Migrate PyTorch to C++17 (#85969)

With CUDA-10.2 gone we can finally do it!

This PR mostly contains build system related changes, invasive functional ones are to be followed.
Among many expected tweaks to the build system, here are few unexpected ones:
 - Force onnx_proto project to be updated to C++17 to avoid `duplicate symbols` error when compiled by gcc-7.5.0, as storage rule for `constexpr` changed in C++17, but gcc does not seem to follow it
 - Do not use `std::apply` on CUDA but rely on the built-in variant, as it results in test failures when CUDA runtime picks host rather than device function when `std::apply` is invoked from CUDA code.
 - `std::decay_t` -> `::std::decay_t` and `std::move`->`::std::move` as VC++ for some reason claims that `std` symbol is ambigious
 - Disable use of `std::aligned_alloc` on Android, as its `libc++` does not implement it.

Some prerequisites:
 - https://github.com/pytorch/pytorch/pull/89297
 - https://github.com/pytorch/pytorch/pull/89605
 - https://github.com/pytorch/pytorch/pull/90228
 - https://github.com/pytorch/pytorch/pull/90389
 - https://github.com/pytorch/pytorch/pull/90379
 - https://github.com/pytorch/pytorch/pull/89570
 - https://github.com/facebookincubator/gloo/pull/336
 - https://github.com/facebookincubator/gloo/pull/343
 - 919676fb32

Fixes https://github.com/pytorch/pytorch/issues/56055

Pull Request resolved: https://github.com/pytorch/pytorch/pull/85969
Approved by: https://github.com/ezyang, https://github.com/kulinseth

torch/utils/cpp_extension.py

@@ -431,7 +431,7 @@ class BuildExtension(build_ext, object):
     A custom :mod:`setuptools` build extension .
 
     This :class:`setuptools.build_ext` subclass takes care of passing the
-    minimum required compiler flags (e.g. ``-std=c++14``) as well as mixed
+    minimum required compiler flags (e.g. ``-std=c++17``) as well as mixed
     C++/CUDA compilation (and support for CUDA files in general).
 
     When using :class:`BuildExtension`, it is allowed to supply a dictionary
@@ -535,12 +535,12 @@ class BuildExtension(build_ext, object):
         else:
             original_compile = self.compiler._compile
 
-        def append_std14_if_no_std_present(cflags) -> None:
+        def append_std17_if_no_std_present(cflags) -> None:
             # NVCC does not allow multiple -std to be passed, so we avoid
             # overriding the option if the user explicitly passed it.
             cpp_format_prefix = '/{}:' if self.compiler.compiler_type == 'msvc' else '-{}='
             cpp_flag_prefix = cpp_format_prefix.format('std')
-            cpp_flag = cpp_flag_prefix + 'c++14'
+            cpp_flag = cpp_flag_prefix + 'c++17'
             if not any(flag.startswith(cpp_flag_prefix) for flag in cflags):
                 cflags.append(cpp_flag)
 
@@ -585,7 +585,7 @@ class BuildExtension(build_ext, object):
                     cflags = cflags['cxx']
                 if IS_HIP_EXTENSION:
                     cflags = COMMON_HIP_FLAGS + cflags
-                append_std14_if_no_std_present(cflags)
+                append_std17_if_no_std_present(cflags)
 
                 original_compile(obj, src, ext, cc_args, cflags, pp_opts)
             finally:
@@ -634,7 +634,7 @@ class BuildExtension(build_ext, object):
                 post_cflags = list(extra_postargs)
             if IS_HIP_EXTENSION:
                 post_cflags = COMMON_HIP_FLAGS + post_cflags
-            append_std14_if_no_std_present(post_cflags)
+            append_std17_if_no_std_present(post_cflags)
 
             cuda_post_cflags = None
             cuda_cflags = None
@@ -649,7 +649,7 @@ class BuildExtension(build_ext, object):
                     cuda_post_cflags = COMMON_HIP_FLAGS + COMMON_HIPCC_FLAGS + cuda_post_cflags
                 else:
                     cuda_post_cflags = unix_cuda_flags(cuda_post_cflags)
-                append_std14_if_no_std_present(cuda_post_cflags)
+                append_std17_if_no_std_present(cuda_post_cflags)
                 cuda_cflags = [shlex.quote(f) for f in cuda_cflags]
                 cuda_post_cflags = [shlex.quote(f) for f in cuda_post_cflags]
 
@@ -785,7 +785,7 @@ class BuildExtension(build_ext, object):
                 post_cflags = extra_postargs['cxx']
             else:
                 post_cflags = list(extra_postargs)
-            append_std14_if_no_std_present(post_cflags)
+            append_std17_if_no_std_present(post_cflags)
 
             cuda_post_cflags = None
             cuda_cflags = None
@@ -1994,7 +1994,7 @@ def _write_ninja_file_to_build_library(path,
         cflags = common_cflags + COMMON_MSVC_FLAGS + extra_cflags
         cflags = _nt_quote_args(cflags)
     else:
-        cflags = common_cflags + ['-fPIC', '-std=c++14'] + extra_cflags
+        cflags = common_cflags + ['-fPIC', '-std=c++17'] + extra_cflags
 
     if with_cuda and IS_HIP_EXTENSION:
         cuda_flags = ['-DWITH_HIP'] + cflags + COMMON_HIP_FLAGS + COMMON_HIPCC_FLAGS
@@ -2013,7 +2013,7 @@ def _write_ninja_file_to_build_library(path,
             cuda_flags += ['--compiler-options', "'-fPIC'"]
             cuda_flags += extra_cuda_cflags
             if not any(flag.startswith('-std=') for flag in cuda_flags):
-                cuda_flags.append('-std=c++14')
+                cuda_flags.append('-std=c++17')
             if os.getenv("CC") is not None:
                 cuda_flags = ['-ccbin', os.getenv("CC")] + cuda_flags
     else: