patch bump (#2324)

* fix module update in strict mode

* allow GELU to be pickled

.circleci/config.yml

@@ -16,6 +16,9 @@ parameters:
   linux_release:
     type: boolean
     default: false
+  cuda_release:
+    type: boolean
+    default: false
 
 jobs:
   build_documentation:
@@ -104,7 +107,7 @@ jobs:
           command: |
             echo "stubs"
             pip install typing_extensions
-            python setup.py generate_stubs 
+            python setup.py generate_stubs
       - run:
           name: Run Python tests
           command: |
@@ -162,7 +165,7 @@ jobs:
           command: |
             source env/bin/activate
             pip install typing_extensions
-            python setup.py generate_stubs 
+            python setup.py generate_stubs
       - run:
           name: Run Python tests
           command: |
@@ -223,7 +226,6 @@ jobs:
           command: |
             sudo apt-get update
             sudo apt-get install libblas-dev liblapack-dev liblapacke-dev
-            sudo apt-get install openmpi-bin openmpi-common libopenmpi-dev
             python -m venv env
             source env/bin/activate
             CMAKE_BUILD_PARALLEL_LEVEL=`nproc` \
@@ -283,7 +285,7 @@ jobs:
           command: |
             source env/bin/activate
             pip install typing_extensions
-            python setup.py generate_stubs 
+            python setup.py generate_stubs
       - run:
           name: Build Python package
           command: |
@@ -342,7 +344,7 @@ jobs:
               CMAKE_BUILD_PARALLEL_LEVEL=`nproc` \
               pip install . -v
             pip install typing_extensions
-            python setup.py generate_stubs 
+            python setup.py generate_stubs
             << parameters.extra_env >> \
               CMAKE_BUILD_PARALLEL_LEVEL=`nproc` \
               python -m build --wheel
@@ -356,6 +358,48 @@ jobs:
       - store_artifacts:
           path: wheelhouse/
 
+  build_cuda_release:
+    parameters:
+      python_version:
+        type: string
+        default: "3.9"
+      extra_env:
+        type: string
+        default: "DEV_RELEASE=1"
+    machine:
+      image: linux-cuda-12:default
+      resource_class: gpu.nvidia.small.gen2
+    steps:
+      - checkout
+      - run:
+          name: Build wheel
+          command: |
+            sudo apt-get update
+            sudo apt-get install libblas-dev liblapack-dev liblapacke-dev
+            python -m venv env
+            source env/bin/activate
+            pip install auditwheel
+            pip install patchelf
+            pip install build
+            pip install twine
+            << parameters.extra_env >> \
+              CMAKE_BUILD_PARALLEL_LEVEL=`nproc` \
+              CMAKE_ARGS="-DMLX_BUILD_CUDA=ON -DCMAKE_CUDA_COMPILER=`which nvcc`" \
+              pip install ".[dev]" -v
+            python setup.py generate_stubs
+            << parameters.extra_env >> \
+              CMAKE_BUILD_PARALLEL_LEVEL=`nproc` \
+              CMAKE_ARGS="-DMLX_BUILD_CUDA=ON -DCMAKE_CUDA_COMPILER=`which nvcc`" \
+              python -m build --wheel
+            bash python/scripts/repair_cuda.sh
+      - run:
+          name: Upload package
+          command: |
+            source env/bin/activate
+            twine upload wheelhouse/*.whl
+      - store_artifacts:
+          path: wheelhouse/
+
 workflows:
   build_and_test:
     when:
@@ -625,3 +669,14 @@ workflows:
             parameters:
               python_version: ["3.9", "3.10", "3.11", "3.12", "3.13"]
               extra_env: ["PYPI_RELEASE=1"]
+  cuda_test_release:
+    when:
+      and:
+        - equal: [ main, << pipeline.git.branch >> ]
+        - << pipeline.parameters.cuda_release >>
+    jobs:
+      - build_cuda_release:
+          matrix:
+            parameters:
+              python_version: ["3.9", "3.10", "3.11", "3.12", "3.13"]
+              extra_env: ["PYPI_RELEASE=1"]

CMakeLists.txt

@@ -35,7 +35,6 @@ option(MLX_BUILD_PYTHON_BINDINGS "Build python bindings for mlx" OFF)
 option(MLX_BUILD_METAL "Build metal backend" ON)
 option(MLX_BUILD_CPU "Build cpu backend" ON)
 option(MLX_BUILD_CUDA "Build cuda backend" OFF)
-option(MLX_BUILD_ROCM "Build ROCm backend" OFF)
 option(MLX_METAL_DEBUG "Enhance metal debug workflow" OFF)
 option(MLX_ENABLE_X64_MAC "Enable building for x64 macOS" OFF)
 option(MLX_BUILD_GGUF "Include support for GGUF format" ON)
@@ -89,10 +88,6 @@ if(MLX_BUILD_CUDA)
   enable_language(CUDA)
 endif()
 
-if(MLX_BUILD_ROCM)
-  enable_language(HIP)
-endif()
-
 if(MLX_BUILD_METAL AND NOT METAL_LIB)
   message(STATUS "Metal not found. Unable to build GPU")
   set(MLX_BUILD_METAL OFF)

benchmarks/python/comparative/bench_torch.py

@@ -5,6 +5,7 @@ import os
 import time
 
 import torch
+import torch.cuda
 import torch.mps
 
 
@@ -44,8 +45,10 @@ def bench(f, *args):
 
 
 def sync_if_needed(x):
-    if x.device != torch.device("cpu"):
+    if x.device == torch.device("mps"):
         torch.mps.synchronize()
+    elif x.device == torch.device("cuda"):
+        torch.cuda.synchronize()
 
 
 @torch.no_grad()
@@ -99,6 +102,14 @@ def reduction(op, axis, x):
     sync_if_needed(x)
 
 
+@torch.no_grad()
+def sum_and_add(axis, x, y):
+    z = x.sum(axis=axis, keepdims=True)
+    for i in range(50):
+        z = (z + y).sum(axis=axis, keepdims=True)
+    sync_if_needed(x)
+
+
 @torch.no_grad()
 def softmax(axis, x):
     ys = []
@@ -340,7 +351,11 @@ if __name__ == "__main__":
         args.axis.pop(0)
 
     torch.set_num_threads(1)
-    device = "cpu" if args.cpu else "mps"
+    device = "mps"
+    if torch.cuda.is_available():
+        device = "cuda"
+    if args.cpu:
+        device = "cpu"
 
     types = args.dtype
     if not types:
@@ -460,5 +475,8 @@ if __name__ == "__main__":
     elif args.benchmark == "selu":
         print(bench(selu, x))
 
+    elif args.benchmark == "sum_and_add":
+        print(bench(sum_and_add, axis, *xs))
+
     else:
         raise ValueError(f"Unknown benchmark `{args.benchmark}`.")

docs/src/install.rst

@@ -30,6 +30,16 @@ MLX is also available on conda-forge. To install MLX with conda do:
 
    conda install conda-forge::mlx
 
+CUDA
+^^^^
+
+MLX has a CUDA backend which you can use on any Linux platform with CUDA 12
+and SM 7.0 (Volta) and up. To install MLX with CUDA support, run:
+
+.. code-block:: shell
+
+    pip install mlx-cuda
+
 
 Troubleshooting
 ^^^^^^^^^^^^^^^
@@ -65,6 +75,8 @@ Build Requirements
 Python API
 ^^^^^^^^^^
 
+.. _python install:
+
 To build and install the MLX python library from source, first, clone MLX from
 `its GitHub repo <https://github.com/ml-explore/mlx>`_:
 
@@ -107,6 +119,8 @@ IDE:
 C++ API
 ^^^^^^^
 
+.. _cpp install:
+
 Currently, MLX must be built and installed from source.
 
 Similarly to the python library, to build and install the MLX C++ library start
@@ -185,6 +199,7 @@ should point to the path to the built metal library.
 
       xcrun -sdk macosx --show-sdk-version
 
+
 Binary Size Minimization
 ~~~~~~~~~~~~~~~~~~~~~~~~
 
@@ -213,6 +228,50 @@ be anwywhere from a few hundred millisecond to a few seconds depending on the
 application. Once a kernel is compiled, it will be cached by the system. The
 Metal kernel cache persists across reboots.
 
+Linux
+^^^^^
+
+To build from source on Linux (CPU only), install the BLAS and LAPACK headers.
+For example on Ubuntu, run the following:
+
+.. code-block:: shell
+
+   apt-get update -y
+   apt-get install libblas-dev liblapack-dev liblapacke-dev -y
+
+From here follow the instructions to install either the :ref:`Python <python
+install>` or :ref:`C++ <cpp install>` APIs.
+
+CUDA
+^^^^
+
+To build from source on Linux with CUDA, install the BLAS and LAPACK headers
+and the CUDA toolkit. For example on Ubuntu, run the following:
+
+.. code-block:: shell
+
+   wget https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2204/x86_64/cuda-keyring_1.1-1_all.deb
+   dpkg -i cuda-keyring_1.1-1_all.deb
+   apt-get update -y
+   apt-get -y install cuda-toolkit-12-9
+   apt-get install libblas-dev liblapack-dev liblapacke-dev -y
+
+
+When building either the Python or C++ APIs make sure to pass the cmake flag
+``MLX_BUILD_CUDA=ON``. For example, to build the Python API run:
+
+.. code-block:: shell
+
+  CMAKE_BUILD_PARALLEL_LEVEL=8 CMAKE_ARGS="-DMLX_BUILD_CUDA=ON" pip install -e ".[dev]"
+
+To build the C++ package run:
+
+.. code-block:: shell
+
+   mkdir -p build && cd build
+   cmake .. -DMLX_BUILD_CUDA=ON && make -j
+
+
 Troubleshooting
 ^^^^^^^^^^^^^^^
 

mlx/CMakeLists.txt

@@ -60,16 +60,7 @@ else()
                  PRIVATE ${CMAKE_CURRENT_SOURCE_DIR}/backend/cuda/no_cuda.cpp)
 endif()
 
-if(MLX_BUILD_ROCM)
-  add_subdirectory(${CMAKE_CURRENT_SOURCE_DIR}/backend/rocm)
-else()
-  target_sources(mlx
-                 PRIVATE ${CMAKE_CURRENT_SOURCE_DIR}/backend/rocm/no_rocm.cpp)
-endif()
-
-if(MLX_BUILD_METAL
-   OR MLX_BUILD_CUDA
-   OR MLX_BUILD_ROCM)
+if(MLX_BUILD_METAL OR MLX_BUILD_CUDA)
   add_subdirectory(${CMAKE_CURRENT_SOURCE_DIR}/backend/gpu)
 else()
   add_subdirectory(${CMAKE_CURRENT_SOURCE_DIR}/backend/no_gpu)

mlx/backend/common/compiled.cpp

@@ -14,6 +14,8 @@ void print_constant(std::ostream& os, const array& x) {
       return print_float_constant<float16_t>(os, x);
     case bfloat16:
       return print_float_constant<bfloat16_t>(os, x);
+    case float64:
+      return print_float_constant<double>(os, x);
     case complex64:
       return print_complex_constant<complex64_t>(os, x);
     case int8:
@@ -50,6 +52,8 @@ std::string get_type_string(Dtype d) {
       return "float16_t";
     case bfloat16:
       return "bfloat16_t";
+    case float64:
+      return "double";
     case complex64:
       return "complex64_t";
     case bool_:

mlx/backend/common/compiled.h

@@ -18,8 +18,12 @@ std::string get_type_string(Dtype d);
 template <typename T>
 void print_float_constant(std::ostream& os, const array& x) {
   auto old_precision = os.precision();
-  os << std::setprecision(std::numeric_limits<float>::digits10 + 1)
-     << x.item<T>() << std::setprecision(old_precision);
+  if constexpr (std::is_same_v<T, double>) {
+    os << std::setprecision(std::numeric_limits<double>::digits10 + 1);
+  } else {
+    os << std::setprecision(std::numeric_limits<float>::digits10 + 1);
+  }
+  os << x.item<T>() << std::setprecision(old_precision);
 }
 
 template <typename T>

mlx/backend/common/reduce.cpp

@@ -5,11 +5,9 @@
 namespace mlx::core {
 
 std::pair<Shape, Strides> shapes_without_reduction_axes(
-    const array& x,
+    Shape shape,
+    Strides strides,
     const std::vector<int>& axes) {
-  auto shape = x.shape();
-  auto strides = x.strides();
-
   for (int i = axes.size() - 1; i >= 0; i--) {
     int a = axes[i];
     shape.erase(shape.begin() + a);
@@ -19,6 +17,15 @@ std::pair<Shape, Strides> shapes_without_reduction_axes(
   return std::make_pair(shape, strides);
 }
 
+std::pair<Shape, Strides> shapes_without_reduction_axes(
+    const array& x,
+    const std::vector<int>& axes) {
+  auto shape = x.shape();
+  auto strides = x.strides();
+  return shapes_without_reduction_axes(
+      std::move(shape), std::move(strides), axes);
+}
+
 ReductionPlan get_reduction_plan(const array& x, const std::vector<int>& axes) {
   // The data is all there and we are reducing over everything
   if (x.size() == x.data_size() && axes.size() == x.ndim() &&

mlx/backend/common/reduce.h

@@ -51,5 +51,9 @@ ReductionPlan get_reduction_plan(const array& x, const std::vector<int>& axes);
 std::pair<Shape, Strides> shapes_without_reduction_axes(
     const array& x,
     const std::vector<int>& axes);
+std::pair<Shape, Strides> shapes_without_reduction_axes(
+    Shape shape,
+    Strides strides,
+    const std::vector<int>& axes);
 
 } // namespace mlx::core

mlx/backend/common/utils.cpp

@@ -199,12 +199,15 @@ Dims get_2d_grid_dims_common(
       }
     }
   }
-  if (grid_y > UINT32_MAX || grid_x > UINT32_MAX || divisor > 1) {
+  if (grid_y > UINT32_MAX || grid_x > UINT32_MAX) {
     throw std::runtime_error("Unable to safely factor shape.");
   }
   if (grid_y > grid_x) {
     std::swap(grid_x, grid_y);
   }
+  if (divisor > 1) {
+    grid_x = ((grid_x + divisor - 1) / divisor) * divisor;
+  }
   return std::make_tuple(
       static_cast<uint32_t>(grid_x), static_cast<uint32_t>(grid_y), 1);
 }

mlx/backend/cuda/CMakeLists.txt

@@ -8,6 +8,7 @@ target_sources(
   PRIVATE ${CMAKE_CURRENT_SOURCE_DIR}/allocator.cpp
           ${CMAKE_CURRENT_SOURCE_DIR}/arg_reduce.cu
           ${CMAKE_CURRENT_SOURCE_DIR}/binary.cu
+          ${CMAKE_CURRENT_SOURCE_DIR}/binary_two.cu
           ${CMAKE_CURRENT_SOURCE_DIR}/compiled.cpp
           ${CMAKE_CURRENT_SOURCE_DIR}/copy.cu
           ${CMAKE_CURRENT_SOURCE_DIR}/copy/copy_contiguous.cu
@@ -28,9 +29,10 @@ target_sources(
           ${CMAKE_CURRENT_SOURCE_DIR}/primitives.cu
           ${CMAKE_CURRENT_SOURCE_DIR}/random.cu
           ${CMAKE_CURRENT_SOURCE_DIR}/reduce.cu
+          ${CMAKE_CURRENT_SOURCE_DIR}/reduce/all_reduce.cu
           ${CMAKE_CURRENT_SOURCE_DIR}/reduce/col_reduce.cu
+          ${CMAKE_CURRENT_SOURCE_DIR}/reduce/init_reduce.cu
           ${CMAKE_CURRENT_SOURCE_DIR}/reduce/row_reduce.cu
-          ${CMAKE_CURRENT_SOURCE_DIR}/reduce/segmented_reduce.cu
           ${CMAKE_CURRENT_SOURCE_DIR}/rms_norm.cu
           ${CMAKE_CURRENT_SOURCE_DIR}/rope.cu
           ${CMAKE_CURRENT_SOURCE_DIR}/slicing.cpp

mlx/backend/cuda/allocator.cpp

@@ -3,6 +3,7 @@
 #include "mlx/backend/cuda/allocator.h"
 #include "mlx/backend/cuda/utils.h"
 #include "mlx/backend/cuda/worker.h"
+#include "mlx/utils.h"
 
 #include <cuda_runtime.h>
 #include <fmt/format.h>
@@ -14,9 +15,11 @@ namespace mlx::core {
 
 namespace cu {
 
+constexpr int page_size = 16384;
+
 CudaAllocator::CudaAllocator()
     : buffer_cache_(
-          getpagesize(),
+          page_size,
           [](CudaBuffer* buf) { return buf->size; },
           [this](CudaBuffer* buf) {
             cuda_free(buf->data);
@@ -31,7 +34,14 @@ CudaAllocator::CudaAllocator()
 
 Buffer CudaAllocator::malloc(size_t size) {
   // Find available buffer from cache.
+  auto orig_size = size;
   std::unique_lock lock(mutex_);
+  if (size < page_size) {
+    size = next_power_of_2(size);
+  } else {
+    size = page_size * ((size + page_size - 1) / page_size);
+  }
+
   CudaBuffer* buf = buffer_cache_.reuse_from_cache(size);
   if (!buf) {
     // If we have a lot of memory pressure or are over the maximum cache size,
@@ -106,7 +116,6 @@ void CudaAllocator::cuda_free(void* buf) {
       return;
     }
   }
-
   cudaFree(buf);
 }
 

mlx/backend/cuda/arg_reduce.cu

@@ -152,35 +152,29 @@ void ArgReduce::eval_gpu(const std::vector<array>& inputs, array& out) {
   encoder.set_input_array(in);
   encoder.set_output_array(out);
   encoder.launch_kernel([&](cudaStream_t stream) {
-    MLX_SWITCH_REAL_TYPES_CHECKED(in.dtype(), "ArgReduce", CTYPE, {
-      using InType = cuda_type_t<CTYPE>;
+    dispatch_real_types(in.dtype(), "ArgReduce", [&](auto type_tag) {
+      using T = cuda_type_t<MLX_GET_TYPE(type_tag)>;
       constexpr uint32_t N_READS = 4;
-      MLX_SWITCH_BLOCK_DIM(cuda::ceil_div(axis_size, N_READS), BLOCK_DIM, {
-        dim3 num_blocks = get_2d_grid_dims(out.shape(), out.strides());
-        dim3 block_dims{BLOCK_DIM, 1, 1};
-        auto kernel = &cu::arg_reduce_general<
-            InType,
-            cu::ArgMax<InType>,
-            BLOCK_DIM,
-            N_READS>;
-        if (reduce_type_ == ArgReduce::ArgMin) {
-          kernel = &cu::arg_reduce_general<
-              InType,
-              cu::ArgMin<InType>,
-              BLOCK_DIM,
-              N_READS>;
-        }
-        kernel<<<num_blocks, block_dims, 0, stream>>>(
-            in.data<InType>(),
-            out.data<uint32_t>(),
-            out.size(),
-            const_param(shape),
-            const_param(in_strides),
-            const_param(out_strides),
-            ndim,
-            axis_stride,
-            axis_size);
-      });
+      dispatch_block_dim(
+          cuda::ceil_div(axis_size, N_READS), [&](auto block_dim) {
+            dim3 num_blocks = get_2d_grid_dims(out.shape(), out.strides());
+            auto kernel =
+                cu::arg_reduce_general<T, cu::ArgMax<T>, block_dim(), N_READS>;
+            if (reduce_type_ == ArgReduce::ArgMin) {
+              kernel = cu::
+                  arg_reduce_general<T, cu::ArgMin<T>, block_dim(), N_READS>;
+            }
+            kernel<<<num_blocks, block_dim(), 0, stream>>>(
+                in.data<T>(),
+                out.data<uint32_t>(),
+                out.size(),
+                const_param(shape),
+                const_param(in_strides),
+                const_param(out_strides),
+                ndim,
+                axis_stride,
+                axis_size);
+          });
     });
   });
 }

mlx/backend/cuda/binary.cu

@@ -125,13 +125,12 @@ constexpr bool supports_binary_op() {
 template <typename Op>
 void binary_op_gpu_inplace(
     const std::vector<array>& inputs,
-    std::vector<array>& outputs,
+    array& out,
     std::string_view op,
     const Stream& s) {
   assert(inputs.size() > 1);
   const auto& a = inputs[0];
   const auto& b = inputs[1];
-  auto& out = outputs[0];
   if (out.size() == 0) {
     return;
   }
@@ -141,55 +140,64 @@ void binary_op_gpu_inplace(
   encoder.set_input_array(b);
   encoder.set_output_array(out);
   encoder.launch_kernel([&](cudaStream_t stream) {
-    MLX_SWITCH_ALL_TYPES(a.dtype(), CTYPE_IN, {
-      MLX_SWITCH_ALL_TYPES(out.dtype(), CTYPE_OUT, {
+    dispatch_all_types(a.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_binary_op<Op, CTYPE_IN, CTYPE_OUT>()) {
           using InType = cuda_type_t<CTYPE_IN>;
           using OutType = cuda_type_t<CTYPE_OUT>;
-
           auto bopt = get_binary_op_type(a, b);
           if (bopt == BinaryOpType::General) {
-            auto [shape, strides] = collapse_contiguous_dims(a, b, out);
-            auto& a_strides = strides[0];
-            auto& b_strides = strides[1];
-            bool large = a.data_size() > INT32_MAX ||
-                b.data_size() > INT32_MAX || out.data_size() > INT32_MAX;
-            MLX_SWITCH_BOOL(large, LARGE, {
-              using IdxT = std::conditional_t<LARGE, int64_t, int32_t>;
-              int ndim = shape.size();
-              if (ndim <= 3) {
-                MLX_SWITCH_1_2_3(ndim, NDIM, {
-                  auto kernel =
-                      &cu::binary_g_nd<Op, InType, OutType, IdxT, NDIM>;
-                  auto [num_blocks, block_dims] =
-                      get_launch_args(kernel, out, large);
-                  kernel<<<num_blocks, block_dims, 0, stream>>>(
-                      a.data<InType>(),
-                      b.data<InType>(),
-                      out.data<OutType>(),
-                      out.size(),
-                      const_param<NDIM>(shape),
-                      const_param<NDIM>(a_strides),
-                      const_param<NDIM>(b_strides));
+            dispatch_bool(
+                a.data_size() > INT32_MAX || b.data_size() > INT32_MAX ||
+                    out.data_size() > INT32_MAX,
+                [&](auto large) {
+                  using IdxT = std::conditional_t<large(), int64_t, int32_t>;
+                  Shape shape;
+                  std::vector<Strides> strides;
+                  std::tie(shape, strides) =
+                      collapse_contiguous_dims(a, b, out);
+                  auto& a_strides = strides[0];
+                  auto& b_strides = strides[1];
+                  int ndim = shape.size();
+                  if (ndim <= 3) {
+                    dispatch_1_2_3(ndim, [&](auto dims_constant) {
+                      auto kernel = cu::binary_g_nd<
+                          Op,
+                          InType,
+                          OutType,
+                          IdxT,
+                          dims_constant()>;
+                      auto [num_blocks, block_dims] =
+                          get_launch_args(kernel, out, large());
+                      kernel<<<num_blocks, block_dims, 0, stream>>>(
+                          a.data<InType>(),
+                          b.data<InType>(),
+                          out.data<OutType>(),
+                          out.size(),
+                          const_param<dims_constant()>(shape),
+                          const_param<dims_constant()>(a_strides),
+                          const_param<dims_constant()>(b_strides));
+                    });
+                  } else {
+                    auto kernel = cu::binary_g<Op, InType, OutType, IdxT>;
+                    auto [num_blocks, block_dims] =
+                        get_launch_args(kernel, out, large());
+                    kernel<<<num_blocks, block_dims, 0, stream>>>(
+                        a.data<InType>(),
+                        b.data<InType>(),
+                        out.data<OutType>(),
+                        out.size(),
+                        const_param(shape),
+                        const_param(a_strides),
+                        const_param(b_strides),
+                        ndim);
+                  }
                 });
-              } else {
-                auto kernel = cu::binary_g<Op, InType, OutType, IdxT>;
-                auto [num_blocks, block_dims] =
-                    get_launch_args(kernel, out, large);
-                kernel<<<num_blocks, block_dims, 0, stream>>>(
-                    a.data<InType>(),
-                    b.data<InType>(),
-                    out.data<OutType>(),
-                    out.size(),
-                    const_param(shape),
-                    const_param(a_strides),
-                    const_param(b_strides),
-                    ndim);
-              }
-            });
           } else {
-            MLX_SWITCH_BOOL(out.data_size() > UINT32_MAX, LARGE, {
-              using IdxT = std::conditional_t<LARGE, int64_t, uint32_t>;
+            dispatch_bool(out.data_size() > INT32_MAX, [&](auto large) {
+              using IdxT = std::conditional_t<large(), int64_t, uint32_t>;
               auto kernel = cu::binary_ss<Op, InType, OutType, IdxT>;
               if (bopt == BinaryOpType::ScalarVector) {
                 kernel = cu::binary_sv<Op, InType, OutType, IdxT>;
@@ -199,7 +207,7 @@ void binary_op_gpu_inplace(
                 kernel = cu::binary_vv<Op, InType, OutType, IdxT>;
               }
               auto [num_blocks, block_dims] = get_launch_args(
-                  kernel, out.data_size(), out.shape(), out.strides(), LARGE);
+                  kernel, out.data_size(), out.shape(), out.strides(), large());
               kernel<<<num_blocks, block_dims, 0, stream>>>(
                   a.data<InType>(),
                   b.data<InType>(),
@@ -219,20 +227,6 @@ void binary_op_gpu_inplace(
   });
 }
 
-template <typename Op>
-void binary_op_gpu(
-    const std::vector<array>& inputs,
-    std::vector<array>& outputs,
-    std::string_view op,
-    const Stream& s) {
-  auto& a = inputs[0];
-  auto& b = inputs[1];
-  auto bopt = get_binary_op_type(a, b);
-  set_binary_op_output_data(a, b, outputs[0], bopt);
-  set_binary_op_output_data(a, b, outputs[1], bopt);
-  binary_op_gpu_inplace<Op>(inputs, outputs, op, s);
-}
-
 template <typename Op>
 void binary_op_gpu(
     const std::vector<array>& inputs,
@@ -243,8 +237,7 @@ void binary_op_gpu(
   auto& b = inputs[1];
   auto bopt = get_binary_op_type(a, b);
   set_binary_op_output_data(a, b, out, bopt);
-  std::vector<array> outputs{out};
-  binary_op_gpu_inplace<Op>(inputs, outputs, op, s);
+  binary_op_gpu_inplace<Op>(inputs, out, op, s);
 }
 
 #define BINARY_GPU(func)                                                 \
@@ -254,14 +247,6 @@ void binary_op_gpu(
     binary_op_gpu<cu::func>(inputs, out, get_primitive_string(this), s); \
   }
 
-#define BINARY_GPU_MULTI(func)                                               \
-  void func::eval_gpu(                                                       \
-      const std::vector<array>& inputs, std::vector<array>& outputs) {       \
-    nvtx3::scoped_range r(#func "::eval_gpu");                               \
-    auto& s = outputs[0].primitive().stream();                               \
-    binary_op_gpu<cu::func>(inputs, outputs, get_primitive_string(this), s); \
-  }
-
 BINARY_GPU(Add)
 BINARY_GPU(ArcTan2)
 BINARY_GPU(Divide)

mlx/backend/cuda/binary_two.cu

@@ -0,0 +1,258 @@
+// Copyright © 2025 Apple Inc.
+
+#include "mlx/backend/common/binary.h"
+#include "mlx/backend/cuda/device.h"
+#include "mlx/backend/cuda/device/binary_ops.cuh"
+#include "mlx/backend/cuda/device/cucomplex_math.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>
+__global__ void
+binary_ss(const In* a, const In* b, Out* out_a, Out* out_b, IdxT size) {
+  IdxT index = cg::this_grid().thread_rank();
+  if (index < size) {
+    auto out = Op{}(a[0], b[0]);
+    out_a[0] = out[0];
+    out_b[0] = out[1];
+  }
+}
+
+template <typename Op, typename In, typename Out, typename IdxT>
+__global__ void
+binary_sv(const In* a, const In* b, Out* out_a, Out* out_b, IdxT size) {
+  IdxT index = cg::this_grid().thread_rank();
+  if (index < size) {
+    auto out = Op{}(a[0], b[index]);
+    out_a[index] = out[0];
+    out_b[index] = out[1];
+  }
+}
+
+template <typename Op, typename In, typename Out, typename IdxT>
+__global__ void
+binary_vs(const In* a, const In* b, Out* out_a, Out* out_b, IdxT size) {
+  IdxT index = cg::this_grid().thread_rank();
+  if (index < size) {
+    auto out = Op{}(a[index], b[0]);
+    out_a[index] = out[0];
+    out_b[index] = out[1];
+  }
+}
+
+template <typename Op, typename In, typename Out, typename IdxT>
+__global__ void
+binary_vv(const In* a, const In* b, Out* out_a, Out* out_b, IdxT size) {
+  IdxT index = cg::this_grid().thread_rank();
+  if (index < size) {
+    auto out = Op{}(a[index], b[index]);
+    out_a[index] = out[0];
+    out_b[index] = out[1];
+  }
+}
+
+template <typename Op, typename In, typename Out, typename IdxT, int NDIM>
+__global__ void binary_g_nd(
+    const In* a,
+    const In* b,
+    Out* out_a,
+    Out* out_b,
+    IdxT size,
+    const __grid_constant__ cuda::std::array<int32_t, NDIM> shape,
+    const __grid_constant__ cuda::std::array<int64_t, NDIM> a_strides,
+    const __grid_constant__ cuda::std::array<int64_t, NDIM> b_strides) {
+  IdxT index = cg::this_grid().thread_rank();
+  if (index < size) {
+    auto [a_idx, b_idx] = elem_to_loc_nd<NDIM>(
+        index, shape.data(), a_strides.data(), b_strides.data());
+    auto out = Op{}(a[a_idx], b[b_idx]);
+    out_a[index] = out[0];
+    out_b[index] = out[1];
+  }
+}
+
+template <typename Op, typename In, typename Out, typename IdxT>
+__global__ void binary_g(
+    const In* a,
+    const In* b,
+    Out* out_a,
+    Out* out_b,
+    IdxT size,
+    const __grid_constant__ Shape shape,
+    const __grid_constant__ Strides a_strides,
+    const __grid_constant__ Strides b_strides,
+    int ndim) {
+  IdxT index = cg::this_grid().thread_rank();
+  if (index < size) {
+    auto [a_idx, b_idx] = elem_to_loc_4d(
+        index, shape.data(), a_strides.data(), b_strides.data(), ndim);
+    auto out = Op{}(a[a_idx], b[b_idx]);
+    out_a[index] = out[0];
+    out_b[index] = out[1];
+  }
+}
+
+template <typename Op, typename In, typename Out>
+constexpr bool supports_binary_op() {
+  if (std::is_same_v<Op, DivMod>) {
+    return std::is_same_v<In, Out> &&
+        (std::is_integral_v<Out> || is_floating_v<Out>);
+  }
+  return false;
+}
+
+} // namespace cu
+
+template <typename Op>
+void binary_op_gpu_inplace(
+    const std::vector<array>& inputs,
+    std::vector<array>& outputs,
+    std::string_view op,
+    const Stream& s) {
+  assert(inputs.size() > 1);
+  const auto& a = inputs[0];
+  const auto& b = inputs[1];
+  auto& out_a = outputs[0];
+  auto& out_b = outputs[1];
+  auto bopt = get_binary_op_type(a, b);
+  set_binary_op_output_data(a, b, out_a, bopt);
+  set_binary_op_output_data(a, b, out_b, bopt);
+
+  if (out_a.size() == 0) {
+    return;
+  }
+
+  auto& encoder = cu::get_command_encoder(s);
+  encoder.set_input_array(a);
+  encoder.set_input_array(b);
+  encoder.set_output_array(out_a);
+  encoder.set_output_array(out_b);
+  encoder.launch_kernel([&](cudaStream_t stream) {
+    dispatch_all_types(a.dtype(), [&](auto in_type_tag) {
+      dispatch_all_types(out_a.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_binary_op<Op, CTYPE_IN, CTYPE_OUT>()) {
+          using InType = cuda_type_t<CTYPE_IN>;
+          using OutType = cuda_type_t<CTYPE_OUT>;
+
+          auto bopt = get_binary_op_type(a, b);
+          if (bopt == BinaryOpType::General) {
+            dispatch_bool(
+                a.data_size() > INT32_MAX || b.data_size() > INT32_MAX ||
+                    out_a.data_size() > INT32_MAX,
+                [&](auto large) {
+                  using IdxT = std::conditional_t<large(), int64_t, int32_t>;
+                  Shape shape;
+                  std::vector<Strides> strides;
+                  std::tie(shape, strides) =
+                      collapse_contiguous_dims(a, b, out_a);
+                  auto& a_strides = strides[0];
+                  auto& b_strides = strides[1];
+                  int ndim = shape.size();
+                  if (ndim <= 3) {
+                    dispatch_1_2_3(ndim, [&](auto dims_constant) {
+                      auto kernel = cu::binary_g_nd<
+                          Op,
+                          InType,
+                          OutType,
+                          IdxT,
+                          dims_constant()>;
+                      auto [num_blocks, block_dims] =
+                          get_launch_args(kernel, out_a, large());
+                      kernel<<<num_blocks, block_dims, 0, stream>>>(
+                          a.data<InType>(),
+                          b.data<InType>(),
+                          out_a.data<OutType>(),
+                          out_b.data<OutType>(),
+                          out_a.size(),
+                          const_param<dims_constant()>(shape),
+                          const_param<dims_constant()>(a_strides),
+                          const_param<dims_constant()>(b_strides));
+                    });
+                  } else {
+                    auto kernel = cu::binary_g<Op, InType, OutType, IdxT>;
+                    auto [num_blocks, block_dims] =
+                        get_launch_args(kernel, out_a, large());
+                    kernel<<<num_blocks, block_dims, 0, stream>>>(
+                        a.data<InType>(),
+                        b.data<InType>(),
+                        out_a.data<OutType>(),
+                        out_b.data<OutType>(),
+                        out_a.size(),
+                        const_param(shape),
+                        const_param(a_strides),
+                        const_param(b_strides),
+                        ndim);
+                  }
+                });
+          } else {
+            dispatch_bool(out_a.data_size() > INT32_MAX, [&](auto large) {
+              using IdxT = std::conditional_t<large(), int64_t, uint32_t>;
+              auto kernel = cu::binary_ss<Op, InType, OutType, IdxT>;
+              if (bopt == BinaryOpType::ScalarVector) {
+                kernel = cu::binary_sv<Op, InType, OutType, IdxT>;
+              } else if (bopt == BinaryOpType::VectorScalar) {
+                kernel = cu::binary_vs<Op, InType, OutType, IdxT>;
+              } else if (bopt == BinaryOpType::VectorVector) {
+                kernel = cu::binary_vv<Op, InType, OutType, IdxT>;
+              }
+              auto [num_blocks, block_dims] = get_launch_args(
+                  kernel,
+                  out_a.data_size(),
+                  out_a.shape(),
+                  out_a.strides(),
+                  large());
+              kernel<<<num_blocks, block_dims, 0, stream>>>(
+                  a.data<InType>(),
+                  b.data<InType>(),
+                  out_a.data<OutType>(),
+                  out_b.data<OutType>(),
+                  out_a.data_size());
+            });
+          }
+        } else {
+          throw std::runtime_error(fmt::format(
+              "Can not do binary op {} on inputs of {} with result of {}.",
+              op,
+              dtype_to_string(a.dtype()),
+              dtype_to_string(out_a.dtype())));
+        }
+      });
+    });
+  });
+}
+
+template <typename Op>
+void binary_op_gpu(
+    const std::vector<array>& inputs,
+    std::vector<array>& outputs,
+    std::string_view op,
+    const Stream& s) {
+  auto& a = inputs[0];
+  auto& b = inputs[1];
+  auto bopt = get_binary_op_type(a, b);
+  set_binary_op_output_data(a, b, outputs[0], bopt);
+  set_binary_op_output_data(a, b, outputs[1], bopt);
+  binary_op_gpu_inplace<Op>(inputs, outputs, op, s);
+}
+
+void DivMod::eval_gpu(
+    const std::vector<array>& inputs,
+    std::vector<array>& outputs) {
+  nvtx3::scoped_range r("DivMod::eval_gpu");
+  auto& s = outputs[0].primitive().stream();
+  binary_op_gpu<cu::DivMod>(inputs, outputs, get_primitive_string(this), s);
+}
+
+} // namespace mlx::core

mlx/backend/cuda/copy.cu

@@ -24,7 +24,6 @@ void copy_gpu_inplace(
   auto& encoder = cu::get_command_encoder(s);
   encoder.set_input_array(in);
   encoder.set_output_array(out);
-
   if (ctype == CopyType::Scalar || ctype == CopyType::Vector) {
     copy_contiguous(encoder, ctype, in, out, offset_in, offset_out);
     return;

mlx/backend/cuda/copy/copy.cuh

@@ -10,15 +10,6 @@
 
 namespace mlx::core {
 
-#define MLX_SWITCH_COPY_TYPES(in, out, InType, OutType, ...) \
-  MLX_SWITCH_ALL_TYPES(in.dtype(), CTYPE_IN, {               \
-    MLX_SWITCH_ALL_TYPES(out.dtype(), CTYPE_OUT, {           \
-      using InType = cuda_type_t<CTYPE_IN>;                  \
-      using OutType = cuda_type_t<CTYPE_OUT>;                \
-      __VA_ARGS__;                                           \
-    });                                                      \
-  })
-
 void copy_contiguous(
     cu::CommandEncoder& encoder,
     CopyType ctype,

mlx/backend/cuda/copy/copy_contiguous.cu

@@ -36,19 +36,23 @@ void copy_contiguous(
     int64_t in_offset,
     int64_t out_offset) {
   encoder.launch_kernel([&](cudaStream_t stream) {
-    MLX_SWITCH_COPY_TYPES(in, out, InType, OutType, {
-      MLX_SWITCH_BOOL(out.data_size() > UINT32_MAX, LARGE, {
-        using IdxT = std::conditional_t<LARGE, int64_t, uint32_t>;
-        auto kernel = cu::copy_s<InType, OutType, IdxT>;
-        if (ctype == CopyType::Vector) {
-          kernel = cu::copy_v<InType, OutType, IdxT>;
-        }
-        auto [num_blocks, block_dims] = get_launch_args(
-            kernel, out.data_size(), out.shape(), out.strides(), LARGE);
-        kernel<<<num_blocks, block_dims, 0, stream>>>(
-            in.data<InType>() + in_offset,
-            out.data<OutType>() + out_offset,
-            out.data_size());
+    dispatch_all_types(in.dtype(), [&](auto in_type_tag) {
+      dispatch_all_types(out.dtype(), [&](auto out_type_tag) {
+        dispatch_bool(out.data_size() > INT32_MAX, [&](auto large) {
+          using InType = cuda_type_t<MLX_GET_TYPE(in_type_tag)>;
+          using OutType = cuda_type_t<MLX_GET_TYPE(out_type_tag)>;
+          using IdxT = std::conditional_t<large(), int64_t, uint32_t>;
+          auto kernel = cu::copy_s<InType, OutType, IdxT>;
+          if (ctype == CopyType::Vector) {
+            kernel = cu::copy_v<InType, OutType, IdxT>;
+          }
+          auto [num_blocks, block_dims] = get_launch_args(
+              kernel, out.data_size(), out.shape(), out.strides(), large());
+          kernel<<<num_blocks, block_dims, 0, stream>>>(
+              in.data<InType>() + in_offset,
+              out.data<OutType>() + out_offset,
+              out.data_size());
+        });
       });
     });
   });

mlx/backend/cuda/copy/copy_general.cu

@@ -56,37 +56,48 @@ void copy_general(
     const Strides& strides_in,
     const Strides& strides_out) {
   encoder.launch_kernel([&](cudaStream_t stream) {
-    MLX_SWITCH_COPY_TYPES(in, out, InType, OutType, {
-      const InType* in_ptr = in.data<InType>() + offset_in;
-      OutType* out_ptr = out.data<OutType>() + offset_out;
-      bool large = in.data_size() > INT32_MAX || out.data_size() > INT32_MAX;
-      MLX_SWITCH_BOOL(large, LARGE, {
-        using IdxT = std::conditional_t<LARGE, int64_t, int32_t>;
-        int ndim = shape.size();
-        if (ndim <= 3) {
-          MLX_SWITCH_1_2_3(ndim, NDIM, {
-            auto kernel = cu::copy_gg_nd<InType, OutType, IdxT, NDIM>;
-            auto [num_blocks, block_dims] = get_launch_args(kernel, out, large);
-            kernel<<<num_blocks, block_dims, 0, stream>>>(
-                in_ptr,
-                out_ptr,
-                out.size(),
-                const_param<NDIM>(shape),
-                const_param<NDIM>(strides_in),
-                const_param<NDIM>(strides_out));
-          });
-        } else { // ndim >= 4
-          auto kernel = cu::copy_gg<InType, OutType, IdxT>;
-          auto [num_blocks, block_dims] = get_launch_args(kernel, out, large);
-          kernel<<<num_blocks, block_dims, 0, stream>>>(
-              in_ptr,
-              out_ptr,
-              out.size(),
-              const_param(shape),
-              const_param(strides_in),
-              const_param(strides_out),
-              ndim);
-        }
+    dispatch_all_types(in.dtype(), [&](auto in_type_tag) {
+      dispatch_all_types(out.dtype(), [&](auto out_type_tag) {
+        dispatch_bool(
+            in.data_size() > INT32_MAX || out.data_size() > INT32_MAX,
+            [&](auto large) {
+              using InType = cuda_type_t<MLX_GET_TYPE(in_type_tag)>;
+              using OutType = cuda_type_t<MLX_GET_TYPE(out_type_tag)>;
+              using IdxT = std::conditional_t<large(), int64_t, int32_t>;
+              const InType* in_ptr = in.data<InType>() + offset_in;
+              OutType* out_ptr = out.data<OutType>() + offset_out;
+              int ndim = shape.size();
+              size_t data_size = 1;
+              for (auto& s : shape)
+                data_size *= s;
+              if (ndim <= 3) {
+                dispatch_1_2_3(ndim, [&](auto ndim_constant) {
+                  auto kernel =
+                      cu::copy_gg_nd<InType, OutType, IdxT, ndim_constant()>;
+                  auto [num_blocks, block_dims] = get_launch_args(
+                      kernel, data_size, shape, out.strides(), large());
+                  kernel<<<num_blocks, block_dims, 0, stream>>>(
+                      in_ptr,
+                      out_ptr,
+                      data_size,
+                      const_param<ndim_constant()>(shape),
+                      const_param<ndim_constant()>(strides_in),
+                      const_param<ndim_constant()>(strides_out));
+                });
+              } else { // ndim >= 4
+                auto kernel = cu::copy_gg<InType, OutType, IdxT>;
+                auto [num_blocks, block_dims] = get_launch_args(
+                    kernel, data_size, shape, out.strides(), large());
+                kernel<<<num_blocks, block_dims, 0, stream>>>(
+                    in_ptr,
+                    out_ptr,
+                    data_size,
+                    const_param(shape),
+                    const_param(strides_in),
+                    const_param(strides_out),
+                    ndim);
+              }
+            });
       });
     });
   });

mlx/backend/cuda/copy/copy_general_dynamic.cu

@@ -62,41 +62,52 @@ void copy_general_dynamic(
     const array& dynamic_offset_in,
     const array& dynamic_offset_out) {
   encoder.launch_kernel([&](cudaStream_t stream) {
-    MLX_SWITCH_COPY_TYPES(in, out, InType, OutType, {
-      const InType* in_ptr = in.data<InType>() + offset_in;
-      OutType* out_ptr = out.data<OutType>() + offset_out;
-      bool large = in.data_size() > INT32_MAX || out.data_size() > INT32_MAX;
-      MLX_SWITCH_BOOL(large, LARGE, {
-        using IdxT = std::conditional_t<LARGE, int64_t, int32_t>;
-        int ndim = shape.size();
-        if (ndim <= 3) {
-          MLX_SWITCH_1_2_3(ndim, NDIM, {
-            auto kernel = cu::copy_gg_dynamic_nd<InType, OutType, IdxT, NDIM>;
-            auto [num_blocks, block_dims] = get_launch_args(kernel, out, large);
-            kernel<<<num_blocks, block_dims, 0, stream>>>(
-                in_ptr,
-                out_ptr,
-                out.size(),
-                const_param<NDIM>(shape),
-                const_param<NDIM>(strides_in),
-                const_param<NDIM>(strides_out),
-                dynamic_offset_in.data<int64_t>(),
-                dynamic_offset_out.data<int64_t>());
-          });
-        } else { // ndim >= 4
-          auto kernel = cu::copy_gg_dynamic<InType, OutType, IdxT>;
-          auto [num_blocks, block_dims] = get_launch_args(kernel, out, large);
-          kernel<<<num_blocks, block_dims, 0, stream>>>(
-              in_ptr,
-              out_ptr,
-              out.size(),
-              const_param(shape),
-              const_param(strides_in),
-              const_param(strides_out),
-              ndim,
-              dynamic_offset_in.data<int64_t>(),
-              dynamic_offset_out.data<int64_t>());
-        }
+    dispatch_all_types(in.dtype(), [&](auto in_type_tag) {
+      dispatch_all_types(out.dtype(), [&](auto out_type_tag) {
+        dispatch_bool(
+            in.data_size() > INT32_MAX || out.data_size() > INT32_MAX,
+            [&](auto large) {
+              using InType = cuda_type_t<MLX_GET_TYPE(in_type_tag)>;
+              using OutType = cuda_type_t<MLX_GET_TYPE(out_type_tag)>;
+              using IdxT = std::conditional_t<large(), int64_t, int32_t>;
+              const InType* in_ptr = in.data<InType>() + offset_in;
+              OutType* out_ptr = out.data<OutType>() + offset_out;
+              int ndim = shape.size();
+              if (ndim <= 3) {
+                dispatch_1_2_3(ndim, [&](auto dims_constant) {
+                  auto kernel = cu::copy_gg_dynamic_nd<
+                      InType,
+                      OutType,
+                      IdxT,
+                      dims_constant()>;
+                  auto [num_blocks, block_dims] =
+                      get_launch_args(kernel, out, large());
+                  kernel<<<num_blocks, block_dims, 0, stream>>>(
+                      in_ptr,
+                      out_ptr,
+                      out.size(),
+                      const_param<dims_constant()>(shape),
+                      const_param<dims_constant()>(strides_in),
+                      const_param<dims_constant()>(strides_out),
+                      dynamic_offset_in.data<int64_t>(),
+                      dynamic_offset_out.data<int64_t>());
+                });
+              } else { // ndim >= 4
+                auto kernel = cu::copy_gg_dynamic<InType, OutType, IdxT>;
+                auto [num_blocks, block_dims] =
+                    get_launch_args(kernel, out, large());
+                kernel<<<num_blocks, block_dims, 0, stream>>>(
+                    in_ptr,
+                    out_ptr,
+                    out.size(),
+                    const_param(shape),
+                    const_param(strides_in),
+                    const_param(strides_out),
+                    ndim,
+                    dynamic_offset_in.data<int64_t>(),
+                    dynamic_offset_out.data<int64_t>());
+              }
+            });
       });
     });
   });

mlx/backend/cuda/copy/copy_general_input.cu

@@ -51,35 +51,43 @@ void copy_general_input(
     const Shape& shape,
     const Strides& strides_in) {
   encoder.launch_kernel([&](cudaStream_t stream) {
-    MLX_SWITCH_COPY_TYPES(in, out, InType, OutType, {
-      const InType* in_ptr = in.data<InType>() + offset_in;
-      OutType* out_ptr = out.data<OutType>() + offset_out;
-      bool large = in.data_size() > INT32_MAX || out.data_size() > INT32_MAX;
-      MLX_SWITCH_BOOL(large, LARGE, {
-        using IdxT = std::conditional_t<LARGE, int64_t, int32_t>;
-        int ndim = shape.size();
-        if (ndim <= 3) {
-          MLX_SWITCH_1_2_3(ndim, NDIM, {
-            auto kernel = cu::copy_g_nd<InType, OutType, IdxT, NDIM>;
-            auto [num_blocks, block_dims] = get_launch_args(kernel, out, large);
-            kernel<<<num_blocks, block_dims, 0, stream>>>(
-                in_ptr,
-                out_ptr,
-                out.size(),
-                const_param<NDIM>(shape),
-                const_param<NDIM>(strides_in));
-          });
-        } else { // ndim >= 4
-          auto kernel = cu::copy_g<InType, OutType, IdxT>;
-          auto [num_blocks, block_dims] = get_launch_args(kernel, out, large);
-          kernel<<<num_blocks, block_dims, 0, stream>>>(
-              in_ptr,
-              out_ptr,
-              out.size(),
-              const_param(shape),
-              const_param(strides_in),
-              ndim);
-        }
+    dispatch_all_types(in.dtype(), [&](auto in_type_tag) {
+      dispatch_all_types(out.dtype(), [&](auto out_type_tag) {
+        dispatch_bool(
+            in.data_size() > INT32_MAX || out.data_size() > INT32_MAX,
+            [&](auto large) {
+              using InType = cuda_type_t<MLX_GET_TYPE(in_type_tag)>;
+              using OutType = cuda_type_t<MLX_GET_TYPE(out_type_tag)>;
+              using IdxT = std::conditional_t<large(), int64_t, int32_t>;
+              const InType* in_ptr = in.data<InType>() + offset_in;
+              OutType* out_ptr = out.data<OutType>() + offset_out;
+              int ndim = shape.size();
+              if (ndim <= 3) {
+                dispatch_1_2_3(ndim, [&](auto dims_constant) {
+                  auto kernel =
+                      cu::copy_g_nd<InType, OutType, IdxT, dims_constant()>;
+                  auto [num_blocks, block_dims] =
+                      get_launch_args(kernel, out, large());
+                  kernel<<<num_blocks, block_dims, 0, stream>>>(
+                      in_ptr,
+                      out_ptr,
+                      out.size(),
+                      const_param<dims_constant()>(shape),
+                      const_param<dims_constant()>(strides_in));
+                });
+              } else { // ndim >= 4
+                auto kernel = cu::copy_g<InType, OutType, IdxT>;
+                auto [num_blocks, block_dims] =
+                    get_launch_args(kernel, out, large());
+                kernel<<<num_blocks, block_dims, 0, stream>>>(
+                    in_ptr,
+                    out_ptr,
+                    out.size(),
+                    const_param(shape),
+                    const_param(strides_in),
+                    ndim);
+              }
+            });
       });
     });
   });

mlx/backend/cuda/device.cpp

@@ -6,6 +6,7 @@
 
 #include <fmt/format.h>
 #include <nvtx3/nvtx3.hpp>
+#include <future>
 
 namespace mlx::core {
 
@@ -107,6 +108,16 @@ void CommandEncoder::commit() {
   worker_.commit(stream_.last_cuda_stream());
 }
 
+void CommandEncoder::synchronize() {
+  stream().synchronize();
+  auto p = std::make_shared<std::promise<void>>();
+  std::future<void> f = p->get_future();
+  add_completed_handler([p = std::move(p)]() { p->set_value(); });
+  worker_.end_batch();
+  commit();
+  f.wait();
+}
+
 Device& device(mlx::core::Device device) {
   static std::unordered_map<int, Device> devices;
   auto it = devices.find(device.index);

mlx/backend/cuda/device.h

@@ -123,6 +123,9 @@ class CommandEncoder {
     return has_gpu_work_;
   }
 
+  // Wait until kernels and completion handlers are finished
+  void synchronize();
+
  private:
   Device& device_;
   DeviceStream& stream_;

mlx/backend/cuda/device/binary_ops.cuh

@@ -22,7 +22,7 @@ struct FloorDivide {
     if constexpr (cuda::std::is_integral_v<T>) {
       return x / y;
     } else {
-      return trunc(x / y);
+      return truncf(x / y);
     }
   }
 };
@@ -132,7 +132,7 @@ struct LogAddExp {
           cuda::std::numeric_limits<float>::quiet_NaN(),
           cuda::std::numeric_limits<float>::quiet_NaN()};
     }
-    constexpr float inf = cuda::std::numeric_limits<float>::infinity();
+    float inf = cuda::std::numeric_limits<float>::infinity();
     auto maxval = x > y ? x : y;
     auto minval = x < y ? x : y;
     if (cuCrealf(minval) == -inf || cuCrealf(maxval) == inf)

mlx/backend/cuda/device/config.h

@@ -5,7 +5,7 @@
 #pragma once
 
 // The maximum dimensions of shape/strides passed as kernel parameters.
-#define MAX_NDIM 8
+#define MAX_NDIM 10
 
 // All existing NVIDIA hardware has a fixed 32 warp size. Though a built-in
 // warpSize variable exists, using it would prevent compile-time optimizations.

mlx/backend/cuda/device/unary_ops.cuh

@@ -27,6 +27,8 @@ struct ArcCos {
   __device__ T operator()(T x) {
     return acos(x);
   }
+
+  __device__ cuComplex operator()(cuComplex x);
 };
 
 struct ArcCosh {
@@ -41,6 +43,8 @@ struct ArcSin {
   __device__ T operator()(T x) {
     return asin(x);
   }
+
+  __device__ cuComplex operator()(cuComplex x);
 };
 
 struct ArcSinh {
@@ -55,6 +59,8 @@ struct ArcTan {
   __device__ T operator()(T x) {
     return atan(x);
   }
+
+  __device__ cuComplex operator()(cuComplex x);
 };
 
 struct ArcTanh {
@@ -261,13 +267,6 @@ struct Round {
   }
 };
 
-struct Rsqrt {
-  template <typename T>
-  __device__ T operator()(T x) {
-    return rsqrt(x);
-  }
-};
-
 struct Sigmoid {
   template <typename T>
   __device__ T operator()(T x) {
@@ -333,6 +332,29 @@ struct Sqrt {
   __device__ T operator()(T x) {
     return sqrt(x);
   }
+
+  __device__ cuComplex operator()(cuComplex x) {
+    auto xr = cuCrealf(x);
+    auto xi = cuCimagf(x);
+    if (xr == 0.0f && xi == 0.0f) {
+      return {0.0f, 0.0f};
+    }
+    auto r = cuCrealf(Abs{}(x));
+    auto a = sqrt((r + xr) / 2.0f);
+    auto b_abs = sqrt((r - xr) / 2.0f);
+    auto b = copysign(b_abs, xi);
+    return {a, b};
+  }
+};
+
+struct Rsqrt {
+  template <typename T>
+  __device__ T operator()(T x) {
+    return rsqrt(x);
+  }
+  __device__ cuComplex operator()(cuComplex x) {
+    return 1.0f / Sqrt{}(x);
+  }
 };
 
 struct Tan {
@@ -365,4 +387,22 @@ struct Tanh {
   }
 };
 
+__device__ cuComplex ArcCos::operator()(cuComplex x) {
+  auto i = cuComplex{0.0, 1.0};
+  auto y = Log{}(x + i * Sqrt{}(1.0 - x * x));
+  return {cuCimagf(y), -cuCrealf(y)};
+};
+
+__device__ cuComplex ArcSin::operator()(cuComplex x) {
+  auto i = cuComplex{0.0f, 1.0f};
+  auto y = Log{}(i * x + Sqrt{}(1.0f - x * x));
+  return {cuCimagf(y), -cuCrealf(y)};
+};
+
+__device__ cuComplex ArcTan::operator()(cuComplex x) {
+  auto i = cuComplex{0.0f, 1.0f};
+  auto ix = i * x;
+  return (1.0f / cuComplex{0.0f, 2.0f}) * Log{}((1.0f + ix) / (1.0f - ix));
+};
+
 } // namespace mlx::core::cu

mlx/backend/cuda/device/utils.cuh

@@ -155,8 +155,8 @@ inline __host__ __device__ cuda::std::tuple<IdxT, IdxT> elem_to_loc_nd(
 #pragma unroll
   for (int i = NDIM - 1; i >= 0; --i) {
     int dim_idx = elem % shape[i];
-    a_loc += dim_idx * a_strides[i];
-    b_loc += dim_idx * b_strides[i];
+    a_loc += dim_idx * IdxT(a_strides[i]);
+    b_loc += dim_idx * IdxT(b_strides[i]);
     elem /= shape[i];
   }
   return cuda::std::make_tuple(a_loc, b_loc);
@@ -175,9 +175,9 @@ inline __host__ __device__ cuda::std::tuple<IdxT, IdxT, IdxT> elem_to_loc_nd(
 #pragma unroll
   for (int i = NDIM - 1; i >= 0; --i) {
     int dim_idx = elem % shape[i];
-    a_loc += dim_idx * a_strides[i];
-    b_loc += dim_idx * b_strides[i];
-    c_loc += dim_idx * c_strides[i];
+    a_loc += dim_idx * IdxT(a_strides[i]);
+    b_loc += dim_idx * IdxT(b_strides[i]);
+    c_loc += dim_idx * IdxT(c_strides[i]);
     elem /= shape[i];
   }
   return cuda::std::make_tuple(a_loc, b_loc, c_loc);
@@ -206,8 +206,8 @@ inline __host__ __device__ cuda::std::tuple<IdxT, IdxT> elem_to_loc_4d(
   IdxT b_loc = 0;
   for (int i = ndim - 1; i >= 0; --i) {
     int dim_idx = elem % shape[i];
-    a_loc += dim_idx * a_strides[i];
-    b_loc += dim_idx * b_strides[i];
+    a_loc += dim_idx * IdxT(a_strides[i]);
+    b_loc += dim_idx * IdxT(b_strides[i]);
     elem /= shape[i];
   }
   return cuda::std::make_tuple(a_loc, b_loc);
@@ -226,9 +226,9 @@ inline __host__ __device__ cuda::std::tuple<IdxT, IdxT, IdxT> elem_to_loc_4d(
   IdxT c_loc = 0;
   for (int i = ndim - 1; i >= 0; --i) {
     int dim_idx = elem % shape[i];
-    a_loc += dim_idx * a_strides[i];
-    b_loc += dim_idx * b_strides[i];
-    c_loc += dim_idx * c_strides[i];
+    a_loc += dim_idx * IdxT(a_strides[i]);
+    b_loc += dim_idx * IdxT(b_strides[i]);
+    c_loc += dim_idx * IdxT(c_strides[i]);
     elem /= shape[i];
   }
   return cuda::std::make_tuple(a_loc, b_loc, c_loc);

mlx/backend/cuda/eval.cpp

@@ -62,7 +62,7 @@ void finalize(Stream s) {
 
 void synchronize(Stream s) {
   nvtx3::scoped_range r("gpu::synchronize");
-  cu::get_stream(s).synchronize();
+  cu::get_command_encoder(s).synchronize();
 }
 
 } // namespace mlx::core::gpu

mlx/backend/cuda/jit_module.cpp

@@ -37,36 +37,46 @@ void check_cu_error(const char* name, CUresult err) {
 }
 
 // Return the location of the CUDA toolkit.
-const char* cuda_home() {
-  const char* home = std::getenv("CUDA_HOME");
-  if (home) {
-    return home;
-  }
-  home = std::getenv("CUDA_PATH");
-  if (home) {
-    return home;
-  }
+const std::string& cuda_home() {
+  static std::string home = []() -> std::string {
+    const char* home = std::getenv("CUDA_HOME");
+    if (home) {
+      return home;
+    }
+    home = std::getenv("CUDA_PATH");
+    if (home) {
+      return home;
+    }
 #if defined(__linux__)
-  home = "/usr/local/cuda";
-  if (std::filesystem::exists(home)) {
-    return home;
-  }
+    home = "/usr/local/cuda";
+    if (std::filesystem::exists(home)) {
+      return home;
+    }
 #endif
-  throw std::runtime_error(
-      "Environment variable CUDA_HOME or CUDA_PATH is not set.");
+    throw std::runtime_error(
+        "Environment variable CUDA_HOME or CUDA_PATH is not set.");
+  }();
+  return home;
 }
 
 // Get the cache directory for storing compiled results.
-bool get_ptx_cache_dir(std::filesystem::path* result) {
-  auto path = std::filesystem::temp_directory_path() / "mlx" / "ptx";
-  if (!std::filesystem::is_directory(path)) {
-    std::error_code error;
-    if (!std::filesystem::create_directories(path, error)) {
-      return false;
+const std::filesystem::path& ptx_cache_dir() {
+  static std::filesystem::path cache = []() -> std::filesystem::path {
+    std::filesystem::path cache;
+    if (auto c = std::getenv("MLX_PTX_CACHE"); c) {
+      cache = c;
+    } else {
+      cache = std::filesystem::temp_directory_path() / "mlx" / "ptx";
     }
-  }
-  *result = path;
-  return true;
+    if (!std::filesystem::exists(cache)) {
+      std::error_code error;
+      if (!std::filesystem::create_directories(cache, error)) {
+        return std::filesystem::path();
+      }
+    }
+    return cache;
+  }();
+  return cache;
 }
 
 // Try to read the cached |ptx| and |ptx_kernels| from |cache_dir|.
@@ -75,6 +85,10 @@ bool read_cached_ptx(
     const std::string& module_name,
     std::vector<char>* ptx,
     std::vector<std::pair<std::string, std::string>>* ptx_kernels) {
+  if (cache_dir.empty()) {
+    return false;
+  }
+
   auto ptx_path = cache_dir / (module_name + ".ptx");
   std::error_code error;
   auto ptx_size = std::filesystem::file_size(ptx_path, error);
@@ -105,6 +119,10 @@ void write_cached_ptx(
     const std::string& module_name,
     const std::vector<char>& ptx,
     const std::vector<std::pair<std::string, std::string>>& ptx_kernels) {
+  if (cache_dir.empty()) {
+    return;
+  }
+
   std::ofstream ptx_file(cache_dir / (module_name + ".ptx"), std::ios::binary);
   if (!ptx.empty()) {
     ptx_file.write(&ptx.front(), ptx.size());
@@ -184,11 +202,9 @@ JitModule::JitModule(
     const std::string& module_name,
     const KernelBuilder& builder) {
   // Check cache.
-  std::filesystem::path cache_dir;
   std::vector<char> ptx;
   std::vector<std::pair<std::string, std::string>> ptx_kernels;
-  if (!get_ptx_cache_dir(&cache_dir) ||
-      !read_cached_ptx(cache_dir, module_name, &ptx, &ptx_kernels)) {
+  if (!read_cached_ptx(ptx_cache_dir(), module_name, &ptx, &ptx_kernels)) {
     // Create program.
     auto [source_code, kernel_names] = builder();
     nvrtcProgram prog;
@@ -246,7 +262,7 @@ JitModule::JitModule(
     } else {
       CHECK_NVRTC_ERROR(nvrtcGetPTX(prog, ptx.data()));
     }
-    write_cached_ptx(cache_dir, module_name, ptx, ptx_kernels);
+    write_cached_ptx(ptx_cache_dir(), module_name, ptx, ptx_kernels);
   }
 
   // Load module.

mlx/backend/cuda/kernel_utils.cuh

@@ -6,6 +6,8 @@
 
 #pragma once
 
+#include <type_traits>
+
 #include "mlx/array.h"
 #include "mlx/backend/cuda/device/utils.cuh"
 
@@ -17,60 +19,46 @@
 
 namespace mlx::core {
 
-// Convert a number between 1~3 to constexpr.
-#define MLX_SWITCH_1_2_3(N, NDIM, ...) \
-  switch (N) {                         \
-    case 1: {                          \
-      constexpr int NDIM = 1;          \
-      __VA_ARGS__;                     \
-      break;                           \
-    }                                  \
-    case 2: {                          \
-      constexpr int NDIM = 2;          \
-      __VA_ARGS__;                     \
-      break;                           \
-    }                                  \
-    case 3: {                          \
-      constexpr int NDIM = 3;          \
-      __VA_ARGS__;                     \
-      break;                           \
-    }                                  \
+template <typename F>
+void dispatch_1_2_3(int n, F&& f) {
+  switch (n) {
+    case 1:
+      f(std::integral_constant<int, 1>{});
+      break;
+    case 2:
+      f(std::integral_constant<int, 2>{});
+      break;
+    case 3:
+      f(std::integral_constant<int, 3>{});
+      break;
   }
+}
 
-// Like MLX_SWITCH_ALL_TYPES but for booleans.
-#define MLX_SWITCH_BOOL(BOOL, BOOL_ALIAS, ...) \
-  if (BOOL) {                                  \
-    constexpr bool BOOL_ALIAS = true;          \
-    __VA_ARGS__;                               \
-  } else {                                     \
-    constexpr bool BOOL_ALIAS = false;         \
-    __VA_ARGS__;                               \
+template <typename F>
+void dispatch_bool(bool v, F&& f) {
+  if (v) {
+    f(std::true_type{});
+  } else {
+    f(std::false_type{});
   }
+}
 
-// Convert a block_dim to constexpr between WARP_SIZE and WARP_SIZE ^ 2.
-#define MLX_SWITCH_BLOCK_DIM(NUM_THREADS, BLOCK_DIM, ...)   \
-  {                                                         \
-    uint32_t _num_threads = NUM_THREADS;                    \
-    if (_num_threads <= WARP_SIZE) {                        \
-      constexpr uint32_t BLOCK_DIM = WARP_SIZE;             \
-      __VA_ARGS__;                                          \
-    } else if (_num_threads <= WARP_SIZE * 2) {             \
-      constexpr uint32_t BLOCK_DIM = WARP_SIZE * 2;         \
-      __VA_ARGS__;                                          \
-    } else if (_num_threads <= WARP_SIZE * 4) {             \
-      constexpr uint32_t BLOCK_DIM = WARP_SIZE * 4;         \
-      __VA_ARGS__;                                          \
-    } else if (_num_threads <= WARP_SIZE * 8) {             \
-      constexpr uint32_t BLOCK_DIM = WARP_SIZE * 8;         \
-      __VA_ARGS__;                                          \
-    } else if (_num_threads <= WARP_SIZE * 16) {            \
-      constexpr uint32_t BLOCK_DIM = WARP_SIZE * 16;        \
-      __VA_ARGS__;                                          \
-    } else {                                                \
-      constexpr uint32_t BLOCK_DIM = WARP_SIZE * WARP_SIZE; \
-      __VA_ARGS__;                                          \
-    }                                                       \
+template <typename F>
+void dispatch_block_dim(int threads, F&& f) {
+  if (threads <= WARP_SIZE) {
+    f(std::integral_constant<int, WARP_SIZE>{});
+  } else if (threads <= WARP_SIZE * 2) {
+    f(std::integral_constant<int, WARP_SIZE * 2>{});
+  } else if (threads <= WARP_SIZE * 4) {
+    f(std::integral_constant<int, WARP_SIZE * 4>{});
+  } else if (threads <= WARP_SIZE * 8) {
+    f(std::integral_constant<int, WARP_SIZE * 8>{});
+  } else if (threads <= WARP_SIZE * 16) {
+    f(std::integral_constant<int, WARP_SIZE * 16>{});
+  } else {
+    f(std::integral_constant<int, WARP_SIZE * 32>{});
   }
+}
 
 // Maps CPU types to CUDA types.
 template <typename T>

mlx/backend/cuda/layer_norm.cu

@@ -259,21 +259,22 @@ void LayerNorm::eval_gpu(
   encoder.set_input_array(b);
   encoder.set_output_array(out);
   encoder.launch_kernel([&](cudaStream_t stream) {
-    MLX_SWITCH_FLOAT_TYPES_CHECKED(out.dtype(), "layernorm", CTYPE, {
-      using DataType = cuda_type_t<CTYPE>;
+    dispatch_float_types(out.dtype(), "layernorm", [&](auto type_tag) {
       constexpr uint32_t N_READS = 4;
-      MLX_SWITCH_BLOCK_DIM(cuda::ceil_div(axis_size, N_READS), BLOCK_DIM, {
-        auto kernel = cu::layer_norm<DataType, BLOCK_DIM, N_READS>;
-        kernel<<<n_rows, BLOCK_DIM, 0, stream>>>(
-            x.data<DataType>(),
-            w.data<DataType>(),
-            b.data<DataType>(),
-            out.data<DataType>(),
-            eps_,
-            axis_size,
-            w_stride,
-            b_stride);
-      });
+      dispatch_block_dim(
+          cuda::ceil_div(axis_size, N_READS), [&](auto block_dim) {
+            using DataType = cuda_type_t<MLX_GET_TYPE(type_tag)>;
+            auto kernel = cu::layer_norm<DataType, block_dim(), N_READS>;
+            kernel<<<n_rows, block_dim(), 0, stream>>>(
+                x.data<DataType>(),
+                w.data<DataType>(),
+                b.data<DataType>(),
+                out.data<DataType>(),
+                eps_,
+                axis_size,
+                w_stride,
+                b_stride);
+          });
     });
   });
 }
@@ -341,8 +342,6 @@ void LayerNormVJP::eval_gpu(
       encoder.add_temporary(gw_temp);
     }
   }
-  gw.set_data(allocator::malloc(gw.nbytes()));
-  gb.set_data(allocator::malloc(gb.nbytes()));
 
   // Finish with the gradient for b in case we had a b.
   if (gb.ndim() == 1 && gb.size() == axis_size) {
@@ -357,22 +356,27 @@ void LayerNormVJP::eval_gpu(
   encoder.set_output_array(gx);
   encoder.set_output_array(gw_temp);
   encoder.launch_kernel([&, x = x, g = g](cudaStream_t stream) {
-    MLX_SWITCH_FLOAT_TYPES_CHECKED(gx.dtype(), "layernorm_vjp", CTYPE, {
-      using DataType = cuda_type_t<CTYPE>;
-      constexpr int N_READS = 4;
-      MLX_SWITCH_BOOL(has_w, HAS_W, {
-        MLX_SWITCH_BLOCK_DIM(cuda::ceil_div(axis_size, N_READS), BLOCK_DIM, {
-          auto kernel = cu::layer_norm_vjp<DataType, HAS_W, BLOCK_DIM, N_READS>;
-          kernel<<<n_rows, BLOCK_DIM, 0, stream>>>(
-              x.data<DataType>(),
-              w.data<DataType>(),
-              g.data<DataType>(),
-              gx.data<DataType>(),
-              gw_temp.data<DataType>(),
-              eps_,
-              axis_size,
-              w_stride);
-        });
+    dispatch_float_types(gx.dtype(), "layernorm_vjp", [&](auto type_tag) {
+      dispatch_bool(has_w, [&](auto has_w_constant) {
+        constexpr int N_READS = 4;
+        dispatch_block_dim(
+            cuda::ceil_div(axis_size, N_READS), [&](auto block_dim) {
+              using DataType = cuda_type_t<MLX_GET_TYPE(type_tag)>;
+              auto kernel = cu::layer_norm_vjp<
+                  DataType,
+                  has_w_constant(),
+                  block_dim(),
+                  N_READS>;
+              kernel<<<n_rows, block_dim(), 0, stream>>>(
+                  x.data<DataType>(),
+                  w.data<DataType>(),
+                  g.data<DataType>(),
+                  gx.data<DataType>(),
+                  gw_temp.data<DataType>(),
+                  eps_,
+                  axis_size,
+                  w_stride);
+            });
       });
     });
   });

mlx/backend/cuda/logsumexp.cu

@@ -144,14 +144,15 @@ void LogSumExp::eval_gpu(const std::vector<array>& inputs, array& out) {
   encoder.set_input_array(in);
   encoder.set_output_array(out);
   encoder.launch_kernel([&](cudaStream_t stream) {
-    MLX_SWITCH_FLOAT_TYPES_CHECKED(out.dtype(), "logsumexp", CTYPE, {
-      using DataType = cuda_type_t<CTYPE>;
+    dispatch_float_types(out.dtype(), "logsumexp", [&](auto type_tag) {
       constexpr int N_READS = 4;
-      MLX_SWITCH_BLOCK_DIM(cuda::ceil_div(axis_size, N_READS), BLOCK_DIM, {
-        auto kernel = cu::logsumexp<DataType, float, BLOCK_DIM, N_READS>;
-        kernel<<<n_rows, BLOCK_DIM, 0, stream>>>(
-            in.data<DataType>(), out.data<DataType>(), axis_size);
-      });
+      dispatch_block_dim(
+          cuda::ceil_div(axis_size, N_READS), [&](auto block_dim) {
+            using DataType = cuda_type_t<MLX_GET_TYPE(type_tag)>;
+            auto kernel = cu::logsumexp<DataType, float, block_dim(), N_READS>;
+            kernel<<<n_rows, block_dim(), 0, stream>>>(
+                in.data<DataType>(), out.data<DataType>(), axis_size);
+          });
     });
   });
 }

mlx/backend/cuda/matmul.cpp

@@ -162,11 +162,15 @@ class MatMul {
       }
     }
 
-    array workspace(
-        allocator::malloc(heuristic_.workspaceSize),
-        {static_cast<int>(heuristic_.workspaceSize)},
-        int8);
-    encoder.add_temporary(workspace);
+    void* workspace_ptr = nullptr;
+    if (heuristic_.workspaceSize > 0) {
+      array workspace(
+          allocator::malloc(heuristic_.workspaceSize),
+          {static_cast<int>(heuristic_.workspaceSize)},
+          int8);
+      encoder.add_temporary(workspace);
+      workspace_ptr = workspace.data<void>();
+    }
 
     encoder.launch_kernel([&](cudaStream_t stream) {
       CHECK_CUBLAS_ERROR(cublasLtMatmul(
@@ -183,8 +187,8 @@ class MatMul {
           out,
           out_desc_,
           &heuristic_.algo,
-          workspace.data<void>(),
-          workspace.nbytes(),
+          workspace_ptr,
+          heuristic_.workspaceSize,
           stream));
     });
   }
@@ -358,9 +362,18 @@ void Matmul::eval_gpu(const std::vector<array>& inputs, array& out) {
       a_batch_strides.back(),
       b_batch_strides.back());
 
+  encoder.set_input_array(a);
+  encoder.set_input_array(b);
+  encoder.set_output_array(out);
+  auto nbatch = batch_count / batch_shape.back();
+  if (nbatch == 1) {
+    matmul.run(encoder, out.data<int8_t>(), a.data<int8_t>(), b.data<int8_t>());
+    return;
+  }
+
   ContiguousIterator a_it(batch_shape, a_batch_strides, batch_shape.size() - 1);
   ContiguousIterator b_it(batch_shape, b_batch_strides, batch_shape.size() - 1);
-  for (size_t i = 0; i < batch_count / batch_shape.back(); ++i) {
+  for (size_t i = 0; i < nbatch; ++i) {
     matmul.run(
         encoder,
         out.data<int8_t>() + out.itemsize() * i * batch_shape.back() * M * N,
@@ -444,10 +457,28 @@ void AddMM::eval_gpu(const std::vector<array>& inputs, array& out) {
       b_batch_strides.back(),
       c_batch_strides.back());
 
+  encoder.set_input_array(a);
+  encoder.set_input_array(b);
+  encoder.set_input_array(c);
+  encoder.set_output_array(out);
+
+  auto nbatch = batch_count / batch_shape.back();
+  if (nbatch == 1) {
+    matmul.run(
+        encoder,
+        out.data<int8_t>(),
+        a.data<int8_t>(),
+        b.data<int8_t>(),
+        c.data<int8_t>(),
+        alpha_,
+        beta_);
+    return;
+  }
+
   ContiguousIterator a_it(batch_shape, a_batch_strides, batch_shape.size() - 1);
   ContiguousIterator b_it(batch_shape, b_batch_strides, batch_shape.size() - 1);
   ContiguousIterator c_it(batch_shape, c_batch_strides, batch_shape.size() - 1);
-  for (size_t i = 0; i < batch_count / batch_shape.back(); ++i) {
+  for (size_t i = 0; i < nbatch; ++i) {
     matmul.run(
         encoder,
         out.data<int8_t>() + out.itemsize() * i * batch_shape.back() * M * N,

mlx/backend/cuda/primitives.cu

@@ -28,7 +28,8 @@ void Arange::eval_gpu(const std::vector<array>& inputs, array& out) {
   auto& encoder = cu::get_command_encoder(s);
   encoder.set_output_array(out);
   encoder.launch_kernel([&, this](cudaStream_t stream) {
-    MLX_SWITCH_INT_FLOAT_TYPES_CHECKED(out.dtype(), "Arange", CTYPE, {
+    dispatch_int_float_types(out.dtype(), "Arange", [&](auto type_tag) {
+      using CTYPE = MLX_GET_TYPE(type_tag);
       using OutType = cuda_type_t<CTYPE>;
       CTYPE step =
           static_cast<CTYPE>(start_ + step_) - static_cast<CTYPE>(start_);
@@ -71,10 +72,8 @@ bool fast::ScaledDotProductAttention::use_fallback(
     throw std::runtime_error(#func " has no CUDA implementation.");   \
   }
 
-NO_GPU(ArgPartition)
 NO_GPU(BlockMaskedMM)
 NO_GPU(Convolution)
-NO_GPU_MULTI(DivMod)
 NO_GPU(DynamicSlice)
 NO_GPU(DynamicSliceUpdate)
 NO_GPU(FFT)
@@ -83,7 +82,6 @@ NO_GPU(GatherQMM)
 NO_GPU(Hadamard)
 NO_GPU(Load)
 NO_GPU_MULTI(LUF)
-NO_GPU(Partition)
 NO_GPU_MULTI(QRF)
 NO_GPU(QuantizedMatmul)
 NO_GPU(Scan)

mlx/backend/cuda/reduce.cu

@@ -21,28 +21,11 @@ void Reduce::eval_gpu(const std::vector<array>& inputs, array& out) {
   assert(!axes_.empty());
   assert(out.size() != in.size());
 
-  out.set_data(allocator::malloc(out.nbytes()));
-
   auto& s = stream();
   auto& encoder = cu::get_command_encoder(s);
-  encoder.set_input_array(in);
-  encoder.set_output_array(out);
 
-  // Fill out with init value.
   if (in.size() == 0) {
-    encoder.launch_kernel([&](cudaStream_t stream) {
-      MLX_SWITCH_ALL_TYPES(in.dtype(), CTYPE, {
-        MLX_SWITCH_REDUCE_OPS(reduce_type_, OP, {
-          using InType = cuda_type_t<CTYPE>;
-          using OutType = cu::ReduceResult<OP, InType>::type;
-          thrust::fill_n(
-              cu::thrust_policy(stream),
-              thrust::device_pointer_cast(out.data<OutType>()),
-              out.data_size(),
-              cu::ReduceInit<OP, InType>::value());
-        });
-      });
-    });
+    init_reduce(encoder, in, out, reduce_type_);
     return;
   }
 
@@ -51,7 +34,19 @@ void Reduce::eval_gpu(const std::vector<array>& inputs, array& out) {
 
   // If it is a general reduce then copy the input to a contiguous array and
   // recompute the plan.
-  if (plan.type == GeneralReduce) {
+  //
+  // TODO: Instead of copying we can use elem-to-loc to deal with broadcasting
+  //       like we do in Metal. When it comes to broadcasted reduction axes
+  //       some can be ignored eg for min/max.
+  bool broadcasted = false;
+  for (int i = 0, j = 0; i < in.ndim() && !broadcasted; i++) {
+    if (j < axes_.size() && axes_[j] == i) {
+      j++;
+    } else {
+      broadcasted = in.strides(i) == 0;
+    }
+  }
+  if (plan.type == GeneralReduce || broadcasted || !in.flags().contiguous) {
     array in_copy(in.shape(), in.dtype(), nullptr, {});
     copy_gpu(in, in_copy, CopyType::General, s);
     encoder.add_temporary(in_copy);
@@ -59,9 +54,8 @@ void Reduce::eval_gpu(const std::vector<array>& inputs, array& out) {
     plan = get_reduction_plan(in, axes_);
   }
 
-  if ((plan.type == ContiguousAllReduce) ||
-      (plan.type == ContiguousReduce && plan.shape.size() == 1)) {
-    segmented_reduce(encoder, in, out, reduce_type_, axes_, plan);
+  if (plan.type == ContiguousAllReduce) {
+    all_reduce(encoder, in, out, reduce_type_);
     return;
   }
 

mlx/backend/cuda/reduce/all_reduce.cu

@@ -0,0 +1,152 @@
+// Copyright © 2025 Apple Inc.
+
+#include "mlx/backend/cuda/device.h"
+#include "mlx/backend/cuda/reduce/reduce.cuh"
+
+#include <cooperative_groups.h>
+#include <cooperative_groups/reduce.h>
+#include <cub/block/block_load.cuh>
+
+namespace mlx::core {
+
+namespace cu {
+
+namespace cg = cooperative_groups;
+
+template <typename T, typename U, typename ReduceOp, int N = 4>
+__global__ void all_reduce(T* in, U* out, size_t block_step, size_t size) {
+  // TODO: Process multiple "rows" in each thread
+  constexpr int M = 1;
+
+  auto grid = cg::this_grid();
+  auto block = cg::this_thread_block();
+  auto warp = cg::tiled_partition<WARP_SIZE>(block);
+
+  const U init = cu::ReduceInit<ReduceOp, T>::value();
+  ReduceOp op;
+
+  T vals[N];
+  U accs[M];
+  accs[0] = init;
+
+  size_t start = grid.block_rank() * block_step;
+  size_t end = start + block_step;
+  size_t check = min(end, size);
+
+  size_t i = start;
+  for (; i + block.size() * N <= check; i += block.size() * N) {
+    cub::LoadDirectBlockedVectorized<T, N>(block.thread_rank(), in + i, vals);
+    for (int j = 0; j < N; j++) {
+      accs[0] = op(accs[0], __cast<U, T>(vals[j]));
+    }
+  }
+
+  if (i < check) {
+    cub::LoadDirectBlocked(
+        block.thread_rank(), in + i, vals, check - i, __cast<T, U>(init));
+    for (int i = 0; i < N; i++) {
+      accs[0] = op(accs[0], __cast<U, T>(vals[i]));
+    }
+  }
+
+  __shared__ U shared_accumulators[32];
+  block_reduce(block, warp, accs, shared_accumulators, op, init);
+
+  if (block.thread_rank() == 0) {
+    out[grid.block_rank()] = accs[0];
+  }
+}
+
+} // namespace cu
+
+void all_reduce(
+    cu::CommandEncoder& encoder,
+    const array& in,
+    array& out,
+    Reduce::ReduceType reduce_type) {
+  constexpr int N_READS = 8;
+
+  out.set_data(allocator::malloc(out.nbytes()));
+
+  auto get_args = [](size_t size, int N) {
+    int threads = std::min(512UL, (size + N - 1) / N);
+    threads = ((threads + WARP_SIZE - 1) / WARP_SIZE) * WARP_SIZE;
+    int reductions_per_step = threads * N;
+    size_t steps_needed =
+        (size + reductions_per_step - 1) / reductions_per_step;
+
+    int blocks;
+    if (steps_needed < 32) {
+      blocks = 1;
+    } else if (steps_needed < 128) {
+      blocks = 32;
+    } else if (steps_needed < 512) {
+      blocks = 128;
+    } else if (steps_needed < 1024) {
+      blocks = 512;
+    } else {
+      blocks = 1024;
+    }
+
+    size_t steps_per_block = (steps_needed + blocks - 1) / blocks;
+    size_t block_step = steps_per_block * reductions_per_step;
+
+    return std::make_tuple(blocks, threads, block_step);
+  };
+
+  int blocks, threads;
+  size_t block_step;
+  size_t insize = in.size();
+  Dtype dt = in.dtype();
+
+  // Cub doesn't like const pointers for load (sigh).
+  void* indata = const_cast<void*>(in.data<void>());
+
+  // Large array so allocate an intermediate and accumulate there
+  std::tie(blocks, threads, block_step) = get_args(insize, N_READS);
+  encoder.set_input_array(in);
+  if (blocks > 1) {
+    array intermediate({blocks}, out.dtype(), nullptr, {});
+    intermediate.set_data(allocator::malloc(intermediate.nbytes()));
+    encoder.add_temporary(intermediate);
+    encoder.set_output_array(intermediate);
+    encoder.launch_kernel([&](cudaStream_t stream) {
+      dispatch_all_types(dt, [&](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;
+          auto kernel = cu::all_reduce<T, U, OP, N_READS>;
+          kernel<<<blocks, threads, 0, stream>>>(
+              static_cast<T*>(indata),
+              intermediate.data<U>(),
+              block_step,
+              insize);
+        });
+      });
+    });
+
+    // Set the input for the next step and recalculate the blocks
+    indata = intermediate.data<void>();
+    dt = intermediate.dtype();
+    insize = intermediate.size();
+    std::tie(blocks, threads, block_step) = get_args(insize, N_READS);
+    encoder.set_input_array(intermediate);
+  }
+
+  encoder.set_output_array(out);
+  encoder.launch_kernel([&](cudaStream_t stream) {
+    dispatch_all_types(dt, [&](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;
+        auto kernel = cu::all_reduce<T, U, OP, N_READS>;
+        kernel<<<blocks, threads, 0, stream>>>(
+            static_cast<T*>(indata), out.data<U>(), block_step, insize);
+      });
+    });
+  });
+}
+
+} // namespace mlx::core

mlx/backend/cuda/reduce/col_reduce.cu

@@ -1,5 +1,7 @@
 // Copyright © 2025 Apple Inc.
 
+#include <numeric>
+
 #include "mlx/backend/cuda/device.h"
 #include "mlx/backend/cuda/device/cast_op.cuh"
 #include "mlx/backend/cuda/reduce/reduce.cuh"
@@ -36,19 +38,36 @@ struct ColReduceArgs {
       const array& in,
       const ReductionPlan& plan,
       const std::vector<int>& axes) {
+    using ShapeVector = decltype(plan.shape);
+    using StridesVector = decltype(plan.strides);
+
+    ShapeVector shape_vec;
+    StridesVector strides_vec;
+
     assert(!plan.shape.empty());
     reduction_size = plan.shape.back();
     reduction_stride = plan.strides.back();
 
     int64_t stride_back = 1;
-    auto [shape_vec, strides_vec] = shapes_without_reduction_axes(in, axes);
+    std::tie(shape_vec, strides_vec) = shapes_without_reduction_axes(in, axes);
     while (!shape_vec.empty() && stride_back < reduction_stride) {
       stride_back *= shape_vec.back();
       shape_vec.pop_back();
       strides_vec.pop_back();
     }
+    std::vector<int> indices(shape_vec.size());
+    std::iota(indices.begin(), indices.end(), 0);
+    std::sort(indices.begin(), indices.end(), [&](int left, int right) {
+      return strides_vec[left] > strides_vec[right];
+    });
+    ShapeVector sorted_shape;
+    StridesVector sorted_strides;
+    for (auto idx : indices) {
+      sorted_shape.push_back(shape_vec[idx]);
+      sorted_strides.push_back(strides_vec[idx]);
+    }
     std::tie(shape_vec, strides_vec) =
-        collapse_contiguous_dims(shape_vec, strides_vec);
+        collapse_contiguous_dims(sorted_shape, sorted_strides);
     shape = const_param(shape_vec);
     strides = const_param(strides_vec);
     ndim = shape_vec.size();
@@ -64,86 +83,6 @@ struct ColReduceArgs {
   }
 };
 
-template <typename T, typename U, typename Op, int NDIM, int N_READS = 4>
-__global__ void col_reduce_small(
-    const T* in,
-    U* out,
-    const __grid_constant__ ColReduceArgs args) {
-  auto grid = cg::this_grid();
-  auto block = cg::this_thread_block();
-
-  int column =
-      grid.block_index().x * block.dim_threads().x + block.thread_index().x;
-  if (column * N_READS >= args.reduction_stride) {
-    return;
-  }
-
-  int out_idx = grid.block_rank() / grid.dim_blocks().x;
-  in += elem_to_loc(out_idx, args.shape.data(), args.strides.data(), args.ndim);
-
-  Op op;
-  U totals[N_READS];
-  for (int i = 0; i < N_READS; i++) {
-    totals[i] = ReduceInit<Op, T>::value();
-  }
-
-  // Read input to local.
-  LoopedElemToLoc<NDIM, (NDIM > 2)> loop(args.reduce_ndim);
-  loop.next(
-      block.thread_index().y,
-      args.reduce_shape.data(),
-      args.reduce_strides.data());
-  for (size_t r = block.thread_index().y;
-       r < args.non_col_reductions * args.reduction_size;
-       r += block.dim_threads().y) {
-    U vals[N_READS];
-    cub::LoadDirectBlocked(
-        column,
-        make_cast_iterator<U>(in + loop.location()),
-        vals,
-        args.reduction_stride,
-        ReduceInit<Op, T>::value());
-    for (int i = 0; i < N_READS; i++) {
-      totals[i] = op(vals[i], totals[i]);
-    }
-    loop.next(
-        block.dim_threads().y,
-        args.reduce_shape.data(),
-        args.reduce_strides.data());
-  }
-
-  // Do block reduce when each column has more than 1 element to reduce.
-  if (block.dim_threads().y > 1) {
-    __shared__ U shared_vals[32 * 8 * N_READS];
-    size_t col =
-        block.thread_index().y * block.dim_threads().x + block.thread_index().x;
-    for (int i = 0; i < N_READS; i++) {
-      shared_vals[col * N_READS + i] = totals[i];
-    }
-    block.sync();
-    if (block.thread_index().y == 0) {
-      for (int i = 0; i < N_READS; i++) {
-        totals[i] = shared_vals[block.thread_index().x * N_READS + i];
-      }
-      for (int j = 1; j < block.dim_threads().y; j++) {
-        col = j * block.dim_threads().x + block.thread_index().x;
-        for (int i = 0; i < N_READS; i++) {
-          totals[i] = op(shared_vals[col * N_READS + i], totals[i]);
-        }
-      }
-    }
-  }
-
-  // Write result.
-  if (block.thread_index().y == 0) {
-    cub::StoreDirectBlocked(
-        column,
-        out + out_idx * args.reduction_stride,
-        totals,
-        args.reduction_stride);
-  }
-}
-
 template <
     typename T,
     typename U,
@@ -152,67 +91,94 @@ template <
     int BM,
     int BN,
     int N_READS = 4>
-__global__ void col_reduce_looped(
-    const T* in,
-    U* out,
-    const __grid_constant__ ColReduceArgs args) {
+__global__ void
+col_reduce_looped(T* in, U* out, const __grid_constant__ ColReduceArgs args) {
   auto grid = cg::this_grid();
   auto block = cg::this_thread_block();
   auto warp = cg::tiled_partition<WARP_SIZE>(block);
 
-  constexpr int n_warps = BN / N_READS;
+  constexpr int threads_per_row = BN / N_READS;
 
-  int out_idx = grid.block_rank() / grid.dim_blocks().x;
-  in += elem_to_loc(out_idx, args.shape.data(), args.strides.data(), args.ndim);
+  // Compute the indices for the tile
+  size_t tile_idx = grid.block_rank();
+  size_t tile_x = tile_idx % ((args.reduction_stride + BN - 1) / BN);
+  size_t tile_y = tile_idx / ((args.reduction_stride + BN - 1) / BN);
 
+  // Compute the indices for the thread within the tile
+  short thread_x = block.thread_rank() % threads_per_row;
+  short thread_y = block.thread_rank() / threads_per_row;
+
+  // Move the input pointer
+  in += elem_to_loc(tile_y, args.shape.data(), args.strides.data(), args.ndim) +
+      tile_x * BN;
+
+  // Initialize the running totals
   Op op;
   U totals[N_READS];
   for (int i = 0; i < N_READS; i++) {
     totals[i] = ReduceInit<Op, T>::value();
   }
 
-  // Read input to local.
-  int r = block.thread_rank() / n_warps;
-  int column = block.thread_rank() % n_warps;
-  int in_offset = grid.block_index().x * BN;
   LoopedElemToLoc<NDIM, (NDIM > 2)> loop(args.reduce_ndim);
-  loop.next(r, args.reduce_shape.data(), args.reduce_strides.data());
-  for (; r < args.non_col_reductions * args.reduction_size; r += BM) {
-    U vals[N_READS];
-    cub::LoadDirectBlocked(
-        column,
-        make_cast_iterator<U>(in + loop.location() + in_offset),
-        vals,
-        args.reduction_stride - in_offset,
-        ReduceInit<Op, T>::value());
-    for (int i = 0; i < N_READS; i++) {
-      totals[i] = op(vals[i], totals[i]);
+  loop.next(thread_y, args.reduce_shape.data(), args.reduce_strides.data());
+  size_t total = args.non_col_reductions * args.reduction_size;
