.h

@@ -24,9 +24,9 @@ void unary_op(const array& a, array& out, Op) {
   auto ndim = a.ndim();
   if (a.flags().contiguous) {
     auto size = a.data_size();
-    constexpr int N = simd::max_size<T>;
+    constexpr int N = std::min(simd::max_size<T>, simd::max_size<U>);
     while (size >= N) {
-      simd::store(dst, Op{}(simd::load<T, N>(src)));
+      simd::store(dst, simd::Simd<U, N>(Op{}(simd::load<T, N>(src))));
       size -= N;
       src += N;
       dst += N;

mlx/backend/cpu/unary_ops.h

@@ -77,7 +77,8 @@ struct Real {
 struct Sigmoid {
   template <int N, typename T>
   Simd<T, N> operator()(Simd<T, N> x) {
-    return 1.0f / (1.0f + simd::exp(-x));
+    auto y = 1.0f / (1.0f + simd::exp(simd::abs(x)));
+    return simd::select(x < Simd<T, N>{0}, y, Simd<T, N>{1} - y);
   }
   SINGLE()
 };
@@ -107,4 +108,73 @@ struct Square {
   SINGLE()
 };
 
+template <int N>
+Simd<float, N> fp32_from_bits(Simd<uint32_t, N> x) {
+  return *(Simd<float, N>*)(&x);
+}
+template <int N>
+Simd<uint32_t, N> fp32_to_bits(Simd<float, N> x) {
+  return *(Simd<uint32_t, N>*)(&x);
+}
+
+struct ToFP8 {
+  template <typename T, int N>
+  Simd<uint8_t, N> operator()(Simd<T, N> f) {
+    uint32_t fp8_max = 543 << 21;
+    auto denorm_mask = Simd<uint32_t, N>(141 << 23);
+    Simd<uint32_t, N> f_bits;
+    Simd<float, N> f32 = f;
+    f_bits = fp32_to_bits(f32);
+    Simd<uint8_t, N> result = 0u;
+    auto sign = f_bits & 0x80000000;
+    f_bits = f_bits ^ sign;
+
+    auto f_bits_low =
+        fp32_to_bits(fp32_from_bits(f_bits) + fp32_from_bits(denorm_mask));
+    auto result_low = Simd<uint8_t, N>(f_bits_low - denorm_mask);
+
+    auto mant_odd = Simd<uint8_t, N>((f_bits >> 20) & 1);
+    auto f_bits_high = f_bits + (((uint32_t)(7 - 127) << 23) + 0x7FFFF);
+    f_bits_high = f_bits_high + Simd<uint32_t, N>(mant_odd);
+
+    auto result_high = Simd<uint8_t, N>(f_bits_high >> 20);
+    result = select(f_bits < (121 << 23), result_low, result_high);
+
+    auto result_sat = Simd<uint8_t, N>(0x7E);
+    result = select(f_bits >= fp8_max, result_sat, result);
+    return result | Simd<uint8_t, N>(sign >> 24);
+  }
+
+  template <typename T>
+  uint8_t operator()(T x) {
+    return (*this)(Simd<T, 1>(x)).value;
+  }
+};
+
+struct FromFP8 {
+  template <int N>
+  Simd<float, N> operator()(Simd<uint8_t, N> x) {
+    auto w = Simd<uint32_t, N>(x) << 24;
+    auto sign = w & 0x80000000;
+    auto nonsign = w & 0x7FFFFFFF;
+
+    auto renorm_shift = clz(nonsign);
+    renorm_shift = simd::select(
+        renorm_shift > Simd<uint32_t, N>{4},
+        renorm_shift - Simd<uint32_t, N>{4},
+        Simd<uint32_t, N>{0});
+
+    Simd<int32_t, N> inf_nan_mask =
+        (Simd<int32_t, N>(nonsign + 0x01000000) >> 8) & 0x7F800000;
+    auto zero_mask = Simd<int32_t, N>(nonsign - 1) >> 31;
+    auto result = sign |
+        ((((nonsign << renorm_shift >> 4) + ((0x78 - renorm_shift) << 23)) |
+          inf_nan_mask) &
+         ~zero_mask);
+    return fp32_from_bits(result);
+  }
+  float operator()(uint8_t x) {
+    return (*this)(Simd<uint8_t, 1>(x)).value;
+  }
+};
 } // namespace mlx::core::detail

mlx/backend/cuda/CMakeLists.txt

@@ -51,12 +51,19 @@ target_sources(
           ${CMAKE_CURRENT_SOURCE_DIR}/ternary.cu
           ${CMAKE_CURRENT_SOURCE_DIR}/utils.cpp
           ${CMAKE_CURRENT_SOURCE_DIR}/quantized/affine_quantize.cu
+          ${CMAKE_CURRENT_SOURCE_DIR}/quantized/fp_quantize.cu
           ${CMAKE_CURRENT_SOURCE_DIR}/quantized/quantized.cpp
+          ${CMAKE_CURRENT_SOURCE_DIR}/quantized/convert_fp8.cu
           ${CMAKE_CURRENT_SOURCE_DIR}/worker.cpp)
 
 add_subdirectory(${CMAKE_CURRENT_SOURCE_DIR}/binary)
 add_subdirectory(${CMAKE_CURRENT_SOURCE_DIR}/unary)
 
+# fp4 is not available on < 12.8
+if(CMAKE_CUDA_COMPILER_VERSION VERSION_LESS 12.8.0)
+  target_include_directories(mlx PRIVATE ${CMAKE_CURRENT_SOURCE_DIR}/quantized/)
+endif()
+
 if(CMAKE_CUDA_COMPILER_VERSION VERSION_GREATER_EQUAL 12.9.0)
   target_sources(
     mlx PRIVATE ${CMAKE_CURRENT_SOURCE_DIR}/gemms/cublas_gemm_batched_12_9.cu)
@@ -170,7 +177,6 @@ target_link_libraries(mlx PRIVATE CUDNN::cudnn_all)
 # Suppress nvcc warnings on MLX headers.
 target_compile_options(mlx PRIVATE $<$<COMPILE_LANGUAGE:CUDA>:-Xcudafe
                                    --diag_suppress=997>)
-
 # Install CCCL headers for JIT.
 install(DIRECTORY ${cccl_SOURCE_DIR}/include/cuda
         DESTINATION ${CMAKE_INSTALL_INCLUDEDIR}/cccl)

mlx/backend/cuda/allocator.cpp

@@ -30,15 +30,20 @@ SmallSizePool::SmallSizePool() {
   next_free_ = buffer_;
 
   CHECK_CUDA_ERROR(cudaMallocManaged(&data_, small_pool_size));
+
+  int device_count = 0;
+  CHECK_CUDA_ERROR(cudaGetDeviceCount(&device_count));
+  for (int i = 0; i < device_count; ++i) {
 #if CUDART_VERSION >= 13000
-  cudaMemLocation loc;
-  loc.type = cudaMemLocationTypeDevice;
-  loc.id = 0;
+    cudaMemLocation loc;
+    loc.type = cudaMemLocationTypeDevice;
+    loc.id = i;
 #else
-  int loc = 0;
+    int loc = i;
 #endif // CUDART_VERSION >= 13000
-  CHECK_CUDA_ERROR(
-      cudaMemAdvise(data_, small_pool_size, cudaMemAdviseSetReadMostly, loc));
+    CHECK_CUDA_ERROR(
+        cudaMemAdvise(data_, small_pool_size, cudaMemAdviseSetAccessedBy, loc));
+  }
 
   auto curr = next_free_;
   for (size_t i = 1; i < num_blocks; ++i) {
@@ -86,13 +91,12 @@ CudaAllocator::CudaAllocator()
   // TODO: Set memory limit for multi-device.
   size_t free, total;
   CHECK_CUDA_ERROR(cudaMemGetInfo(&free, &total));
-  memory_limit_ = total * 0.8;
+  memory_limit_ = total * 0.95;
   max_pool_size_ = memory_limit_;
 }
 
 Buffer CudaAllocator::malloc(size_t size) {
   // Find available buffer from cache.
-  auto orig_size = size;
   std::unique_lock lock(mutex_);
   if (size <= small_block_size) {
     size = 8;
@@ -126,7 +130,7 @@ Buffer CudaAllocator::malloc(size_t size) {
     }
     lock.lock();
   }
-  active_memory_ += size;
+  active_memory_ += buf->size;
   peak_memory_ = std::max(active_memory_, peak_memory_);
 
   // Maintain the cache below the requested limit.

mlx/backend/cuda/compiled.cpp

@@ -332,9 +332,9 @@ void Compiled::eval_gpu(
     encoder.set_output_array(out);
   }
 
-  auto kernel = mod.get_kernel(kernel_name);
+  auto [kernel, max_block_dims] = mod.get_kernel_and_dims(kernel_name);
   auto [num_blocks, block_dims] =
-      get_launch_args(outputs[0], large, work_per_thread);
+      get_launch_args(outputs[0], large, work_per_thread, max_block_dims);
   encoder.add_kernel_node(kernel, num_blocks, block_dims, 0, args.args());
 }
 

mlx/backend/cuda/conv.cpp

@@ -47,7 +47,7 @@ auto& conv_cache() {
       std::pair<
           cudnnBackendDescriptorType_t,
           std::optional<cudnn_frontend::ExecutionPlan>>>
-      cache(/* capacity */ 128);
+      cache("MLX_CUDA_CONV_CACHE_SIZE", /* default_capacity */ 128);
   return cache;
 }
 
@@ -382,20 +382,19 @@ void Convolution::eval_gpu(const std::vector<array>& inputs, array& out_) {
   }
 
   if (op_graph) {
-    // Setup inputs and outputs.
-    register_args(encoder, backend_type, in, wt, out, out_);
-
     // Find a plan for the graph and execute it.
     auto plan = find_cudnn_plan_from_op_graph(
         encoder.device().cudnn_handle(), backend_type, dtype, *op_graph);
-    if (!plan) {
-      throw std::runtime_error("[conv] Unable to find an execution plan.");
-    }
-    auto [x, w, y] = dispatch_args(backend_type, in, wt, out);
-    if (encode_cudnn_plan(encoder, *plan, {'x', 'w', 'y'}, x, w, y)) {
-      conv_cache().emplace(
-          cache_key, std::make_pair(backend_type, std::move(*plan)));
-      return;
+    if (plan) {
+      // Setup inputs and outputs.
+      register_args(encoder, backend_type, in, wt, out, out_);
+
+      auto [x, w, y] = dispatch_args(backend_type, in, wt, out);
+      if (encode_cudnn_plan(encoder, *plan, {'x', 'w', 'y'}, x, w, y)) {
+        conv_cache().emplace(
+            cache_key, std::make_pair(backend_type, std::move(*plan)));
+        return;
+      }
     }
   }
 

mlx/backend/cuda/cudnn_utils.cpp

@@ -210,6 +210,9 @@ std::optional<cudnn_frontend::ExecutionPlan> find_cudnn_plan_from_op_graph(
     Dtype dtype,
     cudnn_frontend::OperationGraph& op_graph) {
   auto engine_configs = get_cudnn_engine_configs(backend_type, dtype, op_graph);
+  if (engine_configs.empty()) {
+    return std::nullopt;
+  }
   return find_cudnn_plan_from_engine_configs(handle, engine_configs, op_graph);
 }
 

mlx/backend/cuda/device.cpp

@@ -14,10 +14,6 @@ namespace mlx::core::cu {
 
 namespace {
 
-// Can be tuned with MLX_MAX_OPS_PER_BUFFER
-// This should be less than 255
-constexpr int default_max_nodes_per_graph = 20;
-
 #define CHECK_CUDNN_ERROR(cmd) check_cudnn_error(#cmd, (cmd))
 
 void check_cudnn_error(const char* name, cudnnStatus_t err) {
@@ -27,13 +23,6 @@ void check_cudnn_error(const char* name, cudnnStatus_t err) {
   }
 }
 
-int cuda_graph_cache_size() {
-  static int cache_size = []() {
-    return env::get_var("MLX_CUDA_GRAPH_CACHE_SIZE", 400);
-  }();
-  return cache_size;
-}
-
 bool use_cuda_graphs() {
   static bool use_graphs = []() {
     return env::get_var("MLX_USE_CUDA_GRAPHS", true);
@@ -75,8 +64,8 @@ Device::~Device() {
 
 void Device::make_current() {
   // We need to set/get current CUDA device very frequently, cache it to reduce
-  // actual calls of CUDA APIs. This function assumes single-thread in host.
-  static int current = 0;
+  // actual calls of CUDA APIs.
+  static thread_local int current = 0;
   if (current != device_) {
     CHECK_CUDA_ERROR(cudaSetDevice(device_));
     current = device_;
@@ -102,6 +91,7 @@ CommandEncoder::CaptureContext::CaptureContext(CommandEncoder& enc) : enc(enc) {
 
 CommandEncoder::CaptureContext::~CaptureContext() {
   if (!use_cuda_graphs()) {
+    enc.node_count_++;
     return;
   }
 
@@ -203,7 +193,8 @@ CommandEncoder::CommandEncoder(Device& d)
     : device_(d),
       stream_(d),
       graph_(d),
-      graph_cache_(cuda_graph_cache_size()) {}
+      worker_(d),
+      graph_cache_("MLX_CUDA_GRAPH_CACHE_SIZE", /* default_capacity */ 400) {}
 
 void CommandEncoder::add_completed_handler(std::function<void()> task) {
   worker_.add_task(std::move(task));
@@ -227,12 +218,6 @@ void CommandEncoder::set_output_array(const array& arr) {
   active_outputs_.push_back(id);
 }
 
-void CommandEncoder::maybe_commit() {
-  if (node_count_ >= env::max_ops_per_buffer(default_max_nodes_per_graph)) {
-    commit();
-  }
-}
-
 void CommandEncoder::add_kernel_node(
     void* func,
     dim3 grid_dim,
@@ -240,6 +225,7 @@ void CommandEncoder::add_kernel_node(
     uint32_t smem_bytes,
     void** params) {
   if (!use_cuda_graphs()) {
+    node_count_++;
     CHECK_CUDA_ERROR(cudaLaunchKernel(
         func, grid_dim, block_dim, params, smem_bytes, stream()));
     return;
@@ -260,6 +246,7 @@ void CommandEncoder::add_kernel_node(
     uint32_t smem_bytes,
     void** params) {
   if (!use_cuda_graphs()) {
+    node_count_++;
     CHECK_CUDA_ERROR(cuLaunchKernel(
         func,
         grid_dim.x,
@@ -302,22 +289,28 @@ void CommandEncoder::add_kernel_node(const CUDA_KERNEL_NODE_PARAMS& params) {
 
 void CommandEncoder::add_graph_node(cudaGraph_t child) {
   if (!use_cuda_graphs()) {
+    node_count_++;
     CudaGraphExec graph_exec;
     graph_exec.instantiate(child);
     device_.make_current();
     CHECK_CUDA_ERROR(cudaGraphLaunch(graph_exec, stream()));
+    return;
   }
   cudaGraphNode_t node;
   CHECK_CUDA_ERROR(cudaGraphAddChildGraphNode(&node, graph_, NULL, 0, child));
   insert_graph_dependencies(GraphNode{node, 'G'});
 }
 
+int CommandEncoder::get_num_ops() {
+  return node_count_;
+}
+
 void CommandEncoder::commit() {
   nvtx3::scoped_range r("CommandEncoder::commit");
   if (!temporaries_.empty()) {
     add_completed_handler([temporaries = std::move(temporaries_)]() {});
   }
-  if (node_count_ > 0) {
+  if (use_cuda_graphs() && node_count_ > 0) {
     if (!from_nodes_.empty()) {
       CHECK_CUDA_ERROR(cudaGraphAddDependencies(
           graph_,
@@ -360,7 +353,6 @@ void CommandEncoder::commit() {
     CHECK_CUDA_ERROR(cudaGraphLaunch(graph_exec, stream_));
 
     // Reset state
-    node_count_ = 0;
     graph_node_count_ = 0;
     empty_node_count_ = 0;
     from_nodes_.clear();
@@ -372,6 +364,7 @@ void CommandEncoder::commit() {
 
   // Put completion handlers in a batch.
   worker_.commit(stream_);
+  node_count_ = 0;
 }
 
 void CommandEncoder::synchronize() {

mlx/backend/cuda/device.h

@@ -83,7 +83,7 @@ class CommandEncoder {
   }
 
   void add_completed_handler(std::function<void()> task);
-  void maybe_commit();
+  int get_num_ops();
   void commit();
 
   Device& device() {
@@ -140,7 +140,7 @@ class Device {
   Device(const Device&) = delete;
   Device& operator=(const Device&) = delete;
 
-  // Make this device the current cuda device, required by some cuda calls.
+  // Make this device the current cuda device, this method is thread-safe.
   void make_current();
 
   CommandEncoder& get_command_encoder(Stream s);

mlx/backend/cuda/device/unary_ops.cuh

@@ -2,6 +2,8 @@
 
 #pragma once
 
+#include <cuda_fp8.h>
+
 #include "mlx/backend/cuda/device/fp16_math.cuh"
 #include "mlx/backend/cuda/device/utils.cuh"
 
@@ -257,8 +259,8 @@ struct Round {
 struct Sigmoid {
   template <typename T>
   __device__ T operator()(T x) {
-    T y = 1 / (1 + exp(-abs(x)));
-    return (x < 0) ? 1 - y : y;
+    T y = 1 / (1 + exp(abs(x)));
+    return (x < 0) ? y : 1 - y;
   }
 };
 
@@ -334,4 +336,17 @@ struct Tanh {
   }
 };
 
+struct ToFP8 {
+  template <typename T>
+  __device__ uint8_t operator()(T x) {
+    return __nv_fp8_e4m3(x).__x;
+  }
+};
+
+struct FromFP8 {
+  __device__ float operator()(uint8_t x) {
+    return float(*(__nv_fp8_e4m3*)(&x));
+  }
+};
+
 } // namespace mlx::core::cu

mlx/backend/cuda/device/utils.cuh

@@ -1,6 +1,6 @@
 // Copyright © 2025 Apple Inc.
 
