Download the full version of the textbook now at textbookfull.

com

Hardware and Software Verification and

Testing 12th International Haifa Verification
Conference HVC 2016 Haifa Israel November 14
17 2016 Proceedings 1st Edition Roderick
Bloem
https://textbookfull.com/product/hardware-and-
software-verification-and-testing-12th-
international-haifa-verification-conference-
hvc-2016-haifa-israel-
november-14-17-2016-proceedings-1st-edition-
roderick-bloem/

Explore and download more textbook at https://textbookfull.com

 Recommended digital products (PDF, EPUB, MOBI) that
you can download immediately if you are interested.

Business Process Management 12th International Conference

BPM 2014 Haifa Israel September 7 11 2014 Proceedings 1st
Edition Shazia Sadiq
https://textbookfull.com/product/business-process-management-12th-
international-conference-bpm-2014-haifa-israel-
september-7-11-2014-proceedings-1st-edition-shazia-sadiq/
textbookfull.com

Computer Aided Verification 28th International Conference

CAV 2016 Toronto ON Canada July 17 23 2016 Proceedings
Part II 1st Edition Swarat Chaudhuri
https://textbookfull.com/product/computer-aided-verification-28th-
international-conference-cav-2016-toronto-on-canada-
july-17-23-2016-proceedings-part-ii-1st-edition-swarat-chaudhuri/
textbookfull.com

Numerical Software Verification 9th International Workshop

NSV 2016 Toronto ON Canada July 17 18 2016 Revised
Selected Papers 1st Edition Sergiy Bogomolov
https://textbookfull.com/product/numerical-software-verification-9th-
international-workshop-nsv-2016-toronto-on-canada-
july-17-18-2016-revised-selected-papers-1st-edition-sergiy-bogomolov/
textbookfull.com

Physiological and Biotechnological Aspects of

Extremophiles 1st Edition Richa Salwan

https://textbookfull.com/product/physiological-and-biotechnological-
aspects-of-extremophiles-1st-edition-richa-salwan/

textbookfull.com
Infection Control in Primary Dental Care Martin R. Fulford

https://textbookfull.com/product/infection-control-in-primary-dental-
care-martin-r-fulford/

textbookfull.com

Heroes and Heroism in British Fiction Since 1800: Case

Studies 1st Edition Barbara Korte

https://textbookfull.com/product/heroes-and-heroism-in-british-
fiction-since-1800-case-studies-1st-edition-barbara-korte/

textbookfull.com

Introduction to Google Analytics: A Guide for Absolute

Beginners Todd Kelsey

https://textbookfull.com/product/introduction-to-google-analytics-a-
guide-for-absolute-beginners-todd-kelsey/

textbookfull.com

Proactive Images for Pre-Service Teachers: Identity,

Expectations, and Avoiding Practice Shock Jeremy
Delamarter
https://textbookfull.com/product/proactive-images-for-pre-service-
teachers-identity-expectations-and-avoiding-practice-shock-jeremy-
delamarter/
textbookfull.com

Black Scholarly Activism between the Academy and

Grassroots A Bridge for Identities and Social Justice
Ornette D. Clennon
https://textbookfull.com/product/black-scholarly-activism-between-the-
academy-and-grassroots-a-bridge-for-identities-and-social-justice-
ornette-d-clennon/
textbookfull.com
Buttermilk Graffiti A Chef s Journey to Discover America s
New Melting Pot Cuisine Edward Lee

https://textbookfull.com/product/buttermilk-graffiti-a-chef-s-journey-
to-discover-america-s-new-melting-pot-cuisine-edward-lee/

textbookfull.com
 Roderick Bloem
Eli Arbel (Eds.)
LNCS 10028

Hardware and Software:

Verification and Testing
12th International Haifa Verification Conference, HVC 2016
Haifa, Israel, November 14–17, 2016
Proceedings

123
Lecture Notes in Computer Science 10028
Commenced Publication in 1973
Founding and Former Series Editors:
Gerhard Goos, Juris Hartmanis, and Jan van Leeuwen

