const (
	_O_RDONLY   = 0x0
	_O_WRONLY   = 0x1
	_O_CREAT    = 0x40
	_O_TRUNC    = 0x200
	_O_NONBLOCK = 0x800
	_O_CLOEXEC  = 0x80000
)

type usigset struct {
	__val [16]uint64
}

type fpxreg struct {
	significand [4]uint16
	exponent    uint16
	padding     [3]uint16
}

type xmmreg struct {
	element [4]uint32
}

type fpstate struct {

// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
// THE SOFTWARE.

package x86

import (
	"cmd/internal/obj"
	"fmt"
)

var Register = []string{
	"AL", // [D_AL]
	"CL",
	"DL",
	"BL",
	"SPB",
	"BPB",
	"SIB",
	"DIB",
	"R8B",
	"R9B",
	"R10B",
	"R11B",
	"R12B",
	"R13B",
	"R14B",
	"R15B",
	"AX", // [D_AX]
	"CX",
	"DX",
	"BX",

          use_variable: If True, creates a variable for weight.
        """
        super(GatherModel, self).__init__()
        w_val = np.random.randn(128, 32).astype('f4')
        if use_variable:
          self.w = variables.Variable(w_val)
        else:
          self.w = w_val

      @def_function.function(
          input_signature=[
              tensor_spec.TensorSpec(

    // number without knowing this, so we need to manually dispatch to each
    // possible member of TensorProto, depening on its dtype.
#define MATCH(DTYPE, FIELD) \
  case DTYPE:               \
    return tensor.FIELD##_val().size()

  switch (tensor.dtype()) {
    MATCH(DT_FLOAT, float);
    MATCH(DT_DOUBLE, double);
    MATCH(DT_INT8, int);
    MATCH(DT_UINT8, int);
    MATCH(DT_INT16, int);
    MATCH(DT_UINT16, int);

  if (shape_x.size() == shape_y.size()) {
    llvm::SmallVector<int64_t, 4> out_shape(shape_x.size());
    for (int i = 0; i < shape_x.size(); i++) {
      auto x_val = shape_x[i];
      auto y_val = shape_y[i];
      out_shape[i] = std::max(x_val, y_val);
    }
    return tensorflow::GetTypeFromTFTensorShape(out_shape, element_type);
  }

  auto shape_large = shape_x.size() > shape_y.size() ? shape_x : shape_y;