ENH: [3/7] enable multi-platform SIMD compiler optimizations

  - Put `CCompilerOpt` in action through add two command line
    arguments that passed directly to `CCompilerOpt`'s
    parameters which explained as follows:

    * `--cpu-baseline` minimal set of required optimizations,
      default is 'min' which provides the minimum CPU features
      that can safely run on a wide range of users platforms.

    * `--cpu-dispatch` dispatched set of additional optimizations,
      default is 'max-xop-fma4' which enables all CPU features,
      except for AMD legacy features.

    the new arguments can be reached from `build`, `build_clib`,
    `build_ext`, if `build_clib` or `build_ext` are not specified
    by the user the arguments of `build` will be used, which also
    hold the default values.

  - Activate the new compiler dispatcher that comes with `CCompilerOpt`,
    through adding a hock inside `build_clib` and `build_ext`
    that works as a filter taking any C source files ends with
    `.dispatch.c` and pass it directly to `CCompilerOpt` and
    then take returned objects and linked to the final C lib.

  - Add a third command-line argument `--disable-optimization` which
    explicitly disable the whole new infrastructure, also
    It adds a new compiler definition called `NPY_DISABLE_OPTIMIZATION`.

    when `--disable-optimization` is enabled the dispatch-able sources
    that end with `.dispatch.c` will be treated as a normal
    C sources, also due to this disabling any C headers that
    generated by `CCompilerOpt` must guard it with `NPY_DISABLE_OPTIMIZATION`,
    otherwise, it will definitely break the build.

  - New auto-generated C header located at `core/include/numpy/_cpu_dispatch.h`,
    the new header contains all definitions and headers of CPU features that
    enabled according to specified configuration in `--cpu-baseline`
    and `--cpu-dispatch`.

5 files changed, 144 insertions, 5 deletions

diff --git a/.gitignore b/.gitignore
index 302968a14..2a77f4f83 100644
--- a/.gitignore
+++ b/.gitignore
@@ -130,6 +130,7 @@ numpy/core/src/common/npy_cpu_features.c
 numpy/core/src/common/npy_partition.h
 numpy/core/src/common/npy_sort.h
 numpy/core/src/common/templ_common.h
+numpy/core/src/common/_cpu_dispatch.h
 numpy/core/src/multiarray/_multiarray_tests.c
 numpy/core/src/multiarray/arraytypes.c
 numpy/core/src/multiarray/einsum.c
diff --git a/numpy/distutils/command/build.py b/numpy/distutils/command/build.py
index a156a7c6e..60ba4c917 100644
--- a/numpy/distutils/command/build.py
+++ b/numpy/distutils/command/build.py
@@ -16,6 +16,12 @@ class build(old_build):
          "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...)"),
         ]
 
     help_options = old_build.help_options + [
@@ -27,6 +33,9 @@ class build(old_build):
         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
 
     def finalize_options(self):
         build_scripts = self.build_scripts
diff --git a/numpy/distutils/command/build_clib.py b/numpy/distutils/command/build_clib.py
index f6a84e351..87345adbc 100644
--- a/numpy/distutils/command/build_clib.py
+++ b/numpy/distutils/command/build_clib.py
@@ -13,6 +13,7 @@ from numpy.distutils.misc_util import (
     filter_sources, get_lib_source_files, get_numpy_include_dirs,
     has_cxx_sources, has_f_sources, is_sequence
 )
+from numpy.distutils.ccompiler_opt import new_ccompiler_opt
 
 # Fix Python distutils bug sf #1718574:
 _l = old_build_clib.user_options
@@ -34,9 +35,16 @@ class build_clib(old_build_clib):
          "number of parallel jobs"),
         ('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...)"),
     ]
 
-    boolean_options = old_build_clib.boolean_options + ['inplace', 'warn-error']
+    boolean_options = old_build_clib.boolean_options + \
+    ['inplace', 'warn-error', 'disable-optimization']
 
     def initialize_options(self):
         old_build_clib.initialize_options(self)
@@ -44,6 +52,10 @@ class build_clib(old_build_clib):
         self.inplace = 0
         self.parallel = None
         self.warn_error = None
+        self.cpu_baseline = None
+        self.cpu_dispatch = None
+        self.disable_optimization = None
+
 
     def finalize_options(self):
         if self.parallel:
@@ -55,6 +67,9 @@ class build_clib(old_build_clib):
         self.set_undefined_options('build',
                                         ('parallel', 'parallel'),
                                         ('warn_error', 'warn_error'),
+                                        ('cpu_baseline', 'cpu_baseline'),
+                                        ('cpu_dispatch', 'cpu_dispatch'),
+                                        ('disable_optimization', 'disable_optimization')
                                   )
 
     def have_f_sources(self):
@@ -102,6 +117,25 @@ class build_clib(old_build_clib):
 
         self.compiler.show_customization()
 
+        if not self.disable_optimization:
+            opt_cache_path = os.path.abspath(
+                os.path.join(self.build_temp, 'ccompiler_opt_cache_clib.py'
+            ))
+            self.compiler_opt = new_ccompiler_opt(
+                compiler=self.compiler, cpu_baseline=self.cpu_baseline,
+                cpu_dispatch=self.cpu_dispatch, cache_path=opt_cache_path
+            )
+            if not self.compiler_opt.is_cached():
+                log.info("Detected changes on compiler optimizations, force rebuilding")
+                self.force = True
+
+            import atexit
+            def report():
+                log.info("\n########### CLIB COMPILER OPTIMIZATION ###########")
+                log.info(self.compiler_opt.report(full=True))
+
+            atexit.register(report)
+
         if self.have_f_sources():
             from numpy.distutils.fcompiler import new_fcompiler
             self._f_compiler = new_fcompiler(compiler=self.fcompiler,
@@ -211,6 +245,8 @@ class build_clib(old_build_clib):
                 'extra_f90_compile_args') or []
 
         macros = build_info.get('macros')
+        if macros is None:
+            macros = []
         include_dirs = build_info.get('include_dirs')
         if include_dirs is None:
             include_dirs = []
@@ -223,6 +259,31 @@ class build_clib(old_build_clib):
         if requiref90:
             self.mkpath(module_build_dir)
 
+        dispatch_objects = []
+        if not self.disable_optimization:
+            dispatch_sources  = [
+                c_sources.pop(c_sources.index(src))
+                for src in c_sources[:] if src.endswith(".dispatch.c")
+            ]
+            if dispatch_sources:
+                if not self.inplace:
+                    build_src = self.get_finalized_command("build_src").build_src
+                else:
+                    build_src = None
+                dispatch_objects = self.compiler_opt.try_dispatch(
+                    dispatch_sources,
+                    output_dir=self.build_temp,
+                    src_dir=build_src,
+                    macros=macros,
+                    include_dirs=include_dirs,
+                    debug=self.debug,
+                    extra_postargs=extra_postargs
+                )
+            extra_args_baseopt = extra_postargs + self.compiler_opt.cpu_baseline_flags()
+        else:
+            extra_args_baseopt = extra_postargs
+            macros.append(("NPY_DISABLE_OPTIMIZATION", 1))
+
         if compiler.compiler_type == 'msvc':
             # this hack works around the msvc compiler attributes
             # problem, msvc uses its own convention :(
@@ -237,7 +298,8 @@ class build_clib(old_build_clib):
                                        macros=macros,
                                        include_dirs=include_dirs,
                                        debug=self.debug,
-                                       extra_postargs=extra_postargs)
+                                       extra_postargs=extra_args_baseopt)
+        objects.extend(dispatch_objects)
 
         if cxx_sources:
             log.info("compiling C++ sources")
diff --git a/numpy/distutils/command/build_ext.py b/numpy/distutils/command/build_ext.py
index d53285c92..b6557fcf6 100644
--- a/numpy/distutils/command/build_ext.py
+++ b/numpy/distutils/command/build_ext.py
@@ -19,7 +19,7 @@ from numpy.distutils.misc_util import (
     has_cxx_sources, has_f_sources, is_sequence
 )
 from numpy.distutils.command.config_compiler import show_fortran_compilers
-
+from numpy.distutils.ccompiler_opt import new_ccompiler_opt
 
 
 class build_ext (old_build_ext):
@@ -33,6 +33,12 @@ class build_ext (old_build_ext):
          "number of parallel jobs"),
         ('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...)"),
     ]
 
     help_options = old_build_ext.help_options + [
@@ -40,13 +46,16 @@ class build_ext (old_build_ext):
          show_fortran_compilers),
     ]
 
-    boolean_options = old_build_ext.boolean_options + ['warn-error']
+    boolean_options = old_build_ext.boolean_options + ['warn-error', 'disable-optimization']
 
     def initialize_options(self):
         old_build_ext.initialize_options(self)
         self.fcompiler = None
         self.parallel = None
         self.warn_error = None
+        self.cpu_baseline = None
+        self.cpu_dispatch = None
+        self.disable_optimization = None
 
     def finalize_options(self):
         if self.parallel:
@@ -75,6 +84,9 @@ class build_ext (old_build_ext):
         self.set_undefined_options('build',
                                         ('parallel', 'parallel'),
                                         ('warn_error', 'warn_error'),
+                                        ('cpu_baseline', 'cpu_baseline'),
+                                        ('cpu_dispatch', 'cpu_dispatch'),
+                                        ('disable_optimization', 'disable_optimization'),
                                   )
 
     def run(self):
@@ -129,6 +141,22 @@ class build_ext (old_build_ext):
 
         self.compiler.show_customization()
 
+        if not self.disable_optimization:
+            opt_cache_path = os.path.abspath(os.path.join(self.build_temp, 'ccompiler_opt_cache_ext.py'))
+            self.compiler_opt = new_ccompiler_opt(compiler=self.compiler,
+                                                  cpu_baseline=self.cpu_baseline,
+                                                  cpu_dispatch=self.cpu_dispatch,
+                                                  cache_path=opt_cache_path)
+            if not self.compiler_opt.is_cached():
+                log.info("Detected changes on compiler optimizations, force rebuilding")
+                self.force = True
+
+            import atexit
+            def report():
+                log.info("\n########### EXT COMPILER OPTIMIZATION ###########")
+                log.info(self.compiler_opt.report(full=True))
+            atexit.register(report)
+
         # Setup directory for storing generated extra DLL files on Windows
         self.extra_dll_dir = os.path.join(self.build_temp, '.libs')
         if not os.path.isdir(self.extra_dll_dir):
@@ -378,6 +406,32 @@ class build_ext (old_build_ext):
 
         include_dirs = ext.include_dirs + get_numpy_include_dirs()
 
+        dispatch_objects = []
+        if not self.disable_optimization:
+            dispatch_sources  = [
+                c_sources.pop(c_sources.index(src))
+                for src in c_sources[:] if src.endswith(".dispatch.c")
+            ]
+            if dispatch_sources:
+                if not self.inplace:
+                    build_src = self.get_finalized_command("build_src").build_src
+                else:
+                    build_src = None
+                dispatch_objects = self.compiler_opt.try_dispatch(
+                    dispatch_sources,
+                    output_dir=output_dir,
+                    src_dir=build_src,
+                    macros=macros,
+                    include_dirs=include_dirs,
+                    debug=self.debug,
+                    extra_postargs=extra_args,
+                    **kws
+                )
+            extra_args_baseopt = extra_args + self.compiler_opt.cpu_baseline_flags()
+        else:
+            extra_args_baseopt = extra_args
+            macros.append(("NPY_DISABLE_OPTIMIZATION", 1))
+
         c_objects = []
         if c_sources:
             log.info("compiling C sources")
@@ -386,8 +440,9 @@ class build_ext (old_build_ext):
                                               macros=macros,
                                               include_dirs=include_dirs,
                                               debug=self.debug,
-                                              extra_postargs=extra_args,
+                                              extra_postargs=extra_args_baseopt,
                                               **kws)
+        c_objects.extend(dispatch_objects)
 
         if cxx_sources:
             log.info("compiling C++ sources")
diff --git a/runtests.py b/runtests.py
index 7f1d55b85..beaf668d6 100755
--- a/runtests.py
+++ b/runtests.py
@@ -114,6 +114,12 @@ def main(argv):
                         help="Number of parallel jobs during build")
     parser.add_argument("--warn-error", action="store_true",
                         help="Set -Werror to convert all compiler warnings to errors")
+    parser.add_argument("--cpu-baseline", default=None,
+                        help="Specify a list of enabled baseline CPU optimizations"),
+    parser.add_argument("--cpu-dispatch", default=None,
+                        help="Specify a list of dispatched CPU optimizations"),
+    parser.add_argument("--disable-optimization", action="store_true",
+                        help="Disable CPU optimized code(dispatch,simd,fast...)"),
     parser.add_argument("--show-build-log", action="store_true",
                         help="Show build output rather than using a log file")
     parser.add_argument("--bench", action="store_true",
@@ -388,6 +394,12 @@ def build_project(args):
         cmd += ["build_src", "--verbose-cfg"]
     if args.warn_error:
         cmd += ["--warn-error"]
+    if args.cpu_baseline:
+        cmd += ["--cpu-baseline", args.cpu_baseline]
+    if args.cpu_dispatch:
+        cmd += ["--cpu-dispatch", args.cpu_dispatch]
+    if args.disable_optimization:
+        cmd += ["--disable-optimization"]
     # Install; avoid producing eggs so numpy can be imported from dst_dir.
     cmd += ['install', '--prefix=' + dst_dir,
             '--single-version-externally-managed',