offset += int(buf[2])
	}

	*t = Time{}
	t.wall = uint64(nsec)
	t.ext = sec

	if offset == -1*60 {
		t.setLoc(&utcLoc)
	} else if _, localoff, _, _, _ := Local.lookup(t.unixSec()); offset == localoff {
		t.setLoc(Local)
	} else {
		t.setLoc(FixedZone("", offset))
	}

	return nil
}

// TODO(rsc): Remove GobEncoder, GobDecoder, MarshalJSON, UnmarshalJSON in Go 2.

		name, offset, _, _, _ := local.lookup(t.unixSec())
		if offset == zoneOffset && (zoneName == "" || name == zoneName) {
			t.setLoc(local)
			return t, nil
		}

		// Otherwise create fake zone to record offset.
		zoneNameCopy := stringslite.Clone(zoneName) // avoid leaking the input value
		t.setLoc(FixedZone(zoneNameCopy, zoneOffset))
		return t, nil
	}

	if zoneName != "" {

          if (TF::TensorListSetItemOp set_op =
                  llvm::dyn_cast<TF::TensorListSetItemOp>(use.getOwner())) {
            element_shape = rewriter.create<TF::ShapeOp>(
                op.getLoc(),
                tensorflow::GetTypeFromTFTensorShape({-1}, shape_dtype),
                set_op.getItem());
            element_shape_acquired = true;

    // Calculates the broadcast shape using BroadcastArgs op.
    Value cond_shape = GetShape(cond, op->getLoc(), rewriter);
    Value lhs_shape = GetShape(lhs, op->getLoc(), rewriter);
    Value rhs_shape = GetShape(rhs, op->getLoc(), rewriter);
    auto broadcast_shape_value =
        rewriter
            .create<TF::BroadcastArgsOp>(op->getLoc(), lhs_shape.getType(),
                                         lhs_shape, rhs_shape)

  auto erf_op = rewriter.create<chlo::ErfOp>(mul_op.getLoc(), mul_op);
  auto add_op =
      rewriter.create<stablehlo::AddOp>(erf_op.getLoc(), erf_op, one_cst);
  auto lhs_mul_op = rewriter.create<stablehlo::MulOp>(
      half_cst.getLoc(), new_func.getArguments()[0], half_cst);
  auto output_mul_op = rewriter.create<stablehlo::MulOp>(lhs_mul_op.getLoc(),

    auto range = rewriter.create<mhlo::SubtractOp>(op.getLoc(), limit, start);
    auto abs = rewriter.create<mhlo::AbsOp>(op.getLoc(), range);

    // Delta is not necessarily the same type as start and limit.
    auto abs_cast =
        rewriter.create<mhlo::ConvertOp>(op.getLoc(), compute_type, abs);
    auto delta_cast =
        rewriter.create<mhlo::ConvertOp>(op.getLoc(), compute_type, delta);

  builder.insert(var_handle_op_in_init);
  auto const_op = builder.create<mlir::arith::ConstantOp>(
      session_init_func.getLoc(), tensor_attr.getType(), tensor_attr);

  builder.create<TF::AssignVariableOp>(
      session_init_func.getLoc(), llvm::ArrayRef<mlir::Type>{},
      llvm::ArrayRef<mlir::Value>{var_handle_op_in_init->getResult(0),
                                  const_op.getResult()});
}

    }
    operand = rewriter.create<mhlo::BitcastConvertOp>(op->getLoc(),
                                                      *output_type, operand);

    Value res_min_clipped = rewriter.create<chlo::BroadcastMaxOp>(
        op->getLoc(), *output_type, operand, *min_or, broadcast_dims);
    Value res_max_clipped = rewriter.create<chlo::BroadcastMinOp>(
        op->getLoc(), *output_type, res_min_clipped, *max_or, broadcast_dims);

        rewriter.create<arith::ConstantOp>(op->getLoc(), type, begin_attr);
    auto end_attr = DenseElementsAttr::get<int32_t>(type, padded_end);
    auto end_op =
        rewriter.create<arith::ConstantOp>(op->getLoc(), type, end_attr);
    auto stride_attr = DenseElementsAttr::get<int32_t>(type, padded_stride);
    auto stride_op =
        rewriter.create<arith::ConstantOp>(op->getLoc(), type, stride_attr);

      block_arg.getLoc(),
      DenseIntElementsAttr::get(
          RankedTensorType::get({1, num_replicas}, builder.getIntegerType(32)),
          group_assignment_val));

  StringAttr reduce_op = builder.getStringAttr("Add");
  StringAttr mode = builder.getStringAttr("CrossReplica");
  return builder.create<XlaAllReduceOp>(block_arg.getLoc(), block_arg.getType(),