public static int log2(double x, RoundingMode mode) {
    checkArgument(x > 0.0 && isFinite(x), "x must be positive and finite");
    int exponent = getExponent(x);
    if (!isNormal(x)) {
      return log2(x * IMPLICIT_BIT, mode) - SIGNIFICAND_BITS;
      // Do the calculation on a normal value.
    }
    // x is positive, finite, and normal
    boolean increment;
    switch (mode) {
      case UNNECESSARY:

    assertThat(TWO_VALUES_STATS.mean()).isWithin(ALLOWED_ERROR).of(TWO_VALUES_MEAN);
    // For datasets of many double values created from an array, we test many combinations of finite
    // and non-finite values:
    for (ManyValues values : ALL_MANY_VALUES) {
      double mean = Stats.of(values.asArray()).mean();
      if (values.hasAnyNaN()) {
        assertWithMessage("mean of " + values).that(mean).isNaN();

      .Case("TANH", tflite::ActivationFunctionType_TANH)
      .Case("SIGN_BIT", tflite::ActivationFunctionType_SIGN_BIT);
}

static tflite::TensorType ConvertDerivedTFLiteTypeAttrForOptionWriter(
    tflite::TensorType type, flatbuffers::FlatBufferBuilder* builder) {
  if (type == tflite::TensorType_INT64) {
    return tflite::TensorType_INT64;
  } else if (type == tflite::TensorType_INT32) {

#include "mlir/Support/LogicalResult.h"  // from @llvm-project
#include "mlir/Transforms/GreedyPatternRewriteDriver.h"  // from @llvm-project
#include "tensorflow/compiler/mlir/lite/ir/tfl_ops.h"
#include "tensorflow/compiler/mlir/lite/transforms/passes.h"
#include "tensorflow/compiler/mlir/lite/utils/validators.h"
#include "tensorflow/compiler/mlir/quantization/common/quantization_lib/quantization_config.h"

yIsNan:			//  Pow(NaN, y) = NaN
ltZeroInt:		// Pow(x, y) = NaN for finite x < 0 and finite non-integer y
	MOVD	$NaN, R2
	MOVD	R2, ret+16(FP)
	RET
negOnePosInf:	// Pow(-1, ±Inf) = 1
negOneNegInf:
	MOVD	$PosOne, R3
	MOVD	R3, ret+16(FP)
	RET
negZeroOddInt:
	MOVD	$NegInf, R3
	MOVD	R3, ret+16(FP)
	RET
zeroNotOdd:		// Pow(±0, y) = +Inf for finite y < 0 and not an odd integer

     *
     * Consequently:
     * 1. If the previous mean is finite and the new value is non-finite then the new mean is that
     *    value (whether it is NaN or infinity).
     * 2. If the new value is finite and the previous mean is non-finite then the mean is unchanged
     *    (whether it is NaN or infinity).
     * 3. If both the previous mean and the new value are non-finite and...

     *
     * Consequently:
     * 1. If the previous mean is finite and the new value is non-finite then the new mean is that
     *    value (whether it is NaN or infinity).
     * 2. If the new value is finite and the previous mean is non-finite then the mean is unchanged
     *    (whether it is NaN or infinity).
     * 3. If both the previous mean and the new value are non-finite and...

        }
      }
    }

    os << "  tflite::" << tflite_option_name << "Builder b(*fbb);\n";
    for (const auto &option : options)
      os << formatv("  b.add_{0}(std::move({0}));\n", option);
    os << "  return b.Finish();\n}\n";
  }
}

// For each TFLite op, emits a builder function that packs the TFLite op into
// the corresponding FlatBuffer object.
//

        .isWithin(ALLOWED_ERROR)
        .of(MANY_VALUES_MEAN);
    // For datasets of many double values created from an iterable, we test many combinations of
    // finite and non-finite values:
    for (ManyValues values : ALL_MANY_VALUES) {
      StatsAccumulator accumulator = new StatsAccumulator();
      StatsAccumulator accumulatorByAddAllStats = new StatsAccumulator();

// (The caller handles Inf before invoking fmtB.)
func (x *Float) fmtB(buf []byte) []byte {
	if x.form == zero {
		return append(buf, '0')
	}

	if debugFloat && x.form != finite {
		panic("non-finite float")
	}
	// x != 0

	// adjust mantissa to use exactly x.prec bits
	m := x.mant
	switch w := uint32(len(x.mant)) * _W; {
	case w < x.prec:
		m = nat(nil).shl(m, uint(x.prec-w))