+  if (tile_x * BN + BN <= args.reduction_stride) {
+    if (args.reduction_stride % N_READS == 0) {
+      for (size_t r = thread_y; r < total; r += BM) {
+        T vals[N_READS];
+        cub::LoadDirectBlockedVectorized(thread_x, in + loop.location(), vals);
+        for (int i = 0; i < N_READS; i++) {
+          totals[i] = op(totals[i], __cast<U, T>(vals[i]));
+        }
+        loop.next(BM, args.reduce_shape.data(), args.reduce_strides.data());
+      }
+    } else {
+      for (size_t r = thread_y; r < total; r += BM) {
+        T vals[N_READS];
+        cub::LoadDirectBlocked(thread_x, in + loop.location(), vals);
+        for (int i = 0; i < N_READS; i++) {
+          totals[i] = op(totals[i], __cast<U, T>(vals[i]));
+        }
+        loop.next(BM, args.reduce_shape.data(), args.reduce_strides.data());
+      }
+    }
+  } else {
+    for (size_t r = thread_y; r < total; r += BM) {
+      T vals[N_READS];
+      cub::LoadDirectBlocked(
+          thread_x,
+          in + loop.location(),
+          vals,
+          args.reduction_stride - tile_x * BN,
+          __cast<T, U>(ReduceInit<Op, T>::value()));
+      for (int i = 0; i < N_READS; i++) {
+        totals[i] = op(totals[i], __cast<U, T>(vals[i]));
+      }
+      loop.next(BM, args.reduce_shape.data(), args.reduce_strides.data());
     }
-    loop.next(BM, args.reduce_shape.data(), args.reduce_strides.data());
   }
 
   // Do warp reduce for each output.
-  constexpr int n_outputs = BN / n_warps;
+  constexpr int n_outputs = BN / threads_per_row;
   static_assert(BM == 32 && n_outputs == N_READS);
   __shared__ U shared_vals[BM * BN];
-  size_t col = block.thread_index().y * BN + block.thread_index().x * N_READS;
+  short s_idx = thread_y * BN + thread_x * N_READS;
   for (int i = 0; i < N_READS; i++) {
-    shared_vals[col + i] = totals[i];
+    shared_vals[s_idx + i] = totals[i];
   }
   block.sync();
-  col = warp.thread_rank() * BN + warp.meta_group_rank() * n_outputs;
+  s_idx = warp.thread_rank() * BN + warp.meta_group_rank() * n_outputs;
   for (int i = 0; i < n_outputs; i++) {
-    totals[i] = cg::reduce(warp, shared_vals[col + i], op);
+    totals[i] = cg::reduce(warp, shared_vals[s_idx + i], op);
   }
 
   // Write result.
   if (warp.thread_rank() == 0) {
-    size_t out_offset = grid.block_index().x * BN;
     cub::StoreDirectBlocked(
         warp.meta_group_rank(),
-        out + out_idx * args.reduction_stride + out_offset,
+        out + tile_y * args.reduction_stride + tile_x * BN,
         totals,
-        args.reduction_stride - out_offset);
+        args.reduction_stride - tile_x * BN);
   }
 }
 
@@ -220,14 +186,57 @@ __global__ void col_reduce_looped(
 
 inline auto output_grid_for_col_reduce(
     const array& out,
-    const cu::ColReduceArgs& args) {
-  auto out_shape = out.shape();
-  auto out_strides = out.strides();
-  while (!out_shape.empty() && out_strides.back() < args.reduction_stride) {
-    out_shape.pop_back();
-    out_strides.pop_back();
+    const cu::ColReduceArgs& args,
+    int bn) {
+  int gx, gy = 1;
+  size_t n_inner_blocks = cuda::ceil_div(args.reduction_stride, bn);
+  size_t n_outer_blocks = out.size() / args.reduction_stride;
+  size_t n_blocks = n_outer_blocks * n_inner_blocks;
+  while (n_blocks / gy > INT32_MAX) {
+    gy *= 2;
   }
-  return get_2d_grid_dims(out_shape, out_strides);
+  gx = cuda::ceil_div(n_blocks, gy);
+
+  return dim3(gx, gy, 1);
+}
+
+void col_reduce_looped(
+    cu::CommandEncoder& encoder,
+    const array& in,
+    array& out,
+    Reduce::ReduceType reduce_type,
+    const std::vector<int>& axes,
+    const ReductionPlan& plan,
+    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);
+  encoder.launch_kernel([&](cudaStream_t stream) {
+    dispatch_all_types(in.dtype(), [&](auto type_tag) {
+      dispatch_reduce_ops(reduce_type, [&](auto reduce_type_tag) {
+        dispatch_reduce_ndim(args.reduce_ndim, [&](auto reduce_ndim) {
+          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 = 4;
+          constexpr int BM = 32;
+          constexpr int BN = 32;
+          dim3 grid = output_grid_for_col_reduce(out, args, BN);
+          int blocks = BM * BN / N_READS;
+          auto kernel =
+              cu::col_reduce_looped<T, U, OP, reduce_ndim(), BM, BN, N_READS>;
+          kernel<<<grid, blocks, 0, stream>>>(indata, out.data<U>(), args);
+        });
+      });
+    });
+  });
 }
 
 void col_reduce(
@@ -237,42 +246,23 @@ void col_reduce(
     Reduce::ReduceType reduce_type,
     const std::vector<int>& axes,
     const ReductionPlan& plan) {
+  // Current col reduce options
+  //
+  // - col_reduce_looped
+  //
+  //   It is a general strided reduce. Each threadblock computes the output for
+  //   a subrow of the fast moving axis. For instance 32 elements.
+  //
+  // Notes: As in row reduce we opt to read as much in order as possible and
+  //        leave transpositions as they are (contrary to our Metal backend).
+  //
+  //        Moreover we need different kernels for short rows and tuning
+
+  // Make the args struct to help route to the best kernel
   cu::ColReduceArgs args(in, plan, axes);
 
-  encoder.launch_kernel([&](cudaStream_t stream) {
-    MLX_SWITCH_ALL_TYPES(in.dtype(), CTYPE, {
-      using InType = cuda_type_t<CTYPE>;
-      MLX_SWITCH_REDUCE_OPS(reduce_type, OP, {
-        using OutType = cu::ReduceResult<OP, InType>::type;
-        MLX_SWITCH_REDUCE_NDIM(args.reduce_ndim, NDIM, {
-          constexpr int N_READS = 4;
-          dim3 block_dims;
-          dim3 num_blocks = output_grid_for_col_reduce(out, args);
-          num_blocks.z = num_blocks.y;
-          num_blocks.y = num_blocks.x;
-          auto kernel =
-              cu::col_reduce_small<InType, OutType, OP, NDIM, N_READS>;
-          size_t total = args.non_col_reductions * args.reduction_size;
-          if (total < 32) {
-            size_t stride_blocks =
-                cuda::ceil_div(args.reduction_stride, N_READS);
-            block_dims.x = std::min(stride_blocks, 32ul);
-            block_dims.y = std::min(total, 8ul);
-            num_blocks.x = cuda::ceil_div(stride_blocks, block_dims.x);
-          } else {
-            constexpr int BM = 32;
-            constexpr int BN = 32;
-            block_dims.x = BM * BN / N_READS;
-            num_blocks.x = cuda::ceil_div(args.reduction_stride, BN);
-            kernel = cu::
-                col_reduce_looped<InType, OutType, OP, NDIM, BM, BN, N_READS>;
-          }
-          kernel<<<num_blocks, block_dims, 0, stream>>>(
-              in.data<InType>(), out.data<OutType>(), args);
-        });
-      });
-    });
-  });
+  // Fallback col reduce
+  col_reduce_looped(encoder, in, out, reduce_type, axes, plan, args);
 }
 
 } // namespace mlx::core

mlx/backend/cuda/reduce/init_reduce.cu

@@ -0,0 +1,51 @@
+// Copyright © 2025 Apple Inc.
+
+#include "mlx/backend/cuda/device.h"
+#include "mlx/backend/cuda/reduce/reduce.cuh"
+
+#include <cooperative_groups.h>
+
+namespace mlx::core {
+
+namespace cu {
+
+namespace cg = cooperative_groups;
+
+template <typename T, typename U, typename Op>
+__global__ void init_reduce(U* out, size_t size) {
+  auto index = cg::this_grid().thread_rank();
+  if (index < size) {
+    out[index] = ReduceInit<Op, T>::value();
+  }
+}
+
+} // namespace cu
+
+void init_reduce(
+    cu::CommandEncoder& encoder,
+    const array& in,
+    array& out,
+    Reduce::ReduceType reduce_type) {
+  // Allocate if needed
+  if (out.data_shared_ptr() == nullptr) {
+    out.set_data(allocator::malloc(out.nbytes()));
+  }
+
+  encoder.set_output_array(out);
+  encoder.launch_kernel([&](cudaStream_t stream) {
+    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;
+        auto kernel = cu::init_reduce<T, U, OP>;
+        dim3 grid = get_2d_grid_dims(out.shape(), out.strides());
+        dim3 block(grid.x < 1024 ? grid.x : 1024, 1, 1);
+        grid.x = (grid.x + 1023) / 1024;
+        kernel<<<grid, block, 0, stream>>>(out.data<U>(), out.size());
+      });
+    });
+  });
+}
+
+} // namespace mlx::core

mlx/backend/cuda/reduce/reduce.cuh

@@ -1,5 +1,7 @@
 // Copyright © 2025 Apple Inc.
 
+#include <type_traits>
+
 #include "mlx/backend/common/reduce.h"
 #include "mlx/backend/cuda/device/cucomplex_math.cuh"
 #include "mlx/backend/cuda/kernel_utils.cuh"
@@ -9,51 +11,41 @@
 
 namespace mlx::core {
 
-// Dispatch dynamic ndim to constexpr.
-// The behavior follows get_kernel_reduce_ndim in metal/reduce.cpp file.
-#define MLX_SWITCH_REDUCE_NDIM(ndim, NDIM, ...) \
-  if (ndim == 1) {                              \
-    constexpr uint32_t NDIM = 1;                \
-    __VA_ARGS__;                                \
-  } else if (ndim == 2) {                       \
-    constexpr uint32_t NDIM = 2;                \
-    __VA_ARGS__;                                \
-  } else {                                      \
-    constexpr uint32_t NDIM = 5;                \
-    __VA_ARGS__;                                \
+template <typename F>
+void dispatch_reduce_ndim(int ndim, F&& f) {
+  if (ndim == 1) {
+    f(std::integral_constant<int, 1>{});
+  } else if (ndim == 2) {
+    f(std::integral_constant<int, 2>{});
+  } else {
+    f(std::integral_constant<int, 5>{});
   }
+}
 
-// Dispatch reduce ops to constexpr.
-#define MLX_SWITCH_REDUCE_OPS(REDUCE, OP, ...)           \
-  if (REDUCE == Reduce::ReduceType::And) {               \
-    using OP = cu::And;                                  \
-    __VA_ARGS__;                                         \
-  } else if (REDUCE == Reduce::ReduceType::Or) {         \
-    using OP = cu::Or;                                   \
-    __VA_ARGS__;                                         \
-  } else if (REDUCE == Reduce::ReduceType::Sum) {        \
-    using OP = cu::Sum;                                  \
-    __VA_ARGS__;                                         \
-  } else if (REDUCE == Reduce::ReduceType::Prod) {       \
-    using OP = cu::Prod;                                 \
-    __VA_ARGS__;                                         \
-  } else if (REDUCE == Reduce::ReduceType::Max) {        \
-    using OP = cu::Max;                                  \
-    __VA_ARGS__;                                         \
-  } else if (REDUCE == Reduce::ReduceType::Min) {        \
-    using OP = cu::Min;                                  \
-    __VA_ARGS__;                                         \
-  } else {                                               \
-    throw std::invalid_argument("Unknown reduce type."); \
+template <typename F>
+void dispatch_reduce_ops(Reduce::ReduceType reduce_type, F&& f) {
+  if (reduce_type == Reduce::ReduceType::And) {
+    f(type_identity<cu::And>{});
+  } else if (reduce_type == Reduce::ReduceType::Or) {
+    f(type_identity<cu::Or>{});
+  } else if (reduce_type == Reduce::ReduceType::Sum) {
+    f(type_identity<cu::Sum>{});
+  } else if (reduce_type == Reduce::ReduceType::Prod) {
+    f(type_identity<cu::Prod>{});
+  } else if (reduce_type == Reduce::ReduceType::Max) {
+    f(type_identity<cu::Max>{});
+  } else if (reduce_type == Reduce::ReduceType::Min) {
+    f(type_identity<cu::Min>{});
+  } else {
+    throw std::invalid_argument("Unknown reduce type.");
   }
+}
 
-void segmented_reduce(
+void all_reduce(
     cu::CommandEncoder& encoder,
     const array& in,
     array& out,
-    Reduce::ReduceType reduce_type,
-    const std::vector<int>& axes,
-    const ReductionPlan& plan);
+    Reduce::ReduceType reduce_type);
 
 void row_reduce(
     cu::CommandEncoder& encoder,
@@ -71,4 +63,10 @@ void col_reduce(
     const std::vector<int>& axes,
     const ReductionPlan& plan);
 
+void init_reduce(
+    cu::CommandEncoder& encoder,
+    const array& in,
+    array& out,
+    Reduce::ReduceType reduce_type);
+
 } // namespace mlx::core

mlx/backend/cuda/reduce/reduce_ops.cuh

@@ -3,48 +3,89 @@
 #pragma once
 
 #include "mlx/backend/cuda/device/utils.cuh"
+#include "mlx/backend/cuda/reduce/reduce_utils.cuh"
 
 namespace mlx::core::cu {
 
 // Reduce ops.
 struct And {
-  __device__ bool operator()(bool a, bool b) {
+  __device__ __forceinline__ bool operator()(bool a, bool b) {
     return a && b;
   }
+
+  __device__ void atomic_update(bool* x, bool y) {
+    atomic_reduce<bool, And>(x, y);
+  }
 };
 
 struct Or {
-  __device__ bool operator()(bool a, bool b) {
+  __device__ __forceinline__ bool operator()(bool a, bool b) {
     return a || b;
   }
+
+  __device__ void atomic_update(bool* x, bool y) {
+    atomic_reduce<bool, Or>(x, y);
+  }
 };
 
 struct Sum {
   template <typename T>
-  __device__ T operator()(T a, T b) {
+  __device__ __forceinline__ T operator()(T a, T b) {
     return a + b;
   }
+
+  template <typename T>
+  __device__ void atomic_update(T* x, T y) {
+    atomic_reduce<T, Sum>(x, y);
+  }
+
+  __device__ void atomic_update(__nv_bfloat16* x, __nv_bfloat16 y) {
+    atomicAdd(x, y);
+  }
+
+  __device__ void atomic_update(int* x, int y) {
+    atomicAdd(x, y);
+  }
+
+  __device__ void atomic_update(float* x, float y) {
+    atomicAdd(x, y);
+  }
 };
 
 struct Prod {
   template <typename T>
-  __device__ T operator()(T a, T b) {
+  __device__ __forceinline__ T operator()(T a, T b) {
     return a * b;
   }
+
+  template <typename T>
+  __device__ void atomic_update(T* x, T y) {
+    atomic_reduce<T, Prod>(x, y);
+  }
 };
 
 struct Min {
   template <typename T>
-  __device__ T operator()(T a, T b) {
+  __device__ __forceinline__ T operator()(T a, T b) {
     return a < b ? a : b;
   }
+
+  template <typename T>
+  __device__ void atomic_update(T* x, T y) {
+    atomic_reduce<T, Min>(x, y);
+  }
 };
 
 struct Max {
   template <typename T>
-  __device__ T operator()(T a, T b) {
+  __device__ __forceinline__ T operator()(T a, T b) {
     return a > b ? a : b;
   }
+
+  template <typename T>
+  __device__ void atomic_update(T* x, T y) {
+    atomic_reduce<T, Max>(x, y);
+  }
 };
 
 // Traits to get the result type of reduce op.
@@ -120,7 +161,7 @@ template <typename T>
 struct ReduceInit<Prod, T> {
   static constexpr __host__ __device__ auto value() {
     if constexpr (cuda::std::is_same_v<T, cuComplex>) {
-      return T{1, 1};
+      return T{1, 0};
     } else {
       return typename ReduceResult<Prod, T>::type{1};
     }

mlx/backend/cuda/reduce/reduce_utils.cuh

@@ -0,0 +1,158 @@
+// Copyright © 2025 Apple Inc.
+
+#pragma once
+
+#include <numeric>
+
+#include "mlx/backend/cuda/device/utils.cuh"
+
+#include <cooperative_groups.h>
+#include <cooperative_groups/reduce.h>
+
+namespace mlx::core {
+
+namespace cu {
+
+namespace cg = cooperative_groups;
+
+template <size_t N>
+struct uint_by_size;
+template <>
+struct uint_by_size<2> {
+  using type = uint16_t;
+};
+template <>
+struct uint_by_size<4> {
+  using type = uint32_t;
+};
+template <>
+struct uint_by_size<8> {
+  using type = unsigned long long int;
+};
+
+template <typename T, typename Op>
+__device__ void atomic_reduce(T* x, T y) {
+  if constexpr (sizeof(T) == 1) {
+    using U = uint16_t;
+    U* x_int = (U*)((char*)x - ((size_t)x % 2));
+    int shift = ((char*)x - (char*)x_int) * 8;
+    int mask = 0xff << shift;
+    U old_val, new_val;
+    do {
+      old_val = *x_int;
+      T result = Op{}(static_cast<T>((old_val >> shift) & 0xff), y);
+      new_val = (old_val & ~mask) | (result << shift);
+    } while (atomicCAS(x_int, old_val, new_val) != old_val);
+  } else {
+    using U = typename uint_by_size<sizeof(T)>::type;
+    U* x_int = (U*)(x);
+    U old_val, new_val;
+    do {
+      old_val = *x_int;
+      T result = Op{}(*((T*)&old_val), y);
+      new_val = *((U*)&result);
+    } while (atomicCAS(x_int, old_val, new_val) != old_val);
+  }
+}
+
+// TODO: Should make a custom complex type
+template <typename U, typename T>
+inline __device__ U __cast(T x) {
+  return static_cast<U>(x);
+}
+
+template <>
+inline __device__ bool __cast<bool, cuComplex>(cuComplex x) {
+  return x.x != 0 && x.y != 0;
+}
+
+template <>
+inline __device__ cuComplex __cast<cuComplex, bool>(bool x) {
+  return x ? make_cuFloatComplex(1, 1) : make_cuFloatComplex(0, 0);
+}
+
+template <typename T, int N, typename Block, typename Warp, typename Op>
+inline __device__ void
+block_reduce(Block block, Warp warp, T (&vals)[N], T* smem, Op op, T init) {
+  // First reduce in the current warp
+  for (int i = 0; i < N; i++) {
+    vals[i] = cg::reduce(warp, vals[i], op);
+  }
+
+  // Reduce across warps
+  if (warp.meta_group_size() > 1) {
+    if (warp.thread_rank() == 0) {
+      for (int i = 0; i < N; i++) {
+        smem[warp.meta_group_rank() * N + i] = vals[i];
+      }
+    }
+    block.sync();
+    if (warp.thread_rank() < warp.meta_group_size()) {
+      for (int i = 0; i < N; i++) {
+        vals[i] = smem[warp.thread_rank() * N + i];
+      }
+    } else {
+      for (int i = 0; i < N; i++) {
+        vals[i] = init;
+      }
+    }
+    for (int i = 0; i < N; i++) {
+      vals[i] = cg::reduce(warp, vals[i], op);
+    }
+  }
+}
+
+} // namespace cu
+
+inline void allocate_same_layout(
+    array& out,
+    const array& in,
+    const std::vector<int>& axes) {
+  if (in.flags().row_contiguous) {
+    out.set_data(allocator::malloc(out.nbytes()));
+    return;
+  }
+
+  if (out.ndim() < in.ndim()) {
+    throw std::runtime_error(
+        "Reduction without keepdims only supported for row-contiguous inputs");
+  }
+
+  // Calculate the transpositions applied to in in order to apply them to out.
+  std::vector<int> axis_order(in.ndim());
+  std::iota(axis_order.begin(), axis_order.end(), 0);
+  std::sort(axis_order.begin(), axis_order.end(), [&](int left, int right) {
+    return in.strides(left) > in.strides(right);
+  });
+
+  // Transpose the shape and calculate the strides
+  Shape out_shape(in.ndim());
+  Strides out_strides(in.ndim(), 1);
+  for (int i = 0; i < in.ndim(); i++) {
+    out_shape[i] = out.shape(axis_order[i]);
+  }
+  for (int i = in.ndim() - 2; i >= 0; i--) {
+    out_strides[i] = out_shape[i + 1] * out_strides[i + 1];
+  }
+
+  // Reverse the axis order to get the final strides
+  Strides final_strides(in.ndim());
+  for (int i = 0; i < in.ndim(); i++) {
+    final_strides[axis_order[i]] = out_strides[i];
+  }
+
+  // Calculate the resulting contiguity and do the memory allocation
+  auto [data_size, rc, cc] = check_contiguity(out.shape(), final_strides);
+  auto fl = in.flags();
+  fl.row_contiguous = rc;
+  fl.col_contiguous = cc;
+  fl.contiguous = true;
+  out.set_data(
+      allocator::malloc(out.nbytes()),
+      data_size,
+      final_strides,
+      fl,
+      allocator::free);
+}
+
+} // namespace mlx::core

mlx/backend/cuda/reduce/row_reduce.cu

@@ -1,5 +1,7 @@
 // Copyright © 2025 Apple Inc.
 
+#include <numeric>
+
 #include "mlx/backend/cuda/device.h"
 #include "mlx/backend/cuda/device/cast_op.cuh"
 #include "mlx/backend/cuda/reduce/reduce.cuh"
@@ -55,84 +57,108 @@ struct RowReduceArgs {
       non_row_reductions *= reduce_shape[i];
     }
   }
+
+  // Convert shape and strides as if in was contiguous
+  void sort_access_pattern(const array& in, const std::vector<int>& axes) {
+    auto shape_vec = in.shape();
+    auto strides_vec = in.strides();
+    std::tie(shape_vec, strides_vec) =
+        shapes_without_reduction_axes(shape_vec, strides_vec, axes);
+    std::vector<int> indices(shape_vec.size());
+    std::iota(indices.begin(), indices.end(), 0);
+    std::sort(indices.begin(), indices.end(), [&](int left, int right) {
+      return strides_vec[left] > strides_vec[right];
+    });
+    decltype(shape_vec) sorted_shape;
+    decltype(strides_vec) sorted_strides;
+    for (auto idx : indices) {
+      sorted_shape.push_back(shape_vec[idx]);
+      sorted_strides.push_back(strides_vec[idx]);
+    }
+    std::tie(shape_vec, strides_vec) =
+        collapse_contiguous_dims(sorted_shape, sorted_strides);
+    shape = const_param(shape_vec);
+    strides = const_param(strides_vec);
+    ndim = shape_vec.size();
+  }
 };
 
-template <typename T, typename U, typename Op, int NDIM, int N_READS = 4>
-__global__ void row_reduce_small(
-    const T* in,
-    U* out,
-    size_t out_size,
-    const __grid_constant__ RowReduceArgs args) {
-  size_t out_idx = cg::this_grid().thread_rank();
-  if (out_idx >= out_size) {
-    return;
-  }
-
-  Op op;
-
-  U total_val = ReduceInit<Op, T>::value();
-  LoopedElemToLoc<NDIM, (NDIM > 2)> loop(args.reduce_ndim);
-
-  in += elem_to_loc(out_idx, args.shape.data(), args.strides.data(), args.ndim);
-
-  for (size_t n = 0; n < args.non_row_reductions; n++) {
-    for (int r = 0; r < cuda::ceil_div(args.row_size, N_READS); r++) {
-      U vals[N_READS];
-      cub::LoadDirectBlocked(
-          r,
-          make_cast_iterator<U>(in + loop.location()),
-          vals,
-          args.row_size,
-          ReduceInit<Op, T>::value());
-      total_val = op(total_val, cub::ThreadReduce(vals, op));
-    }
-    loop.next(args.reduce_shape.data(), args.reduce_strides.data());
-  }
-
-  out[out_idx] = total_val;
-}
-
-template <typename T, typename U, typename Op, int NDIM, int N_READS = 4>
-__global__ void row_reduce_small_warp(
-    const T* in,
-    U* out,
-    size_t out_size,
-    const __grid_constant__ RowReduceArgs args) {
+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) {
   auto grid = cg::this_grid();
   auto block = cg::this_thread_block();
   auto warp = cg::tiled_partition<WARP_SIZE>(block);
 
-  size_t out_idx = grid.thread_rank() / WARP_SIZE;
-  if (out_idx >= out_size) {
-    return;
+  const U init = cu::ReduceInit<ReduceOp, T>::value();
+  ReduceOp op;
+
+  T vals[M][N];
+  U accs[M];
+  for (int i = 0; i < M; i++) {
+    accs[i] = init;
   }
 
-  Op op;
+  const size_t start_row =
+      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;
+  out += start_row;
 
-  U total_val = ReduceInit<Op, T>::value();
-  LoopedElemToLoc<NDIM, (NDIM > 2)> loop(args.reduce_ndim);
-
-  in += elem_to_loc(out_idx, args.shape.data(), args.strides.data(), args.ndim);
-
-  for (size_t n = warp.thread_rank(); n < args.non_row_reductions;
-       n += WARP_SIZE) {
-    for (int r = 0; r < cuda::ceil_div(args.row_size, N_READS); r++) {
-      U vals[N_READS];
-      cub::LoadDirectBlocked(
-          r,
-          make_cast_iterator<U>(in + loop.location()),
-          vals,
-          args.row_size,
-          ReduceInit<Op, T>::value());
-      total_val = op(total_val, cub::ThreadReduce(vals, op));
+  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<U, T>(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<U, T>(vals[k][j]));
+        }
+      }
     }
-    loop.next(WARP_SIZE, args.reduce_shape.data(), args.reduce_strides.data());
   }
 
-  total_val = cg::reduce(warp, total_val, op);
+  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<T, U>(init));
+      for (int j = 0; j < N; j++) {
+        accs[k] = op(accs[k], __cast<U, T>(vals[k][j]));
+      }
+    }
+  }
 
-  if (warp.thread_rank() == 0) {
-    out[out_idx] = total_val;
+  __shared__ U shared_accumulators[32 * M];
+  block_reduce(block, warp, accs, 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];
+      }
+    } else {
+      short offset = grid.block_rank() * M + M - n_rows;
+      for (int i = offset; i < M; i++) {
+        out[i] = accs[i];
+      }
+    }
   }
 }
 
@@ -141,55 +167,173 @@ template <
     typename U,
     typename Op,
     int NDIM,
-    int BLOCK_DIM_X,
+    int BLOCK_DIM,
     int N_READS = 4>
 __global__ void row_reduce_looped(
-    const T* in,
+    T* in,
     U* out,
     size_t out_size,
     const __grid_constant__ RowReduceArgs args) {
   auto grid = cg::this_grid();
   auto block = cg::this_thread_block();
+  auto warp = cg::tiled_partition<WARP_SIZE>(block);
 
-  size_t out_idx = grid.thread_rank() / BLOCK_DIM_X;
-  if (out_idx >= out_size) {
-    return;
-  }
+  size_t out_idx = grid.block_rank();
 
   Op op;
 
-  U total_val = ReduceInit<Op, T>::value();
+  U total[1];
+  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;
 
   in += elem_to_loc(out_idx, args.shape.data(), args.strides.data(), args.ndim);
 
   for (size_t n = 0; n < args.non_row_reductions; n++) {
-    for (size_t r = 0; r < cuda::ceil_div(args.row_size, BLOCK_DIM_X * N_READS);
-         r++) {
-      U vals[N_READS];
-      cub::LoadDirectBlocked(
-          r * BLOCK_DIM_X + block.thread_index().x,
-          make_cast_iterator<U>(in + loop.location()),
-          vals,
-          args.row_size,
-          ReduceInit<Op, T>::value());
-      total_val = op(total_val, cub::ThreadReduce(vals, op));
+    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<U, T>(vals[i]));
+      }
     }
+    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<T, U>(init));
+      for (int i = 0; i < N_READS; i++) {
+        total[0] = op(total[0], __cast<U, T>(vals[i]));
+      }
+    }
+    // TODO: Maybe block.sync() here?
     loop.next(args.reduce_shape.data(), args.reduce_strides.data());
   }
 
-  typedef cub::BlockReduce<U, BLOCK_DIM_X> BlockReduceT;
-  __shared__ typename BlockReduceT::TempStorage temp;
-
-  total_val = BlockReduceT(temp).Reduce(total_val, op);
+  __shared__ U shared_accumulators[32];
+  block_reduce(block, warp, total, shared_accumulators, op, init);
 
   if (block.thread_rank() == 0) {
-    out[out_idx] = total_val;
+    out[out_idx] = total[0];
   }
 }
 
 } // namespace cu
 
+void row_reduce_simple(
+    cu::CommandEncoder& encoder,
+    const array& in,
+    array& out,
+    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);
+
+  // TODO: If out.size() < 1024 which will be a common case then write this in
+  //       2 passes. Something like 32 * out.size() and then do a warp reduce.
+  encoder.set_input_array(in);
+  encoder.set_output_array(out);
+  encoder.launch_kernel([&](cudaStream_t stream) {
+    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;
+
+        // Cub doesn't like const pointers for vectorized loads. (sigh)
+        T* indata = const_cast<T*>(in.data<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;
+        dim3 block(threads, 1, 1);
+
+        // Pick the kernel
+        auto kernel = cu::row_reduce_simple<T, U, OP, N_READS>;
+        if (grid.x >= 1024) {
+          grid.x = (grid.x + 1) / 2;
+          kernel = cu::row_reduce_simple<T, U, OP, N_READS, 2>;
+        }
+
+        // Launch
+        kernel<<<grid, block, 0, stream>>>(
+            indata, out.data<U>(), out.size(), plan.shape.back());
+      });
+    });
+  });
+}
+
+void row_reduce_looped(
+    cu::CommandEncoder& encoder,
+    const array& in,
+    array& out,
+    Reduce::ReduceType reduce_type,
+    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);
+
+  encoder.set_input_array(in);
+  encoder.set_output_array(out);
+  encoder.launch_kernel([&](cudaStream_t stream) {
+    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;
+
+        // Cub doesn't like const pointers for vectorized loads. (sigh)
+        T* indata = const_cast<T*>(in.data<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;
+        dim3 block(threads, 1, 1);
+
+        // Pick the kernel
+        auto kernel = cu::row_reduce_looped<T, U, OP, 1, 32, 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(),
+                threads_constant(),
+                N_READS>;
+            block.x = threads_constant();
+          });
+        });
+
+        // Launch
+        kernel<<<grid, block, 0, stream>>>(
+            indata, out.data<U>(), out.size(), args);
+      });
+    });
+  });
+}
+
 void row_reduce(
     cu::CommandEncoder& encoder,
     const array& in,
@@ -197,54 +341,35 @@ void row_reduce(
     Reduce::ReduceType reduce_type,
     const std::vector<int>& axes,
     const ReductionPlan& plan) {
+  // Current row reduction options
+  //
+  // - row_reduce_simple
+  //
+  //   That means that we are simply reducing across the fastest moving axis.
+  //   We are reducing 1 or 2 rows per threadblock depending on the size of
+  //   output.
+  //
+  // - row_reduce_looped
+  //
+  //   It is a general row reduction. We are computing 1 output per
+  //   threadblock. We read the fastest moving axis vectorized and loop over
+  //   the rest of the axes.
+  //
+  // Notes: We opt to read as much in order as possible and leave
+  //        transpositions as they are (contrary to our Metal backend).
+
+  // Simple row reduce means that we have 1 axis that we are reducing over and
+  // it has stride 1.
+  if (plan.shape.size() == 1) {
+    row_reduce_simple(encoder, in, out, reduce_type, axes, plan);
+    return;
+  }
+
+  // Make the args struct to help route to the best kernel
   cu::RowReduceArgs args(in, plan, axes);
 
-  encoder.launch_kernel([&](cudaStream_t stream) {
-    MLX_SWITCH_ALL_TYPES(in.dtype(), CTYPE, {
-      using InType = cuda_type_t<CTYPE>;
-      MLX_SWITCH_REDUCE_OPS(reduce_type, OP, {
-        using OutType = cu::ReduceResult<OP, InType>::type;
-        MLX_SWITCH_REDUCE_NDIM(args.reduce_ndim, NDIM, {
-          constexpr size_t N_READS = 4;
-          dim3 out_dims = get_2d_grid_dims(out.shape(), out.strides());
-          dim3 block_dims, num_blocks;
-          auto kernel =
-              cu::row_reduce_small<InType, OutType, OP, NDIM, N_READS>;
-          if (args.row_size <= 64) {
-            if ((args.non_row_reductions < 32 && args.row_size <= 8) ||
-                (args.non_row_reductions <= 8)) {
-              block_dims.x = std::min(out_dims.x, 1024u);
-              num_blocks.x = cuda::ceil_div(out_dims.x, block_dims.x);
-              num_blocks.y = out_dims.y;
-            } else {
-              block_dims.x = WARP_SIZE;
-              num_blocks.y = out_dims.x;
-              num_blocks.z = out_dims.y;
-              kernel =
-                  cu::row_reduce_small_warp<InType, OutType, OP, NDIM, N_READS>;
-            }
-          } else {
-            size_t num_threads = cuda::ceil_div(args.row_size, N_READS);
-            num_threads = cuda::ceil_div(num_threads, WARP_SIZE) * WARP_SIZE;
-            MLX_SWITCH_BLOCK_DIM(num_threads, BLOCK_DIM_X, {
-              num_blocks.y = out_dims.x;
-              num_blocks.z = out_dims.y;
-              block_dims.x = BLOCK_DIM_X;
-              kernel = cu::row_reduce_looped<
-                  InType,
-                  OutType,
-                  OP,
-                  NDIM,
-                  BLOCK_DIM_X,
-                  N_READS>;
-            });
-          }
-          kernel<<<num_blocks, block_dims, 0, stream>>>(
-              in.data<InType>(), out.data<OutType>(), out.size(), args);
-        });
-      });
-    });
-  });
+  // Fallback row reduce
+  row_reduce_looped(encoder, in, out, reduce_type, axes, plan, std::move(args));
 }
 
 } // namespace mlx::core

mlx/backend/cuda/reduce/segmented_reduce.cu

@@ -1,84 +0,0 @@
-// Copyright © 2025 Apple Inc.
-
-#include "mlx/backend/cuda/device.h"
-#include "mlx/backend/cuda/device/cast_op.cuh"
-#include "mlx/backend/cuda/reduce/reduce.cuh"
-
-#include <thrust/device_ptr.h>
-#include <cub/device/device_reduce.cuh>
-#include <cub/device/device_segmented_reduce.cuh>
-
-namespace mlx::core {
-
-template <typename... Args>
-void cub_all_reduce(cu::CommandEncoder& encoder, Args&&... args) {
-  // Allocate temporary storage.
-  size_t size;
-  CHECK_CUDA_ERROR(cub::DeviceReduce::Reduce(nullptr, size, args...));
-  array temp(allocator::malloc(size), {static_cast<int>(size)}, uint8);
-  encoder.add_temporary(temp);
-  // Run op.
-  CHECK_CUDA_ERROR(cub::DeviceReduce::Reduce(temp.data<void>(), size, args...));
-}
-
-template <typename... Args>
-void cub_segmented_reduce(cu::CommandEncoder& encoder, Args&&... args) {
-  // Allocate temporary storage.
-  size_t size;
-  CHECK_CUDA_ERROR(cub::DeviceSegmentedReduce::Reduce(nullptr, size, args...));
-  array temp(allocator::malloc(size), {static_cast<int>(size)}, uint8);
-  encoder.add_temporary(temp);
-  // Run op.
-  CHECK_CUDA_ERROR(
-      cub::DeviceSegmentedReduce::Reduce(temp.data<void>(), size, args...));
-}
-
-struct MultiplyOp {
-  int factor;
-  __device__ int operator()(int i) {
-    return i * factor;
-  }
-};
-
-void segmented_reduce(
-    cu::CommandEncoder& encoder,
-    const array& in,
-    array& out,
-    Reduce::ReduceType reduce_type,
-    const std::vector<int>& axes,
-    const ReductionPlan& plan) {
-  encoder.launch_kernel([&](cudaStream_t stream) {
-    MLX_SWITCH_ALL_TYPES(in.dtype(), CTYPE, {
-      MLX_SWITCH_REDUCE_OPS(reduce_type, OP, {
-        using InType = cuda_type_t<CTYPE>;
-        using OutType = cu::ReduceResult<OP, InType>::type;
-        auto in_iter = cu::make_cast_iterator<OutType>(
-            thrust::device_pointer_cast(in.data<InType>()));
-        auto out_ptr = thrust::device_pointer_cast(out.data<OutType>());
-        auto init = cu::ReduceInit<OP, InType>::value();
-
-        if (plan.type == ContiguousAllReduce) {
-          cub_all_reduce(
-              encoder, in_iter, out_ptr, in.data_size(), OP(), init, stream);
-        } else if (plan.type == ContiguousReduce) {
-          auto offsets = thrust::make_transform_iterator(
-              thrust::make_counting_iterator(0), MultiplyOp{plan.shape.back()});
-          cub_segmented_reduce(
-              encoder,
-              in_iter,
-              out_ptr,
-              out.size(),
-              offsets,
-              offsets + 1,
-              OP(),
-              init,
-              stream);
-        } else {
-          throw std::runtime_error("Unsupported plan in segmented_reduce.");
-        }
-      });
-    });
-  });
-}
-
-} // namespace mlx::core

mlx/backend/cuda/rms_norm.cu

@@ -225,19 +225,20 @@ void RMSNorm::eval_gpu(
   encoder.set_input_array(w);
   encoder.set_output_array(out);
   encoder.launch_kernel([&](cudaStream_t stream) {
-    MLX_SWITCH_FLOAT_TYPES_CHECKED(out.dtype(), "rms_norm", CTYPE, {
-      using DataType = cuda_type_t<CTYPE>;
+    dispatch_float_types(out.dtype(), "rms_norm", [&](auto type_tag) {
       constexpr uint32_t N_READS = 4;
-      MLX_SWITCH_BLOCK_DIM(cuda::ceil_div(axis_size, N_READS), BLOCK_DIM, {
-        auto kernel = cu::rms_norm<DataType, BLOCK_DIM, N_READS>;
-        kernel<<<n_rows, BLOCK_DIM, 0, stream>>>(
-            x.data<DataType>(),
-            w.data<DataType>(),
-            out.data<DataType>(),
-            eps_,
-            axis_size,
-            w_stride);
-      });
+      dispatch_block_dim(
+          cuda::ceil_div(axis_size, N_READS), [&](auto block_dim) {
+            using DataType = cuda_type_t<MLX_GET_TYPE(type_tag)>;
+            auto kernel = cu::rms_norm<DataType, block_dim(), N_READS>;
+            kernel<<<n_rows, block_dim(), 0, stream>>>(
+                x.data<DataType>(),
+                w.data<DataType>(),
+                out.data<DataType>(),
+                eps_,
+                axis_size,
+                w_stride);
+          });
     });
   });
 }
@@ -303,7 +304,6 @@ void RMSNormVJP::eval_gpu(
       encoder.add_temporary(gw_temp);
     }
   }
-  gw.set_data(allocator::malloc(gw.nbytes()));
 
   encoder.set_input_array(x);
   encoder.set_input_array(w);
@@ -311,22 +311,28 @@ void RMSNormVJP::eval_gpu(
   encoder.set_output_array(gx);
   encoder.set_output_array(gw_temp);
   encoder.launch_kernel([&, x = x, g = g](cudaStream_t stream) {
-    MLX_SWITCH_FLOAT_TYPES_CHECKED(gx.dtype(), "rms_norm_vjp", CTYPE, {
-      using DataType = cuda_type_t<CTYPE>;
-      constexpr int N_READS = 4;
-      MLX_SWITCH_BOOL(has_w, HAS_W, {
-        MLX_SWITCH_BLOCK_DIM(cuda::ceil_div(axis_size, N_READS), BLOCK_DIM, {
-          auto kernel = cu::rms_norm_vjp<DataType, HAS_W, BLOCK_DIM, N_READS>;
-          kernel<<<n_rows, BLOCK_DIM, 0, stream>>>(
-              x.data<DataType>(),
-              w.data<DataType>(),
-              g.data<DataType>(),
-              gx.data<DataType>(),
-              gw_temp.data<DataType>(),
-              eps_,
-              axis_size,
-              w_stride);
-        });
+    dispatch_float_types(gx.dtype(), "rms_norm_vjp", [&](auto type_tag) {
+      dispatch_bool(has_w, [&](auto has_w_constant) {
+        constexpr int N_READS = 4;
+        dispatch_block_dim(
+            cuda::ceil_div(axis_size, N_READS), [&](auto block_dim) {
+              using DataType = cuda_type_t<MLX_GET_TYPE(type_tag)>;
+              constexpr int N_READS = 4;
+              auto kernel = cu::rms_norm_vjp<
+                  DataType,
+                  has_w_constant(),
+                  block_dim(),
+                  N_READS>;
+              kernel<<<n_rows, block_dim(), 0, stream>>>(
+                  x.data<DataType>(),
+                  w.data<DataType>(),
+                  g.data<DataType>(),
+                  gx.data<DataType>(),
+                  gw_temp.data<DataType>(),
+                  eps_,
+                  axis_size,
+                  w_stride);
+            });
       });
     });
   });

mlx/backend/cuda/rope.cu

@@ -310,12 +310,12 @@ void RoPE::eval_gpu(
   encoder.set_input_array(offset);
   encoder.set_output_array(out);
   encoder.launch_kernel([&](cudaStream_t stream) {
-    MLX_SWITCH_FLOAT_TYPES_CHECKED(in.dtype(), "rope", CTYPE, {
-      using DataType = cuda_type_t<CTYPE>;
-      MLX_SWITCH_BOOL(traditional_, TRADITIONAL, {
-        MLX_SWITCH_BOOL(forward_, FORWARD, {
+    dispatch_float_types(out.dtype(), "rope", [&](auto type_tag) {
+      dispatch_bool(traditional_, [&](auto traditional) {
+        dispatch_bool(forward_, [&](auto forward) {
+          using DataType = cuda_type_t<MLX_GET_TYPE(type_tag)>;
           if (single && !with_freqs) {
-            auto kernel = cu::rope_single<DataType, TRADITIONAL, FORWARD>;
+            auto kernel = cu::rope_single<DataType, traditional(), forward()>;
             uint2 dims = make_uint2(dims_ / 2, in.size() / mat_size);
             auto [grid, block] = get_grid_and_block(dims.x, dims.y, 1);
             kernel<<<grid, block, 0, stream>>>(
@@ -327,7 +327,8 @@ void RoPE::eval_gpu(
                 mat_size,
                 dims);
           } else if (single) {
-            auto kernel = cu::rope_single_freqs<DataType, TRADITIONAL, FORWARD>;
+            auto kernel =
+                cu::rope_single_freqs<DataType, traditional(), forward()>;
             uint2 dims = make_uint2(dims_ / 2, in.size() / mat_size);
             auto [grid, block] = get_grid_and_block(dims.x, dims.y, 1);
             kernel<<<grid, block, 0, stream>>>(
@@ -340,7 +341,7 @@ void RoPE::eval_gpu(
                 dims,
                 inputs[2].strides(0));
           } else if (with_freqs) {
-            auto kernel = cu::rope_freqs<DataType, TRADITIONAL, FORWARD>;
+            auto kernel = cu::rope_freqs<DataType, traditional(), forward()>;
             uint3 dims =
                 make_uint3(dims_ / 2, in.shape(-2), in.size() / mat_size);
             dims.z = (dims.z + 3) / 4;
@@ -358,7 +359,7 @@ void RoPE::eval_gpu(
                 dims,
                 inputs[2].strides(0));
           } else {
-            auto kernel = cu::rope<DataType, TRADITIONAL, FORWARD>;
+            auto kernel = cu::rope<DataType, traditional(), forward()>;
             uint3 dims =
                 make_uint3(dims_ / 2, in.shape(-2), in.size() / mat_size);
             dims.z = (dims.z + 3) / 4;

mlx/backend/cuda/softmax.cu

@@ -51,7 +51,7 @@ __global__ void softmax(const T* in, T* out, int axis_size) {
         make_cast_iterator<AccT>(in),
         vals,
         axis_size,
-        Limits<AccT>::finite_min());
+        Limits<AccT>::min());
     prevmax = maxval;
     maxval = max_op(maxval, cub::ThreadReduce(vals, max_op));
     // Online normalizer calculation for softmax:
@@ -79,7 +79,7 @@ __global__ void softmax(const T* in, T* out, int axis_size) {
   block.sync();
   maxval = warp.thread_rank() < warp.meta_group_size()
       ? local_max[warp.thread_rank()]
-      : Limits<AccT>::finite_min();
+      : Limits<AccT>::min();
   maxval = cg::reduce(warp, maxval, max_op);
   normalizer = normalizer * softmax_exp(prevmax - maxval);
   if (warp.thread_rank() == 0) {
@@ -142,17 +142,18 @@ void Softmax::eval_gpu(const std::vector<array>& inputs, array& out) {
   encoder.set_input_array(in);
   encoder.set_output_array(out);
   encoder.launch_kernel([&](cudaStream_t stream) {
-    MLX_SWITCH_FLOAT_TYPES_CHECKED(out.dtype(), "softmax", CTYPE, {
-      using DataType = cuda_type_t<CTYPE>;
+    dispatch_float_types(out.dtype(), "softmax", [&](auto type_tag) {
       constexpr int N_READS = 4;
-      MLX_SWITCH_BLOCK_DIM(cuda::ceil_div(axis_size, N_READS), BLOCK_DIM, {
-        auto kernel = cu::softmax<DataType, DataType, BLOCK_DIM, N_READS>;
-        if (precise) {
-          kernel = cu::softmax<DataType, float, BLOCK_DIM, N_READS>;
-        }
-        kernel<<<n_rows, BLOCK_DIM, 0, stream>>>(
-            in.data<DataType>(), out.data<DataType>(), axis_size);
-      });
+      dispatch_block_dim(
+          cuda::ceil_div(axis_size, N_READS), [&](auto block_dim) {
+            using DataType = cuda_type_t<MLX_GET_TYPE(type_tag)>;
+            auto kernel = cu::softmax<DataType, DataType, block_dim(), N_READS>;
+            if (precise) {
+              kernel = cu::softmax<DataType, float, block_dim(), N_READS>;
+            }
+            kernel<<<n_rows, block_dim(), 0, stream>>>(
+                in.data<DataType>(), out.data<DataType>(), axis_size);
+          });
     });
   });
 }

mlx/backend/cuda/sort.cu

@@ -76,17 +76,21 @@ void segmented_sort(cu::CommandEncoder& encoder, Args&&... args) {
       temp.data<void>(), size, args...));
 }
 
+struct OffsetTransform {
+  int nsort;
+
+  int __device__ operator()(int i) {
+    return i * nsort;
+  }
+};
+
 void gpu_sort(const Stream& s, array in, array& out_, int axis, bool argsort) {
   array out = out_;
   auto& encoder = cu::get_command_encoder(s);
-  encoder.set_input_array(in);
-  encoder.set_output_array(out);
-
   if (axis < 0) {
     axis += in.ndim();
   }
   int nsort = in.shape(axis);
-  int nsegments = in.data_size() / nsort;
   int last_dim = in.ndim() - 1;
 
   // If we are not sorting the innermost dimension of a contiguous array,
@@ -100,16 +104,22 @@ void gpu_sort(const Stream& s, array in, array& out_, int axis, bool argsort) {
     out = array(allocator::malloc(out.nbytes()), in.shape(), out.dtype());
     encoder.add_temporary(out);
   } else {
-    out.set_data(allocator::malloc(out.nbytes()));
+    out.set_data(
+        allocator::malloc(in.data_size() * out.itemsize()),
+        in.data_size(),
+        in.strides(),
+        in.flags());
   }
 
+  encoder.set_input_array(in);
+  encoder.set_output_array(out);
   encoder.launch_kernel([&](cudaStream_t stream) {
-    MLX_SWITCH_ALL_TYPES(in.dtype(), CTYPE, {
+    dispatch_all_types(in.dtype(), [&](auto type_tag) {
+      using CTYPE = MLX_GET_TYPE(type_tag);
       if constexpr (!std::is_same_v<CTYPE, complex64_t>) {
         using Type = cuda_type_t<CTYPE>;
         auto offsets = thrust::make_transform_iterator(
-            thrust::make_counting_iterator(0),
-            [nsort] __device__(int i) { return i * nsort; });
+            thrust::make_counting_iterator(0), OffsetTransform{nsort});
         if (argsort) {
           // Indices in the sorted dimension.
           array indices(
@@ -134,7 +144,7 @@ void gpu_sort(const Stream& s, array in, array& out_, int axis, bool argsort) {
               indices.data<uint32_t>(),
               out.data<uint32_t>(),
               in.data_size(),
-              nsegments,
+              in.data_size() / nsort,
               offsets,
               offsets + 1,
               stream);
@@ -144,7 +154,7 @@ void gpu_sort(const Stream& s, array in, array& out_, int axis, bool argsort) {
               in.data<Type>(),
               out.data<Type>(),
               in.data_size(),
-              nsegments,
+              in.data_size() / nsort,
               offsets,
               offsets + 1,
               stream);
@@ -177,4 +187,14 @@ void Sort::eval_gpu(const std::vector<array>& inputs, array& out) {
   gpu_sort(stream(), inputs[0], out, axis_, false);
 }
 
+void ArgPartition::eval_gpu(const std::vector<array>& inputs, array& out) {
+  nvtx3::scoped_range r("ArgPartition::eval_gpu");
+  gpu_sort(stream(), inputs[0], out, axis_, true);
+}
+
+void Partition::eval_gpu(const std::vector<array>& inputs, array& out) {
+  nvtx3::scoped_range r("Partition::eval_gpu");
+  gpu_sort(stream(), inputs[0], out, axis_, false);
+}
+
 } // namespace mlx::core

mlx/backend/cuda/ternary.cu

@@ -92,58 +92,63 @@ void ternary_op_gpu_inplace(
   encoder.set_input_array(c);
   encoder.set_output_array(out);
   encoder.launch_kernel([&](cudaStream_t stream) {
-    MLX_SWITCH_ALL_TYPES(out.dtype(), CTYPE, {
-      using DType = cuda_type_t<CTYPE>;
+    dispatch_all_types(out.dtype(), [&](auto type_tag) {
+      using DType = cuda_type_t<MLX_GET_TYPE(type_tag)>;
 
       auto topt = get_ternary_op_type(a, b, c);
       if (topt == TernaryOpType::General) {
-        auto [shape, strides] = collapse_contiguous_dims(a, b, c, out);
-        auto& a_strides = strides[0];
-        auto& b_strides = strides[1];
-        auto& c_strides = strides[2];
-        bool large = a.data_size() > INT32_MAX || b.data_size() > INT32_MAX ||
-            c.data_size() > INT32_MAX || out.data_size() > INT32_MAX;
-        MLX_SWITCH_BOOL(large, LARGE, {
-          using IdxT = std::conditional_t<LARGE, int64_t, int32_t>;
-          int ndim = shape.size();
-          if (ndim <= 3) {
-            MLX_SWITCH_1_2_3(ndim, NDIM, {
-              auto kernel = cu::ternary_g_nd<Op, DType, IdxT, NDIM>;
-              auto [num_blocks, block_dims] =
-                  get_launch_args(kernel, out, large);
-              kernel<<<num_blocks, block_dims, 0, stream>>>(
-                  a.data<bool>(),
-                  b.data<DType>(),
-                  c.data<DType>(),
-                  out.data<DType>(),
-                  out.size(),
-                  const_param<NDIM>(shape),
-                  const_param<NDIM>(a_strides),
-                  const_param<NDIM>(b_strides),
-                  const_param<NDIM>(c_strides));
+        dispatch_bool(
+            a.data_size() > INT32_MAX || b.data_size() > INT32_MAX ||
+                c.data_size() > INT32_MAX || out.data_size() > INT32_MAX,
+            [&](auto large) {
+              using IdxT = std::conditional_t<large(), int64_t, int32_t>;
+              Shape shape;
+              std::vector<Strides> strides;
+              std::tie(shape, strides) = collapse_contiguous_dims(a, b, c, out);
+              auto& a_strides = strides[0];
+              auto& b_strides = strides[1];
+              auto& c_strides = strides[2];
+              int ndim = shape.size();
+              if (ndim <= 3) {
+                dispatch_1_2_3(ndim, [&](auto dims_constant) {
+                  auto kernel =
+                      cu::ternary_g_nd<Op, DType, IdxT, dims_constant()>;
+                  auto [num_blocks, block_dims] =
+                      get_launch_args(kernel, out, large());
+                  kernel<<<num_blocks, block_dims, 0, stream>>>(
+                      a.data<bool>(),
+                      b.data<DType>(),
+                      c.data<DType>(),
+                      out.data<DType>(),
+                      out.size(),
+                      const_param<dims_constant()>(shape),
+                      const_param<dims_constant()>(a_strides),
+                      const_param<dims_constant()>(b_strides),
+                      const_param<dims_constant()>(c_strides));
+                });
+              } else {
+                auto kernel = cu::ternary_g<Op, DType, IdxT>;
+                auto [num_blocks, block_dims] =
+                    get_launch_args(kernel, out, large());
+                kernel<<<num_blocks, block_dims, 0, stream>>>(
+                    a.data<bool>(),
+                    b.data<DType>(),
+                    c.data<DType>(),
+                    out.data<DType>(),
+                    out.data_size(),
+                    const_param(shape),
+                    const_param(a_strides),
+                    const_param(b_strides),
+                    const_param(c_strides),
+                    ndim);
+              }
             });
-          } else {
-            auto kernel = cu::ternary_g<Op, DType, IdxT>;
-            auto [num_blocks, block_dims] = get_launch_args(kernel, out, large);
-            kernel<<<num_blocks, block_dims, 0, stream>>>(
-                a.data<bool>(),
-                b.data<DType>(),
-                c.data<DType>(),
-                out.data<DType>(),
-                out.data_size(),
-                const_param(shape),
-                const_param(a_strides),
-                const_param(b_strides),
-                const_param(c_strides),
-                ndim);
-          }
-        });
       } else {
-        MLX_SWITCH_BOOL(out.data_size() > UINT32_MAX, LARGE, {
-          using IdxT = std::conditional_t<LARGE, int64_t, uint32_t>;
+        dispatch_bool(out.data_size() > INT32_MAX, [&](auto large) {
+          using IdxT = std::conditional_t<large(), int64_t, uint32_t>;
           auto kernel = cu::ternary_v<Op, DType, IdxT>;
           auto [num_blocks, block_dims] = get_launch_args(
-              kernel, out.data_size(), out.shape(), out.strides(), LARGE);
+              kernel, out.data_size(), out.shape(), out.strides(), large());
           kernel<<<num_blocks, block_dims, 0, stream>>>(
               a.data<bool>(),
               b.data<DType>(),

mlx/backend/cuda/unary.cu

@@ -20,38 +20,35 @@ namespace cu {
 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, Sign> || std::is_same_v<Op, Square>) {
     return std::is_same_v<In, Out>;
   }
-  if (std::is_same_v<Op, ArcCos> || std::is_same_v<Op, ArcCosh> ||
-      std::is_same_v<Op, ArcSin> || std::is_same_v<Op, ArcSinh> ||
-      std::is_same_v<Op, ArcTan> || 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> ||
-      std::is_same_v<Op, Sqrt> || std::is_same_v<Op, Rsqrt>) {
+  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, Log> || std::is_same_v<Op, Log2> ||
-      std::is_same_v<Op, Log10> || std::is_same_v<Op, Log1p>) {
-    return std::is_same_v<In, Out> && is_inexact_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> ||
-      std::is_same_v<Op, Square>) {
+  if (std::is_same_v<Op, Ceil> || std::is_same_v<Op, Floor>) {
     return std::is_same_v<In, Out> && !std::is_same_v<In, complex64_t>;
   }
   if (std::is_same_v<Op, Conjugate>) {
     return std::is_same_v<In, Out> && std::is_same_v<In, complex64_t>;
   }
-  if (std::is_same_v<Op, Cos> || std::is_same_v<Op, Cosh> ||
-      std::is_same_v<Op, Exp> || std::is_same_v<Op, Round> ||
-      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_floating_v<In> || std::is_same_v<In, complex64_t>);
+  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 std::is_same_v<In, complex64_t> && std::is_same_v<Out, float>;
@@ -79,8 +76,10 @@ void unary_op_gpu_inplace(
   encoder.set_input_array(in);
   encoder.set_output_array(out);
   encoder.launch_kernel([&](cudaStream_t stream) {
-    MLX_SWITCH_ALL_TYPES(in.dtype(), CTYPE_IN, {
-      MLX_SWITCH_ALL_TYPES(out.dtype(), CTYPE_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>()) {
           using InType = cuda_type_t<CTYPE_IN>;
           using OutType = cuda_type_t<CTYPE_OUT>;

mlx/backend/cuda/utils.cpp

@@ -25,22 +25,38 @@ void check_cuda_error(const char* name, cudaError_t err) {
 }
 
 const char* dtype_to_cuda_type(const Dtype& dtype) {
-  if (dtype == float16) {
-    return "__half";
+  switch (dtype) {
+    case bool_:
+      return "bool";
+    case int8:
+      return "int8_t";
+    case int16:
+      return "int16_t";
+    case int32:
+      return "int32_t";
+    case int64:
+      return "int64_t";
+    case uint8:
+      return "uint8_t";
+    case uint16:
+      return "uint16_t";
+    case uint32:
+      return "uint32_t";
+    case uint64:
+      return "uint64_t";
+    case float16:
+      return "__half";
+    case bfloat16:
+      return "__nv_bfloat16";
+    case float32:
+      return "float";
+    case float64:
+      return "double";
+    case complex64:
+      return "cuComplex";
+    default:
+      return "unknown";
   }
-  if (dtype == bfloat16) {
-    return "__nv_bfloat16";
-  }
-  if (dtype == complex64) {
-    return "cuComplex";
-  }
-#define SPECIALIZE_DtypeToString(CPP_TYPE, DTYPE) \
-  if (dtype == DTYPE) {                           \
-    return #CPP_TYPE;                             \
-  }
-  MLX_FORALL_DTYPES(SPECIALIZE_DtypeToString)
-#undef SPECIALIZE_DtypeToString
-  return nullptr;
 }
 
 } // namespace mlx::core

mlx/backend/cuda/worker.cpp

@@ -80,7 +80,9 @@ void Worker::thread_fn() {
       }
       worker_tasks_.erase(worker_tasks_.begin(), end);
     }
-    for (auto& task : tasks) {
+    // Make sure tasks are cleared before the next wait
+    for (int i = 0; i < tasks.size(); ++i) {
+      auto task = std::move(tasks[i]);
       task();
     }
     worker_event_.wait(batch + 1);

mlx/backend/rocm/CMakeLists.txt

@@ -1,85 +0,0 @@
-# Filename rules in ROCm backend:
-#
-# * Use .hip/.hpp if code contains device code, and .cpp/.h if not.
-# * Device-only code should be put in device/ subdir.
-# * Files in device/ subdir should not include files outside.
-target_sources(
-  mlx
-  PRIVATE ${CMAKE_CURRENT_SOURCE_DIR}/allocator.cpp
-          ${CMAKE_CURRENT_SOURCE_DIR}/arg_reduce.hip
-          ${CMAKE_CURRENT_SOURCE_DIR}/binary.hip
-          ${CMAKE_CURRENT_SOURCE_DIR}/compiled.cpp
-          ${CMAKE_CURRENT_SOURCE_DIR}/copy.hip
-          ${CMAKE_CURRENT_SOURCE_DIR}/device.cpp
-          ${CMAKE_CURRENT_SOURCE_DIR}/eval.cpp
-          ${CMAKE_CURRENT_SOURCE_DIR}/event.hip
-          ${CMAKE_CURRENT_SOURCE_DIR}/fence.cpp
-          ${CMAKE_CURRENT_SOURCE_DIR}/indexing.cpp
-          ${CMAKE_CURRENT_SOURCE_DIR}/kernel_utils.hip
-          ${CMAKE_CURRENT_SOURCE_DIR}/matmul.cpp
-          ${CMAKE_CURRENT_SOURCE_DIR}/layer_norm.hip
-          ${CMAKE_CURRENT_SOURCE_DIR}/logsumexp.hip
-          ${CMAKE_CURRENT_SOURCE_DIR}/primitives.hip
-          ${CMAKE_CURRENT_SOURCE_DIR}/random.hip
-          ${CMAKE_CURRENT_SOURCE_DIR}/reduce.hip
-          ${CMAKE_CURRENT_SOURCE_DIR}/rms_norm.hip
-          ${CMAKE_CURRENT_SOURCE_DIR}/rope.hip
-          ${CMAKE_CURRENT_SOURCE_DIR}/rocm.cpp
-          ${CMAKE_CURRENT_SOURCE_DIR}/slicing.cpp
-          ${CMAKE_CURRENT_SOURCE_DIR}/softmax.hip
-          ${CMAKE_CURRENT_SOURCE_DIR}/sort.hip
-          ${CMAKE_CURRENT_SOURCE_DIR}/ternary.hip
-          ${CMAKE_CURRENT_SOURCE_DIR}/unary.hip
-          ${CMAKE_CURRENT_SOURCE_DIR}/utils.cpp
-          ${CMAKE_CURRENT_SOURCE_DIR}/worker.cpp)
-
-target_compile_definitions(mlx PRIVATE MLX_USE_ROCM)
-
-# Embed kernel sources in binary for JIT compilation.
-file(
-  GLOB MLX_JIT_SOURCES
-  RELATIVE ${CMAKE_CURRENT_SOURCE_DIR}
-  "${CMAKE_CURRENT_SOURCE_DIR}/device/*.h"
-  "${CMAKE_CURRENT_SOURCE_DIR}/device/*.hpp")
-string(JOIN ":" MLX_JIT_SOURCES_ARG ${MLX_JIT_SOURCES})
-add_custom_command(
-  OUTPUT gen/rocm_jit_sources.h
-  COMMAND
-    ${CMAKE_COMMAND} -DMLX_SOURCE_ROOT=${CMAKE_CURRENT_SOURCE_DIR}
-    -DMLX_JIT_SOURCES=${MLX_JIT_SOURCES_ARG} -P
-    "${CMAKE_CURRENT_SOURCE_DIR}/bin2h.cmake"
-  DEPENDS bin2h.cmake ${MLX_JIT_SOURCES})
-add_custom_target(rocm_jit_sources DEPENDS gen/rocm_jit_sources.h)
-add_dependencies(mlx rocm_jit_sources)
-target_include_directories(mlx PRIVATE "${CMAKE_CURRENT_BINARY_DIR}/gen")
-
-# Find ROCm installation
-find_package(hip REQUIRED)
-find_package(rocblas REQUIRED)
-
-# Link with ROCm libraries
-target_link_libraries(mlx PRIVATE hip::device roc::rocblas)
-
-# Set GPU architectures for ROCm Common ROCm architectures: gfx900, gfx906,
-# gfx908, gfx90a, gfx1030, gfx1100
-set(MLX_ROCM_ARCHITECTURES
-    "gfx900;gfx906;gfx908;gfx90a;gfx1030;gfx1100"
-    CACHE STRING "ROCm GPU architectures")
-message(STATUS "ROCm GPU architectures: ${MLX_ROCM_ARCHITECTURES}")
-
-# Set GPU targets for HIP compilation
-set_property(TARGET mlx PROPERTY HIP_ARCHITECTURES "${MLX_ROCM_ARCHITECTURES}")
-
-# Enable HIP language support
-enable_language(HIP)
-
-# Set HIP compiler flags
-target_compile_options(
-  mlx
-  PRIVATE "$<$<COMPILE_LANGUAGE:HIP>:-fgpu-rdc>"
-          "$<$<COMPILE_LANGUAGE:HIP>:-Xcompiler=-Wall>"
-          "$<$<COMPILE_LANGUAGE:HIP>:-Xcompiler=-Wextra>")
-
-# Add ROCm include directories
-target_include_directories(mlx PRIVATE ${hip_INCLUDE_DIRS})
-target_include_directories(mlx PRIVATE ${rocblas_INCLUDE_DIRS})

mlx/backend/rocm/allocator.cpp

@@ -1,206 +0,0 @@
-// Copyright © 2025 Apple Inc.
-
-#include "mlx/backend/rocm/allocator.h"
-#include "mlx/backend/rocm/utils.h"
-#include "mlx/backend/rocm/worker.h"
-
-#include <fmt/format.h>
-#include <hip/hip_runtime.h>
-#include <unistd.h>
-
-#include <cassert>
-
-namespace mlx::core {
-
-namespace rocm {
-
-RocmAllocator::RocmAllocator()
-    : buffer_cache_(
-          getpagesize(),
-          [](RocmBuffer* buf) { return buf->size; },
-          [this](RocmBuffer* buf) {
-            rocm_free(buf->data);
-            delete buf;
-          }) {
-  // TODO: Set memory limit for multi-device.
-  size_t free, total;
-  CHECK_HIP_ERROR(hipMemGetInfo(&free, &total));
-  memory_limit_ = total * 0.8;
-  max_pool_size_ = memory_limit_;
-}
-
-Buffer RocmAllocator::malloc(size_t size) {
-  // Find available buffer from cache.
-  std::unique_lock lock(mutex_);
-  RocmBuffer* buf = buffer_cache_.reuse_from_cache(size);
-  if (!buf) {
-    // If we have a lot of memory pressure or are over the maximum cache size,
-    // try to reclaim memory from the cache.
-    size_t mem_required = get_active_memory() + get_cache_memory() + size;
-    if (mem_required >= memory_limit_) {
-      buffer_cache_.release_cached_buffers(mem_required - memory_limit_);
-    }
-
-    lock.unlock();
-    buf = new RocmBuffer{nullptr, size};
-    hipError_t err = hipMallocManaged(&buf->data, size);
-    if (err != hipSuccess && err != hipErrorMemoryAllocation) {
-      throw std::runtime_error(
-          fmt::format("hipMallocManaged failed: {}.", hipGetErrorString(err)));
-    }
-    lock.lock();
-  }
-  active_memory_ += size;
-  peak_memory_ = std::max(active_memory_, peak_memory_);
-
-  // Maintain the cache below the requested limit.
-  if (get_cache_memory() > max_pool_size_) {
-    buffer_cache_.release_cached_buffers(get_cache_memory() - max_pool_size_);
-  }
-
-  return Buffer{buf};
-}
-
-void RocmAllocator::free(Buffer buffer) {
-  auto* buf = static_cast<RocmBuffer*>(buffer.ptr());
-  if (!buf) {
-    return;
-  }
-
-  std::unique_lock lock(mutex_);
-  active_memory_ -= buf->size;
-  if (get_cache_memory() < max_pool_size_) {
-    buffer_cache_.recycle_to_cache(buf);
-  } else {
-    lock.unlock();
-    rocm_free(buf->data);
-    delete buf;
-  }
-}
-
-size_t RocmAllocator::size(Buffer buffer) const {
-  auto* buf = static_cast<RocmBuffer*>(buffer.ptr());
-  if (!buf) {
-    return 0;
-  }
-  return buf->size;
-}
-
-void RocmAllocator::register_this_thread() {
-  std::lock_guard lock(worker_mutex_);
-  allowed_threads_.insert(std::this_thread::get_id());
-}
-
-void RocmAllocator::rocm_free(void* buf) {
-  // If rocm_free() is called from a unregistered thread, reschedule the call to
-  // worker.
-  {
-    std::lock_guard lock(worker_mutex_);
-    if (allowed_threads_.count(std::this_thread::get_id()) == 0) {
-      if (!worker_) {
-        worker_.reset(new Worker);
-      }
-      worker_->add_task([this, buf]() { this->rocm_free(buf); });
-      worker_->end_batch();
-      worker_->commit();
-      return;
-    }
-  }
-
-  hipFree(buf);
-}
-
-size_t RocmAllocator::get_active_memory() const {
-  return active_memory_;
-}
-
-size_t RocmAllocator::get_peak_memory() const {
-  return peak_memory_;
-}
-
-void RocmAllocator::reset_peak_memory() {
-  std::lock_guard lock(mutex_);
-  peak_memory_ = 0;
-}
-
-size_t RocmAllocator::get_memory_limit() {
-  return memory_limit_;
-}
-
-size_t RocmAllocator::set_memory_limit(size_t limit) {
-  std::lock_guard lock(mutex_);
-  std::swap(limit, memory_limit_);
-  return limit;
-}
-
-size_t RocmAllocator::get_cache_memory() const {
-  return buffer_cache_.cache_size();
-}
-
-size_t RocmAllocator::set_cache_limit(size_t limit) {
-  std::lock_guard lk(mutex_);
-  std::swap(limit, max_pool_size_);
-  return limit;
-}
-
-void RocmAllocator::clear_cache() {
-  std::lock_guard lk(mutex_);
-  buffer_cache_.clear();
-}
-
-RocmAllocator& allocator() {
-  // By creating the |allocator_| on heap, the destructor of RocmAllocator
-  // will not be called on exit and buffers in the cache will be leaked. This
-  // can save some time at program exit.
-  static RocmAllocator* allocator_ = new RocmAllocator;
-  return *allocator_;
-}
-
-} // namespace rocm
-
-namespace allocator {
-
-Allocator& allocator() {
-  return rocm::allocator();
-}
-
-void* Buffer::raw_ptr() {
-  if (!ptr_) {
-    return nullptr;
-  }
-  return static_cast<rocm::RocmBuffer*>(ptr_)->data;
-}
-
-} // namespace allocator
-
-size_t get_active_memory() {
-  return rocm::allocator().get_active_memory();
-}
-size_t get_peak_memory() {
-  return rocm::allocator().get_peak_memory();
-}
-void reset_peak_memory() {
-  return rocm::allocator().reset_peak_memory();
-}
-size_t set_memory_limit(size_t limit) {
-  return rocm::allocator().set_memory_limit(limit);
-}
-size_t get_memory_limit() {
-  return rocm::allocator().get_memory_limit();
-}
-size_t get_cache_memory() {
-  return rocm::allocator().get_cache_memory();
-}
-size_t set_cache_limit(size_t limit) {
-  return rocm::allocator().set_cache_limit(limit);
-}
-void clear_cache() {
-  rocm::allocator().clear_cache();
-}
-
-// Not supported in ROCm.
-size_t set_wired_limit(size_t) {
-  return 0;
-}
-
-} // namespace mlx::core

mlx/backend/rocm/allocator.h

@@ -1,67 +0,0 @@
-// Copyright © 2025 Apple Inc.
-
-#pragma once
-
-#include "mlx/allocator.h"
-#include "mlx/backend/common/buffer_cache.h"
-
-#include <mutex>
-#include <set>
-#include <thread>
-#include <utility>
-
-namespace mlx::core::rocm {
-
-class Worker;
-
-using allocator::Buffer;
-
-// Stores ROCm-managed unified memory.
-struct RocmBuffer {
-  void* data;
-  size_t size;
-};
-
-class RocmAllocator : public allocator::Allocator {
- public:
-  Buffer malloc(size_t size) override;
-  void free(Buffer buffer) override;
-  size_t size(Buffer buffer) const override;
-
-  // Register current thread as safe to free buffers.
-  // In ROCm freeing a buffer implicitly synchronizes stream, and for threads
-  // that may be waited by gpu stream (for example cpu stream threads), freeing
-  // buffers there would result in dead lock.
-  void register_this_thread();
-
-  // Call hipFree in the safe thread.
-  void rocm_free(void* buf);
-
-  size_t get_active_memory() const;
-  size_t get_peak_memory() const;
-  void reset_peak_memory();
-  size_t get_memory_limit();
-  size_t set_memory_limit(size_t limit);
-  size_t get_cache_memory() const;
-  size_t set_cache_limit(size_t limit);
-  void clear_cache();
-
- private:
-  RocmAllocator();
-  friend RocmAllocator& allocator();
-
-  std::mutex worker_mutex_;
-  std::unique_ptr<Worker> worker_;
-  std::set<std::thread::id> allowed_threads_;
-
-  std::mutex mutex_;
-  size_t memory_limit_;
-  size_t max_pool_size_;
-  BufferCache<RocmBuffer> buffer_cache_;
-  size_t active_memory_{0};
-  size_t peak_memory_{0};
-};
-
-RocmAllocator& allocator();
-
-} // namespace mlx::core::rocm

mlx/backend/rocm/arg_reduce.hip

@@ -1,28 +0,0 @@
-// Copyright © 2025 Apple Inc.
-
-#include <hip/hip_runtime.h>
-
-namespace mlx::core::rocm {
-
-__global__ void argmax_kernel(float* input, int* output, int n) {
-  int idx = blockIdx.x * blockDim.x + threadIdx.x;
-  
-  // Simple argmax placeholder
-  if (idx == 0) {
-    int max_idx = 0;
-    float max_val = input[0];
-    for (int i = 1; i < n; i++) {
-      if (input[i] > max_val) {
-        max_val = input[i];
-        max_idx = i;
-      }
-    }
-    output[0] = max_idx;
-  }
-}
-
-void launch_argmax(float* input, int* output, int n, hipStream_t stream) {
-  hipLaunchKernelGGL(argmax_kernel, dim3(1), dim3(1), 0, stream, input, output, n);
-}
-
-} // namespace mlx::core::rocm 

mlx/backend/rocm/bin2h.cmake

@@ -1,47 +0,0 @@
-# Copyright © 2025 Apple Inc.
-
-# Script to embed kernel source files as header for JIT compilation
-
-set(MLX_OUTPUT_FILE "${CMAKE_CURRENT_BINARY_DIR}/gen/rocm_jit_sources.h")
-set(MLX_KERNEL_HEADER
-    "#pragma once\n\n#include <unordered_map>\n#include <string>\n\nnamespace mlx::core::rocm {\n\n"
-)
-set(MLX_KERNEL_FOOTER "\n} // namespace mlx::core::rocm\n")
-
-# Create output directory
-get_filename_component(MLX_OUTPUT_DIR ${MLX_OUTPUT_FILE} DIRECTORY)
-file(MAKE_DIRECTORY ${MLX_OUTPUT_DIR})
-
-# Write header
-file(WRITE ${MLX_OUTPUT_FILE} ${MLX_KERNEL_HEADER})
-
-# Process JIT sources
-string(REPLACE ":" ";" MLX_JIT_SOURCES_LIST ${MLX_JIT_SOURCES})
-
-set(MLX_SOURCE_MAP
-    "const std::unordered_map<std::string, std::string> kernel_sources = {\n")
-
-foreach(source IN LISTS MLX_JIT_SOURCES_LIST)
-  set(source_file "${MLX_SOURCE_ROOT}/${source}")
-  if(EXISTS ${source_file})
-    # Read source file
-    file(READ ${source_file} source_content)
-
-    # Escape content for C++ string literal
-    string(REPLACE "\\" "\\\\" source_content "${source_content}")
-    string(REPLACE "\"" "\\\"" source_content "${source_content}")
-    string(REPLACE "\n" "\\n\"\n\"" source_content "${source_content}")
-
-    # Add to map
-    set(MLX_SOURCE_MAP
-        "${MLX_SOURCE_MAP}  {\"${source}\", \"${source_content}\"},\n")
-  endif()
-endforeach()
-
-set(MLX_SOURCE_MAP "${MLX_SOURCE_MAP}};\n")
-
-# Write source map
-file(APPEND ${MLX_OUTPUT_FILE} ${MLX_SOURCE_MAP})
-
-# Write footer
-file(APPEND ${MLX_OUTPUT_FILE} ${MLX_KERNEL_FOOTER})

mlx/backend/rocm/binary.hip

@@ -1,312 +0,0 @@
-// Copyright © 2025 Apple Inc.
-
-#include "mlx/backend/common/binary.h"
-#include "mlx/backend/rocm/device.h"
-#include "mlx/backend/rocm/device/binary_ops.hpp"
-#include "mlx/backend/rocm/kernel_utils.hpp"
-#include "mlx/dtype_utils.h"
-#include "mlx/primitives.h"
-
-#include <hip/hip_cooperative_groups.h>
-
-namespace mlx::core {
-
-namespace rocm {
-
-namespace cg = cooperative_groups;
-
-template <typename Op, typename In, typename Out, typename IdxT>
-__global__ void binary_ss(const In* a, const In* b, Out* out, IdxT size) {
-  IdxT index = cg::this_grid().thread_rank();
-  if (index < size) {
-    out[index] = Op{}(a[0], b[0]);
-  }
-}
-
-template <typename Op, typename In, typename Out, typename IdxT>
-__global__ void binary_sv(const In* a, const In* b, Out* out, IdxT size) {
-  IdxT index = cg::this_grid().thread_rank();
-  if (index < size) {
-    out[index] = Op{}(a[0], b[index]);
-  }
-}
-
-template <typename Op, typename In, typename Out, typename IdxT>
-__global__ void binary_vs(const In* a, const In* b, Out* out, IdxT size) {
-  IdxT index = cg::this_grid().thread_rank();
-  if (index < size) {
-    out[index] = Op{}(a[index], b[0]);
-  }
-}
-
-template <typename Op, typename In, typename Out, typename IdxT>
-__global__ void binary_vv(const In* a, const In* b, Out* out, IdxT size) {
-  IdxT index = cg::this_grid().thread_rank();
-  if (index < size) {
-    out[index] = Op{}(a[index], b[index]);
-  }
-}
-
-template <typename Op, typename In, typename Out, typename IdxT, int NDIM>
-__global__ void binary_g_nd(
-    const In* a,
-    const In* b,
-    Out* out,
-    IdxT size,
-    const hip_array<int32_t, NDIM> shape,
-    const hip_array<int64_t, NDIM> a_strides,
-    const hip_array<int64_t, NDIM> b_strides) {
-  IdxT index = cg::this_grid().thread_rank();
-  if (index < size) {
-    auto [a_idx, b_idx] = elem_to_loc_nd<NDIM>(
-        index, shape.data(), a_strides.data(), b_strides.data());
-    out[index] = Op{}(a[a_idx], b[b_idx]);
-  }
-}
-
-template <typename Op, typename In, typename Out, typename IdxT>
-__global__ void binary_g(
-    const In* a,
-    const In* b,
-    Out* out,
-    IdxT size,
-    const hip_array<int32_t, MAX_DIMS> shape,
-    const hip_array<int64_t, MAX_DIMS> a_strides,
-    const hip_array<int64_t, MAX_DIMS> b_strides,
-    int ndim) {
-  IdxT index = cg::this_grid().thread_rank();
-  if (index < size) {
-    auto [a_idx, b_idx] = elem_to_loc_4d(
-        index, shape.data(), a_strides.data(), b_strides.data(), ndim);
-    out[index] = Op{}(a[a_idx], b[b_idx]);
-  }
-}
-
-// Binary operation support checking
-template <typename Op, typename In, typename Out>
-constexpr bool supports_binary_op() {
-  if (std::is_same_v<Op, Add> || std::is_same_v<Op, Divide> ||
-      std::is_same_v<Op, Maximum> || std::is_same_v<Op, Minimum> ||
-      std::is_same_v<Op, Multiply> || std::is_same_v<Op, Subtract> ||
-      std::is_same_v<Op, Power> || std::is_same_v<Op, Remainder>) {
-    return std::is_same_v<In, Out>;
-  }
-  if (std::is_same_v<Op, Equal> || std::is_same_v<Op, Greater> ||
-      std::is_same_v<Op, GreaterEqual> || std::is_same_v<Op, Less> ||
-      std::is_same_v<Op, LessEqual> || std::is_same_v<Op, NotEqual>) {
-    return std::is_same_v<Out, bool>;
-  }
-  if (std::is_same_v<Op, LogicalAnd> || std::is_same_v<Op, LogicalOr>) {
-    return std::is_same_v<Out, bool> && std::is_same_v<In, bool>;
-  }
-  if (std::is_same_v<Op, NaNEqual>) {
-    return std::is_same_v<Out, bool> && is_inexact_v<In>;
-  }
-  if (std::is_same_v<Op, LogAddExp>) {
-    return std::is_same_v<In, Out> && is_inexact_v<In>;
-  }
-  if (std::is_same_v<Op, ArcTan2>) {
-    return std::is_same_v<In, Out> && is_floating_v<In>;
-  }
-  if (std::is_same_v<Op, BitwiseAnd> || std::is_same_v<Op, BitwiseOr> ||
-      std::is_same_v<Op, BitwiseXor>) {
-    return std::is_same_v<In, Out> && std::is_integral_v<In>;
-  }
-  if (std::is_same_v<Op, LeftShift> || std::is_same_v<Op, RightShift>) {
-    return std::is_same_v<In, Out> && std::is_integral_v<In> &&
-        !std::is_same_v<In, bool>;
-  }
-  return false;
-}
-
-} // namespace rocm
-
-template <typename Op>
-void binary_op_gpu_inplace(
-    const std::vector<array>& inputs,
-    std::vector<array>& outputs,
-    std::string_view op,
-    const Stream& s) {
-  assert(inputs.size() > 1);
-  const auto& a = inputs[0];
-  const auto& b = inputs[1];
-  auto& out = outputs[0];
-  if (out.size() == 0) {
-    return;
-  }
-
-  auto& encoder = rocm::get_command_encoder(s);
-  encoder.set_input_array(a);
-  encoder.set_input_array(b);
-  encoder.set_output_array(out);
-  encoder.launch_kernel([&](hipStream_t stream) {
-    MLX_SWITCH_ALL_TYPES(a.dtype(), CTYPE_IN, {
-      MLX_SWITCH_ALL_TYPES(out.dtype(), CTYPE_OUT, {
-        if constexpr (rocm::supports_binary_op<Op, CTYPE_IN, CTYPE_OUT>()) {
-          using InType = hip_type_t<CTYPE_IN>;
-          using OutType = hip_type_t<CTYPE_OUT>;
-
-          auto bopt = get_binary_op_type(a, b);
-          if (bopt == BinaryOpType::General) {
-            auto [shape, strides] = collapse_contiguous_dims(a, b, out);
-            auto& a_strides = strides[0];
-            auto& b_strides = strides[1];
-            bool large = a.data_size() > INT32_MAX ||
-                b.data_size() > INT32_MAX || out.data_size() > INT32_MAX;
-            MLX_SWITCH_BOOL(large, LARGE, {
-              using IdxT = std::conditional_t<LARGE, int64_t, int32_t>;
-              int ndim = shape.size();
-              if (ndim <= 3) {
-                MLX_SWITCH_1_2_3(ndim, NDIM, {
-                  auto kernel =
-                      &rocm::binary_g_nd<Op, InType, OutType, IdxT, NDIM>;
-                  auto [num_blocks, block_dims] =
-                      get_launch_args(kernel, out, large);
-                  hipLaunchKernelGGL(kernel, num_blocks, block_dims, 0, stream,
-                      a.data<InType>(),
-                      b.data<InType>(),
-                      out.data<OutType>(),
-                      out.size(),
-                      make_hip_array<NDIM>(shape),
-                      make_hip_array<NDIM>(a_strides),
-                      make_hip_array<NDIM>(b_strides));
-                });
-              } else {
-                auto kernel = rocm::binary_g<Op, InType, OutType, IdxT>;
-                auto [num_blocks, block_dims] =
-                    get_launch_args(kernel, out, large);
-                hipLaunchKernelGGL(kernel, num_blocks, block_dims, 0, stream,
-                    a.data<InType>(),
-                    b.data<InType>(),
-                    out.data<OutType>(),
-                    out.size(),
-                    make_hip_array(shape),
-                    make_hip_array(a_strides),
-                    make_hip_array(b_strides),
-                    ndim);
-              }
-            });
-          } else {
-            MLX_SWITCH_BOOL(out.data_size() > UINT32_MAX, LARGE, {
-              using IdxT = std::conditional_t<LARGE, int64_t, uint32_t>;
-              auto kernel = rocm::binary_ss<Op, InType, OutType, IdxT>;
-              if (bopt == BinaryOpType::ScalarVector) {
-                kernel = rocm::binary_sv<Op, InType, OutType, IdxT>;
-              } else if (bopt == BinaryOpType::VectorScalar) {
-                kernel = rocm::binary_vs<Op, InType, OutType, IdxT>;
-              } else if (bopt == BinaryOpType::VectorVector) {
-                kernel = rocm::binary_vv<Op, InType, OutType, IdxT>;
-              }
-              auto [num_blocks, block_dims] = get_launch_args(
-                  kernel, out.data_size(), out.shape(), out.strides(), LARGE);
-              hipLaunchKernelGGL(kernel, num_blocks, block_dims, 0, stream,
-                  a.data<InType>(),
-                  b.data<InType>(),
-                  out.data<OutType>(),
-                  out.data_size());
-            });
-          }
-        } else {
-          throw std::runtime_error(fmt::format(
-              "Can not do binary op {} on inputs of {} with result of {}.",
-              op,
-              dtype_to_string(a.dtype()),
-              dtype_to_string(out.dtype())));
-        }
-      });
-    });
-  });
-}
-
-template <typename Op>
-void binary_op_gpu(
-    const std::vector<array>& inputs,
-    std::vector<array>& outputs,
-    std::string_view op,
-    const Stream& s) {
-  auto& a = inputs[0];
-  auto& b = inputs[1];
-  auto bopt = get_binary_op_type(a, b);
-  set_binary_op_output_data(a, b, outputs[0], bopt);
-  set_binary_op_output_data(a, b, outputs[1], bopt);
-  binary_op_gpu_inplace<Op>(inputs, outputs, op, s);
-}
-
-template <typename Op>
-void binary_op_gpu(
-    const std::vector<array>& inputs,
-    array& out,
-    std::string_view op,
-    const Stream& s) {
-  auto& a = inputs[0];
-  auto& b = inputs[1];
-  auto bopt = get_binary_op_type(a, b);
-  set_binary_op_output_data(a, b, out, bopt);
-  std::vector<array> outputs{out};
-  binary_op_gpu_inplace<Op>(inputs, outputs, op, s);
-}
-
-#define BINARY_GPU(func)                                                 \
-  void func::eval_gpu(const std::vector<array>& inputs, array& out) {    \
-    auto& s = out.primitive().stream();                                  \
-    binary_op_gpu<rocm::func>(inputs, out, get_primitive_string(this), s); \
-  }
-
-#define BINARY_GPU_MULTI(func)                                               \
-  void func::eval_gpu(                                                       \
-      const std::vector<array>& inputs, std::vector<array>& outputs) {       \
-    auto& s = outputs[0].primitive().stream();                               \
-    binary_op_gpu<rocm::func>(inputs, outputs, get_primitive_string(this), s); \
