numpy 2.4.1__pp311-pypy311_pp73-macosx_14_0_arm64.whl

numpy/distutils/checks/cpu_sve.c

@@ -0,0 +1,14 @@
+#include <arm_sve.h>
+
+int accumulate(svint64_t a, svint64_t b) {
+    svbool_t p = svptrue_b64();
+    return svaddv(p, svmla_z(p, a, a, b));
+}
+
+int main(void)
+{
+    svbool_t p = svptrue_b64();
+    svint64_t a = svdup_s64(1);
+    svint64_t b = svdup_s64(2);
+    return accumulate(a, b);
+}

numpy/distutils/checks/cpu_vsx.c

@@ -0,0 +1,21 @@
+#ifndef __VSX__
+    #error "VSX is not supported"
+#endif
+#include <altivec.h>
+
+#if (defined(__GNUC__) && !defined(vec_xl)) || (defined(__clang__) && !defined(__IBMC__))
+    #define vsx_ld  vec_vsx_ld
+    #define vsx_st  vec_vsx_st
+#else
+    #define vsx_ld  vec_xl
+    #define vsx_st  vec_xst
+#endif
+
+int main(void)
+{
+    unsigned int zout[4];
+    unsigned int z4[] = {0, 0, 0, 0};
+    __vector unsigned int v_z4 = vsx_ld(0, z4);
+    vsx_st(v_z4, 0, zout);
+    return zout[0];
+}

numpy/distutils/checks/cpu_vsx2.c

@@ -0,0 +1,13 @@
+#ifndef __VSX__
+    #error "VSX is not supported"
+#endif
+#include <altivec.h>
+
+typedef __vector unsigned long long v_uint64x2;
+
+int main(void)
+{
+    v_uint64x2 z2 = (v_uint64x2){0, 0};
+    z2 = (v_uint64x2)vec_cmpeq(z2, z2);
+    return (int)vec_extract(z2, 0);
+}

numpy/distutils/checks/cpu_vsx3.c

@@ -0,0 +1,13 @@
+#ifndef __VSX__
+    #error "VSX is not supported"
+#endif
+#include <altivec.h>
+
+typedef __vector unsigned int v_uint32x4;
+
+int main(void)
+{
+    v_uint32x4 z4 = (v_uint32x4){0, 0, 0, 0};
+    z4 = vec_absd(z4, z4);
+    return (int)vec_extract(z4, 0);
+}

numpy/distutils/checks/cpu_vsx4.c

@@ -0,0 +1,14 @@
+#ifndef __VSX__
+    #error "VSX is not supported"
+#endif
+#include <altivec.h>
+
+typedef __vector unsigned int v_uint32x4;
+
+int main(void)
+{
+    v_uint32x4 v1 = (v_uint32x4){2, 4, 8, 16};
+    v_uint32x4 v2 = (v_uint32x4){2, 2, 2, 2};
+    v_uint32x4 v3 = vec_mod(v1, v2);
+    return (int)vec_extractm(v3);
+}

numpy/distutils/checks/cpu_vx.c

@@ -0,0 +1,16 @@
+#if (__VEC__ < 10301) || (__ARCH__ < 11)
+    #error VX not supported
+#endif
+
+#include <vecintrin.h>
+int main(int argc, char **argv)
+{
+    __vector double x = vec_abs(vec_xl(argc, (double*)argv));
+    __vector double y = vec_load_len((double*)argv, (unsigned int)argc);
+
+    x = vec_round(vec_ceil(x) + vec_floor(y));
+    __vector bool long long m = vec_cmpge(x, y);
+    __vector long long i = vec_signed(vec_sel(x, y, m));
+
+    return (int)vec_extract(i, 0);
+}

numpy/distutils/checks/cpu_vxe.c

@@ -0,0 +1,25 @@
+#if (__VEC__ < 10302) || (__ARCH__ < 12)
+    #error VXE not supported
+#endif
+
+#include <vecintrin.h>
+int main(int argc, char **argv)
+{
+    __vector float x = vec_nabs(vec_xl(argc, (float*)argv));
+    __vector float y = vec_load_len((float*)argv, (unsigned int)argc);
+    
+    x = vec_round(vec_ceil(x) + vec_floor(y));
+    __vector bool int m = vec_cmpge(x, y);
+    x = vec_sel(x, y, m);
+
+    // need to test the existence of intrin "vflls" since vec_doublee
+    // is vec_doublee maps to wrong intrin "vfll".
+    // see https://gcc.gnu.org/bugzilla/show_bug.cgi?id=100871
+#if defined(__GNUC__) && !defined(__clang__)
+    __vector long long i = vec_signed(__builtin_s390_vflls(x));
+#else
+    __vector long long i = vec_signed(vec_doublee(x));
+#endif
+
+    return (int)vec_extract(i, 0);
+}