Editorial Board
David Hutchison
Lancaster University, Lancaster, UK
Takeo Kanade
Carnegie Mellon University, Pittsburgh, PA, USA
Josef Kittler
University of Surrey, Guildford, UK
Jon M. Kleinberg
Cornell University, Ithaca, NY, USA
Friedemann Mattern
ETH Zurich, Zurich, Switzerland
John C. Mitchell
Stanford University, Stanford, CA, USA
Moni Naor
Weizmann Institute of Science, Rehovot, Israel
C. Pandu Rangan
Indian Institute of Technology, Madras, India
Bernhard Steffen
TU Dortmund University, Dortmund, Germany
Demetri Terzopoulos
University of California, Los Angeles, CA, USA
Doug Tygar
University of California, Berkeley, CA, USA
Gerhard Weikum
Max Planck Institute for Informatics, Saarbrücken, Germany
More information about this series at http://www.springer.com/series/7408
Roderick Bloem Eli Arbel (Eds.)
•

Hardware and Software:

Verification and Testing
12th International
Haifa Verification Conference, HVC 2016
Haifa, Israel, November 14–17, 2016
Proceedings

123
Editors
Roderick Bloem Eli Arbel
IAIK IBM Research Labs
Graz University of Technology Haifa
Graz Israel
Austria

ISSN 0302-9743 ISSN 1611-3349 (electronic)

Lecture Notes in Computer Science
ISBN 978-3-319-49051-9 ISBN 978-3-319-49052-6 (eBook)
DOI 10.1007/978-3-319-49052-6

Library of Congress Control Number: 2016956611

LNCS Sublibrary: SL2 – Programming and Software Engineering

© Springer International Publishing AG 2016

This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of the
material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation,
broadcasting, reproduction on microfilms or in any other physical way, and transmission or information
storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now
known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication
does not imply, even in the absence of a specific statement, that such names are exempt from the relevant
protective laws and regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this book are
believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors
give a warranty, express or implied, with respect to the material contained herein or for any errors or
omissions that may have been made.

Printed on acid-free paper

This Springer imprint is published by Springer Nature

The registered company is Springer International Publishing AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
 Preface

This volume contains the proceedings of the 12th Haifa Verification Conference (HVC
2016). The conference was hosted by IBM Research Haifa Laboratory and took place
during November 14–17, 2016. It was the 12th event in this series of annual confer-
ences dedicated to advancing the state of the art and state of the practice in verification
and testing. The conference provided a forum for researchers and practitioners from
academia and industry to share their work, exchange ideas, and discuss the future
directions of testing and verification for hardware, software, and complex hybrid
systems.
Overall, HVC 2016 attracted 26 submissions in response to the call for papers. Each
submission was assigned to at least three members of the Program Committee and in
some cases additional reviews were solicited from external experts. The Program
Committee selected 13 papers for presentation. In addition to the contributed papers,
the program included four invited talks, by Swarat Chaudhuri (Rice University),
Markulf Kohlweiss (Microsoft Research), Rajeev Ranjan (Cadence), and Andreas
Veneris (University of Toronto). On the last day of the conference, the HVC award was
presented to Marta Kwiatkowska (University of Oxford), Gethin Norman (University
of Glasgow), and Dave Parker (University of Birmingham), for the invention, devel-
opment and maintenance of the PRISM probabilistic model checker. A special session
about verification and testing challenges of autonomous systems was held on the first
day of the conference. Thanks to Yoav Hollander (Foretellix LTD) for presenting in
this session. On November 13, one day before the conference, we held a tutorial day
with tutorials by Sanjit A. Seshia (University of California, Berkeley) on formal
inductive synthesis, by Hari Mony (IBM) on sequential equivalence checking for
hardware design and verification, by Amir Rahat (Optima Design Automation) on
design reliability, and by Cristian Cadar (Imperial College) on dynamic symbolic
execution and the KLEE infrastructure.
We would like to extend our appreciation and sincere thanks to the local organi-
zation team from IBM Research Haifa Laboratory: Tali Rabetti, the publicity chair,
Revivit Yankovich, the local coordinator, Yair Harry, the Web master, and the
Organizing Committee, which consisted of Laurent Fournier, Sharon Keidar-Barner,
Moshe Levinger, Michael Vinov, Karen Yorav, and Avi Ziv. We would also like to
thank the tutorial chair Natasha Sharygina (University of Lugano), and the HVC Award
Committee, consisting of Armin Biere (Johannes Kepler University), Hana Chockler
(King’s College London), Kerstin Eder (University of Bristol), Andrey Rybalchenko
(Microsoft Research), Ofer Strichman (Technion), and particularly its energetic chair,
Leonardo Mariani (University of Milano Bicocca).
HVC 2016 received sponsorships from IBM, Cadence Design Systems, Mellanox
Technologies, Mentor Graphics, Qualcomm, and Intel. (Thanks!)
VI Preface

Submission and evaluation of papers, as well as the preparation of this proceedings

volume, were handled by the EasyChair conference management system. (Thanks,
Andrei!)
It was a pleasure to organize this conference with so many old friends!

Graz Eli Arbel

September 2016 Roderick Bloem
 Organization

Program Committee
Eli Arbel IBM Research, Israel
Domagoj Babic Google, USA
Aviv Barkai Intel Corporation, Israel
Nikolaj Bjorner Microsoft Research, USA
Roderick Bloem Graz University of Technology, Austria
Hana Chockler King’s College London, UK
Rayna Dimitrova MPI-SWS, Germany
Adrian Evans iRoC Technologies, France
Franco Fummi University of Verona, Italy
Raviv Gal IBM Research, Israel
Warren Hunt University of Texas, USA
Barbara Jobstmann EPFL and Cadence Design Systems, Switzerland
Laura Kovacs Vienna University of Technology, Austria
João Lourenço NOVA LINCS – Universidade Nova de Lisboa, Portugal
Annalisa Massini Sapienza University of Rome, Italy
Hari Mony IBM Corporation, USA
Nir Piterman University of Leicester, UK
Pavithra Prabhakar Kansas State University, USA
Sandip Ray NXP Semiconductors, USA
Orna Raz HRL, IBM Research, Israel
Martina Seidl Johannes Kepler University Linz, Asutria
Sanjit A. Seshia UC Berkeley, USA
A. Prasad Sistla University of Illinois at Chicago, USA
Ufuk Topcu University of Texas at Austin, USA
Eran Yahav Technion, Israel

Additional Reviewers

Arechiga, Nikos Krakovski, Roi Sadigh, Dorsa

Dreossi, Tommaso Lal, Ratan Salvo, Ivano
Fremont, Daniel J. Mari, Federico Soto, Miriam Garcia
Gao, Sicun Rabe, Markus N. Veneris, Andreas
Junges, Sebastian Rabetti, Tali
Abstracts
 Current Trends and Future Direction
in Eco-system of Hardware Formal
Verification: A Technical
and Business Perspective

Rajeev K. Ranjan

Cadence, San Jose, USA

Hardware formal verification is increasingly being adopted in the modern SoC

design and verification flow for architectural specification and verification
through RTL development and debugging through SoC integration – all the way
up to post-silicon debugging. The productivity and quality benefits of adopting
this technology for a gamut of verification tasks are well established. In this talk,
we will cover the current trends and future directions in this area that is shaped
by the technical feasibility of the solutions and the business RoI seen by dif-
ferent stakeholders V- chip companies, design/verification engineers, formal
EDA vendors, and formal solution development engineers.
Visit https://textbookfull.com
now to explore a rich
collection of eBooks, textbook
and enjoy exciting offers!
 Guiding Formal Methods
with Discovered Knowledge

Swarat Chaudhuri

Rice University, Houston, USA

Systems for automated formal reasoning about programs depend on human

specification at multiple levels. Users of such a system must write full specifi-
cations of the tasks that they want performed. The designer of the system is
expected to formalize the domain-specific language in which tasks are descri-
bed, and specify the domain-dependent heuristics that guide automated rea-
soning. The assumption of specifications reflects a common expectation in
formal methods research: that humans hold deep knowledge about problem
domains and instances. In practice, this expectation can be violated and lead to
hard-to-use or brittle tools. In this talk, we describe a new class of formal
methods that address this difficulty through automatic discovery of knowledge
from corpora of pre-existing code, execution traces, and proofs.
The starting point of this work is the observation that a human who describes
or solves a reasoning task does not do so in a vacuum, but using insights
established through prior experiences of their own or others. The thesis is that
such insights can be algorithmically learned from datasets of existing formal
artifacts, and can lead to systems for automated reasoning that demand less
human intervention than traditional tools. The talk will describe multiple
instantiations of this thesis, including a statistical notion of program correctness,
a program synthesis algorithm guided by a “neural” model of program syntax
and semantics, and an approach to program verification that uses pre-existing
formal proofs.
 Bug Wars: Automation Awakens