-// This file must not include any host-only code, utilies that work under both
+// This file must not include any host-only code, utilities that work under both
 // host and device can be put here.
 //
 // See more about the requirements at:
@@ -202,7 +202,7 @@ struct Limits<
   }
 };
 
-// CUDA 11 does not have host side arithmatic operators for half types.
+// CUDA 11 does not have host side arithmetic operators for half types.
 template <typename T>
 struct Limits<
     T,

mlx/backend/cuda/eval.cpp

@@ -5,18 +5,24 @@
 #include "mlx/backend/cuda/device.h"
 #include "mlx/backend/gpu/available.h"
 #include "mlx/primitives.h"
+#include "mlx/scheduler.h"
 
 #include <nvtx3/nvtx3.hpp>
 
 namespace mlx::core::gpu {
 
+// Can be tuned with MLX_MAX_OPS_PER_BUFFER
+constexpr int default_max_nodes_per_graph = 20;
+
 bool is_available() {
   return true;
 }
 
 void new_stream(Stream s) {
-  // Force initalization of cuda, so cuda runtime get destroyed at last.
+  // Force initalization of CUDA, so CUDA runtime get destroyed at last.
   cudaFree(nullptr);
+  // Make sure CUDA event pool get destroyed after device and stream.
+  cu::CudaEvent::init_pool();
   // Ensure the static stream objects get created.
   cu::get_command_encoder(s);
 }
@@ -34,7 +40,8 @@ void eval(array& arr) {
     arr.primitive().eval_gpu(arr.inputs(), outputs);
   }
 
-  auto& encoder = cu::get_command_encoder(arr.primitive().stream());
+  auto& stream = arr.primitive().stream();
+  auto& encoder = cu::get_command_encoder(stream);
   // Keep used buffers alive until kernel finishes running.
   for (auto& in : arr.inputs()) {
     // Except for the donated one.
@@ -45,7 +52,14 @@ void eval(array& arr) {
   for (auto& s : arr.siblings()) {
     encoder.add_temporary(s);
   }
-  encoder.maybe_commit();
+
+  if (encoder.get_num_ops() >=
+      env::max_ops_per_buffer(default_max_nodes_per_graph)) {
+    scheduler::notify_new_task(stream);
+    encoder.add_completed_handler(
+        [stream]() { scheduler::notify_task_completion(stream); });
+    encoder.commit();
+  }
 }
 
 void finalize(Stream s) {

mlx/backend/cuda/event.cu

@@ -3,10 +3,12 @@
 #include "mlx/backend/cuda/allocator.h"
 #include "mlx/backend/cuda/device.h"
 #include "mlx/backend/cuda/event.h"
-#include "mlx/backend/cuda/utils.h"
 #include "mlx/event.h"
 #include "mlx/scheduler.h"
 
+#include <map>
+#include <vector>
+
 #include <nvtx3/nvtx3.hpp>
 
 namespace mlx::core {
@@ -17,104 +19,180 @@ namespace cu {
 // CudaEvent implementations
 ///////////////////////////////////////////////////////////////////////////////
 
-// Cuda event managed with RAII.
-class CudaEventHandle {
+namespace {
+
+// Manage cached cudaEvent_t objects.
+class CudaEventPool {
  public:
-  CudaEventHandle() {
-    CHECK_CUDA_ERROR(cudaEventCreateWithFlags(
-        &event_, cudaEventDisableTiming | cudaEventBlockingSync));
+  CudaEventHandle create(Device& d, int flags) {
+    if (!on_creation_thread()) {
+      return CudaEventHandle(d, flags);
+    }
+    auto& cache = cache_for(d, flags);
+    if (cache.empty()) {
+      return CudaEventHandle(d, flags);
+    } else {
+      CudaEventHandle ret = std::move(cache.back());
+      cache.pop_back();
+      return ret;
+    }
   }
 
-  ~CudaEventHandle() {
-    CHECK_CUDA_ERROR(cudaEventDestroy(event_));
-  }
-
-  CudaEventHandle(const CudaEventHandle&) = delete;
-  CudaEventHandle& operator=(const CudaEventHandle&) = delete;
-
-  operator cudaEvent_t() const {
-    return event_;
+  void release(CudaEventHandle event) {
+    if (!on_creation_thread()) {
+      // Event will be destroyed directly instead of getting moved to cache.
+      return;
+    }
+    cache_for(event.device, event.flags).push_back(std::move(event));
   }
 
  private:
-  cudaEvent_t event_;
+  std::vector<CudaEventHandle>& cache_for(Device& d, int flags) {
+    return cache_[d.cuda_device()][flags];
+  }
+
+  bool on_creation_thread() {
+    return std::this_thread::get_id() == thread_id_;
+  }
+
+  // The CudaEvent may be created and destroyed on different threads (for
+  // example when waiting on GPU work in CPU stream), we don't want to make
+  // the cache thread-safe as it adds overhead, so we just skip cache when
+  // using events in worker threads.
+  std::thread::id thread_id_{std::this_thread::get_id()};
+
+  // {device: {flags: [events]}}
+  std::map<int, std::map<int, std::vector<CudaEventHandle>>> cache_;
 };
 
-CudaEvent::CudaEvent() : event_(std::make_shared<CudaEventHandle>()) {}
+CudaEventPool& cuda_event_pool() {
+  static CudaEventPool pool;
+  return pool;
+}
+
+} // namespace
+
+CudaEventHandle::CudaEventHandle(Device& d, int flags)
+    : device(d), flags(flags) {
+  device.make_current();
+  CHECK_CUDA_ERROR(cudaEventCreateWithFlags(&handle_, flags));
+  assert(handle_ != nullptr);
+}
+
+CudaEvent::CudaEvent(Device& d, int flags)
+    : event_(cuda_event_pool().create(d, flags)) {}
+
+CudaEvent::~CudaEvent() {
+  cuda_event_pool().release(std::move(event_));
+}
 
 void CudaEvent::wait() {
   nvtx3::scoped_range r("cu::CudaEvent::wait");
-  if (!recorded_) {
-    throw std::runtime_error("Should not wait on a CudaEvent before record.");
-  }
-  cudaEventSynchronize(*event_);
+  event_.device.make_current();
+  cudaEventSynchronize(event_);
 }
 
 void CudaEvent::wait(cudaStream_t stream) {
-  if (!recorded_) {
-    throw std::runtime_error("Should not wait on a CudaEvent before record.");
-  }
-  cudaStreamWaitEvent(stream, *event_);
-}
-
-void CudaEvent::wait(Stream s) {
-  if (s.device == mlx::core::Device::cpu) {
-    scheduler::enqueue(s, [*this]() mutable { wait(); });
-  } else {
-    auto& enc = cu::get_command_encoder(s);
-    enc.commit();
-    wait(enc.stream());
-  }
+  event_.device.make_current();
+  cudaStreamWaitEvent(stream, event_);
 }
 
 void CudaEvent::record(cudaStream_t stream) {
-  cudaEventRecord(*event_, stream);
-  recorded_ = true;
-}
-
-void CudaEvent::record(Stream s) {
-  if (s.device == mlx::core::Device::cpu) {
-    throw std::runtime_error("CudaEvent can not wait on cpu stream.");
-  } else {
-    auto& enc = cu::get_command_encoder(s);
-    enc.commit();
-    record(enc.stream());
-  }
+  event_.device.make_current();
+  cudaEventRecord(event_, stream);
 }
 
 bool CudaEvent::completed() const {
-  return cudaEventQuery(*event_) == cudaSuccess;
+  // Note: cudaEventQuery can be safely called from any device.
+  return cudaEventQuery(event_) == cudaSuccess;
 }
 
+// static
+void CudaEvent::init_pool() {
+  cuda_event_pool();
+}
+
+// Wraps CudaEvent with a few features:
+// 1. The class can be copied.
+// 2. Make wait/record work with CPU streams.
+// 3. Add checks for waiting on un-recorded event.
+class CopyableCudaEvent {
+ public:
+  explicit CopyableCudaEvent(Device& d)
+      : event_(std::make_shared<CudaEvent>(
+            d,
+            cudaEventDisableTiming | cudaEventBlockingSync)) {}
+
+  void wait() {
+    event_->wait();
+  }
+
+  void wait(Stream s) {
+    if (s.device == mlx::core::Device::cpu) {
+      scheduler::enqueue(s, [*this]() mutable {
+        check_recorded();
+        event_->wait();
+      });
+    } else {
+      check_recorded();
+      auto& encoder = cu::get_command_encoder(s);
+      encoder.commit();
+      event_->wait(encoder.stream());
+    }
+  }
+
+  void record(Stream s) {
+    if (s.device == mlx::core::Device::cpu) {
+      throw std::runtime_error("CudaEvent can not wait on CPU stream.");
+    } else {
+      auto& encoder = cu::get_command_encoder(s);
+      encoder.commit();
+      event_->record(encoder.stream());
+      recorded_ = true;
+    }
+  }
+
+  bool is_signaled() const {
+    return recorded_ && event_->completed();
+  }
+
+ private:
+  void check_recorded() const {
+    if (!recorded_) {
+      throw std::runtime_error(
+          "Should not wait on a CudaEvent before recording.");
+    }
+  }
+
+  std::shared_ptr<CudaEvent> event_;
+  bool recorded_{false};
+};
+
 ///////////////////////////////////////////////////////////////////////////////
-// SharedEvent implementations
+// AtomicEvent implementations
 ///////////////////////////////////////////////////////////////////////////////
 
-__host__ __device__ void event_wait(SharedEvent::Atomic* ac, uint64_t value) {
+__host__ __device__ void event_wait(AtomicEvent::Atomic* ac, uint64_t value) {
   uint64_t current;
   while ((current = ac->load()) < value) {
     ac->wait(current);
   }
 }
 
-__host__ __device__ void event_signal(SharedEvent::Atomic* ac, uint64_t value) {
+__host__ __device__ void event_signal(AtomicEvent::Atomic* ac, uint64_t value) {
   ac->store(value);
   ac->notify_all();
 }
 
-__global__ void event_wait_kernel(SharedEvent::Atomic* ac, uint64_t value) {
+__global__ void event_wait_kernel(AtomicEvent::Atomic* ac, uint64_t value) {
   event_wait(ac, value);
 }
 
-__global__ void event_signal_kernel(SharedEvent::Atomic* ac, uint64_t value) {
+__global__ void event_signal_kernel(AtomicEvent::Atomic* ac, uint64_t value) {
   event_signal(ac, value);
 }
 
-SharedEvent::Atomic* to_atomic(std::shared_ptr<Buffer> buf) {
-  return static_cast<SharedEvent::Atomic*>(buf->raw_ptr());
-}
-
-SharedEvent::SharedEvent() {
+AtomicEvent::AtomicEvent() {
   buf_ = std::shared_ptr<Buffer>(
       new Buffer{allocator().malloc(sizeof(Atomic))}, [](Buffer* ptr) {
         allocator().free(*ptr);
@@ -123,17 +201,17 @@ SharedEvent::SharedEvent() {
   *static_cast<uint64_t*>(buf_->raw_ptr()) = 0;
 }
 
-void SharedEvent::wait(uint64_t value) {
-  nvtx3::scoped_range r("cu::SharedEvent::wait");
-  event_wait(to_atomic(buf_), value);
+void AtomicEvent::wait(uint64_t value) {
+  nvtx3::scoped_range r("cu::AtomicEvent::wait");
+  event_wait(atomic(), value);
 }
 
-void SharedEvent::wait(cudaStream_t stream, uint64_t value) {
-  event_wait_kernel<<<1, 1, 0, stream>>>(to_atomic(buf_), value);
+void AtomicEvent::wait(cudaStream_t stream, uint64_t value) {
+  event_wait_kernel<<<1, 1, 0, stream>>>(atomic(), value);
 }
 
-void SharedEvent::wait(Stream s, uint64_t value) {
-  nvtx3::scoped_range r("cu::SharedEvent::wait(s)");
+void AtomicEvent::wait(Stream s, uint64_t value) {
+  nvtx3::scoped_range r("cu::AtomicEvent::wait(s)");
   if (s.device == mlx::core::Device::cpu) {
     scheduler::enqueue(s, [*this, value]() mutable { wait(value); });
   } else {
@@ -144,17 +222,17 @@ void SharedEvent::wait(Stream s, uint64_t value) {
   }
 }
 
-void SharedEvent::signal(uint64_t value) {
-  nvtx3::scoped_range r("cu::SharedEvent::signal");
-  event_signal(to_atomic(buf_), value);
+void AtomicEvent::signal(uint64_t value) {
+  nvtx3::scoped_range r("cu::AtomicEvent::signal");
+  event_signal(atomic(), value);
 }
 
-void SharedEvent::signal(cudaStream_t stream, uint64_t value) {
-  event_signal_kernel<<<1, 1, 0, stream>>>(to_atomic(buf_), value);
+void AtomicEvent::signal(cudaStream_t stream, uint64_t value) {
+  event_signal_kernel<<<1, 1, 0, stream>>>(atomic(), value);
 }
 
-void SharedEvent::signal(Stream s, uint64_t value) {
-  nvtx3::scoped_range r("cu::SharedEvent::signal(s)");
+void AtomicEvent::signal(Stream s, uint64_t value) {
+  nvtx3::scoped_range r("cu::AtomicEvent::signal(s)");
   if (s.device == mlx::core::Device::cpu) {
     // Signal through a GPU stream so the atomic is updated in GPU - updating
     // the atomic in CPU sometimes does not get GPU notified.
@@ -168,14 +246,14 @@ void SharedEvent::signal(Stream s, uint64_t value) {
   }
 }
 
-bool SharedEvent::is_signaled(uint64_t value) const {
-  nvtx3::scoped_range r("cu::SharedEvent::is_signaled");
-  return to_atomic(buf_)->load() >= value;
+bool AtomicEvent::is_signaled(uint64_t value) const {
+  nvtx3::scoped_range r("cu::AtomicEvent::is_signaled");
+  return atomic()->load() >= value;
 }
 
-uint64_t SharedEvent::value() const {
-  nvtx3::scoped_range r("cu::SharedEvent::value");
-  return to_atomic(buf_)->load();
+uint64_t AtomicEvent::value() const {
+  nvtx3::scoped_range r("cu::AtomicEvent::value");
+  return atomic()->load();
 }
 
 } // namespace cu
@@ -188,14 +266,14 @@ namespace {
 
 struct EventImpl {
   // CudaEvent is preferred when possible because it is fast, however we have
-  // to fallback to SharedEvent in following cases:
+  // to fallback to AtomicEvent in following cases:
   // 1. the event is used to wait/signal a cpu stream;
   // 2. signal value other than 1 has been specified.
-  std::unique_ptr<cu::CudaEvent> cuda;
-  std::unique_ptr<cu::SharedEvent> shared;
+  std::unique_ptr<cu::CopyableCudaEvent> cuda;
+  std::unique_ptr<cu::AtomicEvent> atomic;
 
   bool is_created() const {
-    return cuda || shared;
+    return cuda || atomic;
   }
 
   void ensure_created(Stream s, uint64_t signal_value) {
@@ -203,10 +281,10 @@ struct EventImpl {
       return;
     }
     if (s.device == mlx::core::Device::cpu || signal_value > 1) {
-      nvtx3::mark("Using slow SharedEvent");
-      shared = std::make_unique<cu::SharedEvent>();
+      nvtx3::mark("Using slow AtomicEvent");
+      atomic = std::make_unique<cu::AtomicEvent>();
     } else {
-      cuda = std::make_unique<cu::CudaEvent>();
+      cuda = std::make_unique<cu::CopyableCudaEvent>(cu::device(s.device));
     }
   }
 };
@@ -225,7 +303,7 @@ void Event::wait() {
     assert(value() == 1);
     event->cuda->wait();
   } else {
-    event->shared->wait(value());
+    event->atomic->wait(value());
   }
 }
 
@@ -236,7 +314,7 @@ void Event::wait(Stream s) {
     assert(value() == 1);
     event->cuda->wait(s);
   } else {
-    event->shared->wait(s, value());
+    event->atomic->wait(s, value());
   }
 }
 
@@ -247,7 +325,7 @@ void Event::signal(Stream s) {
     assert(value() == 1);
     event->cuda->record(s);
   } else {
-    event->shared->signal(s, value());
+    event->atomic->signal(s, value());
   }
 }
 
@@ -258,9 +336,9 @@ bool Event::is_signaled() const {
   }
   if (event->cuda) {
     assert(value() == 1);
-    return event->cuda->recorded() && event->cuda->completed();
+    return event->cuda->is_signaled();
   } else {
-    return event->shared->is_signaled(value());
+    return event->atomic->is_signaled(value());
   }
 }
 

mlx/backend/cuda/event.h

@@ -3,49 +3,60 @@
 #pragma once
 
 #include "mlx/allocator.h"
+#include "mlx/backend/cuda/utils.h"
 #include "mlx/stream.h"
 
+#include <memory>
+
 #include <cuda_runtime.h>
 #include <cuda/atomic>
 
-#include <memory>
-
 namespace mlx::core::cu {
 
-class CudaEventHandle;
+class Device;
+
+// RAII-managed move-only wrapper of cudaEvent_t.
+struct CudaEventHandle : public CudaHandle<cudaEvent_t, cudaEventDestroy> {
+  CudaEventHandle(Device& d, int flags);
+  Device& device;
+  int flags;
+};
 
 // Wrapper of native cuda event. It can synchronize between GPU streams, or wait
 // on GPU stream in CPU stream, but can not wait on CPU stream.
 class CudaEvent {
  public:
-  CudaEvent();
+  CudaEvent(Device& d, int flags);
+  ~CudaEvent();
+
+  CudaEvent(CudaEvent&&) = default;
+  CudaEvent& operator=(CudaEvent&&) = default;
+
+  CudaEvent(const CudaEvent&) = delete;
+  CudaEvent& operator=(const CudaEvent&) = delete;
 
   void wait();
   void wait(cudaStream_t stream);
-  void wait(Stream s);
   void record(cudaStream_t stream);
-  void record(Stream s);
 
   // Return whether the recorded kernels have completed. Note that this method
   // returns true if record() has not been called.
   bool completed() const;
 
-  bool recorded() const {
-    return recorded_;
-  }
+  // Internal: make sure event pool is initialized.
+  static void init_pool();
 
  private:
-  bool recorded_{false};
-  std::shared_ptr<CudaEventHandle> event_;
+  CudaEventHandle event_;
 };
 
 // Event that can synchronize between CPU and GPU. It is much slower than
 // CudaEvent so the latter should always be preferred when possible.
-class SharedEvent {
+class AtomicEvent {
  public:
   using Atomic = cuda::atomic<uint64_t>;
 
-  SharedEvent();
+  AtomicEvent();
 
   void wait(uint64_t value);
   void wait(cudaStream_t stream, uint64_t value);
@@ -57,7 +68,11 @@ class SharedEvent {
   uint64_t value() const;
 
  private:
-  std::shared_ptr<mlx::core::allocator::Buffer> buf_;
+  Atomic* atomic() const {
+    return static_cast<AtomicEvent::Atomic*>(buf_->raw_ptr());
+  }
+
+  std::shared_ptr<allocator::Buffer> buf_;
 };
 
 } // namespace mlx::core::cu

mlx/backend/cuda/fence.cpp

@@ -7,7 +7,7 @@ namespace mlx::core {
 
 struct FenceImpl {
   uint32_t count;
-  cu::SharedEvent event;
+  cu::AtomicEvent event;
 };
 
 Fence::Fence(Stream s) {

mlx/backend/cuda/gemms/cublas_gemm.cpp

@@ -50,8 +50,10 @@ cublasComputeType_t dtype_to_compute_type(Dtype dtype) {
       return mlx::core::env::enable_tf32() ? CUBLAS_COMPUTE_32F_FAST_TF32
                                            : CUBLAS_COMPUTE_32F;
     case float64:
-    case complex64:
       return CUBLAS_COMPUTE_64F;
+    case complex64:
+      return mlx::core::env::enable_tf32() ? CUBLAS_COMPUTE_32F_FAST_TF32
+                                           : CUBLAS_COMPUTE_32F;
     default:
       throw std::runtime_error(fmt::format(
           "Unsupported dtype in CublasGemm: {}.", dtype_to_string(dtype)));
@@ -85,10 +87,10 @@ cublasLtMatrixLayout_t create_matrix_layout(
     int32_t batch_count,
     int64_t batch_stride) {
   cublasLtMatrixLayout_t desc;
+  if (transposed) {
+    std::swap(rows, cols);
+  }
   CHECK_CUBLAS_ERROR(cublasLtMatrixLayoutCreate(&desc, type, rows, cols, ld));
-  cublasLtOrder_t order = transposed ? CUBLASLT_ORDER_COL : CUBLASLT_ORDER_ROW;
-  CHECK_CUBLAS_ERROR(cublasLtMatrixLayoutSetAttribute(
-      desc, CUBLASLT_MATRIX_LAYOUT_ORDER, &order, sizeof(cublasLtOrder_t)));
   if (batch_count > 1) {
     CHECK_CUBLAS_ERROR(cublasLtMatrixLayoutSetAttribute(
         desc,
@@ -126,37 +128,47 @@ CublasGemm::CublasGemm(
       N_(b_cols) {
   heuristic_.state = CUBLAS_STATUS_NOT_INITIALIZED;
 
-  auto scale_type = dtype_to_cublas_type(dtype);
+  scale_type_ = dtype_to_cublas_type(dtype);
   if (dtype == bfloat16 || dtype == float16) {
-    scale_type = CUDA_R_32F;
+    scale_type_ = CUDA_R_32F;
   }
+
   CHECK_CUBLAS_ERROR(cublasLtMatmulDescCreate(
-      &matmul_desc_, dtype_to_compute_type(dtype), scale_type));
+      &matmul_desc_, dtype_to_compute_type(dtype), scale_type_));
   int32_t pointer_mode = CUBLASLT_POINTER_MODE_HOST;
   CHECK_CUBLAS_ERROR(cublasLtMatmulDescSetAttribute(
       matmul_desc_,
       CUBLASLT_MATMUL_DESC_POINTER_MODE,
       &pointer_mode,
       sizeof(int32_t)));
-  cublasOperation_t op = CUBLAS_OP_N;
+
+  // In cublasLt matrices use column-major layout, while it is possible to use
+  // the CUBLASLT_ORDER_ROW option to switch to row-major layout, the bias
+  // epilogue does not work with the option. So instead we swap A and B to make
+  // cublasLt return the row-major result, which works because:
+  // - the data of a matrix in row-major layout is identical to its transpose in
+  //   column-major layout
+  // - C^T = (A @ B)^T = B^T @ A^T
+  cublasOperation_t a_op = b_transposed ? CUBLAS_OP_T : CUBLAS_OP_N;
   CHECK_CUBLAS_ERROR(cublasLtMatmulDescSetAttribute(
       matmul_desc_,
       CUBLASLT_MATMUL_DESC_TRANSA,
-      &op,
+      &a_op,
       sizeof(cublasOperation_t)));
+  cublasOperation_t b_op = a_transposed ? CUBLAS_OP_T : CUBLAS_OP_N;
   CHECK_CUBLAS_ERROR(cublasLtMatmulDescSetAttribute(
       matmul_desc_,
       CUBLASLT_MATMUL_DESC_TRANSB,
-      &op,
+      &b_op,
       sizeof(cublasOperation_t)));
 
   auto type = dtype_to_cublas_type(dtype);
   a_desc_ = create_matrix_layout(
-      type, a_rows, a_cols, a_transposed, lda, batch_count, a_batch_stride);
+      type, b_cols, b_rows, b_transposed, ldb, batch_count, b_batch_stride);
   b_desc_ = create_matrix_layout(
-      type, b_rows, b_cols, b_transposed, ldb, batch_count, b_batch_stride);
+      type, a_cols, a_rows, a_transposed, lda, batch_count, a_batch_stride);
   out_desc_ = create_matrix_layout(
-      type, a_rows, b_cols, false, b_cols, batch_count, a_rows * b_cols);
+      type, b_cols, a_rows, false, b_cols, batch_count, a_rows * b_cols);
 }
 
 CublasGemm::CublasGemm(
@@ -191,7 +203,7 @@ CublasGemm::CublasGemm(
           b_batch_stride) {
   auto type = dtype_to_cublas_type(dtype);
   c_desc_ = create_matrix_layout(
-      type, a_rows, b_cols, false, ldc, batch_count, c_batch_stride);
+      type, b_cols, a_rows, false, ldc, batch_count, c_batch_stride);
 }
 
 CublasGemm::~CublasGemm() {
@@ -213,14 +225,30 @@ void CublasGemm::set_out(
   CHECK_CUBLAS_ERROR(cublasLtMatrixLayoutDestroy(out_desc_));
   out_desc_ = create_matrix_layout(
       dtype_to_cublas_type(dtype),
-      rows,
       cols,
+      rows,
       transposed,
       ld,
       batch_count,
       batch_stride);
 }
 
+void CublasGemm::set_bias(cu::CommandEncoder& encoder, const array& bias) {
+  encoder.set_input_array(bias);
+  cublasLtEpilogue_t epilogue = CUBLASLT_EPILOGUE_BIAS;
+  CHECK_CUBLAS_ERROR(cublasLtMatmulDescSetAttribute(
+      matmul_desc_,
+      CUBLASLT_MATMUL_DESC_EPILOGUE,
+      &epilogue,
+      sizeof(epilogue)));
+  auto* bias_ptr = bias.data<void>();
+  CHECK_CUBLAS_ERROR(cublasLtMatmulDescSetAttribute(
+      matmul_desc_,
+      CUBLASLT_MATMUL_DESC_BIAS_POINTER,
+      &bias_ptr,
+      sizeof(bias_ptr)));
+}
+
 void CublasGemm::run(
     cu::CommandEncoder& encoder,
     array& out,
@@ -228,11 +256,19 @@ void CublasGemm::run(
     const array& b,
     const Shape& batch_shape,
     const Strides& a_batch_strides,
-    const Strides& b_batch_strides) {
+    const Strides& b_batch_strides,
+    float alpha) {
   int batch_count = out.size() / (M_ * N_);
   if (batch_count / batch_shape.back() > 1) {
     run_batched(
-        encoder, out, a, b, batch_shape, a_batch_strides, b_batch_strides);
+        encoder,
+        out,
+        a,
+        b,
+        batch_shape,
+        a_batch_strides,
+        b_batch_strides,
+        alpha);
     return;
   }
 
@@ -240,7 +276,13 @@ void CublasGemm::run(
   encoder.set_input_array(b);
   encoder.set_output_array(out);
 
-  execute(encoder, out.data<void>(), a.data<void>(), b.data<void>(), nullptr);
+  execute(
+      encoder,
+      out.data<void>(),
+      a.data<void>(),
+      b.data<void>(),
+      nullptr,
+      alpha);
 }
 
 void CublasGemm::run(
@@ -313,6 +355,16 @@ void CublasGemm::execute(
     }
   }
 
+  const void* alpha_ptr = &alpha;
+  const void* beta_ptr = &beta;
+  complex64_t alpha_c, beta_c;
+  if (scale_type_ == CUDA_C_32F) {
+    alpha_c = complex64_t{alpha, 0.0f};
+    beta_c = complex64_t{beta, 0.0f};
+    alpha_ptr = &alpha_c;
+    beta_ptr = &beta_c;
+  }
+
   void* workspace_ptr = nullptr;
   if (heuristic_.workspaceSize > 0) {
     // Ensure workspace is 256-byte aligned
@@ -329,12 +381,12 @@ void CublasGemm::execute(
   CHECK_CUBLAS_ERROR(cublasLtMatmul(
       handle_,
       matmul_desc_,
-      &alpha,
-      a,
+      alpha_ptr,
+      b, // a and b are swapped
       a_desc_,
-      b,
+      a,
       b_desc_,
-      &beta,
+      beta_ptr,
       c ? c : out,
       c ? c_desc_ : out_desc_,
       out,

mlx/backend/cuda/gemms/cublas_gemm.h

@@ -55,6 +55,8 @@ class CublasGemm {
       int32_t batch_count,
       int64_t batch_stride);
 
+  void set_bias(cu::CommandEncoder& encoder, const array& bias);
+
   void run(
       cu::CommandEncoder& encoder,
       array& out,
@@ -62,7 +64,8 @@ class CublasGemm {
       const array& b,
       const Shape& batch_shape,
       const Strides& a_batch_strides,
-      const Strides& b_batch_strides);
+      const Strides& b_batch_strides,
+      float alpha = 1.0f);
 
   void run(
       cu::CommandEncoder& encoder,
@@ -85,7 +88,8 @@ class CublasGemm {
       const array& b,
       const Shape& batch_shape,
       const Strides& a_batch_strides,
-      const Strides& b_batch_strides);
+      const Strides& b_batch_strides,
+      float alpha);
 
   void run_batched(
       cu::CommandEncoder& encoder,
@@ -111,6 +115,7 @@ class CublasGemm {
 
   uint64_t M_;
   uint64_t N_;
+  cudaDataType_t scale_type_;
   cublasLtMatmulPreference_t pref_{nullptr};
   cublasLtHandle_t handle_{nullptr};
   cublasLtMatmulDesc_t matmul_desc_{nullptr};

mlx/backend/cuda/gemms/cublas_gemm_batched_12_0.cpp

@@ -13,7 +13,8 @@ void CublasGemm::run_batched(
     const array& b,
     const Shape& batch_shape,
     const Strides& a_batch_strides,
-    const Strides& b_batch_strides) {
+    const Strides& b_batch_strides,
+    float alpha) {
   encoder.set_input_array(a);
   encoder.set_input_array(b);
   encoder.set_output_array(out);
@@ -27,7 +28,8 @@ void CublasGemm::run_batched(
         out.data<int8_t>() + out.itemsize() * i * batch_shape.back() * M_ * N_,
         a.data<int8_t>() + a.itemsize() * a_it.loc,
         b.data<int8_t>() + b.itemsize() * b_it.loc,
-        nullptr);
+        nullptr,
+        alpha);
     a_it.step();
     b_it.step();
   }

mlx/backend/cuda/gemms/cublas_gemm_batched_12_9.cu

@@ -154,7 +154,8 @@ void CublasGemm::run_batched(
     const array& b,
     const Shape& batch_shape,
     const Strides& a_batch_strides,
-    const Strides& b_batch_strides) {
+    const Strides& b_batch_strides,
+    float alpha) {
   int batch_count = out.size() / (M_ * N_);
   set_pointer_mode(a_desc_, batch_count);
   set_pointer_mode(b_desc_, batch_count);
@@ -226,7 +227,8 @@ void CublasGemm::run_batched(
       reinterpret_cast<void*>(out_pointers),
       reinterpret_cast<void*>(a_pointers),
       reinterpret_cast<void*>(b_pointers),
-      nullptr);
+      nullptr,
+      alpha);
 }
 
 void CublasGemm::run_batched(

mlx/backend/cuda/gemms/gemv.cu

@@ -13,6 +13,37 @@ namespace cg = cooperative_groups;
 
 static constexpr int rows_per_block = 8;
 
+// Accumulator type selection per input element type T.
+template <typename T>
+struct GemvAccType {
+  using type = T;
+};
+
+template <>
+struct GemvAccType<__half> {
+  using type = float;
+};
+
+template <>
+struct GemvAccType<__nv_bfloat16> {
+  using type = float;
+};
+
+template <>
+struct GemvAccType<float> {
+  using type = float;
+};
+
+template <>
+struct GemvAccType<double> {
+  using type = double;
+};
+
+template <>
+struct GemvAccType<cu::complex64_t> {
+  using type = cu::complex64_t;
+};
+
 template <typename T, int rows_per_block, int n_per_thread>
 __device__ void
 gemv_impl(const T* mat, const T* vec, T* out, int rows, int cols) {
@@ -24,7 +55,8 @@ gemv_impl(const T* mat, const T* vec, T* out, int rows, int cols) {
   int row = g_idx.x * rows_per_block + t_idx.y;
 
   if (row < rows) {
-    float sum = 0.0f;
+    using Acc = typename GemvAccType<T>::type;
+    Acc sum = Acc(0);
     for (int col = n_per_thread * warp.thread_rank(); col < cols;
          col += (WARP_SIZE * n_per_thread)) {
       auto local_mat =
@@ -32,12 +64,11 @@ gemv_impl(const T* mat, const T* vec, T* out, int rows, int cols) {
       auto local_vec = unsafe_load_vector<n_per_thread>(vec + col, 0);
 #pragma unroll
       for (int j = 0; j < n_per_thread; ++j) {
-        sum +=
-            static_cast<float>(local_mat[j]) * static_cast<float>(local_vec[j]);
+        sum += static_cast<Acc>(local_mat[j]) * static_cast<Acc>(local_vec[j]);
       }
     }
 
-    sum = cg::reduce(warp, sum, cg::plus<float>{});
+    sum = cg::reduce(warp, sum, cg::plus<Acc>{});
     if (warp.thread_rank() == 0) {
       out[row] = static_cast<T>(sum);
     }
@@ -107,7 +138,7 @@ void gemv(
   encoder.set_input_array(a);
   encoder.set_input_array(b);
   encoder.set_output_array(out);
-  dispatch_float_types(out.dtype(), "gemv", [&](auto type_tag) {
+  dispatch_inexact_types(out.dtype(), "gemv", [&](auto type_tag) {
     using DataType = cuda_type_t<MLX_GET_TYPE(type_tag)>;
     dim3 block_dims{WARP_SIZE, rows_per_block};
     const DataType* mat;

mlx/backend/cuda/jit_module.cpp

@@ -99,6 +99,30 @@ const std::filesystem::path& ptx_cache_dir() {
   return cache;
 }
 
+std::filesystem::path get_ptx_path(
+    const std::filesystem::path& cache_dir,
+    const std::string& module_name) {
+#ifdef _WIN32
+  constexpr int max_file_name_length = 140;
+#else
+  constexpr int max_file_name_length = 245;
+#endif
+
+  if (module_name.size() <= max_file_name_length) {
+    return cache_dir / (module_name + ".ptx");
+  }
+
+  auto ptx_path = cache_dir;
+  int offset = 0;
+  while (module_name.size() - offset > max_file_name_length) {
+    ptx_path /= module_name.substr(offset, max_file_name_length);
+    offset += max_file_name_length;
+  }
+  ptx_path /= module_name.substr(offset) + ".ptx";
+
+  return ptx_path;
+}
+
 // Try to read the cached |ptx| and |ptx_kernels| from |cache_dir|.
 bool read_cached_ptx(
     const std::filesystem::path& cache_dir,
@@ -109,7 +133,7 @@ bool read_cached_ptx(
     return false;
   }
 
-  auto ptx_path = cache_dir / (module_name + ".ptx");
+  auto ptx_path = get_ptx_path(cache_dir, module_name);
   std::error_code error;
   auto ptx_size = std::filesystem::file_size(ptx_path, error);
   if (error) {
@@ -122,7 +146,7 @@ bool read_cached_ptx(
   ptx.resize(ptx_size);
   ptx_file.read(ptx.data(), ptx_size);
 
-  std::ifstream txt_file(cache_dir / (module_name + ".txt"), std::ios::binary);
+  std::ifstream txt_file(ptx_path.replace_extension(".txt"), std::ios::binary);
   std::string line;
   while (std::getline(txt_file, line)) {
     auto tab = line.find('\t');
@@ -144,16 +168,26 @@ void write_cached_ptx(
     return;
   }
 
-  std::ofstream ptx_file(cache_dir / (module_name + ".ptx"), std::ios::binary);
+  auto ptx_path = get_ptx_path(cache_dir, module_name);
+
+  // Ensure that the directory exists
+  auto parent = ptx_path.parent_path();
+  if (parent != cache_dir) {
+    std::filesystem::create_directories(parent);
+  }
+
+  // Write the compiled code and mangled names
+  std::ofstream ptx_file(ptx_path, std::ios::binary);
   if (!ptx.empty()) {
     ptx_file.write(&ptx.front(), ptx.size());
   }
-  std::ofstream txt_file(cache_dir / (module_name + ".txt"), std::ios::binary);
+  std::ofstream txt_file(ptx_path.replace_extension(".txt"), std::ios::binary);
   for (const auto& [name, mangled] : ptx_kernels) {
     txt_file << name << "\t" << mangled << std::endl;
   }
 
-  std::ofstream source_file(cache_dir / (module_name + ".cu"));
+  // Write the generated code
+  std::ofstream source_file(ptx_path.replace_extension(".cu"));
   source_file << source_code;
 }
 
@@ -297,7 +331,8 @@ void load_module(
     const std::string& ptx,
     const std::vector<std::pair<std::string, std::string>>& ptx_kernels,
     CUmodule& module_,
-    std::unordered_map<std::string, std::pair<CUfunction, bool>>& kernels) {
+    std::unordered_map<std::string, std::tuple<CUfunction, bool, uint>>&
+        kernels) {
   // Load module.
   char jit_log[4089] = {};
   CUjit_option options[] = {
@@ -314,7 +349,7 @@ void load_module(
   for (const auto& [name, mangled] : ptx_kernels) {
     CUfunction kernel;
     CHECK_CUDA_ERROR(cuModuleGetFunction(&kernel, module_, mangled.c_str()));
-    kernels[name] = std::make_pair(kernel, false);
+    kernels[name] = std::make_tuple(kernel, false, 0);
   }
 }
 
@@ -358,7 +393,7 @@ JitModule::~JitModule() {
   CHECK_CUDA_ERROR(cuModuleUnload(module_));
 }
 
-CUfunction JitModule::get_kernel(
+std::pair<CUfunction, uint> JitModule::get_kernel_and_dims(
     const std::string& kernel_name,
     std::function<void(CUfunction)> configure_kernel) {
   auto it = kernels_.find(kernel_name);
@@ -369,14 +404,22 @@ CUfunction JitModule::get_kernel(
 
   // If it is the first time we run this kernel then configure it. Do it only
   // once!
-  if (!it->second.second) {
+  auto kernel = std::get<0>(it->second);
+  if (!std::get<1>(it->second)) {
     if (configure_kernel) {
-      configure_kernel(it->second.first);
+      configure_kernel(kernel);
     }
-    it->second.second = true;
+    std::get<1>(it->second) = true;
+    std::get<2>(it->second) = max_occupancy_block_dim(kernel);
   }
 
-  return it->second.first;
+  return {kernel, std::get<2>(it->second)};
+}
+
+CUfunction JitModule::get_kernel(
+    const std::string& kernel_name,
+    std::function<void(CUfunction)> configure_kernel) {
+  return get_kernel_and_dims(kernel_name, std::move(configure_kernel)).first;
 }
 
 std::unordered_map<std::string, JitModule>& get_jit_module_cache() {

mlx/backend/cuda/jit_module.h

@@ -99,10 +99,13 @@ class JitModule {
   CUfunction get_kernel(
       const std::string& kernel_name,
       std::function<void(CUfunction)> configure_kernel = nullptr);
+  std::pair<CUfunction, uint> get_kernel_and_dims(
+      const std::string& kernel_name,
+      std::function<void(CUfunction)> configure_kernel = nullptr);
 
  private:
   CUmodule module_{nullptr};
-  std::unordered_map<std::string, std::pair<CUfunction, bool>> kernels_;
+  std::unordered_map<std::string, std::tuple<CUfunction, bool, uint>> kernels_;
 };
 
 std::unordered_map<std::string, JitModule>& get_jit_module_cache();

mlx/backend/cuda/kernel_utils.cu

@@ -35,12 +35,10 @@ std::tuple<dim3, uint> get_launch_args(
     const Shape& shape,
     const Strides& strides,
     bool large,
-    int work_per_thread) {
+    int work_per_thread /* = 1 */,
+    uint max_block_dim /* = 1024 */) {
   size_t nthreads = cuda::ceil_div(size, work_per_thread);
-  uint block_dim = 1024;
-  if (block_dim > nthreads) {
-    block_dim = nthreads;
-  }
+  uint block_dim = max_block_dim < nthreads ? max_block_dim : nthreads;
   dim3 num_blocks;
   if (large) {
     num_blocks = get_2d_grid_dims(shape, strides, work_per_thread);

mlx/backend/cuda/kernel_utils.cuh

@@ -1,8 +1,8 @@
 // Copyright © 2025 Apple Inc.
 
-// This file includes host-only utilies for writing CUDA kernels, the difference
-// from backend/cuda/device/utils.cuh is that the latter file only include
-// device-only code.
+// This file includes host-only utilities for writing CUDA kernels, the
+// difference from backend/cuda/device/utils.cuh is that the latter file only
+// include device-only code.
 
 #pragma once
 
@@ -120,19 +120,28 @@ dim3 get_2d_grid_dims(
     size_t divisor);
 std::pair<dim3, dim3> get_grid_and_block(int dim0, int dim1, int dim2);
 
-// Get the num_blocks and block_dims that maximize occupancy for |kernel|,
-// assuming each thread handles |work_per_thread| elements of |arr|.
+// Get the num_blocks and block_dims assuming each thread handles
+// |work_per_thread| elements of |arr|.
 std::tuple<dim3, uint> get_launch_args(
     size_t size,
     const Shape& shape,
     const Strides& strides,
     bool large,
-    int work_per_thread = 1);
+    int work_per_thread = 1,
+    uint max_block_dim = 1024);
 
-inline std::tuple<dim3, uint>
-get_launch_args(const array& arr, bool large, int work_per_thread = 1) {
+inline std::tuple<dim3, uint> get_launch_args(
+    const array& arr,
+    bool large,
+    int work_per_thread = 1,
+    uint max_block_dim = 1024) {
   return get_launch_args(
-      arr.size(), arr.shape(), arr.strides(), large, work_per_thread);
+      arr.size(),
+      arr.shape(),
+      arr.strides(),
+      large,
+      work_per_thread,
+      max_block_dim);
 }
 
 } // namespace mlx::core

mlx/backend/cuda/lru_cache.h

@@ -2,11 +2,15 @@
 
 #pragma once
 
+#include "mlx/utils.h"
+
 #include <cstring>
 #include <list>
 #include <unordered_map>
 #include <utility>
 
+#include <fmt/format.h>
+
 namespace mlx::core {
 
 template <
@@ -27,6 +31,14 @@ class LRUCache {
     }
   }
 
+  // Initialize with capacity read from |env_name|.
+  LRUCache(const char* env_name, int default_capacity)
+      : LRUCache(env::get_var(env_name, default_capacity)) {
+    if (env::get_var("MLX_ENABLE_CACHE_THRASHING_CHECK", 1)) {
+      env_name_ = env_name;
+    }
+  }
+
   size_t size() const {
     return map_.size();
   }
@@ -76,6 +88,14 @@ class LRUCache {
       return {it->second, false};
     }
 
+    if (env_name_ && ++cache_misses_ > 2 * capacity_) {
+      throw std::runtime_error(fmt::format(
+          "Cache thrashing is happening, please set the environment variable "
+          "{} to a larger value than {} to fix degraded performance.",
+          env_name_,
+          capacity_));
+    }
+
     vlist_.emplace_front(key, std::forward<U>(value));
     map_[key] = vlist_.begin();
 
@@ -106,6 +126,9 @@ class LRUCache {
     }
   }
 
+  const char* env_name_{nullptr};
+  size_t cache_misses_{0};
+
   list_type vlist_;
   map_type map_;
   size_t capacity_;

mlx/backend/cuda/matmul.cpp

@@ -11,6 +11,7 @@
 #include <numeric>
 
 namespace mlx::core {
+
 namespace {
 
 std::tuple<bool, int64_t, array>
@@ -28,6 +29,80 @@ check_transpose(cu::CommandEncoder& enc, const Stream& s, const array& arr) {
   }
 }
 