numpy/distutils/checks/cpu_vxe2.c

@@ -0,0 +1,21 @@
+#if (__VEC__ < 10303) || (__ARCH__ < 13)
+    #error VXE2 not supported
+#endif
+
+#include <vecintrin.h>
+
+int main(int argc, char **argv)
+{
+    int val;
+    __vector signed short large = { 'a', 'b', 'c', 'a', 'g', 'h', 'g', 'o' };
+    __vector signed short search = { 'g', 'h', 'g', 'o' };
+    __vector unsigned char len = { 0 };
+    __vector unsigned char res = vec_search_string_cc(large, search, len, &val);
+    __vector float x = vec_xl(argc, (float*)argv);
+    __vector int i = vec_signed(x);
+
+    i = vec_srdb(vec_sldb(i, i, 2), i, 3);
+    val += (int)vec_extract(res, 1);
+    val += vec_extract(i, 0);
+    return val;
+}

numpy/distutils/checks/cpu_xop.c

@@ -0,0 +1,12 @@
+#include <immintrin.h>
+#ifdef _MSC_VER
+    #include <ammintrin.h>
+#else
+    #include <x86intrin.h>
+#endif
+
+int main(void)
+{
+    __m128i a = _mm_comge_epu32(_mm_setzero_si128(), _mm_setzero_si128());
+    return _mm_cvtsi128_si32(a);
+}

numpy/distutils/checks/extra_avx512bw_mask.c

@@ -0,0 +1,18 @@
+#include <immintrin.h>
+/**
+ * Test BW mask operations due to:
+ *  - MSVC has supported it since vs2019 see,
+ *    https://developercommunity.visualstudio.com/content/problem/518298/missing-avx512bw-mask-intrinsics.html
+ *  - Clang >= v8.0
+ *  - GCC >= v7.1
+ */
+int main(void)
+{
+    __mmask64 m64 = _mm512_cmpeq_epi8_mask(_mm512_set1_epi8((char)1), _mm512_set1_epi8((char)1));
+    m64 = _kor_mask64(m64, m64);
+    m64 = _kxor_mask64(m64, m64);
+    m64 = _cvtu64_mask64(_cvtmask64_u64(m64));
+    m64 = _mm512_kunpackd(m64, m64);
+    m64 = (__mmask64)_mm512_kunpackw((__mmask32)m64, (__mmask32)m64);
+    return (int)_cvtmask64_u64(m64);
+}

numpy/distutils/checks/extra_avx512dq_mask.c

@@ -0,0 +1,16 @@
+#include <immintrin.h>
+/**
+ * Test DQ mask operations due to:
+ *  - MSVC has supported it since vs2019 see,
+ *    https://developercommunity.visualstudio.com/content/problem/518298/missing-avx512bw-mask-intrinsics.html
+ *  - Clang >= v8.0
+ *  - GCC >= v7.1
+ */
+int main(void)
+{
+    __mmask8 m8 = _mm512_cmpeq_epi64_mask(_mm512_set1_epi64(1), _mm512_set1_epi64(1));
+    m8 = _kor_mask8(m8, m8);
+    m8 = _kxor_mask8(m8, m8);
+    m8 = _cvtu32_mask8(_cvtmask8_u32(m8));
+    return (int)_cvtmask8_u32(m8);
+}

numpy/distutils/checks/extra_avx512f_reduce.c

@@ -0,0 +1,41 @@
+#include <immintrin.h>
+/**
+ * The following intrinsics don't have direct native support but compilers
+ * tend to emulate them.
+ * They're usually supported by gcc >= 7.1, clang >= 4 and icc >= 19
+ */
+int main(void)
+{
+    __m512  one_ps = _mm512_set1_ps(1.0f);
+    __m512d one_pd = _mm512_set1_pd(1.0);
+    __m512i one_i64 = _mm512_set1_epi64(1);
+    // add
+    float sum_ps  = _mm512_reduce_add_ps(one_ps);
+    double sum_pd = _mm512_reduce_add_pd(one_pd);
+    int sum_int   = (int)_mm512_reduce_add_epi64(one_i64);
+        sum_int  += (int)_mm512_reduce_add_epi32(one_i64);
+    // mul
+    sum_ps  += _mm512_reduce_mul_ps(one_ps);
+    sum_pd  += _mm512_reduce_mul_pd(one_pd);
+    sum_int += (int)_mm512_reduce_mul_epi64(one_i64);
+    sum_int += (int)_mm512_reduce_mul_epi32(one_i64);
+    // min
+    sum_ps  += _mm512_reduce_min_ps(one_ps);
+    sum_pd  += _mm512_reduce_min_pd(one_pd);
+    sum_int += (int)_mm512_reduce_min_epi32(one_i64);
+    sum_int += (int)_mm512_reduce_min_epu32(one_i64);
+    sum_int += (int)_mm512_reduce_min_epi64(one_i64);
+    // max
+    sum_ps  += _mm512_reduce_max_ps(one_ps);
+    sum_pd  += _mm512_reduce_max_pd(one_pd);
+    sum_int += (int)_mm512_reduce_max_epi32(one_i64);
+    sum_int += (int)_mm512_reduce_max_epu32(one_i64);
+    sum_int += (int)_mm512_reduce_max_epi64(one_i64);
+    // and
+    sum_int += (int)_mm512_reduce_and_epi32(one_i64);
+    sum_int += (int)_mm512_reduce_and_epi64(one_i64);
+    // or
+    sum_int += (int)_mm512_reduce_or_epi32(one_i64);
+    sum_int += (int)_mm512_reduce_or_epi64(one_i64);
+    return (int)sum_ps + (int)sum_pd + sum_int;
+}

numpy/distutils/checks/extra_vsx3_half_double.c

@@ -0,0 +1,12 @@
+/**
+ * Assembler may not fully support the following VSX3 scalar
+ * instructions, even though compilers report VSX3 support.
+ */
+int main(void)
+{
+    unsigned short bits = 0xFF;
+    double f;
+    __asm__ __volatile__("xscvhpdp %x0,%x1" : "=wa"(f) : "wa"(bits));
+    __asm__ __volatile__ ("xscvdphp %x0,%x1" : "=wa" (bits) : "wa" (f));
+    return bits;
+}

numpy/distutils/checks/extra_vsx4_mma.c

@@ -0,0 +1,21 @@
+#ifndef __VSX__
+    #error "VSX is not supported"
+#endif
+#include <altivec.h>
+
+typedef __vector float fv4sf_t;
+typedef __vector unsigned char vec_t;
+
+int main(void)
+{
+    __vector_quad acc0;
+    float a[4] = {0,1,2,3};
+    float b[4] = {0,1,2,3};
+    vec_t *va = (vec_t *) a;
+    vec_t *vb = (vec_t *) b;
+    __builtin_mma_xvf32ger(&acc0, va[0], vb[0]);
+    fv4sf_t result[4];
+    __builtin_mma_disassemble_acc((void *)result, &acc0);
+    fv4sf_t c0 = result[0];
+    return (int)((float*)&c0)[0];
+}

numpy/distutils/checks/extra_vsx_asm.c

@@ -0,0 +1,36 @@
+/**
+ * Testing ASM VSX register number fixer '%x<n>'
+ *
+ * old versions of CLANG doesn't support %x<n> in the inline asm template
+ * which fixes register number when using any of the register constraints wa, wd, wf.
+ *
+ * xref:
+ * - https://bugs.llvm.org/show_bug.cgi?id=31837
+ * - https://gcc.gnu.org/onlinedocs/gcc/Machine-Constraints.html
+ */
+#ifndef __VSX__
+    #error "VSX is not supported"
+#endif
+#include <altivec.h>
+
+#if (defined(__GNUC__) && !defined(vec_xl)) || (defined(__clang__) && !defined(__IBMC__))
+    #define vsx_ld  vec_vsx_ld
+    #define vsx_st  vec_vsx_st
+#else
+    #define vsx_ld  vec_xl
+    #define vsx_st  vec_xst
+#endif
+
+int main(void)
+{
+    float z4[] = {0, 0, 0, 0};
+    signed int zout[] = {0, 0, 0, 0};
+
+    __vector float vz4 = vsx_ld(0, z4);
+    __vector signed int asm_ret = vsx_ld(0, zout);
+
+    __asm__ ("xvcvspsxws %x0,%x1" : "=wa" (vz4) : "wa" (asm_ret));
+
+    vsx_st(asm_ret, 0, zout);
+    return zout[0];
+}