-  }
-
-BINARY_GPU(Add)
-BINARY_GPU(ArcTan2)
-BINARY_GPU(Divide)
-BINARY_GPU(Remainder)
-BINARY_GPU(Greater)
-BINARY_GPU(GreaterEqual)
-BINARY_GPU(Less)
-BINARY_GPU(LessEqual)
-BINARY_GPU(LogicalAnd)
-BINARY_GPU(LogicalOr)
-BINARY_GPU(LogAddExp)
-BINARY_GPU(Maximum)
-BINARY_GPU(Minimum)
-BINARY_GPU(Multiply)
-BINARY_GPU(NotEqual)
-BINARY_GPU(Power)
-BINARY_GPU(Subtract)
-
-void Equal::eval_gpu(const std::vector<array>& inputs, array& out) {
-  auto& s = out.primitive().stream();
-  auto op = get_primitive_string(this);
-  if (equal_nan_) {
-    binary_op_gpu<rocm::NaNEqual>(inputs, out, op, s);
-  } else {
-    binary_op_gpu<rocm::Equal>(inputs, out, op, s);
-  }
-}
-
-void BitwiseBinary::eval_gpu(const std::vector<array>& inputs, array& out) {
-  auto& s = out.primitive().stream();
-  auto op = get_primitive_string(this);
-  switch (op_) {
-    case BitwiseBinary::And:
-      binary_op_gpu<rocm::BitwiseAnd>(inputs, out, op, s);
-      break;
-    case BitwiseBinary::Or:
-      binary_op_gpu<rocm::BitwiseOr>(inputs, out, op, s);
-      break;
-    case BitwiseBinary::Xor:
-      binary_op_gpu<rocm::BitwiseXor>(inputs, out, op, s);
-      break;
-    case BitwiseBinary::LeftShift:
-      binary_op_gpu<rocm::LeftShift>(inputs, out, op, s);
-      break;
-    case BitwiseBinary::RightShift:
-      binary_op_gpu<rocm::RightShift>(inputs, out, op, s);
-      break;
-  }
-}
-
-} // namespace mlx::core 

mlx/backend/rocm/compiled.cpp

@@ -1,9 +0,0 @@
-// Copyright © 2025 Apple Inc.
-
-namespace mlx::core::rocm {
-
-void compile() {
-  // Placeholder for ROCm compilation
-}
-
-} // namespace mlx::core::rocm

mlx/backend/rocm/copy.hip

@@ -1,20 +0,0 @@
-// Copyright © 2025 Apple Inc.
-
-#include <hip/hip_runtime.h>
-
-namespace mlx::core::rocm {
-
-__global__ void copy_kernel(float* src, float* dst, int n) {
-  int idx = blockIdx.x * blockDim.x + threadIdx.x;
-  if (idx < n) {
-    dst[idx] = src[idx];
-  }
-}
-
-void launch_copy(float* src, float* dst, int n, hipStream_t stream) {
-  int threads = 256;
-  int blocks = (n + threads - 1) / threads;
-  hipLaunchKernelGGL(copy_kernel, dim3(blocks), dim3(threads), 0, stream, src, dst, n);
-}
-
-} // namespace mlx::core::rocm 

mlx/backend/rocm/copy/copy.hpp

@@ -1,60 +0,0 @@
-// Copyright © 2025 Apple Inc.
-
-#pragma once
-
-#include <hip/hip_runtime.h>
-#include <cstddef>
-
-namespace mlx::core::rocm {
-
-// Copy function declarations
-void copy_contiguous(
-    const void* src,
-    void* dst,
-    size_t size,
-    hipStream_t stream);
-
-void copy_general(
-    const void* src,
-    void* dst,
-    const int* src_shape,
-    const size_t* src_strides,
-    const int* dst_shape,
-    const size_t* dst_strides,
-    int ndim,
-    size_t size,
-    size_t dtype_size,
-    hipStream_t stream);
-
-void copy_general_dynamic(
-    const void* src,
-    void* dst,
-    const int* src_shape,
-    const size_t* src_strides,
-    const int* dst_shape,
-    const size_t* dst_strides,
-    int ndim,
-    size_t size,
-    size_t dtype_size,
-    hipStream_t stream);
-
-void copy_general_input(
-    const void* src,
-    void* dst,
-    const int* src_shape,
-    const size_t* src_strides,
-    const int* dst_shape,
-    const size_t* dst_strides,
-    int ndim,
-    size_t size,
-    size_t dtype_size,
-    hipStream_t stream);
-
-// Utility functions for element location calculation
-__device__ size_t
-elem_to_loc(size_t elem, const int* shape, const size_t* strides, int ndim);
-
-__device__ size_t
-loc_to_elem(size_t loc, const int* shape, const size_t* strides, int ndim);
-
-} // namespace mlx::core::rocm

mlx/backend/rocm/copy/copy_contiguous.hip

@@ -1,38 +0,0 @@
-// Copyright © 2025 Apple Inc.
-
-#include "mlx/backend/rocm/copy/copy.hpp"
-#include "mlx/backend/rocm/kernel_utils.hpp"
-
-#include <hip/hip_runtime.h>
-
-namespace mlx::core::rocm {
-
-__global__ void copy_contiguous_kernel(
-    const char* src,
-    char* dst,
-    size_t size) {
-  size_t tid = blockIdx.x * blockDim.x + threadIdx.x;
-  if (tid < size) {
-    dst[tid] = src[tid];
-  }
-}
-
-void copy_contiguous(
-    const void* src,
-    void* dst,
-    size_t size,
-    hipStream_t stream) {
-  if (size == 0) {
-    return;
-  }
-
-  const int threads_per_block = 256;
-  const int blocks = (size + threads_per_block - 1) / threads_per_block;
-
-  copy_contiguous_kernel<<<blocks, threads_per_block, 0, stream>>>(
-      static_cast<const char*>(src),
-      static_cast<char*>(dst),
-      size);
-}
-
-} // namespace mlx::core::rocm 

mlx/backend/rocm/device.cpp

@@ -1,130 +0,0 @@
-// Copyright © 2025 Apple Inc.
-
-#include "mlx/backend/rocm/device.h"
-#include "mlx/backend/metal/metal.h"
-#include "mlx/backend/rocm/worker.h"
-
-#include <fmt/format.h>
-
-namespace mlx::core {
-
-namespace rocm {
-
-DeviceStream::DeviceStream(Device& device) : device_(device), stream_(device) {}
-
-void DeviceStream::synchronize() {
-  CHECK_HIP_ERROR(hipStreamSynchronize(stream_));
-}
-
-hipStream_t DeviceStream::schedule_hip_stream() {
-  // TODO: Return a stream that maximizes parallelism.
-  return stream_;
-}
-
-hipStream_t DeviceStream::last_hip_stream() {
-  return stream_;
-}
-
-CommandEncoder& DeviceStream::get_encoder() {
-  if (!encoder_) {
-    encoder_ = std::make_unique<CommandEncoder>(*this);
-  }
-  return *encoder_;
-}
-
-Device::Device(int device) : device_(device) {
-  CHECK_HIP_ERROR(hipDeviceGetAttribute(
-      &compute_capability_major_,
-      hipDeviceAttributeComputeCapabilityMajor,
-      device_));
-  CHECK_HIP_ERROR(hipDeviceGetAttribute(
-      &compute_capability_minor_,
-      hipDeviceAttributeComputeCapabilityMinor,
-      device_));
-
-  // Validate device requirements
-  int attr = 0;
-  CHECK_HIP_ERROR(hipDeviceGetAttribute(
-      &attr, hipDeviceAttributeConcurrentManagedAccess, device_));
-  if (attr != 1) {
-    // ROCm unified memory might not be available on all devices
-    // This is a warning rather than an error for ROCm
-    // TODO: Add proper ROCm unified memory checking
-  }
-
-  // Create rocBLAS handle
-  make_current();
-  CHECK_HIP_ERROR(
-      static_cast<hipError_t>(rocblas_create_handle(&rocblas_handle_)));
-}
-
-Device::~Device() {
-  if (rocblas_handle_) {
-    rocblas_destroy_handle(rocblas_handle_);
-  }
-}
-
-void Device::make_current() {
-  // Cache current device to reduce HIP API calls
-  static int current = 0;
-  if (current != device_) {
-    CHECK_HIP_ERROR(hipSetDevice(device_));
-    current = device_;
-  }
-}
-
-DeviceStream& Device::get_stream(Stream s) {
-  auto it = streams_.find(s.index);
-  if (it == streams_.end()) {
-    it = streams_.try_emplace(s.index, *this).first;
-  }
-  return it->second;
-}
-
-CommandEncoder::CommandEncoder(DeviceStream& s)
-    : device_(s.device()), stream_(s) {}
-
-void CommandEncoder::add_completed_handler(std::function<void()> task) {
-  worker_.add_task(std::move(task));
-}
-
-void CommandEncoder::end_encoding() {
-  if (!temporaries_.empty()) {
-    add_completed_handler([temporaries = std::move(temporaries_)]() {});
-  }
-
-  // There is no kernel running, run completion handlers immediately.
-  if (!has_gpu_work_) {
-    worker_.consume_in_this_thread();
-    return;
-  }
-  has_gpu_work_ = false;
-
-  // Commit tasks
-  commit();
-}
-
-void CommandEncoder::commit() {
-  worker_.commit(stream_.last_hip_stream());
-}
-
-Device& device(mlx::core::Device device) {
-  static std::unordered_map<int, Device> devices;
-  auto it = devices.find(device.index);
-  if (it == devices.end()) {
-    it = devices.try_emplace(device.index, device.index).first;
-  }
-  return it->second;
-}
-
-DeviceStream& get_stream(Stream s) {
-  return device(s.device).get_stream(s);
-}
-
-CommandEncoder& get_command_encoder(Stream s) {
-  return get_stream(s).get_encoder();
-}
-
-} // namespace rocm
-
-} // namespace mlx::core

mlx/backend/rocm/device.h

@@ -1,146 +0,0 @@
-// Copyright © 2025 Apple Inc.
-
-#pragma once
-
-#include "mlx/array.h"
-#include "mlx/backend/rocm/utils.h"
-#include "mlx/backend/rocm/worker.h"
-#include "mlx/stream.h"
-
-#include <hip/hip_runtime.h>
-#include <rocblas/rocblas.h>
-
-#include <unordered_map>
-
-namespace mlx::core {
-
-namespace rocm {
-
-class Device;
-class CommandEncoder;
-
-class DeviceStream {
- public:
-  explicit DeviceStream(Device& device);
-
-  DeviceStream(const DeviceStream&) = delete;
-  DeviceStream& operator=(const DeviceStream&) = delete;
-
-  // Wait until kernels in the stream complete.
-  void synchronize();
-
-  // Return a HIP stream for launching kernels.
-  hipStream_t schedule_hip_stream();
-
-  // Return the last HIP stream used.
-  hipStream_t last_hip_stream();
-
-  CommandEncoder& get_encoder();
-
-  Device& device() {
-    return device_;
-  }
-
- private:
-  Device& device_;
-  HipStream stream_;
-  std::unique_ptr<CommandEncoder> encoder_;
-};
-
-class Device {
- public:
-  explicit Device(int device);
-  ~Device();
-
-  Device(const Device&) = delete;
-  Device& operator=(const Device&) = delete;
-
-  // Make this device the current HIP device, required by some HIP calls.
-  void make_current();
-
-  DeviceStream& get_stream(Stream s);
-
-  int hip_device() const {
-    return device_;
-  }
-  int compute_capability_major() const {
-    return compute_capability_major_;
-  }
-  int compute_capability_minor() const {
-    return compute_capability_minor_;
-  }
-  rocblas_handle rocblas_handle() const {
-    return rocblas_handle_;
-  }
-
- private:
-  int device_;
-  int compute_capability_major_;
-  int compute_capability_minor_;
-  rocblas_handle rocblas_handle_;
-  std::unordered_map<int, DeviceStream> streams_;
-};
-
-class CommandEncoder {
- public:
-  explicit CommandEncoder(DeviceStream& stream);
-
-  CommandEncoder(const CommandEncoder&) = delete;
-  CommandEncoder& operator=(const CommandEncoder&) = delete;
-
-  void set_input_array(const array& arr) {}
-  void set_output_array(const array& arr) {}
-
-  void add_temporary(const array& arr) {
-    temporaries_.push_back(arr.data_shared_ptr());
-  }
-
-  void add_completed_handler(std::function<void()> task);
-  void end_encoding();
-  void commit();
-
-  // Schedule a HIP stream for |fun| to launch kernels, and check error
-  // afterwards.
-  template <typename F>
-  void launch_kernel(F&& fun) {
-    launch_kernel(stream_.schedule_hip_stream(), std::forward<F>(fun));
-  }
-
-  template <typename F>
-  void launch_kernel(hipStream_t stream, F&& fun) {
-    device_.make_current();
-    fun(stream);
-    check_hip_error("kernel launch", hipGetLastError());
-    has_gpu_work_ = true;
-  }
-
-  Device& device() {
-    return device_;
-  }
-
-  DeviceStream& stream() {
-    return stream_;
-  }
-
-  bool has_gpu_work() const {
-    return has_gpu_work_;
-  }
-
- private:
-  Device& device_;
-  DeviceStream& stream_;
-  Worker worker_;
-  bool has_gpu_work_{false};
-  std::vector<std::shared_ptr<array::Data>> temporaries_;
-};
-
-Device& device(mlx::core::Device device);
-DeviceStream& get_stream(Stream s);
-CommandEncoder& get_command_encoder(Stream s);
-
-// Utility function to check HIP errors
-void check_hip_error(const char* msg, hipError_t error);
-
-} // namespace rocm
-
-} // namespace mlx::core

mlx/backend/rocm/device/arange.hpp

@@ -1,17 +0,0 @@
-// Copyright © 2025 Apple Inc.
-
-#pragma once
-
-#include <hip/hip_runtime.h>
-
-namespace mlx::core::rocm {
-
-template <typename T>
-__global__ void arange_kernel(T* out, T start, T step, size_t size) {
-  size_t tid = blockIdx.x * blockDim.x + threadIdx.x;
-  if (tid < size) {
-    out[tid] = start + static_cast<T>(tid) * step;
-  }
-}
-
-} // namespace mlx::core::rocm

mlx/backend/rocm/device/atomic_ops.hpp

@@ -1,36 +0,0 @@
-// Copyright © 2025 Apple Inc.
-
-#pragma once
-
-#include <hip/hip_runtime.h>
-
-namespace mlx::core::rocm {
-
-// Atomic operations for HIP
-__device__ inline float atomicAddFloat(float* address, float val) {
-  return atomicAdd(address, val);
-}
-
-__device__ inline double atomicAddDouble(double* address, double val) {
-  return atomicAdd(address, val);
-}
-
-__device__ inline int atomicAddInt(int* address, int val) {
-  return atomicAdd(address, val);
-}
-
-__device__ inline unsigned int atomicAddUInt(
-    unsigned int* address,
-    unsigned int val) {
-  return atomicAdd(address, val);
-}
-
-__device__ inline float atomicMaxFloat(float* address, float val) {
-  return atomicMax(address, val);
-}
-
-__device__ inline float atomicMinFloat(float* address, float val) {
-  return atomicMin(address, val);
-}
-
-} // namespace mlx::core::rocm

mlx/backend/rocm/device/binary_ops.hpp

@@ -1,217 +0,0 @@
-// Copyright © 2025 Apple Inc.
-
-#pragma once
-
-#include <hip/hip_bfloat16.h>
-#include <hip/hip_fp16.h>
-#include <hip/hip_runtime.h>
-#include <hipcomplex.h>
-
-namespace mlx::core::rocm {
-
-// Arithmetic operations
-struct Add {
-  template <typename T>
-  __device__ T operator()(T a, T b) {
-    return a + b;
-  }
-};
-
-struct Subtract {
-  template <typename T>
-  __device__ T operator()(T a, T b) {
-    return a - b;
-  }
-};
-
-struct Multiply {
-  template <typename T>
-  __device__ T operator()(T a, T b) {
-    return a * b;
-  }
-};
-
-struct Divide {
-  template <typename T>
-  __device__ T operator()(T a, T b) {
-    return a / b;
-  }
-};
-
-struct Power {
-  template <typename T>
-  __device__ T operator()(T a, T b) {
-    return powf(a, b);
-  }
-
-  __device__ double operator()(double a, double b) {
-    return pow(a, b);
-  }
-};
-
-struct Remainder {
-  template <typename T>
-  __device__ T operator()(T a, T b) {
-    return fmodf(a, b);
-  }
-
-  __device__ double operator()(double a, double b) {
-    return fmod(a, b);
-  }
-};
-
-// Comparison operations
-struct Equal {
-  template <typename T>
-  __device__ bool operator()(T a, T b) {
-    return a == b;
-  }
-};
-
-struct NotEqual {
-  template <typename T>
-  __device__ bool operator()(T a, T b) {
-    return a != b;
-  }
-};
-
-struct Greater {
-  template <typename T>
-  __device__ bool operator()(T a, T b) {
-    return a > b;
-  }
-};
-
-struct GreaterEqual {
-  template <typename T>
-  __device__ bool operator()(T a, T b) {
-    return a >= b;
-  }
-};
-
-struct Less {
-  template <typename T>
-  __device__ bool operator()(T a, T b) {
-    return a < b;
-  }
-};
-
-struct LessEqual {
-  template <typename T>
-  __device__ bool operator()(T a, T b) {
-    return a <= b;
-  }
-};
-
-struct NaNEqual {
-  template <typename T>
-  __device__ bool operator()(T a, T b) {
-    return (isnan(a) && isnan(b)) || (a == b);
-  }
-};
-
-// Logic operations
-struct LogicalAnd {
-  __device__ bool operator()(bool a, bool b) {
-    return a && b;
-  }
-};
-
-struct LogicalOr {
-  __device__ bool operator()(bool a, bool b) {
-    return a || b;
-  }
-};
-
-// Math operations
-struct Maximum {
-  template <typename T>
-  __device__ T operator()(T a, T b) {
-    return fmaxf(a, b);
-  }
-
-  __device__ double operator()(double a, double b) {
-    return fmax(a, b);
-  }
-};
-
-struct Minimum {
-  template <typename T>
-  __device__ T operator()(T a, T b) {
-    return fminf(a, b);
-  }
-
-  __device__ double operator()(double a, double b) {
-    return fmin(a, b);
-  }
-};
-
-struct LogAddExp {
-  template <typename T>
-  __device__ T operator()(T a, T b) {
-    T max_val = fmaxf(a, b);
-    T min_val = fminf(a, b);
-    if (isinf(max_val)) {
-      return max_val;
-    }
-    return max_val + log1pf(expf(min_val - max_val));
-  }
-
-  __device__ double operator()(double a, double b) {
-    double max_val = fmax(a, b);
-    double min_val = fmin(a, b);
-    if (isinf(max_val)) {
-      return max_val;
-    }
-    return max_val + log1p(exp(min_val - max_val));
-  }
-};
-
-struct ArcTan2 {
-  template <typename T>
-  __device__ T operator()(T a, T b) {
-    return atan2f(a, b);
-  }
-
-  __device__ double operator()(double a, double b) {
-    return atan2(a, b);
-  }
-};
-
-// Bitwise operations
-struct BitwiseAnd {
-  template <typename T>
-  __device__ T operator()(T a, T b) {
-    return a & b;
-  }
-};
-
-struct BitwiseOr {
-  template <typename T>
-  __device__ T operator()(T a, T b) {
-    return a | b;
-  }
-};
-
-struct BitwiseXor {
-  template <typename T>
-  __device__ T operator()(T a, T b) {
-    return a ^ b;
-  }
-};
-
-struct LeftShift {
-  template <typename T>
-  __device__ T operator()(T a, T b) {
-    return a << b;
-  }
-};
-
-struct RightShift {
-  template <typename T>
-  __device__ T operator()(T a, T b) {
-    return a >> b;
-  }
-};
-
-} // namespace mlx::core::rocm

mlx/backend/rocm/device/cast_op.hpp

@@ -1,21 +0,0 @@
-// Copyright © 2025 Apple Inc.
-
-#pragma once
-
-#include <hip/hip_runtime.h>
-
-namespace mlx::core::rocm {
-
-template <typename To, typename From>
-struct CastOp {
-  __device__ To operator()(From x) const {
-    return static_cast<To>(x);
-  }
-};
-
-template <typename To, typename From>
-__device__ inline To cast_op(From x) {
-  return static_cast<To>(x);
-}
-
-} // namespace mlx::core::rocm

mlx/backend/rocm/device/config.h

@@ -1,14 +0,0 @@
-// Copyright © 2025 Apple Inc.
-
-#pragma once
-
-// ROCm/HIP specific configuration
-#define ROCM_MAX_THREADS_PER_BLOCK 1024
-#define ROCM_WARP_SIZE 64
-#define ROCM_MAX_BLOCKS_PER_GRID 65535
-
-namespace mlx::core::rocm {
-constexpr int kMaxThreadsPerBlock = ROCM_MAX_THREADS_PER_BLOCK;
-constexpr int kWarpSize = ROCM_WARP_SIZE;
-constexpr int kMaxBlocksPerGrid = ROCM_MAX_BLOCKS_PER_GRID;
-} // namespace mlx::core::rocm

mlx/backend/rocm/device/fp16_math.hpp

@@ -1,87 +0,0 @@
-// Copyright © 2025 Apple Inc.
-
-#pragma once
-
-#include <hip/hip_fp16.h>
-#include <hip/hip_runtime.h>
-
-namespace mlx::core::rocm {
-
-// HIP/ROCm equivalents of CUDA half precision math functions
-inline __device__ __half2 h2sin(__half2 x) {
-  return __half2{hsin(x.x), hsin(x.y)};
-}
-
-inline __device__ __half2 h2cos(__half2 x) {
-  return __half2{hcos(x.x), hcos(x.y)};
-}
-
-inline __device__ __half2 h2exp(__half2 x) {
-  return __half2{hexp(x.x), hexp(x.y)};
-}
-
-inline __device__ __half2 h2log(__half2 x) {
-  return __half2{hlog(x.x), hlog(x.y)};
-}
-
-inline __device__ __half2 h2sqrt(__half2 x) {
-  return __half2{hsqrt(x.x), hsqrt(x.y)};
-}
-
-inline __device__ __half2 h2rsqrt(__half2 x) {
-  return __half2{hrsqrt(x.x), hrsqrt(x.y)};
-}
-
-inline __device__ __half2 h2ceil(__half2 x) {
-  return __half2{hceil(x.x), hceil(x.y)};
-}
-
-inline __device__ __half2 h2floor(__half2 x) {
-  return __half2{hfloor(x.x), hfloor(x.y)};
-}
-
-inline __device__ __half2 h2rint(__half2 x) {
-  return __half2{hrint(x.x), hrint(x.y)};
-}
-
-inline __device__ __half2 h2trunc(__half2 x) {
-  return __half2{htrunc(x.x), htrunc(x.y)};
-}
-
-// Additional math functions for half precision
-inline __device__ __half habs(__half x) {
-  return __half{fabsf(__half2float(x))};
-}
-
-inline __device__ __half2 h2abs(__half2 x) {
-  return __half2{habs(x.x), habs(x.y)};
-}
-
-inline __device__ __half hneg(__half x) {
-  return __half{-__half2float(x)};
-}
-
-inline __device__ __half2 h2neg(__half2 x) {
-  return __half2{hneg(x.x), hneg(x.y)};
-}
-
-// BFloat16 support functions
-#ifdef __HIP_BFLOAT16__
-inline __device__ __hip_bfloat16 habs(__hip_bfloat16 x) {
-  return __hip_bfloat16{fabsf(__bfloat162float(x))};
-}
-
-inline __device__ __hip_bfloat162 h2abs(__hip_bfloat162 x) {
-  return __hip_bfloat162{habs(x.x), habs(x.y)};
-}
-
-inline __device__ __hip_bfloat16 hneg(__hip_bfloat16 x) {
-  return __hip_bfloat16{-__bfloat162float(x)};
-}
-
-inline __device__ __hip_bfloat162 h2neg(__hip_bfloat162 x) {
-  return __hip_bfloat162{hneg(x.x), hneg(x.y)};
-}
-#endif
-
-} // namespace mlx::core::rocm

mlx/backend/rocm/device/hip_complex_math.hpp

@@ -1,52 +0,0 @@
-// Copyright © 2025 Apple Inc.
-
-#pragma once
-
-#include <hip/hip_complex.h>
-#include <hip/hip_runtime.h>
-
-namespace mlx::core::rocm {
-
-// HIP complex math functions
-__device__ inline hipFloatComplex hip_complex_add(
-    hipFloatComplex a,
-    hipFloatComplex b) {
-  return make_hipFloatComplex(
-      hipCrealf(a) + hipCrealf(b), hipCimagf(a) + hipCimagf(b));
-}
-
-__device__ inline hipFloatComplex hip_complex_sub(
-    hipFloatComplex a,
-    hipFloatComplex b) {
-  return make_hipFloatComplex(
-      hipCrealf(a) - hipCrealf(b), hipCimagf(a) - hipCimagf(b));
-}
-
-__device__ inline hipFloatComplex hip_complex_mul(
-    hipFloatComplex a,
-    hipFloatComplex b) {
-  float real = hipCrealf(a) * hipCrealf(b) - hipCimagf(a) * hipCimagf(b);
-  float imag = hipCrealf(a) * hipCimagf(b) + hipCimagf(a) * hipCrealf(b);
-  return make_hipFloatComplex(real, imag);
-}
-
-__device__ inline hipFloatComplex hip_complex_div(
-    hipFloatComplex a,
-    hipFloatComplex b) {
-  float denom = hipCrealf(b) * hipCrealf(b) + hipCimagf(b) * hipCimagf(b);
-  float real =
-      (hipCrealf(a) * hipCrealf(b) + hipCimagf(a) * hipCimagf(b)) / denom;
-  float imag =
-      (hipCimagf(a) * hipCrealf(b) - hipCrealf(a) * hipCimagf(b)) / denom;
-  return make_hipFloatComplex(real, imag);
-}
-
-__device__ inline float hip_complex_abs(hipFloatComplex z) {
-  return sqrtf(hipCrealf(z) * hipCrealf(z) + hipCimagf(z) * hipCimagf(z));
-}
-
-__device__ inline hipFloatComplex hip_complex_conj(hipFloatComplex z) {
-  return make_hipFloatComplex(hipCrealf(z), -hipCimagf(z));
-}
-
-} // namespace mlx::core::rocm

mlx/backend/rocm/device/ternary_ops.hpp

@@ -1,16 +0,0 @@
-// Copyright © 2025 Apple Inc.
-
-#pragma once
-
-#include <hip/hip_runtime.h>
-
-namespace mlx::core::rocm {
-
-struct Select {
-  template <typename T>
-  __device__ T operator()(bool condition, T a, T b) const {
-    return condition ? a : b;
-  }
-};
-
-} // namespace mlx::core::rocm

mlx/backend/rocm/device/unary_ops.hpp

@@ -1,368 +0,0 @@
-// Copyright © 2025 Apple Inc.
-
-#pragma once
-
-#include "mlx/backend/rocm/device/fp16_math.hpp"
-#include "mlx/backend/rocm/device/utils.hpp"
-
-#include <hip/hip_runtime.h>
-
-namespace mlx::core::rocm {
-
-struct Abs {
-  template <typename T>
-  __device__ T operator()(T x) {
-    if constexpr (std::is_unsigned_v<T>) {
-      return x;
-    } else if constexpr (std::is_same_v<T, hipFloatComplex>) {
-      return {
-          sqrt(hipCrealf(x) * hipCrealf(x) + hipCimagf(x) * hipCimagf(x)), 0};
-    } else {
-      return abs(x);
-    }
-  }
-};
-
-struct ArcCos {
-  template <typename T>
-  __device__ T operator()(T x) {
-    return acos(x);
-  }
-};
-
-struct ArcCosh {
-  template <typename T>
-  __device__ T operator()(T x) {
-    return acosh(x);
-  }
-};
-
-struct ArcSin {
-  template <typename T>
-  __device__ T operator()(T x) {
-    return asin(x);
-  }
-};
-
-struct ArcSinh {
-  template <typename T>
-  __device__ T operator()(T x) {
-    return asinh(x);
-  }
-};
-
-struct ArcTan {
-  template <typename T>
-  __device__ T operator()(T x) {
-    return atan(x);
-  }
-};
-
-struct ArcTanh {
-  template <typename T>
-  __device__ T operator()(T x) {
-    return atanh(x);
-  }
-};
-
-struct BitwiseInvert {
-  template <typename T>
-  __device__ T operator()(T x) {
-    return ~x;
-  }
-};
-
-struct Ceil {
-  template <typename T>
-  __device__ T operator()(T x) {
-    if constexpr (std::is_integral_v<T>) {
-      return x;
-    } else {
-      return ceil(x);
-    }
-  }
-};
-
-struct Conjugate {
-  __device__ hipFloatComplex operator()(hipFloatComplex x) {
-    return {hipCrealf(x), -hipCimagf(x)};
-  }
-};
-
-struct Cos {
-  template <typename T>
-  __device__ T operator()(T x) {
-    if constexpr (std::is_same_v<T, hipFloatComplex>) {
-      return {
-          cos(hipCrealf(x)) * cosh(hipCimagf(x)),
-          -sin(hipCrealf(x)) * sinh(hipCimagf(x))};
-    } else {
-      return cos(x);
-    }
-  }
-};
-
-struct Cosh {
-  template <typename T>
-  __device__ T operator()(T x) {
-    if constexpr (std::is_same_v<T, hipFloatComplex>) {
-      return {
-          cosh(hipCrealf(x)) * cos(hipCimagf(x)),
-          sinh(hipCrealf(x)) * sin(hipCimagf(x))};
-    } else {
-      return cosh(x);
-    }
-  }
-};
-
-struct Erf {
-  template <typename T>
-  __device__ T operator()(T x) {
-    if constexpr (std::is_same_v<T, __half>) {
-      return erf(__half2float(x));
-    } else if constexpr (std::is_same_v<T, __hip_bfloat16>) {
-      return erf(__bfloat162float(x));
-    } else {
-      return erf(x);
-    }
-  }
-};
-
-struct ErfInv {
-  template <typename T>
-  __device__ T operator()(T x) {
-    if constexpr (std::is_same_v<T, __half>) {
-      return erfinv(__half2float(x));
-    } else if constexpr (std::is_same_v<T, __hip_bfloat16>) {
-      return erfinv(__bfloat162float(x));
-    } else {
-      return erfinv(x);
-    }
-  }
-};
-
-struct Exp {
-  template <typename T>
-  __device__ T operator()(T x) {
-    if constexpr (std::is_same_v<T, hipFloatComplex>) {
-      auto m = exp(hipCrealf(x));
-      return {m * cos(hipCimagf(x)), m * sinh(hipCimagf(x))};
-    } else {
-      return exp(x);
-    }
-  }
-};
-
-struct Expm1 {
-  template <typename T>
-  __device__ T operator()(T x) {
-    if constexpr (std::is_same_v<T, __half>) {
-      return expm1(__half2float(x));
-    } else if constexpr (std::is_same_v<T, __hip_bfloat16>) {
-      return expm1(__bfloat162float(x));
-    } else {
-      return expm1(x);
-    }
-  }
-};
-
-struct Floor {
-  template <typename T>
-  __device__ T operator()(T x) {
-    if constexpr (std::is_integral_v<T>) {
-      return x;
-    } else {
-      return floor(x);
-    }
-  }
-};
-
-struct Imag {
-  __device__ float operator()(hipFloatComplex x) {
-    return hipCimagf(x);
-  }
-};
-
-struct Log {
-  template <typename T>
-  __device__ T operator()(T x) {
-    if constexpr (std::is_same_v<T, hipFloatComplex>) {
-      auto r = log(hipCrealf(Abs{}(x)));
-      auto i = atan2f(hipCimagf(x), hipCrealf(x));
-      return {r, i};
-    } else {
-      return log(x);
-    }
-  }
-};
-
-struct Log2 {
-  template <typename T>
-  __device__ T operator()(T x) {
-    if constexpr (std::is_same_v<T, hipFloatComplex>) {
-      auto y = Log{}(x);
-      return {hipCrealf(y) / M_LN2, hipCimagf(y) / M_LN2};
-    } else {
-      return log2(x);
-    }
-  }
-};
-
-struct Log10 {
-  template <typename T>
-  __device__ T operator()(T x) {
-    if constexpr (std::is_same_v<T, hipFloatComplex>) {
-      auto y = Log{}(x);
-      return {hipCrealf(y) / M_LN10, hipCimagf(y) / M_LN10};
-    } else {
-      return log10(x);
-    }
-  }
-};
-
-struct Log1p {
-  template <typename T>
-  __device__ T operator()(T x) {
-    return log1p(x);
-  }
-};
-
-struct LogicalNot {
-  __device__ bool operator()(bool x) {
-    return !x;
-  }
-};
-
-struct Negative {
-  template <typename T>
-  __device__ T operator()(T x) {
-    if constexpr (std::is_same_v<T, hipFloatComplex>) {
-      return 0 - x;
-    } else {
-      return -x;
-    }
-  }
-};
-
-struct Real {
-  __device__ float operator()(hipFloatComplex x) {
-    return hipCrealf(x);
-  }
-};
-
-struct Round {
-  template <typename T>
-  __device__ T operator()(T x) {
-    if constexpr (std::is_same_v<T, hipFloatComplex>) {
-      return {rint(hipCrealf(x)), rint(hipCimagf(x))};
-    } else {
-      return rint(x);
-    }
-  }
-};
-
-struct Rsqrt {
-  template <typename T>
-  __device__ T operator()(T x) {
-    return rsqrt(x);
-  }
-};
-
-struct Sigmoid {
-  template <typename T>
-  __device__ T operator()(T x) {
-    T y = 1 / (1 + exp(-abs(x)));
-    return (x < 0) ? 1 - y : y;
-  }
-};
-
-struct Sign {
-  template <typename T>
-  __device__ T operator()(T x) {
-    if constexpr (std::is_unsigned_v<T>) {
-      return x != 0;
-    } else if constexpr (std::is_same_v<T, hipFloatComplex>) {
-      if (hipCrealf(x) == 0 && hipCimagf(x) == 0) {
-        return x;
-      } else {
-        return x / Abs()(x);
-      }
-    } else if constexpr (std::is_same_v<T, __hip_bfloat16>) {
-      return static_cast<float>((x > T(0.f)) - (x < T(0.f)));
-    } else {
-      return (x > T(0)) - (x < T(0));
-    }
-  }
-};
-
-struct Sin {
-  template <typename T>
-  __device__ T operator()(T x) {
-    if constexpr (std::is_same_v<T, hipFloatComplex>) {
-      return {
-          sin(hipCrealf(x)) * cosh(hipCimagf(x)),
-          cos(hipCrealf(x)) * sinh(hipCimagf(x))};
-    } else {
-      return sin(x);
-    }
-  }
-};
-
-struct Sinh {
-  template <typename T>
-  __device__ T operator()(T x) {
-    if constexpr (std::is_same_v<T, hipFloatComplex>) {
-      return {
-          sinh(hipCrealf(x)) * cos(hipCimagf(x)),
-          cosh(hipCrealf(x)) * sin(hipCimagf(x))};
-    } else {
-      return sinh(x);
-    }
-  }
-};
-
-struct Square {
-  template <typename T>
-  __device__ T operator()(T x) {
-    return x * x;
-  }
-};
-
-struct Sqrt {
-  template <typename T>
-  __device__ T operator()(T x) {
-    return sqrt(x);
-  }
-};
-
-struct Tan {
-  template <typename T>
-  __device__ T operator()(T x) {
-    if constexpr (std::is_same_v<T, hipFloatComplex>) {
-      float tan_a = tan(hipCrealf(x));
-      float tanh_b = tanh(hipCimagf(x));
-      float t1 = tan_a * tanh_b;
-      float denom = 1. + t1 * t1;
-      return {(tan_a - tanh_b * t1) / denom, (tanh_b + tan_a * t1) / denom};
-    } else {
-      return tan(x);
-    }
-  }
-};
-
-struct Tanh {
-  template <typename T>
-  __device__ T operator()(T x) {
-    if constexpr (std::is_same_v<T, hipFloatComplex>) {
-      float tanh_a = tanh(hipCrealf(x));
-      float tan_b = tan(hipCimagf(x));
-      float t1 = tanh_a * tan_b;
-      float denom = 1. + t1 * t1;
-      return {(tanh_a + tan_b * t1) / denom, (tan_b - tanh_a * t1) / denom};
-    } else {
-      return tanh(x);
-    }
-  }
-};
-
-} // namespace mlx::core::rocm

mlx/backend/rocm/device/utils.hpp

@@ -1,173 +0,0 @@
-// Copyright © 2025 Apple Inc.
-
-#pragma once
-
-#include <hip/hip_complex.h>
-#include <hip/hip_runtime.h>
-
-namespace mlx::core::rocm {
-
-// HIP/ROCm type definitions
-using hip_complex = hipFloatComplex;
-
-// Utility functions for HIP device code
-template <typename T>
-struct hip_type {
-  using type = T;
-};
-
-template <>
-struct hip_type<bool> {
-  using type = bool;
-};
-
-template <>
-struct hip_type<int8_t> {
-  using type = int8_t;
-};
-
-template <>
-struct hip_type<uint8_t> {
-  using type = uint8_t;
-};
-
-template <>
-struct hip_type<int16_t> {
-  using type = int16_t;
-};
-
-template <>
-struct hip_type<uint16_t> {
-  using type = uint16_t;
-};
-
-template <>
-struct hip_type<int32_t> {
-  using type = int32_t;
-};
-
-template <>
-struct hip_type<uint32_t> {
-  using type = uint32_t;
-};
-
-template <>
-struct hip_type<int64_t> {
-  using type = int64_t;
-};
-
-template <>
-struct hip_type<uint64_t> {
-  using type = uint64_t;
-};
-
-template <>
-struct hip_type<float> {
-  using type = float;
-};
-
-template <>
-struct hip_type<double> {
-  using type = double;
-};
-
-#ifdef __HIP_PLATFORM_HCC__
-template <>
-struct hip_type<__half> {
-  using type = __half;
-};
-
-template <>
-struct hip_type<__hip_bfloat16> {
-  using type = __hip_bfloat16;
-};
-#endif
-
-template <typename T>
-using hip_type_t = typename hip_type<T>::type;
-
-// Element-wise operations support
-template <typename T>
-constexpr bool is_floating_point_v = std::is_floating_point_v<T>;
-
-template <typename T>
-constexpr bool is_integral_v = std::is_integral_v<T>;
-
-template <typename T>
-constexpr bool is_signed_v = std::is_signed_v<T>;
-
-template <typename T>
-constexpr bool is_unsigned_v = std::is_unsigned_v<T>;
-
-// Complex number helper functions
-inline __device__ hipFloatComplex make_complex(float real, float imag) {
-  return make_hipFloatComplex(real, imag);
-}
-
-inline __device__ float hip_real(hipFloatComplex z) {
-  return hipCrealf(z);
-}
-
-inline __device__ float hip_imag(hipFloatComplex z) {
-  return hipCimagf(z);
-}
-
-inline __device__ hipFloatComplex hip_conj(hipFloatComplex z) {
-  return make_hipFloatComplex(hipCrealf(z), -hipCimagf(z));
-}
-
-inline __device__ float hip_abs(hipFloatComplex z) {
-  return sqrtf(hipCrealf(z) * hipCrealf(z) + hipCimagf(z) * hipCimagf(z));
-}
-
-// Memory access utilities
-template <typename T>
-inline __device__ T hip_load_global(const T* ptr) {
-  return *ptr;
-}
-
-template <typename T>
-inline __device__ void hip_store_global(T* ptr, T value) {
-  *ptr = value;
-}
-
-// Grid and block utilities
-inline __device__ int hip_thread_idx() {
-  return threadIdx.x;
-}
-
-inline __device__ int hip_block_idx() {
-  return blockIdx.x;
-}
-
-inline __device__ int hip_block_dim() {
-  return blockDim.x;
-}
-
-inline __device__ int hip_grid_dim() {
-  return gridDim.x;
-}
-
-inline __device__ int hip_global_thread_idx() {
-  return blockIdx.x * blockDim.x + threadIdx.x;
-}
-
-// Synchronization
-inline __device__ void hip_sync_threads() {
-  __syncthreads();
-}
-
-// Math constants for HIP (equivalent to CUDA's math_constants.h)
-#ifndef M_PI
-#define M_PI 3.14159265358979323846
-#endif
-
-#ifndef M_LN2
-#define M_LN2 0.693147180559945309417
-#endif
-
-#ifndef M_LN10
-#define M_LN10 2.302585092994045684018
-#endif
-
-} // namespace mlx::core::rocm

mlx/backend/rocm/eval.cpp

@@ -1,11 +0,0 @@
-// Copyright © 2025 Apple Inc.
-
-#include "mlx/backend/rocm/utils.h"
-
-namespace mlx::core::rocm {
-
-void eval() {
-  // Placeholder for ROCm evaluation
-}
-
-} // namespace mlx::core::rocm

mlx/backend/rocm/event.cpp

@@ -1,50 +0,0 @@
-// Copyright © 2025 Apple Inc.
-
-#include "mlx/backend/rocm/event.h"
-#include "mlx/backend/rocm/utils.h"
-
-namespace mlx::core::rocm {
-
-HipEvent::HipEvent() {
-  CHECK_HIP_ERROR(hipEventCreate(&event_));
-}
-
-HipEvent::~HipEvent() {
-  CHECK_HIP_ERROR(hipEventDestroy(event_));
-}
-
-void HipEvent::record(hipStream_t stream) {
-  CHECK_HIP_ERROR(hipEventRecord(event_, stream));
-}
-
-void HipEvent::wait() {
-  CHECK_HIP_ERROR(hipEventSynchronize(event_));
-}
-
-bool HipEvent::query() const {
-  hipError_t status = hipEventQuery(event_);
-  if (status == hipSuccess) {
-    return true;
-  } else if (status == hipErrorNotReady) {
-    return false;
-  } else {
-    CHECK_HIP_ERROR(status);
-    return false;
-  }
-}
-
-SharedEvent::SharedEvent() = default;
-
-void SharedEvent::notify() {
-  std::lock_guard<std::mutex> lock(mutex_);
-  ready_ = true;
-  cv_.notify_one();
-}
-
-void SharedEvent::wait() {
-  std::unique_lock<std::mutex> lock(mutex_);
-  cv_.wait(lock, [this] { return ready_; });
-  ready_ = false;
-}
-
-} // namespace mlx::core::rocm

mlx/backend/rocm/event.h

@@ -1,48 +0,0 @@
-// Copyright © 2025 Apple Inc.
-
-#pragma once
-
-#include <hip/hip_runtime.h>
-
-#include <condition_variable>
-#include <memory>
-#include <mutex>
-
-namespace mlx::core::rocm {
-
-// HIP event managed with RAII.
-class HipEvent {
- public:
-  HipEvent();
-  ~HipEvent();
-
-  HipEvent(const HipEvent&) = delete;
-  HipEvent& operator=(const HipEvent&) = delete;
-
-  void record(hipStream_t stream);
-  void wait();
-  bool query() const;
-
-  operator hipEvent_t() const {
-    return event_;
-  }
-
- private:
-  hipEvent_t event_;
-};
-
-// Shared event for worker thread synchronization.
-class SharedEvent {
- public:
-  SharedEvent();
-
-  void notify();
-  void wait();
-
- private:
-  std::mutex mutex_;
-  std::condition_variable cv_;
-  bool ready_{false};
-};
-
-} // namespace mlx::core::rocm

mlx/backend/rocm/event.hip

@@ -1,32 +0,0 @@
-// Copyright © 2025 Apple Inc.
-
-#include <hip/hip_runtime.h>
-#include "mlx/backend/rocm/utils.h"
-
-namespace mlx::core::rocm {
-
-class Event {
-public:
-  Event() {
-    check_hip_error("hipEventCreate", hipEventCreate(&event_));
-  }
-  
-  ~Event() {
-    hipEventDestroy(event_);
-  }
-  
-  void record(hipStream_t stream) {
-    check_hip_error("hipEventRecord", hipEventRecord(event_, stream));
-  }
-  
-  void wait() {
-    check_hip_error("hipEventSynchronize", hipEventSynchronize(event_));
-  }
-  
-  hipEvent_t event() const { return event_; }
-
-private:
-  hipEvent_t event_;
-};
-
-} // namespace mlx::core::rocm 

mlx/backend/rocm/fence.cpp

@@ -1,9 +0,0 @@
-// Copyright © 2025 Apple Inc.
-
-namespace mlx::core::rocm {
-
-void fence() {
-  // Placeholder for ROCm fence operation
-}
-
-} // namespace mlx::core::rocm

mlx/backend/rocm/indexing.cpp

@@ -1,9 +0,0 @@
-// Copyright © 2025 Apple Inc.
-
-namespace mlx::core::rocm {
-
-void index() {
-  // Placeholder for ROCm indexing operation
-}
-
-} // namespace mlx::core::rocm

mlx/backend/rocm/iterators/general_iterator.hpp

@@ -1,153 +0,0 @@
-// Copyright © 2025 Apple Inc.
-
-#pragma once
-
-#include <hip/hip_runtime.h>
-#include <cstdint>
-
-namespace mlx::core::rocm {
-
-template <typename IdxType>
-struct GeneralIterator {
-  using difference_type = ptrdiff_t;
-  using value_type = IdxType;
-  using pointer = IdxType*;
-  using reference = IdxType&;
-  using iterator_category = std::random_access_iterator_tag;
-
-  const IdxType* base_ptr;
-  IdxType offset;
-  const int* shape;
-  const size_t* strides;
-  int ndim;
-  size_t size;
-
-  __device__ GeneralIterator(
-      const IdxType* base_ptr,
-      IdxType offset,
-      const int* shape,
-      const size_t* strides,
-      int ndim,
-      size_t size)
-      : base_ptr(base_ptr),
-        offset(offset),
-        shape(shape),
-        strides(strides),
-        ndim(ndim),
-        size(size) {}
-
-  __device__ GeneralIterator operator+(difference_type n) const {
-    return GeneralIterator(base_ptr, offset + n, shape, strides, ndim, size);
-  }
-
-  __device__ GeneralIterator operator-(difference_type n) const {
-    return GeneralIterator(base_ptr, offset - n, shape, strides, ndim, size);
-  }
-
-  __device__ difference_type operator-(const GeneralIterator& other) const {
-    return offset - other.offset;
-  }
-
-  __device__ GeneralIterator& operator+=(difference_type n) {
-    offset += n;
-    return *this;
-  }
-
-  __device__ GeneralIterator& operator-=(difference_type n) {
-    offset -= n;
-    return *this;
-  }
-
-  __device__ GeneralIterator& operator++() {
-    ++offset;
-    return *this;
-  }
-
-  __device__ GeneralIterator operator++(int) {
-    GeneralIterator temp = *this;
-    ++offset;
-    return temp;
-  }
-
-  __device__ GeneralIterator& operator--() {
-    --offset;
-    return *this;
-  }
-
-  __device__ GeneralIterator operator--(int) {
-    GeneralIterator temp = *this;
-    --offset;
-    return temp;
-  }
-
-  __device__ bool operator==(const GeneralIterator& other) const {
-    return offset == other.offset;
-  }
-
-  __device__ bool operator!=(const GeneralIterator& other) const {
-    return offset != other.offset;
-  }
-
-  __device__ bool operator<(const GeneralIterator& other) const {
-    return offset < other.offset;
-  }
-
-  __device__ bool operator>(const GeneralIterator& other) const {
-    return offset > other.offset;
-  }
-
-  __device__ bool operator<=(const GeneralIterator& other) const {
-    return offset <= other.offset;
-  }
-
-  __device__ bool operator>=(const GeneralIterator& other) const {
-    return offset >= other.offset;
-  }
-
-  __device__ IdxType operator*() const {
-    return base_ptr[elem_to_loc(offset, shape, strides, ndim)];
-  }
-
-  __device__ IdxType operator[](difference_type n) const {
-    return base_ptr[elem_to_loc(offset + n, shape, strides, ndim)];
-  }
-
- private:
-  __device__ size_t elem_to_loc(
-      size_t elem,
-      const int* shape,
-      const size_t* strides,
-      int ndim) const {
-    size_t loc = 0;
-    for (int i = ndim - 1; i >= 0; --i) {
-      auto q_and_r = div(elem, static_cast<size_t>(shape[i]));
-      loc += q_and_r.rem * strides[i];
-      elem = q_and_r.quot;
-    }
-    return loc;
-  }
-
-  __device__ div_t div(size_t numer, size_t denom) const {
-    div_t result;
-    result.quot = numer / denom;
-    result.rem = numer % denom;
-    return result;
-  }
-};
-
-template <typename IdxType>
-__device__ std::pair<GeneralIterator<IdxType>, GeneralIterator<IdxType>>
-make_general_iterators(
-    const IdxType* base_ptr,
-    size_t size,
-    const int* shape,
-    const size_t* strides,
-    int ndim) {
-  auto begin =
-      GeneralIterator<IdxType>(base_ptr, 0, shape, strides, ndim, size);
-  auto end =
-      GeneralIterator<IdxType>(base_ptr, size, shape, strides, ndim, size);
-  return std::make_pair(begin, end);
-}
-
-} // namespace mlx::core::rocm

mlx/backend/rocm/iterators/strided_iterator.hpp

@@ -1,106 +0,0 @@
-// Copyright © 2025 Apple Inc.
-
-#pragma once
-
-#include <hip/hip_runtime.h>
-#include <cstdint>
-
-namespace mlx::core::rocm {
-
-template <typename T>
-struct StridedIterator {
-  using difference_type = ptrdiff_t;
-  using value_type = T;
-  using pointer = T*;
-  using reference = T&;
-  using iterator_category = std::random_access_iterator_tag;
-
-  T* ptr;
-  size_t stride;
-
-  __device__ StridedIterator(T* ptr, size_t stride)
-      : ptr(ptr), stride(stride) {}
-
-  __device__ StridedIterator operator+(difference_type n) const {
-    return StridedIterator(ptr + n * stride, stride);
-  }
-
-  __device__ StridedIterator operator-(difference_type n) const {
-    return StridedIterator(ptr - n * stride, stride);
-  }
-
-  __device__ difference_type operator-(const StridedIterator& other) const {
-    return (ptr - other.ptr) / stride;
-  }
-
-  __device__ StridedIterator& operator+=(difference_type n) {
-    ptr += n * stride;
-    return *this;
-  }
-
-  __device__ StridedIterator& operator-=(difference_type n) {
-    ptr -= n * stride;
-    return *this;
-  }
-
-  __device__ StridedIterator& operator++() {
-    ptr += stride;
-    return *this;
-  }
-
-  __device__ StridedIterator operator++(int) {
-    StridedIterator temp = *this;
-    ptr += stride;
-    return temp;
-  }
-
-  __device__ StridedIterator& operator--() {
-    ptr -= stride;
-    return *this;
-  }
-
-  __device__ StridedIterator operator--(int) {
-    StridedIterator temp = *this;
-    ptr -= stride;
-    return temp;
-  }
-
-  __device__ bool operator==(const StridedIterator& other) const {
-    return ptr == other.ptr;
-  }
-
-  __device__ bool operator!=(const StridedIterator& other) const {
-    return ptr != other.ptr;
-  }
-
-  __device__ bool operator<(const StridedIterator& other) const {
-    return ptr < other.ptr;
-  }
-
-  __device__ bool operator>(const StridedIterator& other) const {
-    return ptr > other.ptr;
-  }
-
-  __device__ bool operator<=(const StridedIterator& other) const {
-    return ptr <= other.ptr;
-  }
-
-  __device__ bool operator>=(const StridedIterator& other) const {
-    return ptr >= other.ptr;
-  }
-
-  __device__ T& operator*() const {
-    return *ptr;
-  }
-
-  __device__ T& operator[](difference_type n) const {
-    return *(ptr + n * stride);
-  }
-};
-
-template <typename T>
-__device__ StridedIterator<T> make_strided_iterator(T* ptr, size_t stride) {
-  return StridedIterator<T>(ptr, stride);
-}
-
-} // namespace mlx::core::rocm

mlx/backend/rocm/jit_module.cpp

@@ -1,167 +0,0 @@
-// Copyright © 2025 Apple Inc.
-
-#include "mlx/backend/rocm/jit_module.h"
-#include "mlx/backend/rocm/utils.h"
-
-#include <fmt/format.h>
-#include <mutex>
-#include <sstream>
-
-namespace mlx::core::rocm {
-
-JitModule::JitModule(
-    const std::string& kernel_name,
-    const std::string& kernel_source,
-    const std::vector<std::string>& template_args,
-    const std::vector<std::string>& compiler_flags,
-    bool verbose) {
-  compile(kernel_name, kernel_source, template_args, compiler_flags, verbose);
-}
-
-JitModule::~JitModule() {
-  if (kernel_) {
-    // No hipFunctionDestroy equivalent in HIP
-  }
-  if (module_) {
-    CHECK_HIP_ERROR(hipModuleUnload(module_));
-  }
-  if (program_) {
-    hiprtcDestroyProgram(&program_);
-  }
-}
-
-void JitModule::compile(
-    const std::string& kernel_name,
-    const std::string& kernel_source,
-    const std::vector<std::string>& template_args,
-    const std::vector<std::string>& compiler_flags,
-    bool verbose) {
-  // Create HIPRTC program
-  CHECK_HIP_ERROR(hiprtcCreateProgram(
-      &program_,
-      kernel_source.c_str(),
-      kernel_name.c_str(),
-      0,
-      nullptr,
-      nullptr));
-
-  // Build compiler options
-  std::vector<const char*> options;
-  std::vector<std::string> option_strings;
-
-  // Add default options
-  option_strings.push_back("--std=c++17");
-  option_strings.push_back("-O3");
-  option_strings.push_back("-DMLX_USE_ROCM");
-
-  // Add user-provided flags
-  for (const auto& flag : compiler_flags) {
-    option_strings.push_back(flag);
-  }
-
-  // Add template arguments
-  for (const auto& arg : template_args) {
-    option_strings.push_back("-D" + arg);
-  }
-
-  // Convert to char* array
-  for (const auto& option : option_strings) {
-    options.push_back(option.c_str());
-  }
-
-  // Compile the program
-  hiprtcResult compile_result =
-      hiprtcCompileProgram(program_, options.size(), options.data());
-
-  // Get compilation log
-  size_t log_size;
-  CHECK_HIP_ERROR(hiprtcGetProgramLogSize(program_, &log_size));
-
-  if (log_size > 1) {
-    std::vector<char> log(log_size);
-    CHECK_HIP_ERROR(hiprtcGetProgramLog(program_, log.data()));
-
-    if (verbose || compile_result != HIPRTC_SUCCESS) {
-      fmt::print(
-          "HIPRTC compilation log for {}:\n{}\n", kernel_name, log.data());
-    }
-  }
-
-  if (compile_result != HIPRTC_SUCCESS) {
-    throw std::runtime_error(
-        fmt::format("HIPRTC compilation failed for kernel {}", kernel_name));
-  }
-
-  // Get compiled code
-  size_t code_size;
-  CHECK_HIP_ERROR(hiprtcGetCodeSize(program_, &code_size));
-
-  std::vector<char> code(code_size);
-  CHECK_HIP_ERROR(hiprtcGetCode(program_, code.data()));
-
-  // Load module
-  CHECK_HIP_ERROR(hipModuleLoadData(&module_, code.data()));
-
-  // Get kernel function
-  CHECK_HIP_ERROR(hipModuleGetFunction(&kernel_, module_, kernel_name.c_str()));
-}
-
-JitCache& JitCache::instance() {
-  static JitCache cache;
-  return cache;
-}
-
-std::shared_ptr<JitModule> JitCache::get_or_create(
-    const std::string& kernel_name,
-    const std::string& kernel_source,
-    const std::vector<std::string>& template_args,
-    const std::vector<std::string>& compiler_flags) {
-  std::string key =
-      make_key(kernel_name, kernel_source, template_args, compiler_flags);
-
-  std::lock_guard<std::mutex> lock(mutex_);
-
-  auto it = cache_.find(key);
-  if (it != cache_.end()) {
-    if (auto module = it->second.lock()) {
-      return module;
-    } else {
-      cache_.erase(it);
-    }
-  }
-
-  auto module = std::make_shared<JitModule>(
-      kernel_name, kernel_source, template_args, compiler_flags);
-  cache_[key] = module;
-  return module;
-}
-
-std::string JitCache::make_key(
-    const std::string& kernel_name,
-    const std::string& kernel_source,
-    const std::vector<std::string>& template_args,
-    const std::vector<std::string>& compiler_flags) const {
-  std::ostringstream oss;
-  oss << kernel_name << "|" << kernel_source;
-
-  for (const auto& arg : template_args) {
-    oss << "|" << arg;
-  }
-
-  for (const auto& flag : compiler_flags) {
-    oss << "|" << flag;
-  }
-
-  return oss.str();
-}
-
-std::shared_ptr<JitModule> make_jit_kernel(
-    const std::string& kernel_name,
-    const std::string& kernel_source,
-    const std::vector<std::string>& template_args,
-    const std::vector<std::string>& compiler_flags) {
-  return JitCache::instance().get_or_create(
-      kernel_name, kernel_source, template_args, compiler_flags);
-}
-
-} // namespace mlx::core::rocm

mlx/backend/rocm/jit_module.h

@@ -1,100 +0,0 @@
-// Copyright © 2025 Apple Inc.
-
-#pragma once
-
-#include <hip/hip_runtime.h>
-#include <hip/hiprtc.h>
-
-#include <memory>
-#include <string>
-#include <unordered_map>
-#include <vector>
-
-namespace mlx::core::rocm {
-
-// JIT compilation module for ROCm
-class JitModule {
- public:
-  JitModule(
-      const std::string& kernel_name,
-      const std::string& kernel_source,
-      const std::vector<std::string>& template_args = {},
-      const std::vector<std::string>& compiler_flags = {},
-      bool verbose = false);
-
-  ~JitModule();
-
-  JitModule(const JitModule&) = delete;
-  JitModule& operator=(const JitModule&) = delete;
-
-  // Get the compiled kernel function
-  hipFunction_t get_kernel() const {
-    return kernel_;
-  }
-
-  // Launch the kernel with given arguments
-  template <typename... Args>
-  void launch(
-      dim3 grid_dims,
-      dim3 block_dims,
-      size_t shared_memory,
-      hipStream_t stream,
-      Args&&... args) {
-    void* kernel_args[] = {(void*)&args...};
-    CHECK_HIP_ERROR(hipModuleLaunchKernel(
-        kernel_,
-        grid_dims.x,
-        grid_dims.y,
-        grid_dims.z,
-        block_dims.x,
-        block_dims.y,
-        block_dims.z,
-        shared_memory,
-        stream,
-        kernel_args,
-        nullptr));
-  }
-
- private:
-  void compile(
-      const std::string& kernel_name,
-      const std::string& kernel_source,
-      const std::vector<std::string>& template_args,
-      const std::vector<std::string>& compiler_flags,
-      bool verbose);
-
-  hiprtcProgram program_{nullptr};
-  hipModule_t module_{nullptr};
-  hipFunction_t kernel_{nullptr};
-};
-
-// JIT cache for compiled modules
-class JitCache {
- public:
-  static JitCache& instance();
-
-  std::shared_ptr<JitModule> get_or_create(
-      const std::string& kernel_name,
-      const std::string& kernel_source,
-      const std::vector<std::string>& template_args = {},
-      const std::vector<std::string>& compiler_flags = {});
-
- private:
-  std::unordered_map<std::string, std::weak_ptr<JitModule>> cache_;
-  std::mutex mutex_;
-
-  std::string make_key(
-      const std::string& kernel_name,
-      const std::string& kernel_source,
-      const std::vector<std::string>& template_args,
-      const std::vector<std::string>& compiler_flags) const;
-};
-
-// Helper function to create and cache JIT modules
-std::shared_ptr<JitModule> make_jit_kernel(
-    const std::string& kernel_name,
-    const std::string& kernel_source,
-    const std::vector<std::string>& template_args = {},
-    const std::vector<std::string>& compiler_flags = {});
-
-} // namespace mlx::core::rocm

mlx/backend/rocm/kernel_utils.hip

@@ -1,29 +0,0 @@
-// Copyright © 2025 Apple Inc.
-
-#include <hip/hip_runtime.h>
-
-namespace mlx::core::rocm {
-
-// Utility functions for HIP kernels
-
-__device__ inline int get_global_id() {
-  return blockIdx.x * blockDim.x + threadIdx.x;
-}
-
-__device__ inline int get_local_id() {
-  return threadIdx.x;
-}
-
-__device__ inline int get_group_id() {
-  return blockIdx.x;
-}
-
-__device__ inline int get_local_size() {
-  return blockDim.x;
-}
-
-__device__ inline int get_num_groups() {
-  return gridDim.x;
-}
-
-} // namespace mlx::core::rocm 

mlx/backend/rocm/kernel_utils.hpp

@@ -1,135 +0,0 @@
-// Copyright © 2025 Apple Inc.
-
-#pragma once
-
-#include <hip/hip_runtime.h>
-#include <array>
-
-namespace mlx::core::rocm {
-
-// Constants
-constexpr int MAX_DIMS = 8;
-
-// HIP array type for passing arrays to kernels
-template <typename T, int N>
-using hip_array = std::array<T, N>;
-
-// Helper to create hip_array from vector
-template <int N, typename T>
-__host__ hip_array<T, N> make_hip_array(const std::vector<T>& vec) {
-  hip_array<T, N> arr;
-  for (int i = 0; i < N && i < vec.size(); ++i) {
-    arr[i] = vec[i];
-  }
-  return arr;
-}
-
-template <typename T>
-__host__ hip_array<T, MAX_DIMS> make_hip_array(const std::vector<T>& vec) {
-  return make_hip_array<MAX_DIMS>(vec);
-}
-
-// Type mapping from MLX types to HIP types
-template <typename T>
-using hip_type_t = T;
-
-template <>
-using hip_type_t<float16> = __half;
-
-template <>
-using hip_type_t<bfloat16> = __hip_bfloat16;
-
-template <>
-using hip_type_t<complex64> = hipFloatComplex;
-
-// Element to location mapping for general broadcasting
-template <int NDIM>
-__device__ std::pair<int64_t, int64_t> elem_to_loc_nd(
-    int64_t elem,
-    const int32_t* shape,
-    const int64_t* a_strides,
-    const int64_t* b_strides) {
-  int64_t a_idx = 0;
-  int64_t b_idx = 0;
-
-  for (int i = NDIM - 1; i >= 0; --i) {
-    int64_t pos_in_dim = elem % shape[i];
-    elem /= shape[i];
-    a_idx += pos_in_dim * a_strides[i];
-    b_idx += pos_in_dim * b_strides[i];
-  }
-
-  return {a_idx, b_idx};
-}
-
-// 4D specialization for performance
-__device__ inline std::pair<int64_t, int64_t> elem_to_loc_4d(
-    int64_t elem,
-    const int32_t* shape,
-    const int64_t* a_strides,
-    const int64_t* b_strides,
-    int ndim) {
-  int64_t a_idx = 0;
-  int64_t b_idx = 0;
-
-  for (int i = ndim - 1; i >= 0; --i) {
-    int64_t pos_in_dim = elem % shape[i];
-    elem /= shape[i];
-    a_idx += pos_in_dim * a_strides[i];
-    b_idx += pos_in_dim * b_strides[i];
-  }
-
-  return {a_idx, b_idx};
-}
-
-// Launch configuration calculation
-template <typename Kernel>
-std::pair<dim3, dim3>
-get_launch_args(Kernel kernel, const array& out, bool large = false) {
-  int threads_per_block = 256;
-  int64_t total_threads = out.size();
-
-  if (large) {
-    // For large arrays, use more blocks
-    int64_t blocks =
-        (total_threads + threads_per_block - 1) / threads_per_block;
-    return {dim3(blocks), dim3(threads_per_block)};
-  } else {
-    int blocks = (total_threads + threads_per_block - 1) / threads_per_block;
-    return {dim3(blocks), dim3(threads_per_block)};
-  }
-}
-
-template <typename Kernel>
-std::pair<dim3, dim3> get_launch_args(
-    Kernel kernel,
-    int64_t size,
-    const std::vector<int>& shape,
-    const std::vector<size_t>& strides,
-    bool large = false) {
-  int threads_per_block = 256;
-
-  if (large) {
-    int64_t blocks = (size + threads_per_block - 1) / threads_per_block;
-    return {dim3(blocks), dim3(threads_per_block)};
-  } else {
-    int blocks = (size + threads_per_block - 1) / threads_per_block;
-    return {dim3(blocks), dim3(threads_per_block)};
-  }
-}
-
-// Cooperative groups thread rank equivalent
-namespace cooperative_groups {
-class grid_group {
- public:
-  __device__ int64_t thread_rank() const {
-    return blockIdx.x * blockDim.x + threadIdx.x;
-  }
-};
-
-__device__ grid_group this_grid() {
-  return grid_group{};
-}
-} // namespace cooperative_groups
-
-} // namespace mlx::core::rocm

mlx/backend/rocm/layer_norm.hip

@@ -1,437 +0,0 @@
-// Copyright © 2025 Apple Inc.
-
-#include "mlx/backend/rocm/device.h"
-#include "mlx/backend/rocm/iterators/strided_iterator.hpp"
-#include "mlx/backend/rocm/kernel_utils.hpp"
-#include "mlx/backend/rocm/reduce/reduce.hpp"
-#include "mlx/backend/gpu/copy.h"
-#include "mlx/dtype_utils.h"
-#include "mlx/fast_primitives.h"
-
-#include <hip/hip_runtime.h>
-#include <hip/hip_cooperative_groups.h>
-#include <rocprim/block/block_load.hpp>
-#include <rocprim/block/block_reduce.hpp>
-
-namespace mlx::core {
-
-namespace rocm {
-
-namespace cg = cooperative_groups;
-
-inline __device__ float3 plus_f3(const float3& a, const float3& b) {
-  return {a.x + b.x, a.y + b.y, a.z + b.z};
-}
-
-// Similar to rocprim::BlockReduce, but result is broadcasted to every thread.
-template <typename T, int BLOCK_DIM>
-struct BlockBroadcastReduce {
-  static_assert(WARP_SIZE <= BLOCK_DIM && BLOCK_DIM <= WARP_SIZE * WARP_SIZE);
-  static_assert(BLOCK_DIM % WARP_SIZE == 0);
-  using TempStorage = T[BLOCK_DIM / WARP_SIZE];
-
-  cg::thread_block& block;
-  TempStorage& temp;
-
-  template <typename Op>
-  __device__ T Reduce(const T& input, const Op& op, const T& init_value) {
-    auto warp = cg::tiled_partition<WARP_SIZE>(block);
-    T x = cg::reduce(warp, input, op);
-    if (warp.thread_rank() == 0) {
-      temp[warp.meta_group_rank()] = x;
-    }
-    block.sync();
-    x = warp.thread_rank() < warp.meta_group_size() ? temp[warp.thread_rank()]
-                                                    : init_value;
-    return cg::reduce(warp, x, op);
-  }
-
-  __device__ T Sum(const T& input) {
-    return Reduce(input, hip_plus<T>{}, T{});
-  }
-};
-
-template <typename T, int BLOCK_DIM, int N_READS = 4>
-__global__ void layer_norm(
-    const T* x,
-    const T* w,
-    const T* b,
-    T* out,
-    float eps,
-    int32_t axis_size,
-    int64_t w_stride,
-    int64_t b_stride) {
-  auto grid = cg::this_grid();
-  auto block = cg::this_thread_block();
-
-  using BlockReduceT = BlockBroadcastReduce<float, BLOCK_DIM>;
-  __shared__ typename BlockReduceT::TempStorage temp;
-
-  x += grid.block_rank() * axis_size;
-  out += grid.block_rank() * axis_size;
-
-  // Sum.
-  float sum = 0;
-  for (int r = 0; r < hip_ceil_div(axis_size, BLOCK_DIM * N_READS); ++r) {
-    auto index = r * BLOCK_DIM + block.thread_rank();
-    T xn[N_READS] = {};
-    rocprim::block_load_direct_blocked(index, x, xn, axis_size);
-    sum += static_cast<float>(rocprim::thread_reduce(xn, hip_plus<T>{}));
-  }
-  sum = BlockReduceT{block, temp}.Sum(sum);
-
-  // Mean.
-  float mean = sum / axis_size;
-
-  // Normalizer.
-  float normalizer = 0;
-  for (int r = 0; r < hip_ceil_div(axis_size, BLOCK_DIM * N_READS); ++r) {
-    auto index = r * BLOCK_DIM + block.thread_rank();
-    T xn[N_READS];
-    rocprim::block_load_direct_blocked(index, x, xn, axis_size, mean);
-    for (int i = 0; i < N_READS; ++i) {
-      float t = static_cast<float>(xn[i]) - mean;
-      normalizer += t * t;
-    }
-  }
-  normalizer = BlockReduceT{block, temp}.Sum(normalizer);
-  normalizer = rsqrt(normalizer / axis_size + eps);
-
-  // Outputs.
-  for (int r = 0; r < hip_ceil_div(axis_size, BLOCK_DIM * N_READS); ++r) {
-    auto index = r * BLOCK_DIM + block.thread_rank();
-    T xn[N_READS];
-    T wn[N_READS];
-    T bn[N_READS];
-    rocprim::block_load_direct_blocked(index, x, xn, axis_size);
-    rocprim::block_load_direct_blocked(index, strided_iterator(w, w_stride), wn, axis_size);
-    rocprim::block_load_direct_blocked(index, strided_iterator(b, b_stride), bn, axis_size);
-    for (int i = 0; i < N_READS; ++i) {
-      float norm = (static_cast<float>(xn[i]) - mean) * normalizer;
-      xn[i] = wn[i] * static_cast<T>(norm) + bn[i];
-    }
-    rocprim::block_store_direct_blocked(index, out, xn, axis_size);
-  }
-}
-
-template <typename T, bool HAS_W, int BLOCK_DIM, int N_READS = 4>
-__global__ void layer_norm_vjp(
-    const T* x,
-    const T* w,
-    const T* g,
-    T* gx,
-    T* gw,
-    float eps,
-    int32_t axis_size,
-    int64_t w_stride) {
-  auto grid = cg::this_grid();
-  auto block = cg::this_thread_block();
-
-  using BlockReduceF = BlockBroadcastReduce<float, BLOCK_DIM>;
-  using BlockReduceF3 = BlockBroadcastReduce<float3, BLOCK_DIM>;
-  __shared__ union {
-    typename BlockReduceF::TempStorage f;
-    typename BlockReduceF3::TempStorage f3;
-  } temp;
-
-  x += grid.block_rank() * axis_size;
-  g += grid.block_rank() * axis_size;
-  gx += grid.block_rank() * axis_size;
-  gw += grid.block_rank() * axis_size;
-
-  // Sum.
-  float sum = 0;
-  for (int r = 0; r < hip_ceil_div(axis_size, BLOCK_DIM * N_READS); ++r) {
-    auto index = r * BLOCK_DIM + block.thread_rank();
-    T xn[N_READS] = {};
-    rocprim::block_load_direct_blocked(index, x, xn, axis_size);
-    sum += static_cast<float>(rocprim::thread_reduce(xn, hip_plus<T>{}));
-  }
-  sum = BlockReduceF{block, temp.f}.Sum(sum);
-
-  // Mean.
-  float mean = sum / axis_size;
-
-  // Normalizer.
-  float3 factors = {};
-  for (int r = 0; r < hip_ceil_div(axis_size, BLOCK_DIM * N_READS); ++r) {
-    T xn[N_READS];
-    T wn[N_READS] = {};
-    T gn[N_READS] = {};
-    auto index = r * BLOCK_DIM + block.thread_rank();
-    rocprim::block_load_direct_blocked(index, x, xn, axis_size, mean);
-    rocprim::block_load_direct_blocked(index, g, gn, axis_size);
-    rocprim::block_load_direct_blocked(index, strided_iterator(w, w_stride), wn, axis_size);
-    for (int i = 0; i < N_READS; i++) {
-      float t = static_cast<float>(xn[i]) - mean;
-      float wi = wn[i];
-      float gi = gn[i];
-      float wg = wi * gi;
-      factors = plus_f3(factors, {wg, wg * t, t * t});
-    }
-  }
-  factors = BlockReduceF3{block, temp.f3}.Reduce(factors, plus_f3, {});
-  float meanwg = factors.x / axis_size;
-  float meanwgxc = factors.y / axis_size;
-  float normalizer2 = 1 / (factors.z / axis_size + eps);
-  float normalizer = sqrt(normalizer2);
-
-  // Outputs.
-  for (int r = 0; r < hip_ceil_div(axis_size, BLOCK_DIM * N_READS); ++r) {
-    auto index = r * BLOCK_DIM + block.thread_rank();
-    T xn[N_READS];
-    T wn[N_READS];
-    T gn[N_READS];
-    rocprim::block_load_direct_blocked(index, x, xn, axis_size);
-    rocprim::block_load_direct_blocked(index, g, gn, axis_size);
-    rocprim::block_load_direct_blocked(index, strided_iterator(w, w_stride), wn, axis_size);
-    for (int i = 0; i < N_READS; i++) {
-      float xi = (static_cast<float>(xn[i]) - mean) * normalizer;
-      float wi = wn[i];
-      float gi = gn[i];
-      xn[i] = normalizer * (wi * gi - meanwg) - xi * meanwgxc * normalizer2;
-      if constexpr (HAS_W) {
-        wn[i] = gi * xi;
-      }
-    }
-    rocprim::block_store_direct_blocked(index, gx, xn, axis_size);
-    if constexpr (HAS_W) {
-      rocprim::block_store_direct_blocked(index, gw, wn, axis_size);
-    }
-  }
-}
-
-// Utility functions
-template <typename T>
-struct hip_plus {
-  __device__ T operator()(const T& a, const T& b) const {
-    return a + b;
-  }
-};
-
-inline __device__ int hip_ceil_div(int a, int b) {
-  return (a + b - 1) / b;
-}
-
-template <typename T>
-__device__ inline auto strided_iterator(const T* ptr, int64_t stride) {
-  return ptr + stride;  // Simplified strided iterator
-}
-
-} // namespace rocm
-
-namespace fast {
-
-bool LayerNorm::use_fallback(Stream s) {
-  return s.device == Device::cpu;
-}
-
-// TODO: There are duplicate code with backend/metal/normalization.cpp
-void LayerNorm::eval_gpu(
-    const std::vector<array>& inputs,
-    std::vector<array>& outputs) {
-  auto& s = stream();
-  auto& out = outputs[0];
-
-  // Make sure that the last dimension is contiguous.
-  auto set_output = [&s, &out](const array& x) {
-    bool no_copy = x.flags().contiguous && x.strides()[x.ndim() - 1] == 1;
-    if (no_copy && x.ndim() > 1) {
-      auto s = x.strides()[x.ndim() - 2];
-      no_copy &= (s == 0 || s == x.shape().back());
-    }
-    if (no_copy) {
-      if (x.is_donatable()) {
-        out.copy_shared_buffer(x);
-      } else {
-        out.set_data(
-            allocator::malloc(x.data_size() * x.itemsize()),
-            x.data_size(),
-            x.strides(),
-            x.flags());
-      }
-      return x;
-    } else {
-      auto x_copy = array(x.shape(), x.dtype(), nullptr, {});
-      copy_gpu(x, x_copy, CopyType::General, s);
-      out.copy_shared_buffer(x_copy);
-      return x_copy;
-    }
-  };
-
-  const array x = set_output(inputs[0]);
-  const array& w = inputs[1];
-  const array& b = inputs[2];
-
-  int32_t axis_size = x.shape().back();
-  int32_t n_rows = x.data_size() / axis_size;
-  int64_t w_stride = (w.ndim() == 1) ? w.strides()[0] : 0;
-  int64_t b_stride = (b.ndim() == 1) ? b.strides()[0] : 0;
-
-  auto& encoder = rocm::get_command_encoder(s);
-  encoder.set_input_array(x);
-  encoder.set_input_array(w);
-  encoder.set_input_array(b);
-  encoder.set_output_array(out);
-  encoder.launch_kernel([&](hipStream_t stream) {
-    MLX_SWITCH_FLOAT_TYPES_CHECKED(out.dtype(), "layernorm", CTYPE, {
-      using DataType = hip_type_t<CTYPE>;
-      constexpr uint32_t N_READS = 4;
-      MLX_SWITCH_BLOCK_DIM(rocm::hip_ceil_div(axis_size, N_READS), BLOCK_DIM, {
-        auto kernel = rocm::layer_norm<DataType, BLOCK_DIM, N_READS>;
-        hipLaunchKernelGGL(kernel, n_rows, BLOCK_DIM, 0, stream,
-            x.data<DataType>(),
-            w.data<DataType>(),
-            b.data<DataType>(),
-            out.data<DataType>(),
-            eps_,
-            axis_size,
-            w_stride,
-            b_stride);
-      });
-    });
-  });
-}
-
-void LayerNormVJP::eval_gpu(
-    const std::vector<array>& inputs,
-    std::vector<array>& outputs) {
-  auto& s = stream();
-  auto& encoder = rocm::get_command_encoder(s);
-
-  // Ensure row contiguity. We could relax this step by checking that the array
-  // is contiguous (no broadcasts or holes) and that the input strides are the
-  // same as the cotangent strides but for now this is simpler.
-  auto check_input = [&s](const array& x) -> std::pair<array, bool> {
-    if (x.flags().row_contiguous) {
-      return {x, false};
-    }
-    array x_copy(x.shape(), x.dtype(), nullptr, {});
-    copy_gpu(x, x_copy, CopyType::General, s);
-    return {x_copy, true};
-  };
-  bool donate_x = inputs[0].is_donatable();
-  bool donate_g = inputs[3].is_donatable();
-  auto [x, copied] = check_input(inputs[0]);
-  donate_x |= copied;
-  const array& w = inputs[1];
-  const array& b = inputs[2];
-  auto [g, g_copied] = check_input(inputs[3]);
-  donate_g |= g_copied;
-  array& gx = outputs[0];
-  array& gw = outputs[1];
-  array& gb = outputs[2];
-
-  // Check whether we had a weight.
-  bool has_w = w.ndim() != 0;
-
-  // Allocate space for the outputs.
-  bool g_in_gx = false;
-  if (donate_x) {
-    gx.copy_shared_buffer(x);
-  } else if (donate_g) {
-    gx.copy_shared_buffer(g);
-    g_in_gx = true;
-  } else {
-    gx.set_data(allocator::malloc(gx.nbytes()));
-  }
-  if (g_copied && !g_in_gx) {
-    encoder.add_temporary(g);
-  }
-
-  int32_t axis_size = x.shape().back();
-  int32_t n_rows = x.data_size() / axis_size;
-  int64_t w_stride = (w.ndim() == 1) ? w.strides()[0] : 0;
-
-  // Allocate a temporary to store the gradients for w and allocate the output
-  // gradient accumulators.
-  array gw_temp =
-      (has_w) ? array({n_rows, x.shape().back()}, gw.dtype(), nullptr, {}) : w;
-  if (has_w) {
-    if (!g_in_gx && donate_g) {
-      gw_temp.copy_shared_buffer(g);
-    } else {
-      gw_temp.set_data(allocator::malloc(gw_temp.nbytes()));
-      encoder.add_temporary(gw_temp);
-    }
-  }
-  gw.set_data(allocator::malloc(gw.nbytes()));
-  gb.set_data(allocator::malloc(gb.nbytes()));
-
-  // Finish with the gradient for b in case we had a b.
-  if (gb.ndim() == 1 && gb.size() == axis_size) {
-    ReductionPlan plan(
-        ReductionOpType::ContiguousStridedReduce, {n_rows}, {axis_size});
-    col_reduce(encoder, g, gb, Reduce::ReduceType::Sum, {0}, plan);
-  }
-
-  encoder.set_input_array(x);
-  encoder.set_input_array(w);
-  encoder.set_input_array(g);
-  encoder.set_output_array(gx);
-  encoder.set_output_array(gw_temp);
-  encoder.launch_kernel([&, x = x, g = g](hipStream_t stream) {
-    MLX_SWITCH_FLOAT_TYPES_CHECKED(gx.dtype(), "layernorm_vjp", CTYPE, {
-      using DataType = hip_type_t<CTYPE>;
-      constexpr int N_READS = 4;
-      MLX_SWITCH_BOOL(has_w, HAS_W, {
-        MLX_SWITCH_BLOCK_DIM(rocm::hip_ceil_div(axis_size, N_READS), BLOCK_DIM, {
-          auto kernel = rocm::layer_norm_vjp<DataType, HAS_W, BLOCK_DIM, N_READS>;
-          hipLaunchKernelGGL(kernel, n_rows, BLOCK_DIM, 0, stream,
-              x.data<DataType>(),
-              w.data<DataType>(),
-              g.data<DataType>(),
-              gx.data<DataType>(),
-              gw_temp.data<DataType>(),
-              eps_,
-              axis_size,
-              w_stride);
-        });
-      });
-    });
-  });
-
-  if (has_w) {
-    ReductionPlan plan(
-        ReductionOpType::ContiguousStridedReduce, {n_rows}, {axis_size});
-    col_reduce(encoder, gw_temp, gw, Reduce::ReduceType::Sum, {0}, plan);
-  }
-}
-
-} // namespace fast
-
-} // namespace mlx::core
-
-namespace mlx::core::rocm {
-
-__global__ void layer_norm_kernel(
-    float* input, 
-    float* output, 
-    float* gamma, 
-    float* beta, 
-    int n, 
-    float eps) {
-  int idx = blockIdx.x * blockDim.x + threadIdx.x;
-  
-  if (idx < n) {
-    // Simplified layer norm placeholder
-    // Real implementation would compute mean and variance
-    output[idx] = gamma[idx] * input[idx] + beta[idx];
-  }
-}
-
-void launch_layer_norm(
-    float* input, 
-    float* output, 
-    float* gamma, 
-    float* beta, 
-    int n, 
-    float eps, 
-    hipStream_t stream) {
-  int threads = 256;
-  int blocks = (n + threads - 1) / threads;
-  hipLaunchKernelGGL(layer_norm_kernel, dim3(blocks), dim3(threads), 0, stream, 
-                     input, output, gamma, beta, n, eps);
-}
-
-} // namespace mlx::core::rocm 

mlx/backend/rocm/logsumexp.hip

@@ -1,13 +0,0 @@
-// Copyright © 2025 Apple Inc.
-
-#include <hip/hip_runtime.h>
-
-namespace mlx::core::rocm {
-
-__global__ void logsumexp_kernel(float* input, float* output, int n) {
-  // Placeholder implementation
-  int idx = blockIdx.x * blockDim.x + threadIdx.x;
-  (void)input; (void)output; (void)n; (void)idx;
-}
-
-} // namespace mlx::core::rocm 

mlx/backend/rocm/matmul.cpp

@@ -1,30 +0,0 @@
-// Copyright © 2025 Apple Inc.
-
-#include "mlx/backend/rocm/device.h"
-#include "mlx/backend/rocm/utils.h"
-
-namespace mlx::core::rocm {
-
-void matmul_hip(
-    float* a,
-    float* b,
-    float* c,
-    int m,
-    int n,
-    int k,
-    hipStream_t stream) {
-  // This is a placeholder - in a real implementation, this would use rocBLAS
-  // auto& device = get_current_device();
-  // rocblas_sgemm(device.rocblas_handle(), ...);
-
-  // For now, just a placeholder
-  (void)a;
-  (void)b;
-  (void)c;
-  (void)m;
-  (void)n;
-  (void)k;
-  (void)stream;
-}
-
-} // namespace mlx::core::rocm

mlx/backend/rocm/no_rocm.cpp

@@ -1,11 +0,0 @@
-// Copyright © 2025 Apple Inc.
-
-#include "mlx/backend/rocm/rocm.h"
-
-namespace mlx::core::rocm {
-
-bool is_available() {
-  return false;
-}
-
-} // namespace mlx::core::rocm

mlx/backend/rocm/primitives.hip

@@ -1,21 +0,0 @@
-// Copyright © 2025 Apple Inc.
-
-#include <hip/hip_runtime.h>
-
-#include "mlx/backend/rocm/utils.h"
-#include "mlx/backend/common/primitives.h"
-
-namespace mlx::core::rocm {
-
-// Basic kernel implementations will go here
-// This is a placeholder for ROCm-specific primitive operations
-
-void add_hip() {
-  // Placeholder for HIP add operation
-}
-
-void multiply_hip() {
-  // Placeholder for HIP multiply operation  
-}
-
-} // namespace mlx::core::rocm 

mlx/backend/rocm/random.hip

@@ -1,23 +0,0 @@
-// Copyright © 2025 Apple Inc.
-
-#include <hip/hip_runtime.h>
-
-namespace mlx::core::rocm {
-
-__global__ void random_uniform_kernel(float* output, int n, unsigned int seed) {
-  int idx = blockIdx.x * blockDim.x + threadIdx.x;
-  if (idx < n) {
-    // Simple LCG placeholder - real implementation would use rocRAND
-    unsigned int state = seed + idx;
-    state = state * 1103515245 + 12345;
-    output[idx] = (float)(state & 0x7FFFFFFF) / (float)0x7FFFFFFF;
-  }
-}
-
-void launch_random_uniform(float* output, int n, unsigned int seed, hipStream_t stream) {
-  int threads = 256;
-  int blocks = (n + threads - 1) / threads;
-  hipLaunchKernelGGL(random_uniform_kernel, dim3(blocks), dim3(threads), 0, stream, output, n, seed);
-}
-
-} // namespace mlx::core::rocm 

mlx/backend/rocm/reduce.hip

@@ -1,24 +0,0 @@
-// Copyright © 2025 Apple Inc.
-
-#include <hip/hip_runtime.h>
-
-namespace mlx::core::rocm {
-
-__global__ void sum_reduce_kernel(float* input, float* output, int n) {
-  int idx = blockIdx.x * blockDim.x + threadIdx.x;
-  
-  // Simple reduction placeholder
-  if (idx == 0) {
-    float sum = 0.0f;
-    for (int i = 0; i < n; i++) {
-      sum += input[i];
-    }
-    output[0] = sum;
-  }
-}
-
-void launch_sum_reduce(float* input, float* output, int n, hipStream_t stream) {
-  hipLaunchKernelGGL(sum_reduce_kernel, dim3(1), dim3(1), 0, stream, input, output, n);
-}
-
-} // namespace mlx::core::rocm 

mlx/backend/rocm/reduce/col_reduce.hip

@@ -1,311 +0,0 @@
-// Copyright © 2025 Apple Inc.
-
-#include "mlx/backend/rocm/device.h"
-#include "mlx/backend/rocm/device/cast_op.hpp"
-#include "mlx/backend/rocm/reduce/reduce.hpp"
-
-#include <hip/hip_runtime.h>
-#include <hip/hip_cooperative_groups.h>
-#include <rocprim/block/block_load.hpp>
-
-namespace mlx::core {
-
-namespace rocm {
-
-namespace cg = cooperative_groups;
-
-struct ColReduceArgs {
-  // The size of the contiguous column reduction.
-  size_t reduction_size;
-  int64_t reduction_stride;
-
-  // Input shape and strides excluding the reduction axes.
-  Shape shape;
-  Strides strides;
-  int ndim;
-
-  // Input shape and strides of the reduction axes (including last dimension).
-  Shape reduce_shape;
-  Strides reduce_strides;
-  int reduce_ndim;
-
-  // The number of column we are reducing. Namely prod(reduce_shape).
-  size_t non_col_reductions;
-
-  ColReduceArgs(
-      const array& in,
-      const ReductionPlan& plan,
-      const std::vector<int>& axes) {
-    assert(!plan.shape.empty());
-    reduction_size = plan.shape.back();
-    reduction_stride = plan.strides.back();
-
-    int64_t stride_back = 1;
-    auto [shape_vec, strides_vec] = shapes_without_reduction_axes(in, axes);
-    while (!shape_vec.empty() && stride_back < reduction_stride) {
-      stride_back *= shape_vec.back();
-      shape_vec.pop_back();
-      strides_vec.pop_back();
-    }
-    std::tie(shape_vec, strides_vec) =
-        collapse_contiguous_dims(shape_vec, strides_vec);
-    shape = const_param(shape_vec);
-    strides = const_param(strides_vec);
-    ndim = shape_vec.size();
-
-    reduce_shape = const_param(plan.shape);
-    reduce_strides = const_param(plan.strides);
-    reduce_ndim = plan.shape.size();
-
-    non_col_reductions = 1;
-    for (int i = 0; i < reduce_ndim - 1; i++) {
-      non_col_reductions *= reduce_shape[i];
-    }
-  }
-};
-
-template <typename T, typename U, typename Op, int NDIM, int N_READS = 4>
-__global__ void col_reduce_small(
-    const T* in,
-    U* out,
-    const ColReduceArgs args) {
-  auto grid = cg::this_grid();
-  auto block = cg::this_thread_block();
-
-  int column =
-      grid.block_index().x * block.dim_threads().x + block.thread_index().x;
-  if (column * N_READS >= args.reduction_stride) {
-    return;
-  }
-
-  int out_idx = grid.block_rank() / grid.dim_blocks().x;
-  in += elem_to_loc(out_idx, args.shape.data(), args.strides.data(), args.ndim);
-
-  Op op;
-  U totals[N_READS];
-  for (int i = 0; i < N_READS; i++) {
-    totals[i] = ReduceInit<Op, T>::value();
-  }
-
-  // Read input to local.
-  LoopedElemToLoc<NDIM, (NDIM > 2)> loop(args.reduce_ndim);
-  loop.next(
-      block.thread_index().y,
-      args.reduce_shape.data(),
-      args.reduce_strides.data());
-  for (size_t r = block.thread_index().y;
-       r < args.non_col_reductions * args.reduction_size;
-       r += block.dim_threads().y) {
-    U vals[N_READS];
-    rocprim::block_load_direct_blocked(
-        column,
-        make_cast_iterator<U>(in + loop.location()),
-        vals,
-        args.reduction_stride,
-        ReduceInit<Op, T>::value());
-    for (int i = 0; i < N_READS; i++) {
-      totals[i] = op(vals[i], totals[i]);
-    }
-    loop.next(
-        block.dim_threads().y,
-        args.reduce_shape.data(),
-        args.reduce_strides.data());
-  }
-
-  // Do block reduce when each column has more than 1 element to reduce.
-  if (block.dim_threads().y > 1) {
-    __shared__ U shared_vals[32 * 8 * N_READS];
-    size_t col =
-        block.thread_index().y * block.dim_threads().x + block.thread_index().x;
-    for (int i = 0; i < N_READS; i++) {
-      shared_vals[col * N_READS + i] = totals[i];
-    }
-    block.sync();
-    if (block.thread_index().y == 0) {
-      for (int i = 0; i < N_READS; i++) {
-        totals[i] = shared_vals[block.thread_index().x * N_READS + i];
-      }
-      for (int j = 1; j < block.dim_threads().y; j++) {
-        col = j * block.dim_threads().x + block.thread_index().x;
-        for (int i = 0; i < N_READS; i++) {
-          totals[i] = op(shared_vals[col * N_READS + i], totals[i]);