+void gemm_and_bias(
+    cu::CommandEncoder& encoder,
+    int M,
+    int N,
+    int K,
+    bool a_transposed,
+    int64_t lda,
+    bool b_transposed,
+    int64_t ldb,
+    array& out,
+    const array& a,
+    const array& b,
+    const std::optional<array>& bias = std::nullopt,
+    float alpha = 1.0f) {
+  // Check and collapse batch dimensions
+  auto [batch_shape, a_batch_strides, b_batch_strides] = collapse_batches(a, b);
+
+  auto batch_count = out.size() / (M * N);
+
+  // Collapse batches into M if needed
+  if (batch_count > 1 && !a_transposed && batch_shape.size() == 1 &&
+      a.strides()[a.ndim() - 2] == K && a_batch_strides.back() == M * K &&
+      b_batch_strides.back() == 0) {
+    M *= batch_shape.back();
+    batch_count = 1;
+
+    a_batch_strides = {0};
+    b_batch_strides = {0};
+    batch_shape = {1};
+  }
+
+  // Use gemmv when possible
+  if (!bias && cu::can_use_gemv(M, N, K, a_transposed, b_transposed)) {
+    cu::gemv(
+        a,
+        b,
+        out,
+        M,
+        N,
+        K,
+        batch_count,
+        batch_shape,
+        a_batch_strides,
+        b_batch_strides,
+        encoder);
+    return;
+  }
+
+  // Invoke cublasLt
+  CublasGemm gemm(
+      encoder.device(),
+      a.dtype(),
+      a_transposed,
+      M,
+      K,
+      lda,
+      b_transposed,
+      K,
+      N,
+      ldb,
+      batch_shape.back(),
+      a_batch_strides.back(),
+      b_batch_strides.back());
+  if (bias) {
+    if (a.dtype() == complex64) {
+      throw std::runtime_error(
+          "[gemm_and_bias] complex64 bias epilogue isn’t supported in cublasLtMatmul.");
+    }
+    gemm.set_bias(encoder, *bias);
+  }
+  gemm.run(
+      encoder, out, a, b, batch_shape, a_batch_strides, b_batch_strides, alpha);
+}
+
 } // namespace
 
 void Matmul::eval_gpu(const std::vector<array>& inputs, array& out) {
@@ -48,9 +123,6 @@ void Matmul::eval_gpu(const std::vector<array>& inputs, array& out) {
 
   out.set_data(allocator::malloc(out.nbytes()));
 
-  /////////////////////////////////////////////////////////////////////////////
-  // Init checks and prep
-
   int M = a_pre.shape(-2);
   int N = b_pre.shape(-1);
   int K = a_pre.shape(-1);
@@ -60,58 +132,8 @@ void Matmul::eval_gpu(const std::vector<array>& inputs, array& out) {
   auto [a_transposed, lda, a] = check_transpose(encoder, s, a_pre);
   auto [b_transposed, ldb, b] = check_transpose(encoder, s, b_pre);
 
-  /////////////////////////////////////////////////////////////////////////////
-  // Check and collapse batch dimensions
-
-  auto [batch_shape, a_batch_strides, b_batch_strides] = collapse_batches(a, b);
-
-  auto batch_count = out.size() / (M * N);
-
-  // Collapse batches into M if needed
-  if (batch_count > 1 && !a_transposed && batch_shape.size() == 1 &&
-      a.strides()[a.ndim() - 2] == K && a_batch_strides.back() == M * K &&
-      b_batch_strides.back() == 0) {
-    M *= batch_shape.back();
-    batch_count = 1;
-
-    a_batch_strides = {0};
-    b_batch_strides = {0};
-    batch_shape = {1};
-  }
-
-  if (cu::can_use_gemv(M, N, K, a_transposed, b_transposed)) {
-    cu::gemv(
-        a,
-        b,
-        out,
-        M,
-        N,
-        K,
-        batch_count,
-        batch_shape,
-        a_batch_strides,
-        b_batch_strides,
-        encoder);
-    return;
-  }
-
-  /////////////////////////////////////////////////////////////////////////////
-  // Invoke cublasLt
-  CublasGemm gemm(
-      cu::device(s.device),
-      a.dtype(),
-      a_transposed,
-      M,
-      K,
-      lda,
-      b_transposed,
-      K,
-      N,
-      ldb,
-      batch_shape.back(),
-      a_batch_strides.back(),
-      b_batch_strides.back());
-  gemm.run(encoder, out, a, b, batch_shape, a_batch_strides, b_batch_strides);
+  gemm_and_bias(
+      encoder, M, N, K, a_transposed, lda, b_transposed, ldb, out, a, b);
 }
 
 void AddMM::eval_gpu(const std::vector<array>& inputs, array& out) {
@@ -136,6 +158,29 @@ void AddMM::eval_gpu(const std::vector<array>& inputs, array& out) {
   auto [a_transposed, lda, a] = check_transpose(encoder, s, a_pre);
   auto [b_transposed, ldb, b] = check_transpose(encoder, s, b_pre);
 
+  /////////////////////////////////////////////////////////////////////////////
+  // Dispatch to GEMM with epilogue or AddMM
+
+  if (beta_ == 1 && a.dtype() != complex64 && c.strides(-1) == 1 &&
+      c.data_size() == out.shape(-1)) {
+    out.set_data(allocator::malloc(out.nbytes()));
+    gemm_and_bias(
+        encoder,
+        M,
+        N,
+        K,
+        a_transposed,
+        lda,
+        b_transposed,
+        ldb,
+        out,
+        a,
+        b,
+        c,
+        alpha_);
+    return;
+  }
+
   int64_t ldc;
   {
     auto stx = c.strides()[c.ndim() - 2];
@@ -177,7 +222,7 @@ void AddMM::eval_gpu(const std::vector<array>& inputs, array& out) {
   }
 
   /////////////////////////////////////////////////////////////////////////////
-  // Invoke cublasLt
+  // Invoke cublasLt with AddMM settings
 
   CublasGemm gemm(
       cu::device(s.device),

mlx/backend/cuda/quantized/affine_quantize.cu

@@ -306,7 +306,7 @@ void affine_dequantize(
   enc.set_input_array(scales);
   enc.set_input_array(biases);
   enc.set_output_array(w);
-  dispatch_float_types(w.dtype(), "affine_quantize", [&](auto type_tag) {
+  dispatch_float_types(w.dtype(), "affine_dequantize", [&](auto type_tag) {
     dispatch_groups(group_size_, [&](auto group_size) {
       dispatch_bits(bits_, [&](auto bits) {
         using T = cuda_type_t<MLX_GET_TYPE(type_tag)>;

mlx/backend/cuda/quantized/convert_fp8.cu

@@ -0,0 +1,19 @@
+// Copyright © 2025 Apple Inc.
+#include "mlx/backend/cuda/unary/unary.cuh"
+#include "mlx/fast_primitives.h"
+
+namespace mlx::core {
+void fast::ConvertFP8::eval_gpu(
+    const std::vector<array>& inputs,
+    std::vector<array>& outputs) {
+  nvtx3::scoped_range r("ConvertFP8::eval_gpu");
+  auto& in = inputs[0];
+  auto& out = outputs[0];
+  auto& s = out.primitive().stream();
+  if (to_fp8_) {
+    unary_op_gpu<cu::ToFP8>(inputs, out, name(), s);
+  } else {
+    unary_op_gpu<cu::FromFP8>(inputs, out, name(), s);
+  }
+}
+} // namespace mlx::core

mlx/backend/cuda/quantized/cuda_fp4.h

@@ -0,0 +1,83 @@
+#pragma once
+
+struct __nv_fp8_e8m0 {
+  __device__ __nv_fp8_e8m0(float x) {
+    if (!std::isfinite(x)) {
+      __x = 0xFF;
+      return;
+    }
+    if (x < 0.0f) {
+      __x = 0x00;
+      return;
+    }
+    float le = std::log2f(x);
+    int n = static_cast<int>(std::nearbyintf(le));
+
+    n = n < -127 ? -127 : n;
+    n = n > 127 ? 127 : n;
+    __x = static_cast<uint8_t>(n + 127);
+  }
+
+  __device__ operator float() {
+    if (__x == 0xFF) {
+      return std::numeric_limits<float>::quiet_NaN();
+    }
+    return std::ldexp(1.0f, static_cast<int>(__x) - 127);
+  }
+
+  uint8_t __x{0};
+};
+
+struct __nv_fp4_e2m1 {
+  __device__ __nv_fp4_e2m1(float x) {
+    if (std::isnan(x)) {
+      __x = 0x7;
+      return;
+    }
+
+    const uint8_t sign_bit = (std::signbit(x)) ? 0x8 : 0x0;
+    x = std::abs(x);
+
+    if (x > 5.0f) {
+      __x = 0x7;
+    } else if (x >= 3.5f) {
+      __x = 0x6;
+    } else if (x > 2.5f) {
+      __x = 0x5;
+    } else if (x >= 1.75f) {
+      __x = 0x4;
+    } else if (x > 1.25f) {
+      __x = 0x3;
+    } else if (x >= 0.75f) {
+      __x = 0x2;
+    } else if (x > 0.25f) {
+      __x = 0x1;
+    } else {
+      __x = 0x0;
+    }
+    __x |= sign_bit;
+  }
+
+  __device__ operator float() {
+    static const float LUT[16] = {
+        0.0f,
+        0.5f,
+        1.0f,
+        1.5f,
+        2.0f,
+        3.0f,
+        4.0f,
+        6.0f,
+        -0.0f,
+        -0.5f,
+        -1.0f,
+        -1.5f,
+        -2.0f,
+        -3.0f,
+        -4.0f,
+        -6.0f};
+
+    return LUT[__x];
+  }
+  uint8_t __x{0};
+};

mlx/backend/cuda/quantized/fp_quantize.cu

@@ -0,0 +1,216 @@
+// Copyright © 2025 Apple Inc.
+
+#include "mlx/backend/cuda/device.h"
+#include "mlx/backend/cuda/kernel_utils.cuh"
+#include "mlx/backend/cuda/quantized/quantized.h"
+#include "mlx/dtype_utils.h"
+
+#include <cooperative_groups.h>
+#include <cooperative_groups/reduce.h>
+#include <cuda_fp4.h>
+#include <cuda_fp8.h>
+
+namespace mlx::core {
+namespace cu {
+
+template <int bits>
+struct Quantize {
+  __device__ uint8_t operator()(float x) {
+    if constexpr (bits == 8) {
+      return __nv_fp8_e4m3(x).__x;
+    } else {
+      return __nv_fp4_e2m1(x).__x;
+    }
+  }
+};
+
+template <int bits>
+struct Dequantize {
+  __device__ float operator()(uint8_t x) {
+    if constexpr (bits == 8) {
+      return float(*(__nv_fp8_e4m3*)(&x));
+    } else {
+      return float(*(__nv_fp4_e2m1*)(&x));
+    }
+  }
+};
+
+namespace cg = cooperative_groups;
+
+template <typename T, int group_size, int bits, bool use_mx_scale>
+__global__ void
+fp_quantize(const T* w, uint8_t* out, uint8_t* scales, size_t size) {
+  auto block_size = cg::this_thread_block().dim_threads();
+  auto block_idx = cg::this_thread_block().group_index();
+  auto idx_in_block = cg::this_thread_block().thread_index();
+
+  auto tidx = block_idx.x * block_size.x + idx_in_block.x;
+  auto tidy = block_idx.y * block_size.y + idx_in_block.y;
+
+  auto grid_dim_x =
+      cg::this_grid().dim_blocks().x * cg::this_grid().block_index().x;
+  size_t index = tidx + grid_dim_x * size_t(tidy);
+  if (index >= size) {
+    return;
+  }
+
+  float w_thread = w[index];
+
+  cg::greater<float> max_op;
+  auto warp = cg::tiled_partition<group_size>(cg::this_thread_block());
+
+  float scale = cg::reduce(warp, abs(w_thread), max_op);
+  scale /= bits == 4 ? 6.0f : 448.0f;
+  // Convert to mx scale or nv scale
+  using ScaleType =
+      std::conditional_t<use_mx_scale, __nv_fp8_e8m0, __nv_fp8_e4m3>;
+  auto s = ScaleType(scale);
+  uint8_t q_scale = s.__x;
+  scale = float(s);
+
+  // Write out the scales
+  size_t gindex = index / group_size;
+  if (index % group_size == 0) {
+    scales[gindex] = q_scale;
+  }
+
+  uint8_t output = Quantize<bits>{}(scale == 0 ? 0.0f : w_thread / scale);
+  if (bits == 4) {
+    uint8_t sval = warp.shfl_down(output, 1);
+    output |= sval << bits;
+  }
+  constexpr int pack_factor = bits == 8 ? 1 : 2;
+  if (index % pack_factor == 0) {
+    out[index / pack_factor] = output;
+  }
+}
+
+template <typename T, int group_size, int bits, bool use_mx_scale>
+__global__ void
+fp_dequantize(const uint8_t* w, const uint8_t* scales, T* out, size_t size) {
+  auto block_size = cg::this_thread_block().dim_threads();
+  auto block_idx = cg::this_thread_block().group_index();
+  auto idx_in_block = cg::this_thread_block().thread_index();
+
+  auto tidx = block_idx.x * block_size.x + idx_in_block.x;
+  auto tidy = block_idx.y * block_size.y + idx_in_block.y;
+
+  auto grid_dim_x =
+      cg::this_grid().dim_blocks().x * cg::this_grid().block_index().x;
+
+  constexpr int pack_factor = bits == 8 ? 1 : 2;
+  size_t offset = tidx + grid_dim_x * size_t(tidy);
+  size_t oindex = offset * pack_factor;
+
+  if (oindex >= size) {
+    return;
+  }
+
+  size_t gindex = oindex / group_size;
+  using ScaleType =
+      std::conditional_t<use_mx_scale, __nv_fp8_e8m0, __nv_fp8_e4m3>;
+  auto scale = float(((ScaleType*)(scales))[gindex]);
+
+  out += oindex;
+
+  uint val = w[offset];
+#pragma clang loop unroll(full)
+  for (int i = 0; i < pack_factor; i++) {
+    uint8_t d;
+    if (bits == 4) {
+      d = (val >> (bits * i)) & 0x0f;
+    } else if (bits == 8) {
+      d = val;
+    }
+    out[i] = static_cast<T>(scale * Dequantize<bits>{}(d));
+  }
+}
+
+} // namespace cu
+
+void fp_quantize(
+    const array& w,
+    array& wq,
+    array& scales,
+    int group_size,
+    int bits,
+    cu::CommandEncoder& enc,
+    const Stream& s) {
+  enc.set_input_array(w);
+  enc.set_output_array(wq);
+  enc.set_output_array(scales);
+  dispatch_float_types(w.dtype(), "fp_quantize", [&](auto type_tag) {
+    using T = cuda_type_t<MLX_GET_TYPE(type_tag)>;
+    if constexpr (!std::is_same_v<T, double>) {
+      auto kernel = cu::fp_quantize<T, 32, 4, true>;
+      if (bits == 8) {
+        kernel = cu::fp_quantize<T, 32, 8, true>;
+      } else if (group_size == 16) {
+        kernel = cu::fp_quantize<T, 16, 4, false>;
+      }
+      bool large = w.size() > UINT_MAX;
+      auto [num_blocks, block_dims] =
+          get_launch_args(w.size(), w.shape(), w.strides(), large);
+      enc.add_kernel_node(
+          kernel,
+          num_blocks,
+          block_dims,
+          0,
+          w.data<T>(),
+          wq.data<uint8_t>(),
+          scales.data<uint8_t>(),
+          w.size());
+    } else {
+      throw std::runtime_error(
+          "[Quantize::eval_gpu] Can not quantize input with type float64.");
+    }
+  });
+}
+
+void fp_dequantize(
+    const array& wq,
+    const array& scales,
+    array& w,
+    int group_size,
+    int bits,
+    cu::CommandEncoder& enc,
+    const Stream& s) {
+  constexpr int uint8_per_uint32 = 4;
+  int packs_per_int = 8 / bits;
+
+  size_t size = w.size() / packs_per_int;
+  bool large = size > UINT_MAX;
+  auto grid_shape = w.shape();
+  grid_shape.back() *= uint8_per_uint32;
+
+  enc.set_input_array(wq);
+  enc.set_input_array(scales);
+  enc.set_output_array(w);
+  dispatch_float_types(w.dtype(), "fp_dequantize", [&](auto type_tag) {
+    using T = cuda_type_t<MLX_GET_TYPE(type_tag)>;
+    if constexpr (!std::is_same_v<T, double>) {
+      auto kernel = cu::fp_dequantize<T, 32, 4, true>;
+      if (bits == 8) {
+        kernel = cu::fp_dequantize<T, 32, 8, true>;
+      } else if (group_size == 16) {
+        kernel = cu::fp_dequantize<T, 16, 4, false>;
+      }
+      auto [num_blocks, block_dims] =
+          get_launch_args(size, grid_shape, w.strides(), large);
+      enc.add_kernel_node(
+          kernel,
+          num_blocks,
+          block_dims,
+          0,
+          wq.data<uint8_t>(),
+          scales.data<T>(),
+          w.data<T>(),
+          w.size());
+    } else {
+      throw std::runtime_error(
+          "[Quantize::eval_gpu] Can not dequantize to output with type float64.");
+    }
+  });
+}
+
+} // namespace mlx::core

mlx/backend/cuda/quantized/quantized.cpp

@@ -57,23 +57,30 @@ void fast::Quantize::eval_gpu(
   if (dequantize_) {
     auto wq = ensure_row_contiguous(inputs[0], enc, s);
     auto scales = ensure_row_contiguous(inputs[1], enc, s);
-    auto biases = ensure_row_contiguous(inputs[2], enc, s);
     auto& w = outputs[0];
 
     w.set_data(allocator::malloc(w.nbytes()));
 
-    affine_dequantize(wq, scales, biases, w, group_size_, bits_, enc, s);
+    if (mode_ == QuantizationMode::Affine) {
+      auto biases = ensure_row_contiguous(inputs[2], enc, s);
+      affine_dequantize(wq, scales, biases, w, group_size_, bits_, enc, s);
+    } else {
+      fp_dequantize(wq, scales, w, group_size_, bits_, enc, s);
+    }
   } else {
     auto w = ensure_row_contiguous(inputs[0], enc, s);
     auto& wq = outputs[0];
     auto& scales = outputs[1];
-    auto& biases = outputs[2];
 
     wq.set_data(allocator::malloc(wq.nbytes()));
     scales.set_data(allocator::malloc(scales.nbytes()));
-    biases.set_data(allocator::malloc(biases.nbytes()));
-
-    affine_quantize(w, wq, scales, biases, group_size_, bits_, enc, s);
+    if (mode_ == QuantizationMode::Affine) {
+      auto& biases = outputs[2];
+      biases.set_data(allocator::malloc(biases.nbytes()));
+      affine_quantize(w, wq, scales, biases, group_size_, bits_, enc, s);
+    } else {
+      fp_quantize(w, wq, scales, group_size_, bits_, enc, s);
+    }
   }
 }
 

mlx/backend/cuda/quantized/quantized.h

@@ -24,4 +24,22 @@ void affine_dequantize(
     cu::CommandEncoder& enc,
     const Stream& s);
 
+void fp_quantize(
+    const array& w,
+    array& wq,
+    array& scales,
+    int group_size,
+    int bits,
+    cu::CommandEncoder& enc,
+    const Stream& s);
+
+void fp_dequantize(
+    const array& wq,
+    const array& scales,
+    array& w,
+    int group_size,
+    int bits,
+    cu::CommandEncoder& enc,
+    const Stream& s);
+
 } // namespace mlx::core

mlx/backend/cuda/reduce/col_reduce.cu

@@ -181,6 +181,47 @@ col_reduce_looped(T* in, U* out, const __grid_constant__ ColReduceArgs args) {
   }
 }
 
+template <typename T, typename U, typename Op, int N_READS = 4>
+__global__ void col_reduce_small(
+    const T* in,
+    U* out,
+    const __grid_constant__ ColReduceArgs args,
+    size_t total) {
+  Op op;
+  auto grid = cg::this_grid();
+  auto block = cg::this_thread_block();
+
+  const auto idx = grid.thread_rank() * N_READS;
+  const auto before_axis = idx / args.reduction_stride;
+  const auto after_axis = idx % args.reduction_stride;
+  const auto offset =
+      before_axis * args.reduction_stride * args.reduction_size + after_axis;
+
+  if (idx >= total) {
+    return;
+  }
+
+  in += offset;
+  out += idx;
+
+  AlignedVector<U, N_READS> accumulator;
+  for (int i = 0; i < N_READS; i++) {
+    accumulator[i] = ReduceInit<Op, T>::value();
+  }
+
+  for (int i = 0; i < args.reduction_size; i++) {
+    auto values = load_vector<N_READS>(in, 0);
+
+    for (int j = 0; j < N_READS; j++) {
+      accumulator[j] = op(accumulator[j], cast_to<U>(values[j]));
+    }
+
+    in += args.reduction_stride;
+  }
+
+  store_vector(out, 0, accumulator);
+}
+
 } // namespace cu
 
 inline auto output_grid_for_col_reduce(
@@ -206,7 +247,7 @@ void col_reduce_looped(
     Reduce::ReduceType reduce_type,
     const std::vector<int>& axes,
     const ReductionPlan& plan,
-    cu::ColReduceArgs args) {
+    const cu::ColReduceArgs& args) {
   // Allocate data for the output using in's layout to access them as
   // contiguously as possible.
   allocate_same_layout(out, in, axes);
@@ -230,12 +271,55 @@ void col_reduce_looped(
         auto kernel =
             cu::col_reduce_looped<T, U, OP, reduce_ndim(), BM, BN, N_READS>;
         encoder.add_kernel_node(
-            kernel, grid, blocks, 0, indata, out.data<U>(), args);
+            kernel,
+            grid,
+            blocks,
+            0,
+            indata,
+            out.data<U>(),
+            static_cast<cu::ColReduceArgs>(args));
       });
     });
   });
 }
 