Andreas Veneris

Department of Electrical and Computer Engineering, and Department of

Computer Science, University of Toronto, Toronto, Canada

Verification is the undisputed bottleneck in the design cycle consuming two

thirds of the total chip development effort. This is in part because of the com-
plexity of modern designs, the impact of geographical dispersed teams inte-
grating new components with third-party/legacy IP under tight time-to-market
goals, the evolving role of verification engineers to not only discover bugs but
also aid correct them and the ever-evolving nature of the task itself. Today
verification has stretched itself beyond its traditional boundaries into validation
as most of silicon re-spins are not due to physical defects but because of
functional errors not discovered or fixed earlier in the design cycle. Although
parts of verification have been automated the core issue driving this gap remains
debugging as it consumes more than half of the overall effort being a pre-
dominantly manual task.
In this talk, we revisit automation efforts in functional debug from late 1980s,
when it was first introduced, to recent formal advances placing it into context as
we recount new directions. In more detail, we will first outline early method-
ologies stemming from test and fault diagnosis to more recent techniques based
on Boolean satisfiability. We will examine different angles of the debug problem
and respective solutions for its various manifestations in the verification cycle.
This will allow us to appraise theoretical and practical parallels in the founda-
tions of those two tasks. As we evaluate the progress in debug automation, we
will point out emerging deficiencies of existing methodologies more notably
during regression verification. To that end, we will present novel techniques in
debugging triage where statistical solutions, a radical departure from existing
debug efforts, need complement traditional simulation/formal methods to not
only take into account the design response but also the human factor. We will
conclude with a mantra that research in debugging in the past 30 years points to
a direction where verification and test prove once again to be fields deeply
intertwined, and we will provide guidance for methodologies in silicon debug
rooted on existing functional debug procedures.
miTLS: Can Cryptography, Formal Methods,
and Applied Security be Friends?

Markulf Kohlweiss

Microsoft Research, Cambridge, UK

TLS was designed as a transparent channel abstraction to allow developers with

no cryptographic expertise to protect their application against attackers that may
control some clients, some servers, and may have the capability to tamper with
network connections. However, the security guarantees of TLS fall short of
those of a secure channel, leading to a variety of attacks. The Triple Handshake
attack exploits combinations of RSA and Diffie-Hellman key exchange, session
resumption, and renegotiation to bypass many recent countermeasures.
At the same time we study the provable security of TLS, as it is implemented
and deployed. To capture the details of the standard and its main extensions, we
rely on miTLS, a verified reference implementation of the protocol. miTLS
inter-operates with mainstream browsers and servers for many protocol versions,
configurations, and ciphersuites; and it provides application-level, provable
security for some. This leads to the strange case of how something provable
secure can be insecure.
In this talk I will play Dr Jekyll and Mr Hyde by playing off our CRYPTO
proof and our S&P attack against each other.
 Contents

SAT-Based Combinational and Sequential Dependency Computation. . . . . . . 1

Mathias Soeken, Pascal Raiola, Baruch Sterin, Bernd Becker,
Giovanni De Micheli, and Matthias Sauer

Multi-core SCC-Based LTL Model Checking . . . . . . . . . . . . . . . . . . . . . . . 18

Vincent Bloemen and Jaco van de Pol

Gating Aware Error Injection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34

Eli Arbel, Erez Barak, Bodo Hoppe, Shlomit Koyfman, Udo Krautz,
and Shiri Moran

ddNF: An Efficient Data Structure for Header Spaces . . . . . . . . . . . . . . . . . 49

Nikolaj Bjørner, Garvit Juniwal, Ratul Mahajan, Sanjit A. Seshia,
and George Varghese

Probabilistic Fault Localisation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65

David Landsberg, Hana Chockler, and Daniel Kroening

Iterative User-Driven Fault Localization . . . . . . . . . . . . . . . . . . . . . . . . . . . 82

Xiangyu Li, Marcelo d’Amorim, and Alessandro Orso

Improving Efficiency and Accuracy of Formula-Based Debugging . . . . . . . . 99

Wei Jin and Alessandro Orso

Improving Priority Promotion for Parity Games . . . . . . . . . . . . . . . . . . . . . 117

Massimo Benerecetti, Daniele Dell’Erba, and Fabio Mogavero

Synthesis of Admissible Shields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134

Laura Humphrey, Bettina Könighofer, Robert Könighofer,
and Ufuk Topcu

Probabilistic Hybrid Systems Verification via SMT and Monte

Carlo Techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152
Fedor Shmarov and Paolo Zuliani

Formula Slicing: Inductive Invariants from Preconditions . . . . . . . . . . . . . . . 169

Egor George Karpenkov and David Monniaux
XVI Contents

Advancing Software Model Checking Beyond Linear Arithmetic Theories . . . 186

Ahmed Mahdi, Karsten Scheibler, Felix Neubauer, Martin Fränzle,
and Bernd Becker

Predator Shape Analysis Tool Suite. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202

