Make MLX build on x64 macOS (#901)

The arm64 macbook pros are heavy and I usually care my intel one for
mobile, it would be nice if I can play with MLX on it.

To build with x64, user must pass `MLX_ENABLE_X64_MAC` to cmake:
CMAKE_ARGS='-DMLX_ENABLE_X64_MAC=ON' python setup.py

CMakeLists.txt

@@ -15,6 +15,7 @@ option(MLX_BUILD_EXAMPLES "Build examples for mlx" ON)
 option(MLX_BUILD_BENCHMARKS "Build benchmarks for mlx" OFF)
 option(MLX_BUILD_PYTHON_BINDINGS "Build python bindings for mlx" OFF)
 option(MLX_BUILD_METAL "Build metal backend" ON)
+option(MLX_ENABLE_X64_MAC "Enable building for x64 macOS" OFF)
 option(BUILD_SHARED_LIBS "Build mlx as a shared library" OFF)
 
 if(NOT MLX_VERSION)
@@ -23,23 +24,23 @@ endif()
 
 # --------------------- Processor tests -------------------------
 
-message(STATUS "Building MLX for ${CMAKE_HOST_SYSTEM_PROCESSOR} processor on ${CMAKE_SYSTEM_NAME}")
+message(STATUS "Building MLX for ${CMAKE_SYSTEM_PROCESSOR} processor on ${CMAKE_SYSTEM_NAME}")
 
 set(MLX_BUILD_ARM OFF)
 
 if (${CMAKE_SYSTEM_NAME} MATCHES "Darwin")
-  if (${CMAKE_HOST_SYSTEM_PROCESSOR} MATCHES "x86_64" AND ${CMAKE_HOST_APPLE})
+  if(${CMAKE_SYSTEM_PROCESSOR} MATCHES "x86_64")
-    message(FATAL_ERROR
+    if(NOT MLX_ENABLE_X64_MAC)
-      "Building for x86_64 on macOS is not supported."
+      message(FATAL_ERROR
-      " If you are on an Apple silicon system, check the build"
+        "Building for x86_64 on macOS is not supported."
-      " documentation for possible fixes: "
+        " If you are on an Apple silicon system, check the build"
-      "https://ml-explore.github.io/mlx/build/html/install.html#build-from-source")
+        " documentation for possible fixes: "
-  elseif (${CMAKE_HOST_SYSTEM_PROCESSOR} MATCHES "x86_64")
+        "https://ml-explore.github.io/mlx/build/html/install.html#build-from-source")
-    message(WARNING
+    else()
-      "Building for x86_64 on macOS is not supported."
+      message(WARNING "Building for x86_64 arch is not officially supported.")
-      " If you are on an Apple silicon system, "
+    endif()
-      " make sure you are building for arm64.")
+    set(MLX_BUILD_METAL OFF)
-  elseif(${CMAKE_HOST_SYSTEM_PROCESSOR} MATCHES "arm64")
+  elseif(${CMAKE_SYSTEM_PROCESSOR} MATCHES "arm64")
     set(MLX_BUILD_ARM ON)
   endif()
 
@@ -108,7 +109,27 @@ if (MLX_BUILD_ARM AND ACCELERATE_LIBRARY)
 else()
   message(STATUS "Accelerate or arm neon not found, using default backend.")
   set(MLX_BUILD_ACCELERATE OFF)
-  #set(BLA_VENDOR Generic)
+  if(${CMAKE_HOST_APPLE})
+    # The blas shipped in macOS SDK is not supported, search homebrew for
+    # openblas instead.
+    set(BLA_VENDOR OpenBLAS)
+    set(LAPACK_ROOT "${LAPACK_ROOT};$ENV{LAPACK_ROOT};/usr/local/opt/openblas")
+  endif()
+  # Search and link with lapack.
+  find_package(LAPACK REQUIRED)
+  if (NOT LAPACK_FOUND)
+    message(FATAL_ERROR "Must have LAPACK installed")
+  endif()
+  find_path(LAPACK_INCLUDE_DIRS lapacke.h
+    /usr/include
+    /usr/local/include
+    /usr/local/opt/openblas/include)
+  message(STATUS "Lapack lib " ${LAPACK_LIBRARIES})
+  message(STATUS "Lapack include " ${LAPACK_INCLUDE_DIRS})
+  target_include_directories(mlx PRIVATE ${LAPACK_INCLUDE_DIRS})
+  target_link_libraries(mlx ${LAPACK_LIBRARIES})
+  # List blas after lapack otherwise we may accidentally incldue an old version
+  # of lapack.h from the include dirs of blas.
   find_package(BLAS REQUIRED)
   if (NOT BLAS_FOUND)
     message(FATAL_ERROR "Must have BLAS installed")
@@ -122,17 +143,6 @@ else()
   message(STATUS "Blas include " ${BLAS_INCLUDE_DIRS})
   target_include_directories(mlx PRIVATE ${BLAS_INCLUDE_DIRS})
   target_link_libraries(mlx ${BLAS_LIBRARIES})
-  find_package(LAPACK REQUIRED)
-  if (NOT LAPACK_FOUND)
-      message(FATAL_ERROR "Must have LAPACK installed")
-  endif()
-  find_path(LAPACK_INCLUDE_DIRS lapacke.h
-    /usr/include
-    /usr/local/include)
-  message(STATUS "Lapack lib " ${LAPACK_LIBRARIES})
-  message(STATUS "Lapack include " ${LAPACK_INCLUDE_DIRS})
-  target_include_directories(mlx PRIVATE ${LAPACK_INCLUDE_DIRS})
-  target_link_libraries(mlx ${LAPACK_LIBRARIES})
 endif()
 
 add_subdirectory(${CMAKE_CURRENT_LIST_DIR}/mlx)

python/tests/mlx_tests.py

@@ -11,8 +11,8 @@ import numpy as np
 
 class MLXTestCase(unittest.TestCase):
     @property
-    def is_linux(self):
+    def is_apple_silicon(self):
-        return platform.system() == "Linux"
+        return platform.machine() == "arm64" and platform.system() == "Darwin"
 
     def setUp(self):
         self.default = mx.default_device()

python/tests/test_fast_sdpa.py

@@ -62,7 +62,7 @@ class TestFastSDPA(mlx_tests.MLXTestCase):
         B = 1
         H = 32
         dtypes = [np.float32]
-        if not self.is_linux:
+        if self.is_apple_silicon:
             dtypes.append(np.half)
 
         for SEQUENCE_LENGTH in [1, 7, 9, 32, 63, 67, 129, 400, 2000]:

python/tests/test_ops.py

@@ -1772,8 +1772,8 @@ class TestOps(mlx_tests.MLXTestCase):
         )
 
     def test_tensordot(self):
-        # No fp16 matmuls on linux
+        # No fp16 matmuls on common cpu backend
-        if self.is_linux:
+        if not self.is_apple_silicon:
             dtypes = [mx.float32]
         else:
             dtypes = [mx.float16, mx.float32]