numpy/distutils/checks/test_flags.c

@@ -0,0 +1 @@
+int test_flags;

numpy/distutils/command/__init__.py

@@ -0,0 +1,41 @@
+"""distutils.command
+
+Package containing implementation of all the standard Distutils
+commands.
+
+"""
+def test_na_writable_attributes_deletion():
+    a = np.NA(2)
+    attr =  ['payload', 'dtype']
+    for s in attr:
+        assert_raises(AttributeError, delattr, a, s)
+
+
+__revision__ = "$Id: __init__.py,v 1.3 2005/05/16 11:08:49 pearu Exp $"
+
+distutils_all = [  #'build_py',
+                   'clean',
+                   'install_clib',
+                   'install_scripts',
+                   'bdist',
+                   'bdist_dumb',
+                   'bdist_wininst',
+                ]
+
+__import__('distutils.command', globals(), locals(), distutils_all)
+
+__all__ = ['build',
+           'config_compiler',
+           'config',
+           'build_src',
+           'build_py',
+           'build_ext',
+           'build_clib',
+           'build_scripts',
+           'install',
+           'install_data',
+           'install_headers',
+           'install_lib',
+           'bdist_rpm',
+           'sdist',
+          ] + distutils_all

numpy/distutils/command/autodist.py

@@ -0,0 +1,148 @@
+"""This module implements additional tests ala autoconf which can be useful.
+
+"""
+import textwrap
+
+# We put them here since they could be easily reused outside numpy.distutils
+
+def check_inline(cmd):
+    """Return the inline identifier (may be empty)."""
+    cmd._check_compiler()
+    body = textwrap.dedent("""
+        #ifndef __cplusplus
+        static %(inline)s int static_func (void)
+        {
+            return 0;
+        }
+        %(inline)s int nostatic_func (void)
+        {
+            return 0;
+        }
+        #endif""")
+
+    for kw in ['inline', '__inline__', '__inline']:
+        st = cmd.try_compile(body % {'inline': kw}, None, None)
+        if st:
+            return kw
+
+    return ''
+
+
+def check_restrict(cmd):
+    """Return the restrict identifier (may be empty)."""
+    cmd._check_compiler()
+    body = textwrap.dedent("""
+        static int static_func (char * %(restrict)s a)
+        {
+            return 0;
+        }
+        """)
+
+    for kw in ['restrict', '__restrict__', '__restrict']:
+        st = cmd.try_compile(body % {'restrict': kw}, None, None)
+        if st:
+            return kw
+
+    return ''
+
+
+def check_compiler_gcc(cmd):
+    """Check if the compiler is GCC."""
+
+    cmd._check_compiler()
+    body = textwrap.dedent("""
+        int
+        main()
+        {
+        #if (! defined __GNUC__)
+        #error gcc required
+        #endif
+            return 0;
+        }
+        """)
+    return cmd.try_compile(body, None, None)
+
+
+def check_gcc_version_at_least(cmd, major, minor=0, patchlevel=0):
+    """
+    Check that the gcc version is at least the specified version."""
+
+    cmd._check_compiler()
+    version = '.'.join([str(major), str(minor), str(patchlevel)])
+    body = textwrap.dedent("""
+        int
+        main()
+        {
+        #if (! defined __GNUC__) || (__GNUC__ < %(major)d) || \\
+                (__GNUC_MINOR__ < %(minor)d) || \\
+                (__GNUC_PATCHLEVEL__ < %(patchlevel)d)
+        #error gcc >= %(version)s required
+        #endif
+            return 0;
+        }
+        """)
+    kw = {'version': version, 'major': major, 'minor': minor,
+          'patchlevel': patchlevel}
+
+    return cmd.try_compile(body % kw, None, None)
+
+
+def check_gcc_function_attribute(cmd, attribute, name):
+    """Return True if the given function attribute is supported."""
+    cmd._check_compiler()
+    body = textwrap.dedent("""
+        #pragma GCC diagnostic error "-Wattributes"
+        #pragma clang diagnostic error "-Wattributes"
+
+        int %s %s(void* unused)
+        {
+            return 0;
+        }
+
+        int
+        main()
+        {
+            return 0;
+        }
+        """) % (attribute, name)
+    return cmd.try_compile(body, None, None) != 0
+
+
+def check_gcc_function_attribute_with_intrinsics(cmd, attribute, name, code,
+                                                include):
+    """Return True if the given function attribute is supported with
+    intrinsics."""
+    cmd._check_compiler()
+    body = textwrap.dedent("""
+        #include<%s>
+        int %s %s(void)
+        {
+            %s;
+            return 0;
+        }
+
+        int
+        main()
+        {
+            return 0;
+        }
+        """) % (include, attribute, name, code)
+    return cmd.try_compile(body, None, None) != 0
+
+
+def check_gcc_variable_attribute(cmd, attribute):
+    """Return True if the given variable attribute is supported."""
+    cmd._check_compiler()
+    body = textwrap.dedent("""
+        #pragma GCC diagnostic error "-Wattributes"
+        #pragma clang diagnostic error "-Wattributes"
+
+        int %s foo;
+
+        int
+        main()
+        {
+            return 0;
+        }
+        """) % (attribute, )
+    return cmd.try_compile(body, None, None) != 0

numpy/distutils/command/bdist_rpm.py

@@ -0,0 +1,22 @@
+import os
+import sys
+if 'setuptools' in sys.modules:
+    from setuptools.command.bdist_rpm import bdist_rpm as old_bdist_rpm
+else:
+    from distutils.command.bdist_rpm import bdist_rpm as old_bdist_rpm
+
+class bdist_rpm(old_bdist_rpm):
+
+    def _make_spec_file(self):
+        spec_file = old_bdist_rpm._make_spec_file(self)
+
+        # Replace hardcoded setup.py script name
+        # with the real setup script name.
+        setup_py = os.path.basename(sys.argv[0])
+        if setup_py == 'setup.py':
+            return spec_file
+        new_spec_file = []
+        for line in spec_file:
+            line = line.replace('setup.py', setup_py)
+            new_spec_file.append(line)
+        return new_spec_file

numpy/distutils/command/build.py

@@ -0,0 +1,62 @@
+import os
+import sys
+from distutils.command.build import build as old_build
+from distutils.util import get_platform
+from numpy.distutils.command.config_compiler import show_fortran_compilers
+
+class build(old_build):
+
+    sub_commands = [('config_cc',     lambda *args: True),
+                    ('config_fc',     lambda *args: True),
+                    ('build_src',     old_build.has_ext_modules),
+                    ] + old_build.sub_commands
+
+    user_options = old_build.user_options + [
+        ('fcompiler=', None,
+         "specify the Fortran compiler type"),
+        ('warn-error', None,
+         "turn all warnings into errors (-Werror)"),
+        ('cpu-baseline=', None,
+         "specify a list of enabled baseline CPU optimizations"),
+        ('cpu-dispatch=', None,
+         "specify a list of dispatched CPU optimizations"),
+        ('disable-optimization', None,
+         "disable CPU optimized code(dispatch,simd,fast...)"),
+        ('simd-test=', None,
+         "specify a list of CPU optimizations to be tested against NumPy SIMD interface"),
+        ]
+
+    help_options = old_build.help_options + [
+        ('help-fcompiler', None, "list available Fortran compilers",
+         show_fortran_compilers),
+        ]
+
+    def initialize_options(self):
+        old_build.initialize_options(self)
+        self.fcompiler = None
+        self.warn_error = False
+        self.cpu_baseline = "min"
+        self.cpu_dispatch = "max -xop -fma4" # drop AMD legacy features by default
+        self.disable_optimization = False
+        """
+        the '_simd' module is a very large. Adding more dispatched features
+        will increase binary size and compile time. By default we minimize
+        the targeted features to those most commonly used by the NumPy SIMD interface(NPYV),
+        NOTE: any specified features will be ignored if they're:
+            - part of the baseline(--cpu-baseline)
+            - not part of dispatch-able features(--cpu-dispatch)
+            - not supported by compiler or platform
+        """
+        self.simd_test = "BASELINE SSE2 SSE42 XOP FMA4 (FMA3 AVX2) AVX512F " \
+                         "AVX512_SKX VSX VSX2 VSX3 VSX4 NEON ASIMD VX VXE VXE2"
+
+    def finalize_options(self):
+        build_scripts = self.build_scripts
+        old_build.finalize_options(self)
+        plat_specifier = ".{}-{}.{}".format(get_platform(), *sys.version_info[:2])
+        if build_scripts is None:
+            self.build_scripts = os.path.join(self.build_base,
+                                              'scripts' + plat_specifier)
+
+    def run(self):
+        old_build.run(self)