Build for Linux AArch64 and other miscellaneous cleanups

.github/workflows/ci.yaml

@@ -13,8 +13,8 @@ jobs:
       fail-fast: false
       matrix:
         python-version: ["3.11", "3.12", "3.13"]
-        # macos-13 is an intel runner, macos-14 is a arm64 runner
-        platform: [ubuntu-latest, windows-latest, macos-13, macos-14]
+        # macos-13 is an intel runner, macos-14 is an arm64 runner
+        platform: [ubuntu-latest, ubuntu-22.04-arm, windows-latest, macos-13, macos-14]
 
     defaults:
       run:
@@ -28,7 +28,7 @@ jobs:
           fetch-depth: 0 # required for version resolution
 
       - name: Set up Conda
-        uses: conda-incubator/[email protected].0
+        uses: conda-incubator/[email protected].1
         with:
           channels: conda-forge
           miniforge-version: latest
@@ -56,8 +56,7 @@ jobs:
         # Since zarr v3 requires numpy >= 1.25, on Python 3.11 leave it out
         # so we can have some tests of our minimum version of numpy (1.24)
         if: matrix.python-version != '3.11'
-        # TODO: remove --pre option when zarr v3 is out
-        run: python -m pip install --pre zarr>=3.0.0b2
+        run: python -m pip install zarr>=3
 
       - name: List installed packages
         run: python -m pip list

.github/workflows/wheel.yaml

@@ -13,8 +13,8 @@ jobs:
     strategy:
       fail-fast: false
       matrix:
-        # macos-13 is an intel runner, macos-14 is a arm64 runner
-        os: [ubuntu-latest, windows-latest, macos-13, macos-14]
+        # macos-13 is an intel runner, macos-14 is an arm64 runner
+        os: [ubuntu-latest, ubuntu-22.04-arm, windows-latest, macos-13, macos-14]
     env:
       CIBW_TEST_COMMAND: python -c "import numcodecs"
       CIBW_BUILD: "cp311-* cp312-* cp313-*"

docs/release.rst

@@ -17,6 +17,12 @@ Release notes
 Unreleased
 ----------
 
+Enhancements
+~~~~~~~~~~~~
+
+* Add support for the Linux AArch64 architecture.
+  By :user:`Agriya Khetarpal <agriyakhetarpal>`, :issue:`288`.
+
 Improvements
 ~~~~~~~~~~~~
 * Raise a custom `UnknownCodecError` when trying to retrieve an unavailable codec.
@@ -40,7 +46,7 @@ Breaking changes
 
 Deprecations
 ~~~~~~~~~~~~
-The following ``blosc`` funcitons are deprecated, with no replacement.
+The following ``blosc`` functions are deprecated, with no replacement.
 This is because they are not intended to be public API.
 
 - ``numcodecs.blosc.init``

pyproject.toml

@@ -157,7 +157,9 @@ filterwarnings = [
 [tool.cibuildwheel]
 environment = { DISABLE_NUMCODECS_AVX2=1 }
 [tool.cibuildwheel.macos]
-environment = { MACOSX_DEPLOYMENT_TARGET=10.9, DISABLE_NUMCODECS_AVX2=1, CFLAGS="$CFLAGS -Wno-implicit-function-declaration" }
+# cibuildwheel uses 3.12 for the Python driver, which supports High Sierra and later
+# https://github.com/pypa/cibuildwheel/blob/ee63bf16da6cddfb925f542f2c7b59ad50e93969/action.yml#L31
+environment = { MACOSX_DEPLOYMENT_TARGET=10.13, DISABLE_NUMCODECS_AVX2=1, CFLAGS="$CFLAGS -Wno-implicit-function-declaration" }
 [[tool.cibuildwheel.overrides]]
 select = "*-macosx_arm64"
 environment = { DISABLE_NUMCODECS_AVX2=1, DISABLE_NUMCODECS_SSE2=1 }

setup.py

@@ -13,8 +13,15 @@
 # determine CPU support for SSE2 and AVX2
 cpu_info = cpuinfo.get_cpu_info()
 flags = cpu_info.get('flags', [])
-have_sse2 = 'sse2' in flags
-have_avx2 = 'avx2' in flags
+machine = cpuinfo.platform.machine()
+
+# only check for x86 features on x86_64 arch
+have_sse2 = False
+have_avx2 = False
+if machine == 'x86_64':
+    have_sse2 = 'sse2' in flags
+    have_avx2 = 'avx2' in flags
+
 disable_sse2 = 'DISABLE_NUMCODECS_SSE2' in os.environ
 disable_avx2 = 'DISABLE_NUMCODECS_AVX2' in os.environ
 
@@ -24,7 +31,7 @@
 if have_cflags:
     # respect compiler options set by user
     pass
-elif os.name == 'posix':
+elif os.name == 'posix' and machine == 'x86_64':
     if disable_sse2:
         base_compile_args.append('-mno-sse2')
     elif have_sse2:
@@ -53,8 +60,14 @@ def blosc_extension():
     info('setting up Blosc extension')
 
     extra_compile_args = base_compile_args.copy()
+    extra_link_args = []
     define_macros = []
 
+    # ensure pthread is properly linked on POSIX systems
+    if os.name == 'posix':
+        extra_compile_args.append('-pthread')
+        extra_link_args.append('-pthread')
+
     # setup blosc sources
     blosc_sources = [f for f in glob('c-blosc/blosc/*.c') if 'avx2' not in f and 'sse2' not in f]
     include_dirs = [os.path.join('c-blosc', 'blosc')]
@@ -118,6 +131,7 @@ def blosc_extension():
             include_dirs=include_dirs,
             define_macros=define_macros,
             extra_compile_args=extra_compile_args,
+            extra_link_args=extra_link_args,
             extra_objects=extra_objects,
         ),
     ]
@@ -302,13 +316,24 @@ class BuildFailed(Exception):
     pass
 
 
+def get_arch_specific_objects():
+    machine = cpuinfo.platform.machine()
+    if machine == 'x86_64':
+        return [S[:-1] + 'o' for S in glob("c-blosc/internal-complibs/zstd*/decompress/*amd64.S")]
+    elif machine == 'aarch64':
+        return [S[:-1] + 'o' for S in glob("c-blosc/internal-complibs/zstd*/decompress/*aarch64.S")]
+    return []
+
+
 class ve_build_ext(build_ext):
     # This class allows C extension building to fail.
 
     def run(self):
         try:
-            if cpuinfo.platform.machine() == 'x86_64':
-                S_files = glob('c-blosc/internal-complibs/zstd*/decompress/*amd64.S')
+            machine = cpuinfo.platform.machine()
+            if machine in ('x86_64', 'aarch64'):
+                pattern = '*amd64.S' if machine == 'x86_64' else '*aarch64.S'
+                S_files = glob(f'c-blosc/internal-complibs/zstd*/decompress/{pattern}')
                 compiler = ccompiler.new_compiler()
                 customize_compiler(compiler)
                 compiler.src_extensions.append('.S')