{"OpPlus", Const, 0},
		{"OpQuest", Const, 0},
		{"OpRepeat", Const, 0},
		{"OpStar", Const, 0},
		{"OpWordBoundary", Const, 0},
		{"POSIX", Const, 0},
		{"Parse", Func, 0},
		{"Perl", Const, 0},
		{"PerlX", Const, 0},
		{"Prog", Type, 0},
		{"Prog.Inst", Field, 0},
		{"Prog.NumCap", Field, 0},
		{"Prog.Start", Field, 0},
		{"Regexp", Type, 0},
		{"Regexp.Cap", Field, 0},

* <SIMILAR TO ABOVE SECTION, BUT FOR OTHER IMPORTANT CHANGES / BUG FIXES>
* <IF A CHANGE CLOSES A GITHUB ISSUE, IT SHOULD BE DOCUMENTED HERE>
* <NOTES SHOULD BE GROUPED PER AREA>

* `tf.lite`
    * Quantization for `FullyConnected` layer is switched from per-tensor to
      per-channel scales for dynamic range quantization use case (`float32`
      inputs / outputs and `int8` weights). The change enables new quantization

The output `y` has the same rank as `x`. The shapes of `x` and `y` satisfy:
  `y.shape[i] == x.shape[perm[i]] for i in [0, 1, ..., rank(x) - 1]`
  `y[i,j,k,...,s,t,u] == conj(x[perm[i], perm[j], perm[k],...,perm[s], perm[t], perm[u]])`
  }];

  let arguments = (ins
    TF_Tensor:$x,
    TF_I32OrI64Tensor:$perm
  );

  let results = (outs
    TF_Tensor:$y
  );

  - `kubectl` outputs warnings to stderr, and accepts a `--warnings-as-errors` option to treat warnings as fatal errors
  - `k8s.io/client-go` outputs warnings to stderr by default; override this per-client by setting `config.WarningHandler`, or per-process with `rest.SetDefaultWarningHandler()`