-        }
-      }
-    }
-  }
-
-  // Write result.
-  if (block.thread_index().y == 0) {
-    rocprim::block_store_direct_blocked(
-        column,
-        out + out_idx * args.reduction_stride,
-        totals,
-        args.reduction_stride);
-  }
-}
-
-template <
-    typename T,
-    typename U,
-    typename Op,
-    int NDIM,
-    int BM,
-    int BN,
-    int N_READS = 4>
-__global__ void col_reduce_looped(
-    const T* in,
-    U* out,
-    const ColReduceArgs args) {
-  auto grid = cg::this_grid();
-  auto block = cg::this_thread_block();
-  auto warp = cg::tiled_partition<WARP_SIZE>(block);
-
-  constexpr int n_warps = BN / N_READS;
-
-  int out_idx = grid.block_rank() / grid.dim_blocks().x;
-  in += elem_to_loc(out_idx, args.shape.data(), args.strides.data(), args.ndim);
-
-  Op op;
-  U totals[N_READS];
-  for (int i = 0; i < N_READS; i++) {
-    totals[i] = ReduceInit<Op, T>::value();
-  }
-
-  // Read input to local.
-  int r = block.thread_rank() / n_warps;
-  int column = block.thread_rank() % n_warps;
-  int in_offset = grid.block_index().x * BN;
-  LoopedElemToLoc<NDIM, (NDIM > 2)> loop(args.reduce_ndim);
-  loop.next(r, args.reduce_shape.data(), args.reduce_strides.data());
-  for (; r < args.non_col_reductions * args.reduction_size; r += BM) {
-    U vals[N_READS];
-    rocprim::block_load_direct_blocked(
-        column,
-        make_cast_iterator<U>(in + loop.location() + in_offset),
-        vals,
-        args.reduction_stride - in_offset,
-        ReduceInit<Op, T>::value());
-    for (int i = 0; i < N_READS; i++) {
-      totals[i] = op(vals[i], totals[i]);
-    }
-    loop.next(BM, args.reduce_shape.data(), args.reduce_strides.data());
-  }
-
-  // Do warp reduce for each output.
-  constexpr int n_outputs = BN / n_warps;
-  static_assert(BM == 32 && n_outputs == N_READS);
-  __shared__ U shared_vals[BM * BN];
-  size_t col = block.thread_index().y * BN + block.thread_index().x * N_READS;
-  for (int i = 0; i < N_READS; i++) {
-    shared_vals[col + i] = totals[i];
-  }
-  block.sync();
-  col = warp.thread_rank() * BN + warp.meta_group_rank() * n_outputs;
-  for (int i = 0; i < n_outputs; i++) {
-    totals[i] = cg::reduce(warp, shared_vals[col + i], op);
-  }
-
-  // Write result.
-  if (warp.thread_rank() == 0) {
-    size_t out_offset = grid.block_index().x * BN;
-    rocprim::block_store_direct_blocked(
-        warp.meta_group_rank(),
-        out + out_idx * args.reduction_stride + out_offset,
-        totals,
-        args.reduction_stride - out_offset);
-  }
-}
-
-// Utility functions and templates
-template <int NDIM, bool USE_FAST_PATH>
-struct LoopedElemToLoc {
-  size_t location;
-  
-  __device__ LoopedElemToLoc(int reduce_ndim) : location(0) {}
-  
-  __device__ void next(size_t step, const int* shape, const size_t* strides) {
-    // Simplified implementation - actual would handle multi-dimensional indexing
-    location += step;
-  }
-};
-
-template <typename T>
-__device__ inline T* make_cast_iterator(const T* ptr) {
-  return const_cast<T*>(ptr);
-}
-
-__device__ inline size_t elem_to_loc(
-    size_t elem,
-    const int* shape,
-    const size_t* strides,
-    int ndim) {
-  size_t loc = 0;
-  for (int i = ndim - 1; i >= 0; --i) {
-    size_t q = elem / shape[i];
-    size_t r = elem % shape[i];
-    loc += r * strides[i];
-    elem = q;
-  }
-  return loc;
-}
-
-} // namespace rocm
-
-inline auto output_grid_for_col_reduce(
-    const array& out,
-    const rocm::ColReduceArgs& args) {
-  auto out_shape = out.shape();
-  auto out_strides = out.strides();
-  while (!out_shape.empty() && out_strides.back() < args.reduction_stride) {
-    out_shape.pop_back();
-    out_strides.pop_back();
-  }
-  return get_2d_grid_dims(out_shape, out_strides);
-}
-
-void col_reduce(
-    rocm::CommandEncoder& encoder,
-    const array& in,
-    array& out,
-    Reduce::ReduceType reduce_type,
-    const std::vector<int>& axes,
-    const ReductionPlan& plan) {
-  rocm::ColReduceArgs args(in, plan, axes);
-
-  encoder.launch_kernel([&](hipStream_t stream) {
-    MLX_SWITCH_ALL_TYPES(in.dtype(), CTYPE, {
-      using InType = hip_type_t<CTYPE>;
-      MLX_SWITCH_REDUCE_OPS(reduce_type, OP, {
-        using OutType = rocm::ReduceResult<OP, InType>::type;
-        MLX_SWITCH_REDUCE_NDIM(args.reduce_ndim, NDIM, {
-          constexpr int N_READS = 4;
-          dim3 block_dims;
-          dim3 num_blocks = output_grid_for_col_reduce(out, args);
-          num_blocks.z = num_blocks.y;
-          num_blocks.y = num_blocks.x;
-          auto kernel =
-              rocm::col_reduce_small<InType, OutType, OP, NDIM, N_READS>;
-          size_t total = args.non_col_reductions * args.reduction_size;
-          if (total < 32) {
-            size_t stride_blocks =
-                hip_ceil_div(args.reduction_stride, N_READS);
-            block_dims.x = std::min(stride_blocks, 32ul);
-            block_dims.y = std::min(total, 8ul);
-            num_blocks.x = hip_ceil_div(stride_blocks, block_dims.x);
-          } else {
-            constexpr int BM = 32;
-            constexpr int BN = 32;
-            block_dims.x = BM * BN / N_READS;
-            num_blocks.x = hip_ceil_div(args.reduction_stride, BN);
-            kernel = rocm::
-                col_reduce_looped<InType, OutType, OP, NDIM, BM, BN, N_READS>;
-          }
-          hipLaunchKernelGGL(kernel, num_blocks, block_dims, 0, stream,
-              in.data<InType>(), out.data<OutType>(), args);
-        });
-      });
-    });
-  });
-}
-
-} // namespace mlx::core 

mlx/backend/rocm/reduce/reduce.hpp

@@ -1,119 +0,0 @@
-// Copyright © 2025 Apple Inc.
-
-#pragma once
-
-#include <hip/hip_runtime.h>
-#include <cstddef>
-
-namespace mlx::core::rocm {
-
-// Reduction operation types
-template <typename Op, typename T>
-struct ReduceInit {
-  static constexpr T value();
-};
-
-template <typename T>
-struct ReduceInit<struct Sum, T> {
-  static constexpr T value() {
-    return T(0);
-  }
-};
-
-template <typename T>
-struct ReduceInit<struct Max, T> {
-  static constexpr T value() {
-    return -std::numeric_limits<T>::infinity();
-  }
-};
-
-template <typename T>
-struct ReduceInit<struct Min, T> {
-  static constexpr T value() {
-    return std::numeric_limits<T>::infinity();
-  }
-};
-
-// Reduction operations
-struct Sum {
-  template <typename T>
-  __device__ T operator()(T a, T b) const {
-    return a + b;
-  }
-};
-
-struct Max {
-  template <typename T>
-  __device__ T operator()(T a, T b) const {
-    return fmax(a, b);
-  }
-};
-
-struct Min {
-  template <typename T>
-  __device__ T operator()(T a, T b) const {
-    return fmin(a, b);
-  }
-};
-
-struct Prod {
-  template <typename T>
-  __device__ T operator()(T a, T b) const {
-    return a * b;
-  }
-};
-
-// Utility functions for reductions
-template <typename T>
-__device__ T warp_reduce(T val, T (*op)(T, T)) {
-  for (int offset = warpSize / 2; offset > 0; offset /= 2) {
-    val = op(val, __shfl_down(val, offset));
-  }
-  return val;
-}
-
-template <typename T>
-__device__ T block_reduce(T val, T (*op)(T, T)) {
-  static __shared__ T shared[32];
-  int lane = threadIdx.x % warpSize;
-  int wid = threadIdx.x / warpSize;
-
-  val = warp_reduce(val, op);
-
-  if (lane == 0)
-    shared[wid] = val;
-  __syncthreads();
-
-  val = (threadIdx.x < blockDim.x / warpSize) ? shared[lane] : 0;
-  if (wid == 0)
-    val = warp_reduce(val, op);
-
-  return val;
-}
-
-// Column reduction arguments
-struct ColReduceArgs {
-  size_t reduction_size;
-  int64_t reduction_stride;
-  int* shape;
-  size_t* strides;
-  int ndim;
-  int* reduce_shape;
-  size_t* reduce_strides;
-  int reduce_ndim;
-  size_t non_col_reductions;
-};
-
-// Row reduction arguments
-struct RowReduceArgs {
-  size_t reduction_size;
-  int64_t reduction_stride;
-  int* shape;
-  size_t* strides;
-  int ndim;
-  int* reduce_shape;
-  size_t* reduce_strides;
-  int reduce_ndim;
-};
-
-} // namespace mlx::core::rocm

mlx/backend/rocm/rms_norm.hip

@@ -1,375 +0,0 @@
-// Copyright © 2025 Apple Inc.
-
-#include "mlx/backend/rocm/device.h"
-#include "mlx/backend/rocm/iterators/strided_iterator.hpp"
-#include "mlx/backend/rocm/kernel_utils.hpp"
-#include "mlx/backend/rocm/reduce/reduce.hpp"
-#include "mlx/backend/gpu/copy.h"
-#include "mlx/dtype_utils.h"
-#include "mlx/fast_primitives.h"
-
-#include <hip/hip_runtime.h>
-#include <hip/hip_cooperative_groups.h>
-#include <rocprim/block/block_load.hpp>
-#include <rocprim/block/block_reduce.hpp>
-
-namespace mlx::core {
-
-namespace rocm {
-
-namespace cg = cooperative_groups;
-
-// Similar to rocprim::BlockReduce, but result is broadcasted to every thread.
-template <typename T, int BLOCK_DIM>
-struct BlockBroadcastReduce {
-  static_assert(WARP_SIZE <= BLOCK_DIM && BLOCK_DIM <= WARP_SIZE * WARP_SIZE);
-  static_assert(BLOCK_DIM % WARP_SIZE == 0);
-  using TempStorage = T[BLOCK_DIM / WARP_SIZE];
-
-  cg::thread_block& block;
-  TempStorage& temp;
-
-  template <typename Op>
-  __device__ T Reduce(const T& input, const Op& op, const T& init_value) {
-    auto warp = cg::tiled_partition<WARP_SIZE>(block);
-    T x = cg::reduce(warp, input, op);
-    if (warp.thread_rank() == 0) {
-      temp[warp.meta_group_rank()] = x;
-    }
-    block.sync();
-    x = warp.thread_rank() < warp.meta_group_size() ? temp[warp.thread_rank()]
-                                                    : init_value;
-    return cg::reduce(warp, x, op);
-  }
-
-  __device__ T Sum(const T& input) {
-    return Reduce(input, hip_plus<T>{}, T{});
-  }
-};
-
-template <typename T, int BLOCK_DIM, int N_READS = 4>
-__global__ void rms_norm(
-    const T* x,
-    const T* w,
-    T* out,
-    float eps,
-    int32_t axis_size,
-    int64_t w_stride) {
-  auto grid = cg::this_grid();
-  auto block = cg::this_thread_block();
-
-  using BlockReduceT = BlockBroadcastReduce<float, BLOCK_DIM>;
-  __shared__ typename BlockReduceT::TempStorage temp;
-
-  x += grid.block_rank() * axis_size;
-  out += grid.block_rank() * axis_size;
-
-  // Sum of squares.
-  float sum_sq = 0;
-  for (int r = 0; r < hip_ceil_div(axis_size, BLOCK_DIM * N_READS); ++r) {
-    auto index = r * BLOCK_DIM + block.thread_rank();
-    T xn[N_READS] = {};
-    rocprim::block_load_direct_blocked(index, x, xn, axis_size);
-    for (int i = 0; i < N_READS; ++i) {
-      float val = static_cast<float>(xn[i]);
-      sum_sq += val * val;
-    }
-  }
-  sum_sq = BlockReduceT{block, temp}.Sum(sum_sq);
-
-  // RMS normalizer.
-  float rms_normalizer = rsqrt(sum_sq / axis_size + eps);
-
-  // Outputs.
-  for (int r = 0; r < hip_ceil_div(axis_size, BLOCK_DIM * N_READS); ++r) {
-    auto index = r * BLOCK_DIM + block.thread_rank();
-    T xn[N_READS];
-    T wn[N_READS];
-    rocprim::block_load_direct_blocked(index, x, xn, axis_size);
-    rocprim::block_load_direct_blocked(index, strided_iterator(w, w_stride), wn, axis_size);
-    for (int i = 0; i < N_READS; ++i) {
-      float norm = static_cast<float>(xn[i]) * rms_normalizer;
-      xn[i] = wn[i] * static_cast<T>(norm);
-    }
-    rocprim::block_store_direct_blocked(index, out, xn, axis_size);
-  }
-}
-
-template <typename T, bool HAS_W, int BLOCK_DIM, int N_READS = 4>
-__global__ void rms_norm_vjp(
-    const T* x,
-    const T* w,
-    const T* g,
-    T* gx,
-    T* gw,
-    float eps,
-    int32_t axis_size,
-    int64_t w_stride) {
-  auto grid = cg::this_grid();
-  auto block = cg::this_thread_block();
-
-  using BlockReduceF = BlockBroadcastReduce<float, BLOCK_DIM>;
-  using BlockReduceF2 = BlockBroadcastReduce<float2, BLOCK_DIM>;
-  __shared__ union {
-    typename BlockReduceF::TempStorage f;
-    typename BlockReduceF2::TempStorage f2;
-  } temp;
-
-  x += grid.block_rank() * axis_size;
-  g += grid.block_rank() * axis_size;
-  gx += grid.block_rank() * axis_size;
-  gw += grid.block_rank() * axis_size;
-
-  // Sum of squares.
-  float sum_sq = 0;
-  for (int r = 0; r < hip_ceil_div(axis_size, BLOCK_DIM * N_READS); ++r) {
-    auto index = r * BLOCK_DIM + block.thread_rank();
-    T xn[N_READS] = {};
-    rocprim::block_load_direct_blocked(index, x, xn, axis_size);
-    for (int i = 0; i < N_READS; ++i) {
-      float val = static_cast<float>(xn[i]);
-      sum_sq += val * val;
-    }
-  }
-  sum_sq = BlockReduceF{block, temp.f}.Sum(sum_sq);
-
-  // RMS normalizer.
-  float rms_normalizer = rsqrt(sum_sq / axis_size + eps);
-
-  // Compute gradient terms.
-  float2 factors = {};
-  for (int r = 0; r < hip_ceil_div(axis_size, BLOCK_DIM * N_READS); ++r) {
-    T xn[N_READS];
-    T wn[N_READS] = {};
-    T gn[N_READS] = {};
-    auto index = r * BLOCK_DIM + block.thread_rank();
-    rocprim::block_load_direct_blocked(index, x, xn, axis_size);
-    rocprim::block_load_direct_blocked(index, g, gn, axis_size);
-    rocprim::block_load_direct_blocked(index, strided_iterator(w, w_stride), wn, axis_size);
-    for (int i = 0; i < N_READS; i++) {
-      float xi = static_cast<float>(xn[i]);
-      float wi = wn[i];
-      float gi = gn[i];
-      float wg = wi * gi;
-      factors.x += wg;
-      factors.y += wg * xi;
-    }
-  }
-  auto plus_f2 = [] __device__ (const float2& a, const float2& b) -> float2 {
-    return {a.x + b.x, a.y + b.y};
-  };
-  factors = BlockReduceF2{block, temp.f2}.Reduce(factors, plus_f2, {});
-  float mean_wg = factors.x / axis_size;
-  float mean_wgx = factors.y / axis_size;
-  float rms3 = rms_normalizer * rms_normalizer * rms_normalizer;
-
-  // Outputs.
-  for (int r = 0; r < hip_ceil_div(axis_size, BLOCK_DIM * N_READS); ++r) {
-    auto index = r * BLOCK_DIM + block.thread_rank();
-    T xn[N_READS];
-    T wn[N_READS];
-    T gn[N_READS];
-    rocprim::block_load_direct_blocked(index, x, xn, axis_size);
-    rocprim::block_load_direct_blocked(index, g, gn, axis_size);
-    rocprim::block_load_direct_blocked(index, strided_iterator(w, w_stride), wn, axis_size);
-    for (int i = 0; i < N_READS; i++) {
-      float xi = static_cast<float>(xn[i]);
-      float wi = wn[i];
-      float gi = gn[i];
-      float norm = xi * rms_normalizer;
-      xn[i] = rms_normalizer * (wi * gi - mean_wg) - norm * mean_wgx * rms3;
-      if constexpr (HAS_W) {
-        wn[i] = gi * norm;
-      }
-    }
-    rocprim::block_store_direct_blocked(index, gx, xn, axis_size);
-    if constexpr (HAS_W) {
-      rocprim::block_store_direct_blocked(index, gw, wn, axis_size);
-    }
-  }
-}
-
-// Utility functions
-template <typename T>
-struct hip_plus {
-  __device__ T operator()(const T& a, const T& b) const {
-    return a + b;
-  }
-};
-
-inline __device__ int hip_ceil_div(int a, int b) {
-  return (a + b - 1) / b;
-}
-
-template <typename T>
-__device__ inline auto strided_iterator(const T* ptr, int64_t stride) {
-  return ptr + stride;  // Simplified strided iterator
-}
-
-} // namespace rocm
-
-namespace fast {
-
-bool RMSNorm::use_fallback(Stream s) {
-  return s.device == Device::cpu;
-}
-
-void RMSNorm::eval_gpu(
-    const std::vector<array>& inputs,
-    std::vector<array>& outputs) {
-  auto& s = stream();
-  auto& out = outputs[0];
-
-  // Make sure that the last dimension is contiguous.
-  auto set_output = [&s, &out](const array& x) {
-    bool no_copy = x.flags().contiguous && x.strides()[x.ndim() - 1] == 1;
-    if (no_copy && x.ndim() > 1) {
-      auto s = x.strides()[x.ndim() - 2];
-      no_copy &= (s == 0 || s == x.shape().back());
-    }
-    if (no_copy) {
-      if (x.is_donatable()) {
-        out.copy_shared_buffer(x);
-      } else {
-        out.set_data(
-            allocator::malloc(x.data_size() * x.itemsize()),
-            x.data_size(),
-            x.strides(),
-            x.flags());
-      }
-      return x;
-    } else {
-      auto x_copy = array(x.shape(), x.dtype(), nullptr, {});
-      copy_gpu(x, x_copy, CopyType::General, s);
-      out.copy_shared_buffer(x_copy);
-      return x_copy;
-    }
-  };
-
-  const array x = set_output(inputs[0]);
-  const array& w = inputs[1];
-
-  int32_t axis_size = x.shape().back();
-  int32_t n_rows = x.data_size() / axis_size;
-  int64_t w_stride = (w.ndim() == 1) ? w.strides()[0] : 0;
-
-  auto& encoder = rocm::get_command_encoder(s);
-  encoder.set_input_array(x);
-  encoder.set_input_array(w);
-  encoder.set_output_array(out);
-  encoder.launch_kernel([&](hipStream_t stream) {
-    MLX_SWITCH_FLOAT_TYPES_CHECKED(out.dtype(), "rmsnorm", CTYPE, {
-      using DataType = hip_type_t<CTYPE>;
-      constexpr uint32_t N_READS = 4;
-      MLX_SWITCH_BLOCK_DIM(rocm::hip_ceil_div(axis_size, N_READS), BLOCK_DIM, {
-        auto kernel = rocm::rms_norm<DataType, BLOCK_DIM, N_READS>;
-        hipLaunchKernelGGL(kernel, n_rows, BLOCK_DIM, 0, stream,
-            x.data<DataType>(),
-            w.data<DataType>(),
-            out.data<DataType>(),
-            eps_,
-            axis_size,
-            w_stride);
-      });
-    });
-  });
-}
-
-void RMSNormVJP::eval_gpu(
-    const std::vector<array>& inputs,
-    std::vector<array>& outputs) {
-  auto& s = stream();
-  auto& encoder = rocm::get_command_encoder(s);
-
-  // Ensure row contiguity. We could relax this step by checking that the array
-  // is contiguous (no broadcasts or holes) and that the input strides are the
-  // same as the cotangent strides but for now this is simpler.
-  auto check_input = [&s](const array& x) -> std::pair<array, bool> {
-    if (x.flags().row_contiguous) {
-      return {x, false};
-    }
-    array x_copy(x.shape(), x.dtype(), nullptr, {});
-    copy_gpu(x, x_copy, CopyType::General, s);
-    return {x_copy, true};
-  };
-  bool donate_x = inputs[0].is_donatable();
-  bool donate_g = inputs[2].is_donatable();
-  auto [x, copied] = check_input(inputs[0]);
-  donate_x |= copied;
-  const array& w = inputs[1];
-  auto [g, g_copied] = check_input(inputs[2]);
-  donate_g |= g_copied;
-  array& gx = outputs[0];
-  array& gw = outputs[1];
-
-  // Check whether we had a weight.
-  bool has_w = w.ndim() != 0;
-
-  // Allocate space for the outputs.
-  bool g_in_gx = false;
-  if (donate_x) {
-    gx.copy_shared_buffer(x);
-  } else if (donate_g) {
-    gx.copy_shared_buffer(g);
-    g_in_gx = true;
-  } else {
-    gx.set_data(allocator::malloc(gx.nbytes()));
-  }
-  if (g_copied && !g_in_gx) {
-    encoder.add_temporary(g);
-  }
-
-  int32_t axis_size = x.shape().back();
-  int32_t n_rows = x.data_size() / axis_size;
-  int64_t w_stride = (w.ndim() == 1) ? w.strides()[0] : 0;
-
-  // Allocate a temporary to store the gradients for w and allocate the output
-  // gradient accumulators.
-  array gw_temp =
-      (has_w) ? array({n_rows, x.shape().back()}, gw.dtype(), nullptr, {}) : w;
-  if (has_w) {
-    if (!g_in_gx && donate_g) {
-      gw_temp.copy_shared_buffer(g);
-    } else {
-      gw_temp.set_data(allocator::malloc(gw_temp.nbytes()));
-      encoder.add_temporary(gw_temp);
-    }
-  }
-  gw.set_data(allocator::malloc(gw.nbytes()));
-
-  encoder.set_input_array(x);
-  encoder.set_input_array(w);
-  encoder.set_input_array(g);
-  encoder.set_output_array(gx);
-  encoder.set_output_array(gw_temp);
-  encoder.launch_kernel([&, x = x, g = g](hipStream_t stream) {
-    MLX_SWITCH_FLOAT_TYPES_CHECKED(gx.dtype(), "rmsnorm_vjp", CTYPE, {
-      using DataType = hip_type_t<CTYPE>;
-      constexpr int N_READS = 4;
-      MLX_SWITCH_BOOL(has_w, HAS_W, {
-        MLX_SWITCH_BLOCK_DIM(rocm::hip_ceil_div(axis_size, N_READS), BLOCK_DIM, {
-          auto kernel = rocm::rms_norm_vjp<DataType, HAS_W, BLOCK_DIM, N_READS>;
-          hipLaunchKernelGGL(kernel, n_rows, BLOCK_DIM, 0, stream,
-              x.data<DataType>(),
-              w.data<DataType>(),
-              g.data<DataType>(),
-              gx.data<DataType>(),
-              gw_temp.data<DataType>(),
-              eps_,
-              axis_size,
-              w_stride);
-        });
-      });
-    });
-  });
-
-  if (has_w) {
-    ReductionPlan plan(
-        ReductionOpType::ContiguousStridedReduce, {n_rows}, {axis_size});
-    col_reduce(encoder, gw_temp, gw, Reduce::ReduceType::Sum, {0}, plan);
-  }
-}
-
-} // namespace fast
-
-} // namespace mlx::core 

mlx/backend/rocm/rocm.cpp

@@ -1,11 +0,0 @@
-// Copyright © 2025 Apple Inc.
-
-#include "mlx/backend/rocm/rocm.h"
-
-namespace mlx::core::rocm {
-
-bool is_available() {
-  return true;
-}
-
-} // namespace mlx::core::rocm

mlx/backend/rocm/rocm.h

@@ -1,10 +0,0 @@
-// Copyright © 2025 Apple Inc.
-
-#pragma once
-
-namespace mlx::core::rocm {
-
-/* Check if the ROCm backend is available. */
-bool is_available();
-
-} // namespace mlx::core::rocm

mlx/backend/rocm/rope.hip

@@ -1,383 +0,0 @@
-// Copyright © 2025 Apple Inc.
-
-#include "mlx/backend/rocm/device.h"
-#include "mlx/backend/rocm/kernel_utils.hpp"
-#include "mlx/backend/gpu/copy.h"
-#include "mlx/dtype_utils.h"
-#include "mlx/fast_primitives.h"
-
-#include <hip/hip_runtime.h>
-
-namespace mlx::core {
-
-namespace rocm {
-
-template <typename T, bool traditional, bool forward>
-__device__ void rope_single_impl(
-    const T* in,
-    T* out,
-    int32_t offset,
-    float inv_freq,
-    float scale,
-    int64_t stride,
-    uint2 pos,
-    uint2 dims) {
-  float L = scale * static_cast<float>(offset);
-
-  // Compute costheta, sintheta
-  float theta = L * inv_freq;
-  float costheta = cos(theta);
-  float sintheta = sin(theta);
-
-  // Compute the input and output indices
-  uint index_1, index_2;
-  if (traditional) {
-    index_1 = 2 * pos.x + pos.y * stride;
-    index_2 = index_1 + 1;
-  } else {
-    index_1 = pos.x + pos.y * stride;
-    index_2 = index_1 + dims.x;
-  }
-
-  // Read and write the output
-  float x1 = static_cast<float>(in[index_1]);
-  float x2 = static_cast<float>(in[index_2]);
-  float rx1;
-  float rx2;
-  if (forward) {
-    rx1 = x1 * costheta - x2 * sintheta;
-    rx2 = x1 * sintheta + x2 * costheta;
-  } else {
-    rx1 = x2 * sintheta + x1 * costheta;
-    rx2 = x2 * costheta - x1 * sintheta;
-  }
-  out[index_1] = static_cast<T>(rx1);
-  out[index_2] = static_cast<T>(rx2);
-}
-
-template <typename T, bool traditional, bool forward>
-__global__ void rope_single(
-    const T* in,
-    T* out,
-    const int32_t* offset,
-    float scale,
-    float base,
-    int64_t stride,
-    uint2 dims) {
-  uint2 pos = make_uint2(
-      blockIdx.x * blockDim.x + threadIdx.x,
-      blockIdx.y * blockDim.y + threadIdx.y);
-  if (pos.x >= dims.x || pos.y >= dims.y) {
-    return;
-  }
-
-  float d = static_cast<float>(pos.x) / static_cast<float>(dims.x);
-  float inv_freq = exp2(-d * base);
-  rope_single_impl<T, traditional, forward>(
-      in, out, *offset, inv_freq, scale, stride, pos, dims);
-}
-
-template <typename T, bool traditional, bool forward>
-__global__ void rope_single_freqs(
-    const T* in,
-    T* out,
-    const int32_t* offset,
-    const float* freqs,
-    float scale,
-    int64_t stride,
-    uint2 dims,
-    int64_t freq_stride) {
-  uint2 pos = make_uint2(
-      blockIdx.x * blockDim.x + threadIdx.x,
-      blockIdx.y * blockDim.y + threadIdx.y);
-  if (pos.x >= dims.x || pos.y >= dims.y) {
-    return;
-  }
-
-  float inv_freq = 1.0 / freqs[freq_stride * pos.x];
-  rope_single_impl<T, traditional, forward>(
-      in, out, *offset, inv_freq, scale, stride, pos, dims);
-}
-
-template <typename T, bool traditional, bool forward, int N = 4>
-__device__ void rope_impl(
-    const T* in,
-    T* out,
-    int offset,
-    float inv_freq,
-    float scale,
-    const hip_array<int64_t, 3> strides,
-    const hip_array<int64_t, 3> out_strides,
-    int64_t n_batch,
-    uint3 pos,
-    uint3 dims) {
-  float L = scale * static_cast<float>(pos.y + offset);
-
-  // Compute costheta, sintheta
-  float theta = L * inv_freq;
-  float costheta = cos(theta);
-  float sintheta = sin(theta);
-
-  // Compute the input and output indices
-  size_t in_index_1, in_index_2;
-  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];
-    out_index_2 = out_index_1 + 1;
-    in_index_1 =
-        2 * pos.x * strides[2] + pos.y * strides[1] + N * pos.z * 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];
-    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_2 = in_index_1 + dims.x * strides[2];
-  }
-  for (int i = 0; i < N && pos.z * N + i < n_batch; ++i) {
-    // Read and write the output
-    float x1 = static_cast<float>(in[in_index_1]);
-    float x2 = static_cast<float>(in[in_index_2]);
-    float rx1;
-    float rx2;
-    if (forward) {
-      rx1 = x1 * costheta - x2 * sintheta;
-      rx2 = x1 * sintheta + x2 * costheta;
-    } else {
-      rx1 = x2 * sintheta + x1 * costheta;
-      rx2 = x2 * costheta - x1 * sintheta;
-    }
-    out[out_index_1] = static_cast<T>(rx1);
-    out[out_index_2] = static_cast<T>(rx2);
-    in_index_1 += strides[0];
-    in_index_2 += strides[0];
-    out_index_1 += out_strides[0];
-    out_index_2 += out_strides[0];
-  }
-}
-
-template <typename T, bool traditional, bool forward>
-__global__ void rope(
-    const T* in,
-    T* out,
-    const int32_t* offset,
-    float scale,
-    float base,
-    const hip_array<int64_t, 3> strides,
-    const hip_array<int64_t, 3> out_strides,
-    int64_t n_batch,
-    uint3 dims) {
-  uint3 pos = make_uint3(
-      blockIdx.x * blockDim.x + threadIdx.x,
-      blockIdx.y * blockDim.y + threadIdx.y,
-      blockIdx.z * blockDim.z + threadIdx.z);
-  if (pos.x >= dims.x || pos.y >= dims.y || pos.z >= dims.z) {
-    return;
-  }
-
-  float d = static_cast<float>(pos.x) / static_cast<float>(dims.x);
-  float inv_freq = exp2(-d * base);
-  rope_impl<T, traditional, forward>(
-      in,
-      out,
-      *offset,
-      inv_freq,
-      scale,
-      strides,
-      out_strides,
-      n_batch,
-      pos,
-      dims);
-}
-
-template <typename T, bool traditional, bool forward>
-__global__ void rope_freqs(
-    const T* in,
-    T* out,
-    const int32_t* offset,
-    const float* freqs,
-    float scale,
-    float base,
-    const hip_array<int64_t, 3> strides,
-    const hip_array<int64_t, 3> out_strides,
-    int64_t n_batch,
-    uint3 dims,
-    int64_t freq_stride) {
-  uint3 pos = make_uint3(
-      blockIdx.x * blockDim.x + threadIdx.x,
-      blockIdx.y * blockDim.y + threadIdx.y,
-      blockIdx.z * blockDim.z + threadIdx.z);
-  if (pos.x >= dims.x || pos.y >= dims.y || pos.z >= dims.z) {
-    return;
-  }
-
-  float inv_freq = 1.0 / freqs[freq_stride * pos.x];
-  rope_impl<T, traditional, forward>(
-      in,
-      out,
-      *offset,
-      inv_freq,
-      scale,
-      strides,
-      out_strides,
-      n_batch,
-      pos,
-      dims);
-}
-
-} // namespace rocm
-
-namespace fast {
-
-bool RoPE::use_fallback(Stream s) {
-  return s.device == Device::cpu;
-}
-
-void RoPE::eval_gpu(
-    const std::vector<array>& inputs,
-    std::vector<array>& outputs) {
-  auto& s = stream();
-  auto& in = inputs[0];
-  auto& offset = inputs[1];
-  auto& out = outputs[0];
-
-  if (in.ndim() < 3) {
-    throw std::runtime_error("[RoPE] Input must have at least 3 dimensions");
-  }
-
-  hip_array<int64_t, 3> strides;
-  hip_array<int64_t, 3> out_strides;
-  bool donated = false;
-  int ndim = in.ndim();
-  int dispatch_ndim = in.ndim();
-  while (in.shape(-dispatch_ndim) == 1 && dispatch_ndim > 3) {
-    dispatch_ndim--;
-  }
-  size_t mat_size = in.shape(-2) * in.shape(-1);
-
-  // We apply rope to less that the whole vector so copy to output and then
-  // apply in-place.
-  if (dims_ < in.shape(-1)) {
-    donated = true;
-    auto ctype =
-        (in.flags().row_contiguous) ? CopyType::Vector : CopyType::General;
-    copy_gpu(in, out, ctype, s);
-    strides[0] = mat_size;
-    strides[1] = out.strides()[ndim - 2];
-    strides[2] = out.strides()[ndim - 1];
-  }
-
-  // Either copy or apply in-place
-  else if (in.flags().row_contiguous) {
-    if (in.is_donatable()) {
-      donated = true;
-      out.copy_shared_buffer(in);
-    } else {
-      out.set_data(allocator::malloc(out.nbytes()));
-    }
-    strides[0] = mat_size;
-    strides[1] = in.strides()[ndim - 2];
-    strides[2] = in.strides()[ndim - 1];
-  } else if (dispatch_ndim == 3) {
-    // Handle non-contiguous 3D inputs
-    out.set_data(allocator::malloc(out.nbytes()));
-    strides[0] = in.strides()[ndim - 3];
-    strides[1] = in.strides()[ndim - 2];
-    strides[2] = in.strides()[ndim - 1];
-  } else {
-    // Copy non-contiguous > 3D inputs into the output and treat
-    // input as donated
-    donated = true;
-    copy_gpu(in, out, CopyType::General, s);
-    strides[0] = mat_size;
-    strides[1] = out.strides()[ndim - 2];
-    strides[2] = out.strides()[ndim - 1];
-  }
-  out_strides[0] = mat_size;
-  out_strides[1] = out.strides()[ndim - 2];
-  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 with_freqs = inputs.size() == 3;
-
-  auto& encoder = rocm::get_command_encoder(s);
-  encoder.set_input_array(donated ? out : in);
-  encoder.set_input_array(offset);
-  encoder.set_output_array(out);
-  encoder.launch_kernel([&](hipStream_t stream) {
-    MLX_SWITCH_FLOAT_TYPES_CHECKED(in.dtype(), "rope", CTYPE, {
-      using DataType = hip_type_t<CTYPE>;
-      MLX_SWITCH_BOOL(traditional_, TRADITIONAL, {
-        MLX_SWITCH_BOOL(forward_, FORWARD, {
-          if (single && !with_freqs) {
-            auto kernel = rocm::rope_single<DataType, TRADITIONAL, FORWARD>;
-            uint2 dims = make_uint2(dims_ / 2, in.size() / mat_size);
-            auto [grid, block] = get_grid_and_block(dims.x, dims.y, 1);
-            hipLaunchKernelGGL(kernel, grid, block, 0, stream,
-                (donated ? out : in).data<DataType>(),
-                out.data<DataType>(),
-                offset.data<int32_t>(),
-                scale_,
-                std::log2(base_),
-                mat_size,
-                dims);
-          } else if (single) {
-            auto kernel = rocm::rope_single_freqs<DataType, TRADITIONAL, FORWARD>;
-            uint2 dims = make_uint2(dims_ / 2, in.size() / mat_size);
-            auto [grid, block] = get_grid_and_block(dims.x, dims.y, 1);
-            hipLaunchKernelGGL(kernel, grid, block, 0, stream,
-                (donated ? out : in).data<DataType>(),
-                out.data<DataType>(),
-                offset.data<int32_t>(),
-                inputs[2].data<float>(),
-                scale_,
-                mat_size,
-                dims,
-                inputs[2].strides(0));
-          } else if (with_freqs) {
-            auto kernel = rocm::rope_freqs<DataType, TRADITIONAL, FORWARD>;
-            uint3 dims =
-                make_uint3(dims_ / 2, in.shape(-2), in.size() / mat_size);
-            dims.z = (dims.z + 3) / 4;
-            auto [grid, block] = get_grid_and_block(dims.x, dims.y, dims.z);
-            hipLaunchKernelGGL(kernel, grid, block, 0, stream,
-                (donated ? out : in).data<DataType>(),
-                out.data<DataType>(),
-                offset.data<int32_t>(),
-                inputs[2].data<float>(),
-                scale_,
-                std::log2(base_),
-                strides,
-                out_strides,
-                in.size() / mat_size,
-                dims,
-                inputs[2].strides(0));
-          } else {
-            auto kernel = rocm::rope<DataType, TRADITIONAL, FORWARD>;
-            uint3 dims =
-                make_uint3(dims_ / 2, in.shape(-2), in.size() / mat_size);
-            dims.z = (dims.z + 3) / 4;
-            auto [grid, block] = get_grid_and_block(dims.x, dims.y, dims.z);
-            hipLaunchKernelGGL(kernel, grid, block, 0, stream,
-                (donated ? out : in).data<DataType>(),
-                out.data<DataType>(),
-                offset.data<int32_t>(),
-                scale_,
-                std::log2(base_),
-                strides,
-                out_strides,
-                in.size() / mat_size,
-                dims);
-          }
-        });
-      });
-    });
-  });
-}
-
-} // namespace fast
-
-} // namespace mlx::core 

mlx/backend/rocm/slicing.cpp

@@ -1,9 +0,0 @@
-// Copyright © 2025 Apple Inc.
-
-namespace mlx::core::rocm {
-
-void slice() {
-  // Placeholder for ROCm slicing operation
-}
-
-} // namespace mlx::core::rocm

mlx/backend/rocm/softmax.hip

@@ -1,179 +0,0 @@
-// Copyright © 2025 Apple Inc.
-
-#include "mlx/backend/rocm/device.h"
-#include "mlx/backend/rocm/device/cast_op.hpp"
-#include "mlx/backend/rocm/device/fp16_math.hpp"
-#include "mlx/backend/rocm/kernel_utils.hpp"
-#include "mlx/backend/gpu/copy.h"
-#include "mlx/dtype_utils.h"
-#include "mlx/primitives.h"
-
-#include <hip/hip_runtime.h>
-#include <hip/hip_cooperative_groups.h>
-#include <rocprim/block/block_load.hpp>
-
-#include <cassert>
-
-namespace mlx::core {
-
-namespace rocm {
-
-namespace cg = cooperative_groups;
-
-template <typename T>
-inline __device__ T softmax_exp(T x) {
-  // Softmax doesn't need high precision exponential cause x is gonna be in
-  // (-oo, 0] anyway and subsequently it will be divided by sum(exp(x_i)).
-  return __expf(x);
-}
-
-template <typename T, typename AccT, int BLOCK_DIM, int N_READS = 4>
-__global__ void softmax(const T* in, T* out, int axis_size) {
-  auto grid = cg::this_grid();
-  auto block = cg::this_thread_block();
-  auto warp = cg::tiled_partition<WARP_SIZE>(block);
-
-  in += grid.block_rank() * axis_size;
-  out += grid.block_rank() * axis_size;
-
-  // Thread reduce.
-  AccT prevmax;
-  AccT maxval = -INFINITY;
-  AccT normalizer = 0;
-  for (int r = 0; r < hip_ceil_div(axis_size, BLOCK_DIM * N_READS); r++) {
-    AccT vals[N_READS];
-    rocprim::block_load_direct_blocked(
-        r * BLOCK_DIM + block.thread_rank(),
-        make_cast_iterator<AccT>(in),
-        vals,
-        axis_size,
-        -INFINITY);
-    prevmax = maxval;
-    maxval = fmax(maxval, rocprim::thread_reduce(vals, hip_max<AccT>()));
-    // Online normalizer calculation for softmax:
-    // https://github.com/NVIDIA/online-softmax
-    normalizer = normalizer * softmax_exp(prevmax - maxval);
-    for (int i = 0; i < N_READS; i++) {
-      normalizer = normalizer + softmax_exp(vals[i] - maxval);
-    }
-  }
-
-  // First warp reduce.
-  prevmax = maxval;
-  maxval = cg::reduce(warp, maxval, hip_max<AccT>());
-  normalizer = normalizer * softmax_exp(prevmax - maxval);
-  normalizer = cg::reduce(warp, normalizer, hip_plus<AccT>());
-
-  __shared__ AccT local_max[WARP_SIZE];
-  __shared__ AccT local_normalizer[WARP_SIZE];
-
-  // Write to shared memory and do second warp reduce.
-  prevmax = maxval;
-  if (warp.thread_rank() == 0) {
-    local_max[warp.meta_group_rank()] = maxval;
-  }
-  block.sync();
-  maxval = warp.thread_rank() < warp.meta_group_size()
-      ? local_max[warp.thread_rank()]
-      : -INFINITY;
-  maxval = cg::reduce(warp, maxval, hip_max<AccT>());
-  normalizer = normalizer * softmax_exp(prevmax - maxval);
-  if (warp.thread_rank() == 0) {
-    local_normalizer[warp.meta_group_rank()] = normalizer;
-  }
-  block.sync();
-  normalizer = warp.thread_rank() < warp.meta_group_size()
-      ? local_normalizer[warp.thread_rank()]
-      : AccT{};
-  normalizer = cg::reduce(warp, normalizer, hip_plus<AccT>());
-  normalizer = 1 / normalizer;
-
-  // Write output.
-  for (int r = 0; r < hip_ceil_div(axis_size, BLOCK_DIM * N_READS); r++) {
-    auto index = r * BLOCK_DIM + block.thread_rank();
-    T vals[N_READS];
-    rocprim::block_load_direct_blocked(index, in, vals, axis_size);
-    for (int i = 0; i < N_READS; i++) {
-      vals[i] = softmax_exp(static_cast<AccT>(vals[i]) - maxval) * normalizer;
-    }
-    rocprim::block_store_direct_blocked(index, out, vals, axis_size);
-  }
-}
-
-// Utility functions for ROCm
-template <typename T>
-struct hip_max {
-  __device__ T operator()(const T& a, const T& b) const {
-    return fmax(a, b);
-  }
-};
-
-template <typename T>
-struct hip_plus {
-  __device__ T operator()(const T& a, const T& b) const {
-    return a + b;
-  }
-};
-
-inline __device__ int hip_ceil_div(int a, int b) {
-  return (a + b - 1) / b;
-}
-
-template <typename T>
-__device__ inline T* make_cast_iterator(const T* ptr) {
-  return const_cast<T*>(ptr);
-}
-
-} // namespace rocm
-
-void Softmax::eval_gpu(const std::vector<array>& inputs, array& out) {
-  assert(inputs.size() == 1);
-  auto& s = stream();
-
-  // Make sure that the last dimension is contiguous.
-  auto set_output = [&s, &out](const array& x) {
-    if (x.flags().contiguous && x.strides()[x.ndim() - 1] == 1) {
-      if (x.is_donatable()) {
-        out.copy_shared_buffer(x);
-      } else {
-        out.set_data(
-            allocator::malloc(x.data_size() * x.itemsize()),
-            x.data_size(),
-            x.strides(),
-            x.flags());
-      }
-      return x;
-    } else {
-      auto x_copy = array(x.shape(), x.dtype(), nullptr, {});
-      copy_gpu(x, x_copy, CopyType::General, s);
-      out.copy_shared_buffer(x_copy);
-      return x_copy;
-    }
-  };
-
-  array in = set_output(inputs[0]);
-  bool precise = in.dtype() != float32 && precise_;
-
-  int axis_size = in.shape().back();
-  int n_rows = in.data_size() / axis_size;
-
-  auto& encoder = rocm::get_command_encoder(s);
-  encoder.set_input_array(in);
-  encoder.set_output_array(out);
-  encoder.launch_kernel([&](hipStream_t stream) {
-    MLX_SWITCH_FLOAT_TYPES_CHECKED(out.dtype(), "softmax", CTYPE, {
-      using DataType = hip_type_t<CTYPE>;
-      constexpr int N_READS = 4;
-      MLX_SWITCH_BLOCK_DIM(rocm::hip_ceil_div(axis_size, N_READS), BLOCK_DIM, {
-        auto kernel = rocm::softmax<DataType, DataType, BLOCK_DIM, N_READS>;
-        if (precise) {
-          kernel = rocm::softmax<DataType, float, BLOCK_DIM, N_READS>;
-        }
-        hipLaunchKernelGGL(kernel, n_rows, BLOCK_DIM, 0, stream,
-            in.data<DataType>(), out.data<DataType>(), axis_size);
-      });
-    });
-  });
-}
-
-} // namespace mlx::core