+void col_reduce_small(
+    cu::CommandEncoder& encoder,
+    const array& in,
+    array& out,
+    Reduce::ReduceType reduce_type,
+    const std::vector<int>& axes,
+    const ReductionPlan& plan,
+    const cu::ColReduceArgs& args) {
+  // Allocate data for the output using in's layout to access them as
+  // contiguously as possible.
+  allocate_same_layout(out, in, axes);
+
+  encoder.set_input_array(in);
+  encoder.set_output_array(out);
+  dispatch_all_types(in.dtype(), [&](auto type_tag) {
+    dispatch_reduce_ops(reduce_type, [&](auto reduce_type_tag) {
+      using OP = MLX_GET_TYPE(reduce_type_tag);
+      using T = cuda_type_t<MLX_GET_TYPE(type_tag)>;
+      using U = typename cu::ReduceResult<OP, T>::type;
+
+      constexpr int N_READS = 16 / sizeof(T);
+      auto tmp_grid = get_2d_grid_dims(out.shape(), out.strides());
+      auto [grid, block] = get_grid_and_block(tmp_grid.x, tmp_grid.y, 1);
+      auto kernel = cu::col_reduce_small<T, U, OP, N_READS>;
+      encoder.add_kernel_node(
+          kernel,
+          grid,
+          block,
+          0,
+          in.data<T>(),
+          out.data<U>(),
+          static_cast<cu::ColReduceArgs>(args),
+          out.size());
+    });
+  });
+}
+
 void col_reduce(
     cu::CommandEncoder& encoder,
     const array& in,
@@ -258,6 +342,13 @@ void col_reduce(
   // Make the args struct to help route to the best kernel
   cu::ColReduceArgs args(in, plan, axes);
 
+  // Small col reduce with a single or contiguous reduction axis
+  if (args.non_col_reductions == 1 && args.reduction_size <= 32 &&
+      args.reduction_stride % (16 / in.itemsize()) == 0) {
+    col_reduce_small(encoder, in, out, reduce_type, axes, plan, args);
+    return;
+  }
+
   // Fallback col reduce
   col_reduce_looped(encoder, in, out, reduce_type, axes, plan, args);
 }

mlx/backend/cuda/reduce/row_reduce.cu

@@ -7,8 +7,6 @@
 
 #include <cooperative_groups.h>
 #include <cooperative_groups/reduce.h>
-#include <cub/block/block_load.cuh>
-#include <cub/block/block_reduce.cuh>
 
 namespace mlx::core {
 
@@ -83,7 +81,8 @@ struct RowReduceArgs {
 };
 
 template <typename T, typename U, typename ReduceOp, int N = 4, int M = 1>
-__global__ void row_reduce_simple(T* in, U* out, size_t n_rows, int size) {
+__global__ void
+row_reduce_simple(const T* in, U* out, size_t n_rows, int size) {
   auto grid = cg::this_grid();
   auto block = cg::this_thread_block();
   auto warp = cg::tiled_partition<WARP_SIZE>(block);
@@ -91,8 +90,8 @@ __global__ void row_reduce_simple(T* in, U* out, size_t n_rows, int size) {
   const U init = cu::ReduceInit<ReduceOp, T>::value();
   ReduceOp op;
 
-  T vals[M][N];
-  U accs[M];
+  AlignedVector<T, N> vals[M];
+  AlignedVector<U, M> accs;
   for (int i = 0; i < M; i++) {
     accs[i] = init;
   }
@@ -101,43 +100,31 @@ __global__ void row_reduce_simple(T* in, U* out, size_t n_rows, int size) {
       min(n_rows - M, static_cast<size_t>(grid.block_rank() * M));
   const size_t full_blocks = size / (block.size() * N);
   const size_t final_offset = full_blocks * (block.size() * N);
-  in += start_row * size;
+  in += start_row * size + block.thread_rank() * N;
   out += start_row;
 
-  if (size % N == 0) {
-    for (size_t r = 0; r < full_blocks; r++) {
-      for (int k = 0; k < M; k++) {
-        cub::LoadDirectBlockedVectorized<T, N>(
-            block.thread_rank(),
-            in + k * size + r * (block.size() * N),
-            vals[k]);
-        for (int j = 0; j < N; j++) {
-          accs[k] = op(accs[k], cast_to<U>(vals[k][j]));
-        }
-      }
-    }
-  } else {
-    for (size_t r = 0; r < full_blocks; r++) {
-      for (int k = 0; k < M; k++) {
-        cub::LoadDirectBlocked(
-            block.thread_rank(),
-            in + k * size + r * (block.size() * N),
-            vals[k]);
-        for (int j = 0; j < N; j++) {
-          accs[k] = op(accs[k], cast_to<U>(vals[k][j]));
-        }
+  for (size_t r = 0; r < full_blocks; r++) {
+    for (int k = 0; k < M; k++) {
+      vals[k] = load_vector<N>(in + k * size, 0);
+    }
+    for (int k = 0; k < M; k++) {
+      for (int j = 0; j < N; j++) {
+        accs[k] = op(accs[k], cast_to<U>(vals[k][j]));
       }
     }
+
+    in += block.size() * N;
   }
 
   if (final_offset < size) {
     for (int k = 0; k < M; k++) {
-      cub::LoadDirectBlocked(
-          block.thread_rank(),
-          in + k * size + final_offset,
-          vals[k],
-          size,
-          cast_to<T>(init));
+      for (int i = 0; i < N; i++) {
+        vals[k][i] = ((final_offset + block.thread_rank() * N + i) < size)
+            ? in[k * size + i]
+            : cast_to<T>(init);
+      }
+    }
+    for (int k = 0; k < M; k++) {
       for (int j = 0; j < N; j++) {
         accs[k] = op(accs[k], cast_to<U>(vals[k][j]));
       }
@@ -145,13 +132,11 @@ __global__ void row_reduce_simple(T* in, U* out, size_t n_rows, int size) {
   }
 
   __shared__ U shared_accumulators[32 * M];
-  block_reduce(block, warp, accs, shared_accumulators, op, init);
+  block_reduce(block, warp, accs.val, shared_accumulators, op, init);
 
   if (block.thread_rank() == 0) {
     if (grid.block_rank() * M + M <= n_rows) {
-      for (int i = 0; i < M; i++) {
-        out[i] = accs[i];
-      }
+      store_vector(out, 0, accs);
     } else {
       short offset = grid.block_rank() * M + M - n_rows;
       for (int i = offset; i < M; i++) {
@@ -161,17 +146,10 @@ __global__ void row_reduce_simple(T* in, U* out, size_t n_rows, int size) {
   }
 }
 
-template <
-    typename T,
-    typename U,
-    typename Op,
-    int NDIM,
-    int BLOCK_DIM,
-    int N_READS = 4>
+template <typename T, typename U, typename Op, int NDIM, int N_READS = 4>
 __global__ void row_reduce_looped(
-    T* in,
+    const T* in,
     U* out,
-    size_t out_size,
     const __grid_constant__ RowReduceArgs args) {
   auto grid = cg::this_grid();
   auto block = cg::this_thread_block();
@@ -185,36 +163,60 @@ __global__ void row_reduce_looped(
   U init = ReduceInit<Op, T>::value();
   total[0] = init;
   LoopedElemToLoc<NDIM, (NDIM > 2)> loop(args.reduce_ndim);
-  size_t full_blocks = args.row_size / (BLOCK_DIM * N_READS);
-  size_t final_offset = full_blocks * BLOCK_DIM * N_READS;
+  const size_t full_blocks = args.row_size / (block.size() * N_READS);
+  const size_t final_offset = full_blocks * (block.size() * N_READS);
 
   in += elem_to_loc(out_idx, args.shape.data(), args.strides.data(), args.ndim);
+  in += block.thread_rank() * N_READS;
 
-  for (size_t n = 0; n < args.non_row_reductions; n++) {
-    for (size_t r = 0; r < full_blocks; r++) {
-      T vals[N_READS];
-      cub::LoadDirectBlockedVectorized<T, N_READS>(
-          block.thread_rank(),
-          in + loop.location() + r * BLOCK_DIM * N_READS,
-          vals);
-      for (int i = 0; i < N_READS; i++) {
-        total[0] = op(total[0], cast_to<U>(vals[i]));
-      }
+  // Unaligned reduce
+  if (final_offset < args.row_size) {
+    bool mask[N_READS];
+    for (int i = 0; i < N_READS; i++) {
+      mask[i] =
+          (final_offset + block.thread_rank() * N_READS + i) < args.row_size;
     }
-    if (final_offset < args.row_size) {
-      T vals[N_READS];
-      cub::LoadDirectBlocked(
-          block.thread_rank(),
-          in + loop.location() + final_offset,
-          vals,
-          args.row_size - final_offset,
-          cast_to<T>(init));
-      for (int i = 0; i < N_READS; i++) {
-        total[0] = op(total[0], cast_to<U>(vals[i]));
+
+    for (size_t n = 0; n < args.non_row_reductions; n++) {
+      const T* inlocal = in + loop.location();
+
+      for (size_t r = 0; r < full_blocks; r++) {
+        auto vals = load_vector<N_READS>(inlocal, 0);
+        for (int i = 0; i < N_READS; i++) {
+          total[0] = op(total[0], cast_to<U>(vals[i]));
+        }
+        inlocal += block.size() * N_READS;
       }
+
+      {
+        T vals[N_READS];
+        for (int i = 0; i < N_READS; i++) {
+          vals[i] = mask[i] ? inlocal[i] : cast_to<T>(init);
+        }
+        for (int i = 0; i < N_READS; i++) {
+          total[0] = op(total[0], cast_to<U>(vals[i]));
+        }
+      }
+
+      loop.next(args.reduce_shape.data(), args.reduce_strides.data());
+    }
+  }
+
+  // Aligned case
+  else {
+    for (size_t n = 0; n < args.non_row_reductions; n++) {
+      const T* inlocal = in + loop.location();
+
+      for (size_t r = 0; r < full_blocks; r++) {
+        auto vals = load_vector<N_READS>(inlocal, 0);
+        for (int i = 0; i < N_READS; i++) {
+          total[0] = op(total[0], cast_to<U>(vals[i]));
+        }
+        inlocal += block.size() * N_READS;
+      }
+
+      loop.next(args.reduce_shape.data(), args.reduce_strides.data());
     }
-    // TODO: Maybe block.sync() here?
-    loop.next(args.reduce_shape.data(), args.reduce_strides.data());
   }
 
   __shared__ U shared_accumulators[32];
@@ -234,8 +236,6 @@ void row_reduce_simple(
     Reduce::ReduceType reduce_type,
     const std::vector<int>& axes,
     const ReductionPlan& plan) {
-  constexpr int N_READS = 8;
-
   // Allocate data for the output using in's layout to avoid elem_to_loc in the
   // kernel.
   allocate_same_layout(out, in, axes);
@@ -250,14 +250,15 @@ void row_reduce_simple(
       using T = cuda_type_t<MLX_GET_TYPE(type_tag)>;
       using U = typename cu::ReduceResult<OP, T>::type;
 
-      // Cub doesn't like const pointers for vectorized loads. (sigh)
-      T* indata = const_cast<T*>(in.data<T>());
+      constexpr int N_READS = 16 / sizeof(T);
 
       // Calculate the grid and block dims
       size_t reductions = (plan.shape.back() + N_READS - 1) / N_READS;
       dim3 grid = get_2d_grid_dims(out.shape(), out.strides());
-      int threads = std::min(1024UL, reductions);
-      threads = ((threads + WARP_SIZE - 1) / WARP_SIZE) * WARP_SIZE;
+      int warps = (reductions + WARP_SIZE - 1) / WARP_SIZE;
+      warps /= 4;
+      warps = std::max(std::min(warps, 32), 1);
+      int threads = warps * WARP_SIZE;
       dim3 block(threads, 1, 1);
 
       // Pick the kernel
@@ -267,6 +268,7 @@ void row_reduce_simple(
         kernel = cu::row_reduce_simple<T, U, OP, N_READS, 2>;
       }
 
+      T* indata = const_cast<T*>(in.data<T>());
       int size = plan.shape.back();
       encoder.add_kernel_node(
           kernel, grid, block, 0, indata, out.data<U>(), out.size(), size);
@@ -282,8 +284,6 @@ void row_reduce_looped(
     const std::vector<int>& axes,
     const ReductionPlan& plan,
     cu::RowReduceArgs args) {
-  constexpr int N_READS = 8;
-
   // Allocate data for the output using in's layout to access them as
   // contiguously as possible.
   allocate_same_layout(out, in, axes);
@@ -295,34 +295,27 @@ void row_reduce_looped(
       using OP = MLX_GET_TYPE(reduce_type_tag);
       using T = cuda_type_t<MLX_GET_TYPE(type_tag)>;
       using U = typename cu::ReduceResult<OP, T>::type;
-      // Cub doesn't like const pointers for vectorized loads. (sigh)
-      T* indata = const_cast<T*>(in.data<T>());
+
+      constexpr int N_READS = 16 / sizeof(T);
 
       // Calculate the grid and block dims
       args.sort_access_pattern(in, axes);
       dim3 grid = get_2d_grid_dims(out.shape(), out.strides());
       size_t reductions = (args.row_size + N_READS - 1) / N_READS;
-      int threads = std::min(1024UL, reductions);
-      threads = ((threads + WARP_SIZE - 1) / WARP_SIZE) * WARP_SIZE;
+      int warps = (reductions + WARP_SIZE - 1) / WARP_SIZE;
+      warps /= 4;
+      warps = std::max(std::min(warps, 32), 1);
+      int threads = warps * WARP_SIZE;
       dim3 block(threads, 1, 1);
 
       // Pick the kernel
-      auto kernel = cu::row_reduce_looped<T, U, OP, 1, 32, N_READS>;
+      auto kernel = cu::row_reduce_looped<T, U, OP, 1, N_READS>;
       dispatch_reduce_ndim(args.reduce_ndim, [&](auto reduce_ndim) {
-        dispatch_block_dim(threads, [&](auto threads_constant) {
-          kernel = cu::row_reduce_looped<
-              T,
-              U,
-              OP,
-              reduce_ndim.value,
-              threads_constant.value,
-              N_READS>;
-          block.x = threads_constant.value;
-        });
+        kernel = cu::row_reduce_looped<T, U, OP, reduce_ndim.value, N_READS>;
       });
 
       encoder.add_kernel_node(
-          kernel, grid, block, 0, indata, out.data<U>(), out.size(), args);
+          kernel, grid, block, 0, in.data<T>(), out.data<U>(), args);
     });
   });
 }

mlx/backend/cuda/rope.cu

@@ -103,15 +103,21 @@ template <typename T, bool traditional, bool forward, int N = 4>
 __device__ void rope_impl(
     const T* in,
     T* out,
-    int offset,
+    const int* offset,
     float inv_freq,
     float scale,
     const cuda::std::array<int64_t, 3> strides,
     const cuda::std::array<int64_t, 3> out_strides,
-    int64_t n_batch,
+    int64_t offset_stride,
+    int n_head,
     uint3 pos,
     uint3 dims) {
-  float L = scale * static_cast<float>(pos.y + offset);
+  auto n_head_up = N * ((n_head + N - 1) / N);
+  auto head_idx = static_cast<int>((pos.z * N) % n_head_up);
+  auto batch_idx = (pos.z * N) / n_head_up;
+  auto batch_offset = offset[batch_idx * offset_stride];
+  float L = scale * static_cast<float>(pos.y + batch_offset);
+  auto mat_idx = batch_idx * n_head + head_idx;
 
   // Compute costheta, sintheta
   float theta = L * inv_freq;
@@ -123,20 +129,19 @@ __device__ void rope_impl(
   size_t out_index_1, out_index_2;
   if (traditional) {
     out_index_1 = 2 * pos.x * out_strides[2] + pos.y * out_strides[1] +
-        N * pos.z * out_strides[0];
+        mat_idx * out_strides[0];
     out_index_2 = out_index_1 + 1;
     in_index_1 =
-        2 * pos.x * strides[2] + pos.y * strides[1] + N * pos.z * strides[0];
+        2 * pos.x * strides[2] + pos.y * strides[1] + mat_idx * strides[0];
     in_index_2 = in_index_1 + strides[2];
   } else {
     out_index_1 = pos.x * out_strides[2] + pos.y * out_strides[1] +
-        N * pos.z * out_strides[0];
+        mat_idx * out_strides[0];
     out_index_2 = out_index_1 + dims.x * out_strides[2];
-    in_index_1 =
-        pos.x * strides[2] + pos.y * strides[1] + N * pos.z * strides[0];
+    in_index_1 = pos.x * strides[2] + pos.y * strides[1] + mat_idx * strides[0];
     in_index_2 = in_index_1 + dims.x * strides[2];
   }
-  for (int i = 0; i < N && pos.z * N + i < n_batch; ++i) {
+  for (int i = 0; i < N && head_idx + i < n_head; ++i) {
     // Read and write the output
     float x1 = static_cast<float>(in[in_index_1]);
     float x2 = static_cast<float>(in[in_index_2]);
@@ -167,7 +172,8 @@ __global__ void rope(
     float base,
     const __grid_constant__ cuda::std::array<int64_t, 3> strides,
     const __grid_constant__ cuda::std::array<int64_t, 3> out_strides,
-    int64_t n_batch,
+    int64_t offset_stride,
+    int n_head,
     uint3 dims) {
   uint3 pos = make_uint3(
       blockIdx.x * blockDim.x + threadIdx.x,
@@ -182,12 +188,13 @@ __global__ void rope(
   rope_impl<T, traditional, forward>(
       in,
       out,
-      *offset,
+      offset,
       inv_freq,
       scale,
       strides,
       out_strides,
-      n_batch,
+      offset_stride,
+      n_head,
       pos,
       dims);
 }
@@ -202,7 +209,8 @@ __global__ void rope_freqs(
     float base,
     const __grid_constant__ cuda::std::array<int64_t, 3> strides,
     const __grid_constant__ cuda::std::array<int64_t, 3> out_strides,
-    int64_t n_batch,
+    int64_t offset_stride,
+    int n_head,
     uint3 dims,
     int64_t freq_stride) {
   uint3 pos = make_uint3(
@@ -217,12 +225,13 @@ __global__ void rope_freqs(
   rope_impl<T, traditional, forward>(
       in,
       out,
-      *offset,
+      offset,
       inv_freq,
       scale,
       strides,
       out_strides,
-      n_batch,
+      offset_stride,
+      n_head,
       pos,
       dims);
 }
@@ -245,23 +254,28 @@ void RoPE::eval_gpu(
   auto& offset = inputs[1];
   auto& out = outputs[0];
 
-  if (in.ndim() < 3) {
-    throw std::runtime_error("[RoPE] Input must have at least 3 dimensions");
-  }
-
   cuda::std::array<int64_t, 3> strides;
   cuda::std::array<int64_t, 3> out_strides;
   bool donated = false;
   int ndim = in.ndim();
-  int dispatch_ndim = in.ndim();
+
+  int B = in.shape(0);
+  int T = in.shape(-2);
+  int D = in.shape(-1);
+  size_t mat_size = T * D;
+  int dispatch_ndim = ndim;
   while (in.shape(-dispatch_ndim) == 1 && dispatch_ndim > 3) {
     dispatch_ndim--;
   }
-  size_t mat_size = in.shape(-2) * in.shape(-1);
+
+  int N = 1;
+  for (int i = 1; i < (ndim - 2); ++i) {
+    N *= in.shape(i);
+  }
 
   // We apply rope to less that the whole vector so copy to output and then
   // apply in-place.
-  if (dims_ < in.shape(-1)) {
+  if (dims_ < D) {
     donated = true;
     auto ctype =
         (in.flags().row_contiguous) ? CopyType::Vector : CopyType::General;
@@ -302,7 +316,7 @@ void RoPE::eval_gpu(
   out_strides[2] = out.strides()[ndim - 1];
 
