}

func withSizes(sizes Sizes) func(*Config) {
	return func(cfg *Config) {
		cfg.Sizes = sizes
	}
}

// TestIndexRepresentability tests that constant index operands must
// be representable as int even if they already have a type that can
// represent larger values.
func TestIndexRepresentability(t *testing.T) {

    return %0 : tensor<1x3xf32>
  }
// CHECK: @uniform_dequantize_0
}

// -----

// Tests that a quantized dot_general op is composed when both operands are
// actiavations.

// CHECK-LABEL: dot_general_with_two_activations
// CHECK-SAME: %[[ARG_0:.*]]: tensor<8x16x16xf32>
// CHECK-SAME: %[[ARG_1:.*]]: tensor<8x16x4xf32>
module {

  }
  return absl::OkStatus();
}

// Returns true if the executor/control dialect op should map to Ref node in
// TensorFlow Graph. For control dialect NextIteration it uses the 1st operand
// type. For executor dialect NextIteration it uses the 2nd operand type. For
// all others (Enter/Exit/Merge/Switch), if the output type is ref, they
// correspond to the Ref equivalent op in TF Graph.
static bool IsRefTypeControlOp(mlir::Operation* op) {

// There is no fundamental theory behind this number.
const tmpStringBufSize = 32

type tmpBuf [tmpStringBufSize]byte

// concatstrings implements a Go string concatenation x+y+z+...
// The operands are passed in the slice a.
// If buf != nil, the compiler has determined that the result does not
// escape the calling function, so the string data can be stored in buf
// if small enough.

	var a []complex64
	_ = imag(a... /* ERROR "invalid use of ..." */ )

	// if argument is untyped, result is untyped
	const _ byte = imag(1.2 + 3i)
	const _ complex128 = imag(1.2 + 3i)

	// lhs constant shift operands are typed as complex128
	var s uint
	_ = imag(1 /* ERROR "must be integer" */ << s)
}

func imag2() {
	f1 := func() (x complex128) { return }
	f2 := func() (x, y complex128) { return }

		typ := new(Map)
		setDefType(def, typ)

		typ.key = check.varType(e.Key)
		typ.elem = check.varType(e.Value)

		// spec: "The comparison operators == and != must be fully defined
		// for operands of the key type; thus the key type must not be a
		// function, map, or slice."
		//
		// Delay this check because it requires fully setup types;
		// it is safe to continue in any case (was go.dev/issue/6667).

    for (Value operand : nested_op->getOperands()) {
      Operation* defining_op = operand.getDefiningOp();
      if (!defining_op) continue;
      Operation* producer_in_block = block->findAncestorOpInBlock(*defining_op);
      if (producer_in_block && producer_in_block != &op &&
          unscheduled_ops.count(producer_in_block)) {
        // Found an operand that isn't scheduled yet, interrupt the walk.

          _QuantizationComponent.COMPONENT_ACTIVATION,
      ]:
        if component_spec.tensor_type != _TensorType.TENSORTYPE_INT_8:
          raise ValueError(
              'Only int8 precision is supported for input operands.'
          )
      else:
        if component_spec.tensor_type != _TensorType.TENSORTYPE_INT_32:
          raise ValueError('Only int32 precision is supported for bias.')
      # Update with the custom spec.

      llvm::cl::desc("Force data format for all layout sensitive ops")};
  Option<bool> skip_fold_transpose_in_ops{
      *this, "skip-fold-transpose-in-ops",
      llvm::cl::desc("Skip folding transpose operands in Ops which can support "
                     "different layouts.")};
};

// Layout optimization assigns optimal data layout for layout sensitive
// operations, and cancels all redundant transposes.

// type parameter. Otherwise the result is nil. Errors are reported through the err parameter.
// Note: infer may fail (return nil) due to invalid args operands without reporting additional errors.
func (check *Checker) infer(posn positioner, tparams []*TypeParam, targs []Type, params *Tuple, args []*operand, reverse bool, err *error_) (inferred []Type) {
	// Don't verify result conditions if there's no error handler installed: