Previously, large builds could see 100%+ build time increase from enabling PGO.
In Go 1.23, overhead should be in the single digit percentages.

## Assembler {#assembler}

# the shared library, only export the core TF API functions to avoid
# causing library conflicts (e.g., those reported in github issue 1924).
# On Linux, tell the linker (-Wl,<option>) to use a version script that
# excludes all but a subset of function names.
# On MacOS, the linker does not support version_script, but has an
# an "-exported_symbols_list" command.  -z defs disallows undefined
# symbols in object files.

ideally referring to a person with the responsibility to complete the note.

Use the following forms in your markdown:

	[http.Request]                     # symbol documentation; auto-linked as in Go doc strings
	[#12345](/issue/12345)             # GitHub issues
	[CL 6789](/cl/6789)                # Gerrit changelists

## For the release team

  fun repeat(
    c: Char,
    count: Int,
  ): String {
    val array = CharArray(count)
    Arrays.fill(array, c)
    return String(array)
  }

  /**
   * Okio buffers are internally implemented as a linked list of arrays. Usually this implementation
   * detail is invisible to the caller, but subtle use of certain APIs may depend on these internal
   * structures.
   *

build:release_linux_base --copt=-Wno-error=array-parameter
build:release_linux_base --copt=-Wno-error=unused-command-line-argument
# Set lld as the linker.
build:release_linux_base --linkopt="-fuse-ld=lld"
build:release_linux_base --linkopt="-lm"

# We have some invalid linker scripts in the build,
# so we need to disable this check
build:release_linux_base --linkopt=-Wl,--undefined-version

  if is_ppc64le():
    # Enable MMA Dynamic Dispatch support if 'gcc' and if linker >= 2.35
    gcc_env = get_gcc_compiler(environ_cp)
    if gcc_env is not None:

      # Use gold linker if 'gcc' and if 'ppc64le'
      write_to_bazelrc('build --linkopt="-fuse-ld=gold"')

      # Get the linker version
      ld_version = run_shell([gcc_env, '-Wl,-version']).split()

precommit: format lint

format: fmt ## Auto formats all code. This should be run before sending a PR.

fmt: format-go format-python tidy-go

ifeq ($(DEBUG),1)
# gobuild script uses custom linker flag to set the variables.
RELEASE_LDFLAGS=''
else
RELEASE_LDFLAGS='-extldflags -static -s -w'
endif

# List of all binaries to build

      "PluggableDevice plugin functionality is not supported on mobile");
  return nullptr;
#else
  TF_Library* lib_handle = new TF_Library;
  static tensorflow::mutex mu(tensorflow::LINKER_INITIALIZED);
  static std::unordered_map<std::string, void*>* loaded_libs =
      new std::unordered_map<std::string, void*>();
  tensorflow::Env* env = tensorflow::Env::Default();
  {

      if (count == 0 && !isLinkedHash) {
        contents = new ObjectCountLinkedHashMap<E>(contents);
        isLinkedHash = true;
        // to preserve insertion order through deletions, we have to switch to an actual linked
        // implementation at least for now, but this should be a super rare case
      } else if (buildInvoked) {
        contents = new ObjectCountHashMap<E>(contents);
        isLinkedHash = false;
      }

            AVX512_BF16, AMX, etc.
            ([Intel Cascade Lake](https://www.intel.com/content/www/us/en/products/platforms/details/cascade-lake.html)
            and newer CPUs.)
            *   [Example performance speedups.](https://medium.com/intel-analytics-software/leverage-intel-deep-learning-optimizations-in-tensorflow-129faa80ee07)