   // Some flags to help us dispatch below
-  bool single = in.flags().row_contiguous && (mat_size == in.shape(-1));
+  bool single = in.flags().row_contiguous && B == 1 && T == 1;
   bool with_freqs = inputs.size() == 3;
 
   auto& encoder = cu::get_command_encoder(s);
@@ -319,7 +333,7 @@ void RoPE::eval_gpu(
         if (single && !with_freqs) {
           auto kernel =
               cu::rope_single<DataType, traditional.value, forward.value>;
-          uint2 dims = make_uint2(dims_ / 2, in.size() / mat_size);
+          uint2 dims = make_uint2(dims_ / 2, N);
           auto [grid, block] = get_grid_and_block(dims.x, dims.y, 1);
           encoder.add_kernel_node(
               kernel,
@@ -336,7 +350,7 @@ void RoPE::eval_gpu(
         } else if (single) {
           auto kernel =
               cu::rope_single_freqs<DataType, traditional.value, forward.value>;
-          uint2 dims = make_uint2(dims_ / 2, in.size() / mat_size);
+          uint2 dims = make_uint2(dims_ / 2, N);
           auto [grid, block] = get_grid_and_block(dims.x, dims.y, 1);
           encoder.add_kernel_node(
               kernel,
@@ -354,10 +368,14 @@ void RoPE::eval_gpu(
         } else if (with_freqs) {
           auto kernel =
               cu::rope_freqs<DataType, traditional.value, forward.value>;
-          uint3 dims =
-              make_uint3(dims_ / 2, in.shape(-2), in.size() / mat_size);
-          dims.z = (dims.z + 3) / 4;
+          int n_per_thread = 4;
+          uint32_t dimz = B * ((N + n_per_thread - 1) / n_per_thread);
+          uint3 dims = make_uint3(dims_ / 2, T, dimz);
           auto [grid, block] = get_grid_and_block(dims.x, dims.y, dims.z);
+          int64_t offset_stride = 0;
+          if (inputs[1].ndim() > 0) {
+            offset_stride = inputs[1].strides()[0];
+          }
           encoder.add_kernel_node(
               kernel,
               grid,
@@ -371,15 +389,20 @@ void RoPE::eval_gpu(
               std::log2(base_),
               strides,
               out_strides,
-              in.size() / mat_size,
+              offset_stride,
+              N,
               dims,
               inputs[2].strides(0));
         } else {
           auto kernel = cu::rope<DataType, traditional.value, forward.value>;
-          uint3 dims =
-              make_uint3(dims_ / 2, in.shape(-2), in.size() / mat_size);
-          dims.z = (dims.z + 3) / 4;
+          int n_per_thread = 4;
+          uint32_t dimz = B * ((N + n_per_thread - 1) / n_per_thread);
+          uint3 dims = make_uint3(dims_ / 2, T, dimz);
           auto [grid, block] = get_grid_and_block(dims.x, dims.y, dims.z);
+          int64_t offset_stride = 0;
+          if (inputs[1].ndim() > 0) {
+            offset_stride = inputs[1].strides()[0];
+          }
           encoder.add_kernel_node(
               kernel,
               grid,
@@ -392,7 +415,8 @@ void RoPE::eval_gpu(
               std::log2(base_),
               strides,
               out_strides,
-              in.size() / mat_size,
+              offset_stride,
+              N,
               dims);
         }
       });

mlx/backend/cuda/scaled_dot_product_attention.cu

@@ -4,7 +4,6 @@
 #include "mlx/backend/cuda/device/config.h"
 #include "mlx/backend/cuda/device/utils.cuh"
 #include "mlx/backend/cuda/kernel_utils.cuh"
-#include "mlx/backend/cuda/lru_cache.h"
 #include "mlx/backend/gpu/copy.h"
 #include "mlx/dtype_utils.h"
 #include "mlx/fast_primitives.h"
@@ -46,6 +45,7 @@ __global__ void kernel_sdpav_1pass(
     const T* K,
     const T* V,
     T* O,
+    const T* sinks,
     __grid_constant__ const AttnParams params) {
   constexpr int BN = 32;
   constexpr int BD = 32;
@@ -65,7 +65,7 @@ __global__ void kernel_sdpav_1pass(
   __shared__ U max_scores[BN];
   __shared__ U sum_exp_scores[BN];
 
-  const U scale_log2 = params.scale * 1.44269504089f;
+  const U scale_log2 = params.scale * M_LOG2E;
 
   auto block = cg::this_thread_block();
   auto warp = cg::tiled_partition<32>(block);
@@ -108,8 +108,12 @@ __global__ void kernel_sdpav_1pass(
     o[i] = 0.f;
   }
 
-  U max_score = -INFINITY;
+  U max_score = Limits<U>::finite_min();
   U sum_exp_score = 0.f;
+  if (sinks && warp_idx == 0) {
+    max_score = M_LOG2E * static_cast<U>(sinks[head_idx]);
+    sum_exp_score = 1.f;
+  }
 
   // For each key
   for (int i = kv_seq_idx; i < params.kL; i += BN) {
@@ -167,7 +171,7 @@ __global__ void kernel_sdpav_1pass(
   U factor = exp2f(max_score - new_max);
   sum_exp_score =
       cg::reduce(warp, sum_exp_scores[lane_idx] * factor, cg::plus<U>());
-  sum_exp_score = __frcp_rn(sum_exp_score);
+  sum_exp_score = sum_exp_score == 0 ? 0 : __frcp_rn(sum_exp_score);
 
   // Now we need to aggregate all the outputs
   PRAGMA_LOOP_UNROLL
@@ -193,6 +197,7 @@ __global__ void kernel_sdpav_2pass_1(
     const T* Q,
     const T* K,
     const T* V,
+    const T* sinks,
     float* partials,
     float* sums,
     float* maxs,
@@ -268,8 +273,12 @@ __global__ void kernel_sdpav_2pass_1(
     o[i] = 0.f;
   }
 
-  U max_score = -1e9;
+  U max_score = Limits<U>::finite_min();
   U sum_exp_score = 0.f;
+  if (sinks && warp_idx == 0 && block_idx == 0) {
+    max_score = M_LOG2E * static_cast<U>(sinks[head_idx]);
+    sum_exp_score = 1.f;
+  }
 
   // For each key
   for (int i = kv_seq_idx; i < params.kL; i += blocks * BN) {
@@ -410,7 +419,7 @@ __global__ void kernel_sdpav_2pass_2(
   U new_max = cg::reduce(warp, max_score, cg::greater<U>());
   U factor = exp2f(max_score - new_max);
   U sum_exp_score = cg::reduce(warp, sums[lane_idx] * factor, cg::plus<U>());
-  sum_exp_score = __frcp_rn(sum_exp_score);
+  sum_exp_score = sum_exp_score == 0 ? 0 : __frcp_rn(sum_exp_score);
 
   PRAGMA_LOOP_UNROLL
   for (int i = 0; i < v_per_thread; i++) {
@@ -463,10 +472,14 @@ void sdpa_vector_1pass_fallback(
     const array& v,
     const float scale,
     array& o,
-    bool do_causal_ = false) {
+    bool do_causal,
+    const std::optional<array>& sinks) {
   encoder.set_input_array(q);
   encoder.set_input_array(k);
   encoder.set_input_array(v);
+  if (sinks) {
+    encoder.set_input_array(*sinks);
+  }
   encoder.set_output_array(o);
 
   cu::AttnParams params{
@@ -489,7 +502,7 @@ void sdpa_vector_1pass_fallback(
   dim3 block_dim(1024, 1, 1);
 
   dispatch_float_types(o.dtype(), "kernel_sdpav_1pass", [&](auto type_tag) {
-    dispatch_bool(do_causal_, [&](auto do_causal) {
+    dispatch_bool(do_causal, [&](auto do_causal) {
       dispatch_headdim(params.D, [&](auto headdim) {
         using DataType = cuda_type_t<MLX_GET_TYPE(type_tag)>;
 
@@ -504,6 +517,7 @@ void sdpa_vector_1pass_fallback(
             k.data<DataType>(),
             v.data<DataType>(),
             o.data<DataType>(),
+            sinks ? (*sinks).data<DataType>() : nullptr,
             params);
       });
     });
@@ -518,7 +532,8 @@ void sdpa_vector_2pass_fallback(
     const array& v,
     const float scale,
     array& o,
-    bool do_causal_ = false) {
+    bool do_causal,
+    const std::optional<array>& sinks) {
   cu::AttnParams params{
       /* int B = */ q.shape(0),
       /* int H = */ q.shape(1),
@@ -559,7 +574,7 @@ void sdpa_vector_2pass_fallback(
   encoder.add_temporary(maxs);
 
   dispatch_float_types(o.dtype(), "kernel_sdpav_2pass", [&](auto type_tag) {
-    dispatch_bool(do_causal_, [&](auto do_causal) {
+    dispatch_bool(do_causal, [&](auto do_causal) {
       dispatch_headdim(params.D, [&](auto headdim) {
         using DataType = cuda_type_t<MLX_GET_TYPE(type_tag)>;
 
@@ -570,6 +585,10 @@ void sdpa_vector_2pass_fallback(
           encoder.set_input_array(q);
           encoder.set_input_array(k);
           encoder.set_input_array(v);
+          if (sinks) {
+            encoder.set_input_array(*sinks);
+          }
+
           encoder.set_output_array(intermediate);
           encoder.set_output_array(sums);
           encoder.set_output_array(maxs);
@@ -585,6 +604,7 @@ void sdpa_vector_2pass_fallback(
               q.data<DataType>(),
               k.data<DataType>(),
               v.data<DataType>(),
+              sinks ? (*sinks).data<DataType>() : nullptr,
               intermediate.data<float>(),
               sums.data<float>(),
               maxs.data<float>(),
@@ -627,15 +647,16 @@ void sdpa_vector_fallback(
     const array& v,
     const float scale,
     array& o,
-    bool do_causal_ = false) {
+    bool do_causal,
+    const std::optional<array>& sinks) {
   int kL = k.shape(2);
 
   if (kL > 1024) {
     return sdpa_vector_2pass_fallback(
-        s, encoder, q, k, v, scale, o, do_causal_);
+        s, encoder, q, k, v, scale, o, do_causal, sinks);
   } else {
     return sdpa_vector_1pass_fallback(
-        s, encoder, q, k, v, scale, o, do_causal_);
+        s, encoder, q, k, v, scale, o, do_causal, sinks);
   }
 }
 
@@ -691,7 +712,7 @@ void ScaledDotProductAttention::eval_gpu(
 
   // Define some copy functions to ensure the layout of the inputs is as
   // expected.
-  copies.reserve(3);
+  copies.reserve(inputs.size());
   auto copy_unless = [&copies, &s](
                          auto predicate, const array& arr) -> const array& {
     if (!predicate(arr)) {
@@ -703,6 +724,16 @@ void ScaledDotProductAttention::eval_gpu(
     }
   };
 
+  // Checks that the headdim dimension has stride 1.
+  auto is_matrix_contiguous = [](const array& arr) {
+    return arr.strides(-1) == 1;
+  };
+
+  std::optional<array> sinks = std::nullopt;
+  if (has_sinks_) {
+    sinks = copy_unless(is_matrix_contiguous, inputs.back());
+  }
+
   // We are in vector mode ie single query
   if (q_pre.shape(2) < 4) {
     auto q_copy_unless = [](const array& arr) {
@@ -740,10 +771,6 @@ void ScaledDotProductAttention::eval_gpu(
     const auto& k = copy_unless(kv_copy_unless, k_pre);
     const auto& v = copy_unless(kv_copy_unless, v_pre);
 
-    for (const auto& cp : copies) {
-      encoder.add_temporary(cp);
-    }
-
     // Donate the query if possible
     if (q.is_donatable() && q.flags().row_contiguous && q.size() == o.size()) {
       o.copy_shared_buffer(q);
@@ -752,22 +779,26 @@ void ScaledDotProductAttention::eval_gpu(
       int64_t str_oH = o.shape(3);
       int64_t str_oL = o.shape(1) * str_oH;
       int64_t str_oB = o.shape(2) * str_oL;
-      size_t data_size = o.shape(0) * str_oB;
 
       array::Flags flags{
           /* bool contiguous = */ 1,
           /* bool row_contiguous = */ o.shape(2) == 1,
-          /* bool col_contiguous = */ 0,
+          /* bool col_contiguous = */ o.size() == o.shape(3),
       };
 
       o.set_data(
           allocator::malloc(o.nbytes()),
-          data_size,
+          o.size(),
           {str_oB, str_oH, str_oL, str_oD},
           flags);
     }
 
-    return sdpa_vector_fallback(s, encoder, q, k, v, scale_, o, do_causal_);
+    for (const auto& cp : copies) {
+      encoder.add_temporary(cp);
+    }
+
+    return sdpa_vector_fallback(
+        s, encoder, q, k, v, scale_, o, do_causal_, sinks);
   }
 
   // Full attention mode should never reach here

mlx/backend/cuda/ternary.cu

@@ -156,7 +156,25 @@ void ternary_op_gpu_inplace(
     using DType = cuda_type_t<MLX_GET_TYPE(type_tag)>;
 
     auto topt = get_ternary_op_type(a, b, c);
-    if (topt == TernaryOpType::General) {
+    if (topt == TernaryOpType::VectorVectorVector ||
+        topt == TernaryOpType::ScalarScalarScalar) {
+      dispatch_bool(out.data_size() > UINT32_MAX, [&](auto large) {
+        using IdxT = std::conditional_t<large(), int64_t, uint32_t>;
+        constexpr int N_READS = 16 / sizeof(DType);
+        auto [num_blocks, block_dims] = get_launch_args(
+            out.data_size(), out.shape(), out.strides(), large(), N_READS);
+        encoder.add_kernel_node(
+            cu::ternary_v<Op, DType, IdxT, N_READS>,
+            num_blocks,
+            block_dims,
+            0,
+            a.data<bool>(),
+            b.data<DType>(),
+            c.data<DType>(),
+            out.data<DType>(),
+            out.data_size());
+      });
+    } else {
       dispatch_bool(
           a.data_size() > INT32_MAX || b.data_size() > INT32_MAX ||
               c.data_size() > INT32_MAX || out.data_size() > INT32_MAX,
@@ -225,23 +243,6 @@ void ternary_op_gpu_inplace(
                   ndim);
             }
           });
-    } else {
-      dispatch_bool(out.data_size() > UINT32_MAX, [&](auto large) {
-        using IdxT = std::conditional_t<large(), int64_t, uint32_t>;
-        constexpr int N_READS = 16 / sizeof(DType);
-        auto [num_blocks, block_dims] = get_launch_args(
-            out.data_size(), out.shape(), out.strides(), large(), N_READS);
-        encoder.add_kernel_node(
-            cu::ternary_v<Op, DType, IdxT, N_READS>,
-            num_blocks,
-            block_dims,
-            0,
-            a.data<bool>(),
-            b.data<DType>(),
-            c.data<DType>(),
-            out.data<DType>(),
-            out.data_size());
-      });
     }
   });
 }

mlx/backend/cuda/unary.cu

@@ -1,284 +0,0 @@
-// Copyright © 2025 Apple Inc.
-
-#include "mlx/backend/common/unary.h"
-#include "mlx/backend/cuda/device.h"
-#include "mlx/backend/cuda/device/unary_ops.cuh"
-#include "mlx/backend/cuda/kernel_utils.cuh"
-#include "mlx/dtype_utils.h"
-#include "mlx/primitives.h"
-
-#include <cooperative_groups.h>
-#include <nvtx3/nvtx3.hpp>
-
-namespace mlx::core {
-
-namespace cu {
-
-namespace cg = cooperative_groups;
-
-template <typename Op, typename In, typename Out, typename IdxT, int N_READS>
-__global__ void unary_v(const In* in, Out* out, IdxT size) {
-  IdxT index = cg::this_grid().thread_rank();
-
-  if ((index + 1) * N_READS > size) {
-    for (IdxT i = index * N_READS; i < size; ++i) {
-      out[i] = Op{}(in[i]);
-    }
-  } else {
-    auto in_vec = load_vector<N_READS>(in, index);
-
-    AlignedVector<Out, N_READS> out_vec;
-#pragma unroll
-    for (int i = 0; i < N_READS; ++i) {
-      out_vec[i] = Op{}(in_vec[i]);
-    }
-
-    store_vector<N_READS>(out, index, out_vec);
-  }
-}
-
-template <typename Op, typename In, typename Out, typename IdxT, int N_READS>
-__global__ void unary_g(
-    const In* in,
-    Out* out,
-    IdxT size_rest,
-    const __grid_constant__ Shape shape,
-    const __grid_constant__ Strides strides,
-    int ndim) {
-  auto block = cg::this_thread_block();
-  auto grid = cg::this_grid();
-  IdxT index_rest =
-      grid.block_index().y * block.dim_threads().y + block.thread_index().y;
-  if (index_rest >= size_rest) {
-    return;
-  }
-
-  auto shape_x = shape[ndim - 1];
-  auto stride_x = strides[ndim - 1];
-  IdxT index_x =
-      grid.block_index().x * block.dim_threads().x + block.thread_index().x;
-  auto idx =
-      elem_to_loc(index_rest * shape_x, shape.data(), strides.data(), ndim);
-  auto in_vec =
-      load_vector<N_READS>(in + idx, index_x, shape_x, stride_x, In(0));
-  AlignedVector<Out, N_READS> out_vec;
-#pragma unroll
-  for (int i = 0; i < N_READS; ++i) {
-    out_vec[i] = Op{}(in_vec[i]);
-  }
-  store_vector(out + shape_x * index_rest, index_x, out_vec, shape_x);
-}
-
-template <typename Op, typename In, typename Out>
-constexpr bool supports_unary_op() {
-  if (std::is_same_v<Op, Abs> || std::is_same_v<Op, Negative> ||
-      std::is_same_v<Op, Sign> || std::is_same_v<Op, Square>) {
-    return std::is_same_v<In, Out>;
-  }
-  if (std::is_same_v<Op, ArcCosh> || std::is_same_v<Op, ArcSinh> ||
-      std::is_same_v<Op, ArcTanh> || std::is_same_v<Op, Erf> ||
-      std::is_same_v<Op, ErfInv> || std::is_same_v<Op, Expm1> ||
-      std::is_same_v<Op, Sigmoid>) {
-    return std::is_same_v<In, Out> && is_floating_v<In>;
-  }
-  if (std::is_same_v<Op, BitwiseInvert>) {
-    return std::is_same_v<In, Out> && std::is_integral_v<In> &&
-        !std::is_same_v<In, bool>;
-  }
-  if (std::is_same_v<Op, Ceil> || std::is_same_v<Op, Floor>) {
-    return std::is_same_v<In, Out> && !mlx::core::is_complex_v<In>;
-  }
-  if (std::is_same_v<Op, Conjugate>) {
-    return std::is_same_v<In, Out> && mlx::core::is_complex_v<In>;
-  }
-  if (std::is_same_v<Op, ArcCos> || std::is_same_v<Op, ArcSin> ||
-      std::is_same_v<Op, ArcTan> || std::is_same_v<Op, Cos> ||
-      std::is_same_v<Op, Cosh> || std::is_same_v<Op, Exp> ||
-      std::is_same_v<Op, Log> || std::is_same_v<Op, Log2> ||
-      std::is_same_v<Op, Log10> || std::is_same_v<Op, Log1p> ||
-      std::is_same_v<Op, Round> || std::is_same_v<Op, Rsqrt> ||
-      std::is_same_v<Op, Sqrt> || std::is_same_v<Op, Sin> ||
-      std::is_same_v<Op, Sinh> || std::is_same_v<Op, Tan> ||
-      std::is_same_v<Op, Tanh>) {
-    return std::is_same_v<In, Out> && is_inexact_v<In>;
-  }
-  if (std::is_same_v<Op, Imag> || std::is_same_v<Op, Real>) {
-    return mlx::core::is_complex_v<In> && std::is_same_v<Out, float>;
-  }
-  if (std::is_same_v<Op, LogicalNot>) {
-    return std::is_same_v<In, Out> && std::is_same_v<In, bool>;
-  }
-  return false;
-}
-
-} // namespace cu
-
-template <typename Op>
-void unary_op_gpu_inplace(
-    const std::vector<array>& inputs,
-    array& out,
-    const char* op,
-    const Stream& s) {
-  auto& in = inputs[0];
-  if (in.size() == 0) {
-    return;
-  }
-  bool contig = in.flags().contiguous;
-  bool large;
-  if (!contig) {
-    large = in.data_size() > INT32_MAX || out.size() > INT32_MAX;
-  } else {
-    large = in.data_size() > UINT32_MAX;
-  }
-
-  auto& encoder = cu::get_command_encoder(s);
-  encoder.set_input_array(in);
-  encoder.set_output_array(out);
-  dispatch_all_types(in.dtype(), [&](auto in_type_tag) {
-    dispatch_all_types(out.dtype(), [&](auto out_type_tag) {
-      using CTYPE_IN = MLX_GET_TYPE(in_type_tag);
-      using CTYPE_OUT = MLX_GET_TYPE(out_type_tag);
-      if constexpr (cu::supports_unary_op<Op, CTYPE_IN, CTYPE_OUT>()) {
-        dispatch_bool(large, [&](auto large) {
-          using InType = cuda_type_t<CTYPE_IN>;
-          using OutType = cuda_type_t<CTYPE_OUT>;
-          if (contig) {
-            using IdxT = std::conditional_t<large(), int64_t, uint32_t>;
-            constexpr int N_READS = 16 / sizeof(OutType);
-            auto [num_blocks, block_dims] = get_launch_args(
-                out.data_size(), out.shape(), out.strides(), large, N_READS);
-            encoder.add_kernel_node(
-                cu::unary_v<Op, InType, OutType, IdxT, N_READS>,
-                num_blocks,
-                block_dims,
-                0,
-                in.data<InType>(),
-                out.data<OutType>(),
-                out.data_size());
-          } else {
-            using IdxT = std::conditional_t<large(), int64_t, int32_t>;
-            auto [shape, strides] = collapse_contiguous_dims(in);
-            auto ndim = shape.size();
-            int work_per_thread = 1;
-            auto kernel = cu::unary_g<Op, InType, OutType, IdxT, 1>;
-            auto dim0 = ndim > 0 ? shape.back() : 1;
-            auto rest = out.size() / dim0;
-            if (dim0 >= 4) {
-              kernel = cu::unary_g<Op, InType, OutType, IdxT, 4>;
-              work_per_thread = 4;
-            }
-            dim0 = (dim0 + work_per_thread - 1) / work_per_thread;
-            auto block_dims = get_block_dims(dim0, rest, 1);
-            uint32_t num_blocks_x = cuda::ceil_div(dim0, block_dims.x);
-            uint32_t num_blocks_y = cuda::ceil_div(rest, block_dims.y);
-            encoder.add_kernel_node(
-                kernel,
-                {num_blocks_x, num_blocks_y},
-                block_dims,
-                0,
-                in.data<InType>(),
-                out.data<OutType>(),
-                rest,
-                const_param(shape),
-                const_param(strides),
-                ndim);
-          }
-        });
-      } else {
-        throw std::runtime_error(fmt::format(
-            "Can not do unary op {} on input of {} with output of {}.",
-            op,
-            dtype_to_string(in.dtype()),
-            dtype_to_string(out.dtype())));
-      }
-    });
-  });
-}
-
-template <typename Op>
-void unary_op_gpu(
-    const std::vector<array>& inputs,
-    array& out,
-    const char* op,
-    const Stream& s) {
-  set_unary_output_data(inputs[0], out);
-  unary_op_gpu_inplace<Op>(inputs, out, op, s);
-}
-
-#define UNARY_GPU(func)                                               \
-  void func::eval_gpu(const std::vector<array>& inputs, array& out) { \
-    nvtx3::scoped_range r(#func "::eval_gpu");                        \
-    auto& s = out.primitive().stream();                               \
-    unary_op_gpu<cu::func>(inputs, out, name(), s);                   \
-  }
-
-UNARY_GPU(Abs)
-UNARY_GPU(ArcCos)
-UNARY_GPU(ArcCosh)
-UNARY_GPU(ArcSin)
-UNARY_GPU(ArcSinh)
-UNARY_GPU(ArcTan)
-UNARY_GPU(ArcTanh)
-UNARY_GPU(BitwiseInvert)
-UNARY_GPU(Ceil)
-UNARY_GPU(Conjugate)
-UNARY_GPU(Cos)
-UNARY_GPU(Cosh)
-UNARY_GPU(Erf)
-UNARY_GPU(ErfInv)
-UNARY_GPU(Exp)
-UNARY_GPU(Expm1)
-UNARY_GPU(Floor)
-UNARY_GPU(Imag)
-UNARY_GPU(Log1p)
-UNARY_GPU(LogicalNot)
-UNARY_GPU(Negative)
-UNARY_GPU(Real)
-UNARY_GPU(Sigmoid)
-UNARY_GPU(Sign)
-UNARY_GPU(Sin)
-UNARY_GPU(Sinh)
-UNARY_GPU(Square)
-UNARY_GPU(Tan)
-UNARY_GPU(Tanh)
-
-void Log::eval_gpu(const std::vector<array>& inputs, array& out) {
-  nvtx3::scoped_range r("Log::eval_gpu");
-  auto& s = out.primitive().stream();
-  switch (base_) {
-    case Base::e:
-      unary_op_gpu<cu::Log>(inputs, out, name(), s);
-      break;
-    case Base::two:
-      unary_op_gpu<cu::Log2>(inputs, out, name(), s);
-      break;
-    case Base::ten:
-      unary_op_gpu<cu::Log10>(inputs, out, name(), s);
-      break;
-  }
-}
-
-void Round::eval_gpu(const std::vector<array>& inputs, array& out) {
-  nvtx3::scoped_range r("Round::eval_gpu");
-  assert(inputs.size() == 1);
-  const auto& in = inputs[0];
-  auto& s = out.primitive().stream();
-  if (issubdtype(in.dtype(), inexact)) {
-    unary_op_gpu<cu::Round>(inputs, out, name(), s);
-  } else {
-    // No-op integer types
-    out.copy_shared_buffer(in);
-  }
-}
-
-void Sqrt::eval_gpu(const std::vector<array>& inputs, array& out) {
-  nvtx3::scoped_range r("Sort::eval_gpu");
-  auto& s = out.primitive().stream();
-  if (recip_) {
-    unary_op_gpu<cu::Rsqrt>(inputs, out, "Rsqrt", s);
-  } else {
-    unary_op_gpu<cu::Sqrt>(inputs, out, "Sqrt", s);
-  }
-}
-
-} // namespace mlx::core

mlx/backend/cuda/unary/unary.cuh

@@ -108,6 +108,12 @@ constexpr bool supports_unary_op() {
   if (std::is_same_v<Op, LogicalNot>) {
     return std::is_same_v<In, Out> && std::is_same_v<In, bool>;
   }
+  if (std::is_same_v<Op, ToFP8>) {
+    return std::is_same_v<Out, uint8_t> && is_floating_v<In>;
+  }
+  if (std::is_same_v<Op, FromFP8>) {
+    return std::is_same_v<In, uint8_t> && is_floating_v<Out>;
+  }
   return false;
 }
 

mlx/backend/cuda/utils.h

@@ -1,6 +1,6 @@
 // Copyright © 2025 Apple Inc.
 
-// This file include utilies that are used by C++ code (i.e. .cpp files).
+// This file include utilities that are used by C++ code (i.e. .cpp files).
 
 #pragma once
 
@@ -12,6 +12,7 @@ namespace mlx::core {
 
 namespace cu {
 class Device;
+
 }
 
 struct Dtype;
@@ -86,4 +87,17 @@ class CudaStream : public CudaHandle<cudaStream_t, cudaStreamDestroy> {
   explicit CudaStream(cu::Device& device);
 };
 
+template <typename T>
+inline uint max_occupancy_block_dim(T kernel) {
+  int _, block_dim;
+  if constexpr (std::is_same_v<T, CUfunction>) {
+    CHECK_CUDA_ERROR(
+        cuOccupancyMaxPotentialBlockSize(&_, &block_dim, kernel, 0, 0, 0));
+  } else {
+    CHECK_CUDA_ERROR(
+        cudaOccupancyMaxPotentialBlockSize(&_, &block_dim, kernel));
+  }
+  return block_dim;
+}
+
 } // namespace mlx::core

mlx/backend/cuda/worker.cpp

@@ -5,8 +5,9 @@
 
 namespace mlx::core::cu {
 
-Worker::Worker()
-    : signal_stream_(device(mlx::core::Device::gpu)),
+Worker::Worker(Device& d)
+    : signal_stream_(d),
+      signal_event_(d, cudaEventDisableTiming | cudaEventBlockingSync),
       worker_(&Worker::thread_fn, this) {}
 
 Worker::~Worker() {

mlx/backend/cuda/worker.h

@@ -3,7 +3,6 @@
 #pragma once
 
 #include "mlx/backend/cuda/event.h"
-#include "mlx/backend/cuda/utils.h"
 
 #include <condition_variable>
 #include <functional>
@@ -16,7 +15,7 @@ namespace mlx::core::cu {
 // Run tasks in worker thread, synchronized with cuda stream.
 class Worker {
  public:
-  Worker();
+  explicit Worker(Device& d);
   ~Worker();
 
   Worker(const Worker&) = delete;

mlx/backend/metal/CMakeLists.txt

@@ -29,7 +29,7 @@ make_jit_source(
   kernels/bf16_math.h
   kernels/complex.h
   kernels/defines.h)
-make_jit_source(unary_ops kernels/erf.h kernels/expm1f.h)
+make_jit_source(unary_ops kernels/erf.h kernels/expm1f.h kernels/fp8.h)
 make_jit_source(binary_ops)
 make_jit_source(ternary_ops)
 make_jit_source(reduce_utils kernels/atomic.h kernels/reduction/ops.h)
@@ -81,7 +81,8 @@ if(MLX_METAL_JIT)
 
   make_jit_source(quantized_utils)
   make_jit_source(quantized kernels/quantized_utils.h)
-  make_jit_source(fp4_quantized kernels/quantized_utils.h)
+  make_jit_source(fp_quantized kernels/quantized_utils.h kernels/fp8.h
+                  kernels/fp4.h)
   make_jit_source(gemv_masked)
 else()
   target_sources(mlx PRIVATE ${CMAKE_CURRENT_SOURCE_DIR}/nojit_kernels.cpp)

mlx/backend/metal/allocator.cpp

@@ -32,7 +32,6 @@ namespace metal {
 
 MetalAllocator::MetalAllocator()
     : device_(device(mlx::core::Device::gpu).mtl_device()),
-      residency_set_(device_),
       buffer_cache_(
           vm_page_size,
           [](MTL::Buffer* buf) { return buf->length(); },
@@ -41,7 +40,8 @@ MetalAllocator::MetalAllocator()
               residency_set_.erase(buf);
             }
             buf->release();
-          }) {
+          }),
+      residency_set_(device_) {
   auto pool = metal::new_scoped_memory_pool();
   auto memsize = std::get<size_t>(device_info().at("memory_size"));
   auto max_rec_size =

mlx/backend/metal/allocator.h

@@ -65,7 +65,6 @@ class MetalAllocator : public allocator::Allocator {
   size_t peak_memory_{0};
   size_t max_pool_size_;
   size_t wired_limit_{0};
-  bool relaxed_{true};
   size_t num_resources_{0};
   size_t resource_limit_{0};
 

mlx/backend/metal/conv.cpp

@@ -2,7 +2,6 @@
 #include <algorithm>
 #include <cassert>
 #include <numeric>
-#include <sstream>
 
 #include "mlx/backend/gpu/copy.h"
 #include "mlx/backend/metal/device.h"
@@ -39,10 +38,11 @@ void explicit_gemm_conv_ND_gpu(
   in_unfolded.set_data(allocator::malloc(in_unfolded.nbytes()));
 
   // Prepare unfolding kernel
-  std::ostringstream kname;
-  kname << "naive_unfold_nd_" << type_to_name(in_unfolded) << "_" << N;
+  std::string kname;
+  kname.reserve(32);
+  concatenate(kname, "naive_unfold_nd_", type_to_name(in_unfolded), "_", N);
   auto& compute_encoder = d.get_command_encoder(s.index);
-  auto kernel = d.get_kernel(kname.str());
+  auto kernel = d.get_kernel(kname);
   compute_encoder.set_compute_pipeline_state(kernel);
 
   compute_encoder.set_input_array(in, 0);
@@ -117,11 +117,12 @@ void explicit_gemm_conv_group_ND_gpu(
   in_unfolded.set_data(allocator::malloc(in_unfolded.nbytes()));
 
   // Prepare unfolding kernel
-  std::ostringstream kname;
-  kname << "naive_unfold_transpose_nd_" << type_to_name(in_unfolded) << "_"
-        << N;
+  std::string kname;
+  kname.reserve(32);
+  concatenate(
+      kname, "naive_unfold_transpose_nd_", type_to_name(in_unfolded), "_", N);
   auto& compute_encoder = d.get_command_encoder(s.index);
-  auto kernel = d.get_kernel(kname.str());
+  auto kernel = d.get_kernel(kname);
   compute_encoder.set_compute_pipeline_state(kernel);
 
   compute_encoder.set_input_array(in, 0);
@@ -252,18 +253,32 @@ void implicit_gemm_conv_2D_gpu(
       /* const int swizzle_log = */ swizzle_log};
 
   // Determine kernel
-  std::ostringstream kname;
-  kname << "implicit_gemm_conv_2d_" << type_to_name(out) << "_bm" << bm << "_bn"
-        << bn << "_bk" << bk << "_wm" << wm << "_wn" << wn << "_channel_"
-        << (n_channel_specialization ? std::to_string(n_channel_specialization)
-                                     : "l")
-        << "_filter_" << (small_filter ? 's' : 'l');
+  std::string kname;
+  kname.reserve(64);
+  concatenate(
+      kname,
+      "implicit_gemm_conv_2d_",
+      type_to_name(out),
+      "_bm",
+      bm,
+      "_bn",
+      bn,
+      "_bk",
+      bk,
+      "_wm",
+      wm,
+      "_wn",
+      wn,
+      "_channel_",
+      n_channel_specialization ? std::to_string(n_channel_specialization) : "l",
+      "_filter_",
+      small_filter ? 's' : 'l');
 
   // Encode and dispatch kernel
   auto& compute_encoder = d.get_command_encoder(s.index);
   auto kernel = get_steel_conv_kernel(
       d,
-      kname.str(),
+      kname,
       out,
       bm,
       bn,
@@ -559,11 +574,16 @@ void winograd_conv_2D_gpu(
   {
     int bc = 32;
     int bo = 4;
-    std::ostringstream kname;
-    kname << "winograd_conv_2d_weight_transform_" << type_to_name(out) << "_bc"
-          << bc;
+    std::string kname;
+    kname.reserve(32);
+    concatenate(
+        kname,
+        "winograd_conv_2d_weight_transform_",
+        type_to_name(out),
+        "_bc",
+        bc);
     auto& compute_encoder = d.get_command_encoder(s.index);
-    auto kernel = d.get_kernel(kname.str());
+    auto kernel = d.get_kernel(kname);
     compute_encoder.set_compute_pipeline_state(kernel);
 
     compute_encoder.set_input_array(wt, 0);
@@ -587,11 +607,16 @@ void winograd_conv_2D_gpu(
     int bc = 32;
     int wm = 2;
     int wn = 2;
-    std::ostringstream kname;
-    kname << "winograd_conv_2d_input_transform_" << type_to_name(out) << "_bc"
-          << bc;
+    std::string kname;
+    kname.reserve(32);
+    concatenate(
+        kname,
+        "winograd_conv_2d_input_transform_",
+        type_to_name(out),
+        "_bc",
+        bc);
     auto& compute_encoder = d.get_command_encoder(s.index);
-    auto kernel = d.get_kernel(kname.str());
+    auto kernel = d.get_kernel(kname);
     compute_encoder.set_compute_pipeline_state(kernel);
 
     compute_encoder.set_input_array(in_padded, 0);
@@ -634,11 +659,16 @@ void winograd_conv_2D_gpu(
     int bc = 32;
     int wm = 2;
     int wn = 2;
-    std::ostringstream kname;
-    kname << "winograd_conv_2d_output_transform_" << type_to_name(out) << "_bo"
-          << bc;
+    std::string kname;
+    kname.reserve(32);
+    concatenate(
+        kname,
+        "winograd_conv_2d_output_transform_",
+        type_to_name(out),
+        "_bo",
+        bc);
     auto& compute_encoder = d.get_command_encoder(s.index);
-    auto kernel = d.get_kernel(kname.str());
+    auto kernel = d.get_kernel(kname);
     compute_encoder.set_compute_pipeline_state(kernel);
 
     compute_encoder.set_input_array(out_wg, 0);
@@ -660,9 +690,9 @@ void depthwise_conv_2D_gpu(
     const array& wt,
     array out,
     const MLXConvParams<2>& conv_params) {
-  std::ostringstream kname;
-  kname << "depthwise_conv_2d_" << type_to_name(out);
-  std::string base_name = kname.str();
+  std::string base_name;
+  base_name.reserve(32);
+  concatenate(base_name, "depthwise_conv_2d_", type_to_name(out));
 
   const int N = conv_params.N;
   const int ker_h = conv_params.wS[0];
@@ -685,15 +715,18 @@ void depthwise_conv_2D_gpu(
   };
 
   // clang-format off
-  kname << "_ker_h_" << ker_h
-        << "_ker_w_" << ker_w
-        << "_str_h_" << str_h
-        << "_str_w_" << str_w
-        << "_tgp_h_" << th
-        << "_tgp_w_" << tw
-        << "_do_flip_" << (do_flip ? 't' : 'n'); // clang-format on
-
-  std::string hash_name = kname.str();
+  std::string hash_name;
+  hash_name.reserve(64);
+  concatenate(
+      hash_name,
+      base_name,
+  "_ker_h_", ker_h,
+  "_ker_w_", ker_w,
+  "_str_h_", str_h,
+  "_str_w_", str_w,
+  "_tgp_h_", th,
+  "_tgp_w_", tw,
+  "_do_flip_", do_flip ? 't' : 'n'); // clang-format on
 
   auto& compute_encoder = d.get_command_encoder(s.index);
   auto kernel = d.get_kernel(base_name, hash_name, func_consts);
@@ -774,6 +807,56 @@ void dispatch_conv_2D_gpu(
   }
 }
 
+void depthwise_conv_1D_gpu(
+    const Stream& s,
+    metal::Device& d,
+    const array& in,
+    array wt,
+    array out) {
+  bool large = in.size() > INT32_MAX || in.data_size() > INT32_MAX;
+  std::string base_name;
+  base_name.reserve(32);
+  concatenate(
+      base_name,
+      "depthwise_conv_1d_",
+      large ? "_large" : "",
+      type_to_name(out));
+
+  if (!wt.flags().row_contiguous) {
+    wt = contiguous_copy_gpu(wt, s);
+    d.add_temporary(wt, s.index);
+  }
+  auto& compute_encoder = d.get_command_encoder(s.index);
+  auto kernel = d.get_kernel(base_name);
+  compute_encoder.set_compute_pipeline_state(kernel);
+
+  auto B = in.shape(0);
+  auto Tout = out.shape(1);
+  auto D = in.shape(2);
+  auto K = wt.shape(1);
+
+  compute_encoder.set_input_array(in, 0);
+  compute_encoder.set_input_array(wt, 1);
+  compute_encoder.set_output_array(out, 2);
+  if (large) {
+    int64_t strides[3] = {in.strides(0), in.strides(1), in.strides(2)};
+    compute_encoder.set_bytes(strides, 3, 3);
+
+  } else {
+    int strides[3] = {
+        static_cast<int>(in.strides(0)),
+        static_cast<int>(in.strides(1)),
+        static_cast<int>(in.strides(2))};
+    compute_encoder.set_bytes(strides, 3, 3);
+  }
+
+  compute_encoder.set_bytes(K, 4);
+  auto group_dims = get_block_dims(D, Tout, B);
+  MTL::Size grid_dims = MTL::Size(D, Tout, B);
+
+  compute_encoder.dispatch_threads(grid_dims, group_dims);
+}
+
 void conv_1D_gpu(
     const Stream& s,
     metal::Device& d,
@@ -790,8 +873,15 @@ void conv_1D_gpu(
   bool is_idil_one = in_dilation[0] == 1;
   int C = in.shape(2);
   int O = wt.shape(0);
-  const int C_per_group = in.shape(2) / groups;
-  const int O_per_group = wt.shape(0) / groups;
+  // Fast path for fully separable 1D convolution
+  if (is_idil_one && (groups == C) && groups == O && wt_strides[0] == 1 &&
+      wt_dilation[0] == 1 && padding[0] == 0 && !flip) {
+    depthwise_conv_1D_gpu(s, d, in, wt, out);
+    return;
+  }
+
+  const int C_per_group = C / groups;
+  const int O_per_group = O / groups;
 
   // Direct to implicit gemm conv
   if (is_idil_one && (C_per_group <= 4 || C_per_group % 16 == 0) &&

mlx/backend/metal/custom_kernel.cpp

@@ -327,6 +327,10 @@ CustomKernelFunction metal_kernel(
 void CustomKernel::eval_gpu(
     const std::vector<array>& inputs,
     std::vector<array>& outputs) {
+  // silence some warnings
+  (void)is_precompiled_;
+  (void)shared_memory_;
+
   auto& s = stream();
 
   std::vector<array> copies;

mlx/backend/metal/device.cpp

@@ -72,6 +72,19 @@ MTL::Library* try_load_bundle(
   }
   return nullptr;
 }
+
+MTL::Library* try_load_framework(
+    MTL::Device* device,
+    NS::URL* url,
+    const std::string& lib_name) {
+  std::string resource_path = std::string(url->fileSystemRepresentation()) +
+      "/" + lib_name + ".metallib";
+  auto [lib, error] = load_library_from_path(device, resource_path.c_str());
+  if (lib) {
+    return lib;
+  }
+  return nullptr;
+}
 #endif
 
 // Firstly, search for the metallib in the same path as this binary
@@ -103,12 +116,23 @@ std::pair<MTL::Library*, NS::Error*> load_swiftpm_library(
       return {library, nullptr};
     }
   }
+  // if SWIFTPM_BUNDLE is a framework identifier, try loading from that
+  auto frameworks = NS::Bundle::allFrameworks();
+  for (int i = 0, c = (int)frameworks->count(); i < c; i++) {
+    auto bundle = reinterpret_cast<NS::Bundle*>(frameworks->object(i));
+    if (!strcmp(bundle->bundleIdentifier()->utf8String(), SWIFTPM_BUNDLE)) {
+      library = try_load_framework(device, bundle->resourceURL(), lib_name);
+      if (library != nullptr) {
+        return {library, nullptr};
+      }
+    }
+  }
 #endif
   return {nullptr, nullptr};
 }
 
