SmallVector<int32_t> shape;
  shape.reserve(output_shape.size());

  bool needs_truncation = true;
  for (size_t dim_idx = 0; dim_idx < output_shape.size(); ++dim_idx) {
    int64_t dim = output_shape[dim_idx];
    if (truncate && needs_truncation && dim == 1) {
      continue;
    } else if (needs_truncation && dim != 1) {
      needs_truncation = false;
    }

    for (int64_t dim_idx = 0; dim_idx < value_rank; ++dim_idx) {
      int64_t dim_size = value_type.getDimSize(dim_idx);
      if (dim_size == ShapedType::kDynamic) {
        Value dim_i = rewriter.create<arith::IndexCastOp>(
            loc, shape_scalar_type,
            rewriter.create<tensor::DimOp>(loc, op.getOperand(), dim_idx));
        end_indices.push_back(dim_i);
        if (dim_idx != axis) {

  auto value_shape = value_shape_type.getShape();
  SmallVector<int32_t> return_squeeze_shape;
  return_squeeze_shape.reserve(value_shape.size());

  for (size_t dim_idx = 0; dim_idx < value_shape.size(); ++dim_idx) {
    int64_t dim = value_shape[dim_idx];
    if (dim == 1) {
      continue;
    }
    return_squeeze_shape.push_back(
        ShapedType::isDynamic(dim) ? -1 : static_cast<int32_t>(dim));
  }

  }

  const int64_t rank = input_type.getRank();

  IntegerAttr dim_int = ExtractSingleElementAsInteger(split_dim_attr);

  // "axis" operand could be a i64 tensor. Resolve it here.
  IntegerAttr dim_i32;
  if (failed(ConvertToI32Attr(dim_int, &dim_i32))) return failure();

  int dim = dim_i32.getInt();
  if (dim < 0) dim += rank;

  const SmallVector<Value, 4>& slice_outputs = SliceOutputs(

			delete(s.mm.pod.pmm, k)
		}
	}
	s.imm = nil
}

func (s *sstate) BeginCounterDataFile(cdf string, cdr *decodecounter.CounterDataReader, dirIdx int) {
	dbgtrace(2, "visiting counter data file %s diridx %d", cdf, dirIdx)
	if s.inidx != dirIdx {
		if s.inidx > dirIdx {
			// We're relying on having data files presented in
			// the order they appear in the inputs (e.g. first all

	m.mm.endPod(*pcombineflag)
}

func (m *mstate) BeginCounterDataFile(cdf string, cdr *decodecounter.CounterDataReader, dirIdx int) {
	dbgtrace(2, "visit counter data file %s dirIdx %d", cdf, dirIdx)
	m.mm.beginCounterDataFile(cdr)
}

func (m *mstate) EndCounterDataFile(cdf string, cdr *decodecounter.CounterDataReader, dirIdx int) {
}

func (m *mstate) VisitFuncCounterData(data decodecounter.FuncPayload) {
	m.mm.visitFuncCounterData(data)

    TFE_TensorDebugInfo* debug_info) {
  return debug_info->dev_dims.size();
}

TF_CAPI_EXPORT extern int64_t TFE_TensorDebugInfoOnDeviceDim(
    TFE_TensorDebugInfo* debug_info, int dim_index) {
  return debug_info->dev_dims[dim_index];
}

// Return the number of dimensions that the tensor has.
TF_CAPI_EXPORT extern int TF_NumDims(const TF_Tensor*);

// Return the length of the tensor in the "dim_index" dimension.
// REQUIRES: 0 <= dim_index < TF_NumDims(tensor)
TF_CAPI_EXPORT extern int64_t TF_Dim(const TF_Tensor* tensor, int dim_index);

// Return the size of the underlying data in bytes.
TF_CAPI_EXPORT extern size_t TF_TensorByteSize(const TF_Tensor*);

      int64_t dim_index = dim_index_attr.getValues<APInt>()[0].getSExtValue();

      int64_t rank = operand_ty.getRank();
      if (dim_index < 0 || dim_index >= rank) {
        return emitOptionalError(location, "dim_index (", dim_index,
                                 ") is out of range [0, ", rank, ")");
      }
      shape[dim_index] = ShapedType::kDynamic;
    } else {

		return 0, err
	}
	if f.entries == nil {
		return 0, EINVAL
	}

	n := 0
	for f.dirIdx < len(f.entries) {
		entry := f.entries[f.dirIdx]
		l := 2 + len(entry)
		if l > len(buf) {
			break
		}
		buf[0] = byte(l)
		buf[1] = byte(l >> 8)
		copy(buf[2:], entry)
		buf = buf[l:]
		n += l
		f.dirIdx++
	}

	return n, nil
}

func setStat(st *Stat_t, jsSt js.Value) {