 MTL::Library* load_default_library(MTL::Device* device) {
-  NS::Error* error[4];
+  NS::Error* error[5];
   MTL::Library* lib;
   // First try the colocated mlx.metallib
   std::tie(lib, error[0]) = load_colocated_library(device, "mlx");
@@ -127,12 +151,19 @@ MTL::Library* load_default_library(MTL::Device* device) {
     return lib;
   }
 
+  // Try lo load resources from Framework resources if SwiftPM wrapped as a
+  // dynamic framework.
+  std::tie(lib, error[3]) = load_colocated_library(device, "Resources/default");
+  if (lib) {
+    return lib;
+  }
+
   // Finally try default_mtllib_path
-  std::tie(lib, error[3]) = load_library_from_path(device, default_mtllib_path);
+  std::tie(lib, error[4]) = load_library_from_path(device, default_mtllib_path);
   if (!lib) {
     std::ostringstream msg;
     msg << "Failed to load the default metallib. ";
-    for (int i = 0; i < 4; i++) {
+    for (int i = 0; i < 5; i++) {
       if (error[i] != nullptr) {
         msg << error[i]->localizedDescription()->utf8String() << " ";
       }
@@ -464,6 +495,10 @@ void Device::end_encoding(int index) {
 CommandEncoder& Device::get_command_encoder(int index) {
   auto& stream = get_stream_(index);
   if (stream.encoder == nullptr) {
+    // Ensure there is an active command buffer
+    if (stream.buffer == nullptr) {
+      get_command_buffer(index);
+    }
     stream.encoder = std::make_unique<CommandEncoder>(stream);
     stream.fence = std::make_shared<Fence>(device_->newFence());
   }
@@ -717,7 +752,7 @@ MTL::ComputePipelineState* Device::get_kernel_(
   mtl_linked_funcs->release();
 
   // Add kernel to cache
-  auto inserted = kernel_map_.insert({hash_name, kernel});
+  kernel_map_.insert({hash_name, kernel});
 
   return kernel;
 }

mlx/backend/metal/eval.cpp

@@ -71,7 +71,7 @@ void eval(array& arr) {
     d.get_command_buffer(s.index);
   } else {
     command_buffer->addCompletedHandler(
-        [s, buffers = std::move(buffers)](MTL::CommandBuffer* cbuf) {
+        [buffers = std::move(buffers)](MTL::CommandBuffer* cbuf) {
           check_error(cbuf);
         });
   }
@@ -82,7 +82,7 @@ void finalize(Stream s) {
   auto& d = metal::device(s.device);
   auto cb = d.get_command_buffer(s.index);
   d.end_encoding(s.index);
-  cb->addCompletedHandler([s](MTL::CommandBuffer* cbuf) { check_error(cbuf); });
+  cb->addCompletedHandler([](MTL::CommandBuffer* cbuf) { check_error(cbuf); });
   d.commit_command_buffer(s.index);
   d.get_command_buffer(s.index);
 }

mlx/backend/metal/fft.cpp

@@ -150,7 +150,6 @@ FFTPlan plan_fft(int n) {
       }
       // See if we can use Rader's algorithm to Stockham decompose n - 1
       auto rader_factors = prime_factors(factor - 1);
-      int last_factor = -1;
       for (int rf : rader_factors) {
         // We don't nest Rader's algorithm so if `factor - 1`
         // isn't Stockham decomposable we give up and do Bluestein's.
@@ -313,8 +312,6 @@ std::pair<array, array> compute_bluestein_constants(int n, int bluestein_n) {
   // w_k = np.exp(-1j * np.pi / N * (np.arange(-N + 1, N) ** 2))
   // w_q = np.fft.fft(1/w_k)
   // return w_k, w_q
-  int length = 2 * n - 1;
-
   std::vector<std::complex<float>> w_k_vec(n);
   std::vector<std::complex<float>> w_q_vec(bluestein_n, 0);
 
@@ -484,8 +481,6 @@ void four_step_fft(
     std::vector<array>& copies,
     const Stream& s,
     bool in_place) {
-  auto& d = metal::device(s.device);
-
   if (plan.bluestein_n == -1) {
     // Fast no transpose implementation for powers of 2.
     FourStepParams four_step_params = {
@@ -786,7 +781,6 @@ void nd_fft_op(
     // Mirror np.fft.(i)rfftn and perform a real transform
     // only on the final axis.
     bool step_real = (real && index == axes.size() - 1);
-    auto step_shape = inverse ? out.shape(axis) : in.shape(axis);
     const array& in_arr = i == axes.size() - 1 ? in : temp_arrs[1 - i % 2];
     array& out_arr = i == 0 ? out : temp_arrs[i % 2];
     fft_op(in_arr, out_arr, axis, inverse, step_real, inplace, s);

mlx/backend/metal/indexing.cpp

@@ -378,7 +378,7 @@ void Scatter::eval_gpu(const std::vector<array>& inputs, array& out) {
   }
 
   if (upd_ndim == 0) {
-    // Need placeholders so Metal doesn't compalain
+    // Need placeholders so Metal doesn't complain
     int shape_ = 0;
     int64_t stride_ = 0;
     compute_encoder.set_bytes(shape_, 3);
@@ -393,7 +393,7 @@ void Scatter::eval_gpu(const std::vector<array>& inputs, array& out) {
   // Set output info
   size_t out_ndim = out.ndim();
   if (out_ndim == 0) {
-    // Need placeholders so Metal doesn't compalain
+    // Need placeholders so Metal doesn't complain
     int shape_ = 0;
     int64_t stride_ = 0;
     compute_encoder.set_bytes(shape_, 7);

mlx/backend/metal/jit/includes.h

@@ -24,7 +24,7 @@ const char* hadamard();
 const char* logsumexp();
 const char* quantized_utils();
 const char* quantized();
-const char* fp4_quantized();
+const char* fp_quantized();
 const char* ternary();
 const char* scan();
 const char* scatter_axis();

mlx/backend/metal/jit_kernels.cpp

@@ -144,8 +144,7 @@ MTL::ComputePipelineState* get_ternary_kernel(
     auto t_str = get_type_string(type);
     std::string kernel_source = metal::utils();
     concatenate(kernel_source, metal::ternary_ops(), metal::ternary());
-    const std::array<std::pair<std::string, std::string>, 4> kernel_types = {{
-        {"v2", "ternary_v2"},
+    const std::array<std::pair<std::string, std::string>, 3> kernel_types = {{
         {"g1large", "ternary_g_nd1"},
         {"g2large", "ternary_g_nd2"},
         {"g3large", "ternary_g_nd3"},
@@ -154,13 +153,29 @@ MTL::ComputePipelineState* get_ternary_kernel(
       kernel_source +=
           get_template_definition(name + "_" + lib_name, func, t_str, op);
     }
+
+    kernel_source += get_template_definition(
+        "v2_" + lib_name, "ternary_v2", t_str, op, false, false);
+    kernel_source += get_template_definition(
+        "sv2_" + lib_name, "ternary_v2", t_str, op, true, false);
+    kernel_source += get_template_definition(
+        "vs2_" + lib_name, "ternary_v2", t_str, op, false, true);
+
     if (get_work_per_thread(type) > 1) {
-      kernel_source +=
-          get_template_definition("vn_" + lib_name, "ternary_v", t_str, op);
+      kernel_source += get_template_definition(
+          "vn_" + lib_name, "ternary_v", t_str, op, false, false);
+      kernel_source += get_template_definition(
+          "svn_" + lib_name, "ternary_v", t_str, op, true, false);
+      kernel_source += get_template_definition(
+          "vsn_" + lib_name, "ternary_v", t_str, op, false, true);
     }
 
-    kernel_source +=
-        get_template_definition("v_" + lib_name, "ternary_v", t_str, op, 1);
+    kernel_source += get_template_definition(
+        "v_" + lib_name, "ternary_v", t_str, op, false, false, 1);
+    kernel_source += get_template_definition(
+        "sv_" + lib_name, "ternary_v", t_str, op, true, false, 1);
+    kernel_source += get_template_definition(
+        "vs_" + lib_name, "ternary_v", t_str, op, false, true, 1);
     kernel_source += get_template_definition(
         "g1_" + lib_name, "ternary_g_nd1", t_str, op, "int");
     kernel_source += get_template_definition(
@@ -814,7 +829,7 @@ MTL::ComputePipelineState* get_quantized_kernel(
         metal::utils(),
         metal::gemm(),
         metal::quantized_utils(),
-        (mode == "affine") ? metal::quantized() : metal::fp4_quantized(),
+        (mode == "affine") ? metal::quantized() : metal::fp_quantized(),
         template_def);
     return kernel_source;
   });
@@ -841,39 +856,22 @@ MTL::ComputePipelineState* get_gather_qmm_kernel(
     std::string kernel_source;
     concatenate(
         kernel_source, metal::utils(), metal::quantized_utils(), metal::gemm());
-    if (mode == "affine") {
-      concatenate(
-          kernel_source,
-          metal::quantized(),
-          get_template_definition(
-              lib_name,
-              mode + "_gather_qmm_rhs",
-              get_type_string(x.dtype()),
-              group_size,
-              bits,
-              bm,
-              bn,
-              bk,
-              wm,
-              wn,
-              transpose));
-    } else {
-      concatenate(
-          kernel_source,
-          metal::fp4_quantized(),
-          get_template_definition(
-              lib_name,
-              mode + "_gather_qmm_rhs",
-              get_type_string(x.dtype()),
-              group_size,
-              "uint8_t",
-              bm,
-              bn,
-              bk,
-              wm,
-              wn,
-              transpose));
-    }
+    bool is_affine = mode == "affine";
+    concatenate(
+        kernel_source,
+        is_affine ? metal::quantized() : metal::fp_quantized(),
+        get_template_definition(
+            lib_name,
+            (is_affine ? "affine" : "fp") + std::string("_gather_qmm_rhs"),
+            get_type_string(x.dtype()),
+            group_size,
+            bits,
+            bm,
+            bn,
+            bk,
+            wm,
+            wn,
+            transpose));
     return kernel_source;
   });
   return d.get_kernel(kernel_name, lib, hash_name, func_consts);

mlx/backend/metal/kernels/CMakeLists.txt

@@ -6,6 +6,7 @@ set(BASE_HEADERS
     defines.h
     erf.h
     expm1f.h
+    fp8.h
     utils.h)
 
 function(build_kernel_base TARGET SRCFILE DEPS)
@@ -109,7 +110,8 @@ if(NOT MLX_METAL_JIT)
     reduction/reduce_col.h
     reduction/reduce_row.h)
   build_kernel(quantized quantized.h quantized_utils.h ${STEEL_HEADERS})
-  build_kernel(fp4_quantized fp4_quantized.h quantized_utils.h ${STEEL_HEADERS})
+  build_kernel(fp_quantized fp4.h fp_quantized.h quantized_utils.h
+               ${STEEL_HEADERS})
   build_kernel(scan scan.h)
   build_kernel(softmax softmax.h)
   build_kernel(logsumexp logsumexp.h)

mlx/backend/metal/kernels/complex.h

@@ -104,6 +104,27 @@ constexpr bool operator==(complex64_t a, complex64_t b) {
 constexpr complex64_t operator+(complex64_t a, complex64_t b) {
   return {a.real + b.real, a.imag + b.imag};
 }
+
+constexpr thread complex64_t& operator+=(thread complex64_t& a, complex64_t b) {
+  a.real += b.real;
+  a.imag += b.imag;
+  return a;
+}
+
+constexpr threadgroup complex64_t& operator+=(
+    threadgroup complex64_t& a,
+    complex64_t b) {
+  a.real += b.real;
+  a.imag += b.imag;
+  return a;
+}
+
+constexpr device complex64_t& operator+=(device complex64_t& a, complex64_t b) {
+  a.real += b.real;
+  a.imag += b.imag;
+  return a;
+}
+
 constexpr complex64_t operator+(float a, complex64_t b) {
   return {a + b.real, b.imag};
 }

mlx/backend/metal/kernels/conv.metal

@@ -288,6 +288,40 @@ instantiate_depthconv2d(float32, float);
 instantiate_depthconv2d(float16, half);
 instantiate_depthconv2d(bfloat16, bfloat16_t);
 
+template <typename T, typename IdxT>
+[[kernel]] void depthwise_conv_1d(
+    const device T* in [[buffer(0)]],
+    const device T* w [[buffer(1)]],
+    device T* out [[buffer(2)]],
+    constant const IdxT strides[3],
+    constant const int& kernel_size,
+    uint3 tid [[thread_position_in_grid]],
+    uint3 grid_dim [[threads_per_grid]]) {
+  out += (tid.z * static_cast<IdxT>(grid_dim.y) + tid.y) * grid_dim.x + tid.x;
+  in += tid.z * strides[0] + tid.y * strides[1] + tid.x * strides[2];
+  w += tid.x * kernel_size;
+
+  float acc = 0.0;
+  for (int i = 0; i < kernel_size; ++i) {
+    acc += static_cast<float>(in[0]) * w[i];
+    in += strides[1];
+  }
+  *out = static_cast<T>(acc);
+}
+
+#define instantiate_depthconv1d(iname, itype)                         \
+  instantiate_kernel(                                                 \
+      "depthwise_conv_1d_" #iname, depthwise_conv_1d, itype, int32_t) \
+      instantiate_kernel(                                             \
+          "depthwise_conv_1d_" #iname "_large",                       \
+          depthwise_conv_1d,                                          \
+          itype,                                                      \
+          int64_t)
+
+instantiate_depthconv1d(float32, float);
+instantiate_depthconv1d(float16, half);
+instantiate_depthconv1d(bfloat16, bfloat16_t);
+
 ///////////////////////////////////////////////////////////////////////////////
 /// Winograd kernels
 ///////////////////////////////////////////////////////////////////////////////

mlx/backend/metal/kernels/fp4.h

@@ -0,0 +1,56 @@
+#pragma once
+
+constexpr constant static float FP4_LUT[16] = {
+    +0.0f,
+    +0.5f,
+    +1.0f,
+    +1.5f,
+    +2.0f,
+    +3.0f,
+    +4.0f,
+    +6.0f,
+    -0.0f,
+    -0.5f,
+    -1.0f,
+    -1.5f,
+    -2.0f,
+    -3.0f,
+    -4.0f,
+    -6.0f};
+
+struct fp4_e2m1 {
+  fp4_e2m1(float x) {
+    if (metal::isnan(x)) {
+      bits = 0x7;
+      return;
+    }
+
+    const uint8_t sign_bit = (metal::signbit(x)) ? 0x8 : 0x0;
+    x = metal::abs(x);
+
+    if (x > 5.0f) {
+      bits = 0x7;
+    } else if (x >= 3.5f) {
+      bits = 0x6;
+    } else if (x > 2.5f) {
+      bits = 0x5;
+    } else if (x >= 1.75f) {
+      bits = 0x4;
+    } else if (x > 1.25f) {
+      bits = 0x3;
+    } else if (x >= 0.75f) {
+      bits = 0x2;
+    } else if (x > 0.25f) {
+      bits = 0x1;
+    } else {
+      bits = 0x0;
+    }
+    bits |= sign_bit;
+  }
+
+  operator float() {
+    return FP4_LUT[bits];
+  }
+
+  uint8_t bits;
+};

mlx/backend/metal/kernels/fp4_quantized.metal

@@ -1,127 +0,0 @@
-// Copyright © 2025 Apple Inc.
-
-// clang-format off
-#include "mlx/backend/metal/kernels/utils.h"
-#include "mlx/backend/metal/kernels/steel/gemm/gemm.h"
-#include "mlx/backend/metal/kernels/quantized_utils.h"
-#include "mlx/backend/metal/kernels/fp4_quantized.h"
-
-#define instantiate_quantized(name, type) \
-  instantiate_kernel( \
-      #name "_" #type "_gs_32_b_4", \
-      name, \
-      type, \
-      32,   \
-      uint8_t)
-
-#define instantiate_quantized_batched(name, type, batched) \
-  instantiate_kernel( \
-      #name "_" #type "_gs_32_b_4_batch_" #batched, \
-      name,    \
-      type,    \
-      32,      \
-      uint8_t, \
-      batched)
-
-#define instantiate_quantized_aligned(name, type, aligned) \
-  instantiate_kernel( \
-      #name "_" #type "_gs_32_b_4_alN_" #aligned, \
-      name,    \
-      type,    \
-      32,      \
-      uint8_t, \
-      aligned)
-
-#define instantiate_quantized_aligned_batched(name, type, aligned, batched) \
-  instantiate_kernel( \
-      #name "_" #type "_gs_32_b_4_alN_" #aligned "_batch_" #batched, \
-      name,    \
-      type,    \
-      32,      \
-      uint8_t, \
-      aligned, \
-      batched)
-
-#define instantiate_quantized_quad(name, type, D, batched) \
-  instantiate_kernel( \
-      #name "_" #type "_gs_32_b_4_d_" #D "_batch_" #batched, \
-      name,    \
-      type,    \
-      32,      \
-      uint8_t, \
-      D,       \
-      batched)
-
-#define instantiate_quantized_split_k(name, type, split_k) \
-  instantiate_kernel( \
-      #name "_" #type "_gs_32_b_4_spk_" #split_k, \
-      name,    \
-      type,    \
-      32,      \
-      uint8_t, \
-      split_k)
-
-#define instantiate_gather_qmm_rhs(func, name, type, bm, bn, bk, wm, wn, transpose) \
-  instantiate_kernel( \
-      #name "_" #type "_gs_32_b_4_bm_" #bm "_bn_" #bn "_bk_" #bk "_wm_" #wm "_wn_" #wn, \
-      func,    \
-      type,    \
-      32,      \
-      uint8_t, \
-      bm,      \
-      bn,      \
-      bk,      \
-      wm,      \
-      wn,      \
-      transpose)
-
-#define instantiate_quantized_batched_wrap(name, type) \
-  instantiate_quantized_batched(name, type, 1)         \
-  instantiate_quantized_batched(name, type, 0)
-
-#define instantiate_quantized_all_batched(type) \
-  instantiate_quantized_batched_wrap(mxfp4_qmv_fast, type) \
-  instantiate_quantized_batched_wrap(mxfp4_qmv, type)      \
-  instantiate_quantized_batched_wrap(mxfp4_qvm, type)      \
-  instantiate_quantized_batched_wrap(mxfp4_qmm_n, type)
-
-#define instantiate_quantized_all_single(type) \
-  instantiate_quantized(mxfp4_gather_qmv_fast, type) \
-  instantiate_quantized(mxfp4_gather_qmv, type)      \
-  instantiate_quantized(mxfp4_gather_qvm, type)      \
-  instantiate_quantized(mxfp4_gather_qmm_n, type)
-
-#define instantiate_quantized_all_aligned(type) \
-  instantiate_quantized_aligned(mxfp4_gather_qmm_t, type, true)      \
-  instantiate_quantized_aligned(mxfp4_gather_qmm_t, type, false)     \
-  instantiate_quantized_aligned_batched(mxfp4_qmm_t, type, true, 1)  \
-  instantiate_quantized_aligned_batched(mxfp4_qmm_t, type, true, 0)  \
-  instantiate_quantized_aligned_batched(mxfp4_qmm_t, type, false, 1) \
-  instantiate_quantized_aligned_batched(mxfp4_qmm_t, type, false, 0)
-
-#define instantiate_quantized_all_quad(type) \
-  instantiate_quantized_quad(mxfp4_qmv_quad, type, 64, 1)  \
-  instantiate_quantized_quad(mxfp4_qmv_quad, type, 64, 0)  \
-  instantiate_quantized_quad(mxfp4_qmv_quad, type, 128, 1) \
-  instantiate_quantized_quad(mxfp4_qmv_quad, type, 128, 0)
-
-#define instantiate_quantized_all_splitk(type) \
-  instantiate_quantized_split_k(mxfp4_qvm_split_k, type, 8) \
-  instantiate_quantized_split_k(mxfp4_qvm_split_k, type, 32)
-
-#define instantiate_quantized_all_rhs(type) \
-  instantiate_gather_qmm_rhs(mxfp4_gather_qmm_rhs, mxfp4_gather_qmm_rhs_nt, type, 16, 32, 32, 1, 2, true) \
-  instantiate_gather_qmm_rhs(mxfp4_gather_qmm_rhs, mxfp4_gather_qmm_rhs_nn, type, 16, 32, 32, 1, 2, false)
-
-#define instantiate_quantized_types(type) \
-  instantiate_quantized_all_batched(type) \
-  instantiate_quantized_all_quad(type)    \
-  instantiate_quantized_all_splitk(type)  \
-  instantiate_quantized_all_single(type)  \
-  instantiate_quantized_all_aligned(type) \
-  instantiate_quantized_all_rhs(type)
-
-instantiate_quantized_types(float)
-instantiate_quantized_types(bfloat16_t)
-instantiate_quantized_types(float16_t)
-    // clang-format on