Lukáš Holík, Michal Kotoun, Petr Peringer, Veronika Šoková,
Marek Trtík, and Tomáš Vojnar

Author Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211

 SAT-Based Combinational and Sequential
Dependency Computation

Mathias Soeken1(B) , Pascal Raiola2 , Baruch Sterin3 , Bernd Becker2 ,

Giovanni De Micheli1 , and Matthias Sauer2
1
EPFL, Lausanne, Switzerland
mathias.soeken@epfl.ch
2
University of Freiburg, Freiburg im Breisgau, Germany
3
UC Berkeley, Berkeley, CA, USA

Abstract. We present an algorithm for computing both functional

dependency and unateness of combinational and sequential Boolean func-
tions represented as logic networks. The algorithm uses SAT-based tech-
niques from Combinational Equivalence Checking (CEC) and Automatic
Test Pattern Generation (ATPG) to compute the dependency matrix of
multi-output Boolean functions. Additionally, the classical dependency
definitions are extended to sequential functions and a fast approximation
is presented to efficiently yield a sequential dependency matrix. Exten-
sive experiments show the applicability of the methods and the improved
robustness compared to existing approaches.

1 Introduction
In this paper we present an algorithm to compute the dependency matrix D(f )
for a given combinational or sequential multi-output function f . For every input-
output pair, the combinational dependency matrix indicates whether the output
depends on the input, and whether the output is positive or negative unate in
that input [21].
Several algorithms in logic design use the dependency matrix as a signa-
ture [23] to speed up computation, e.g., Boolean matching [14], functional veri-
fication [11,12], or reverse engineering [29]. Although most of these algorithms
make implicit use of the dependency matrix, the name has been used in this paper
for the first time. The name is inspired by the output format of functional depen-
dence and unateness properties in the state-of-the-art academic logic synthesis
tool ABC [4]. Functional dependency is also related to transparent logic [19,24].
Given a set of inputs Xd and a set of outputs Yd , the problem is to find a set
of inputs Xc that is disjoint from Xd and that distinguishes the output values
at Yd for different input assignments to Xd . In contrast, we consider functional
dependence without constraints for all input-output pairs.
Existing algorithms for computing the dependency matrix are based on
Binary Decision Diagrams (BDDs, [5]) and have been implemented in ABC [4].
It is important to point out that the term functional dependence is used

c Springer International Publishing AG 2016
R. Bloem and E. Arbel (Eds.): HVC 2016, LNCS 10028, pp. 1–17, 2016.
DOI: 10.1007/978-3-319-49052-6 1
2 M. Soeken et al.

to describe a different property in a related context: In [12,13,16], the

authors refer to functional dependence as the question whether given a set
of Boolean functions {f1 , . . . , fn }, there exists an fi , that can be written as
h(f1 , . . . , fi−1 , fi+1 , . . . , fn ). In other words, functional dependence is defined as
a Boolean function w.r.t. to a set of Boolean functions. In contrast, we con-
sider the functional dependence of a Boolean function w.r.t. a single variable as
functional dependence.
Our algorithm uses techniques from Combinational Equivalence Checking
(CEC, e.g., [22]) and Automatic Test Pattern Generation (ATPG, e.g., [15,26,
27]). We employ efficient incremental SAT-based solving techniques and extract
incidental information from solved instances to reduce runtime consumption on
complex functions.
We furthermore present an extension of the combinational dependency defi-
nition to sequential functions. We account the sequential counterpart of a func-
tional dependence relation to an input-output pair if the given relation constantly
holds after some finite number of steps. As an example, some output f may be
always positive unate in some input x after a certain number of iteration steps
of the circuit. In this case, we call f sequential positive unate in x, even if this
relation is not guaranteed in the first steps.
An established method to prove properties on sequential circuits is bounded
model checking (BMC) as first introduced in [1], used, e.g., in [8,25]. In BMC
a circuit is modelled iteratively for k steps as a combinational circuit. With
approaches such as k-induction [18] and Craig interpolation [20] BMC becomes
a complete model checking method. However, as such complete methods are
rather computationally expensive, we rely on an iterative approximation to com-
pute the sequential dependency matrix solely based on the combinational depen-
dency matrix. By iteratively analyzing the combinational dependency until a
fixed point is derived, we can accurately conclude structural dependency and
unateness.
In an extensive experimental evaluation we demonstrate the applicability of
our methods to various combinational and sequential benchmark sets. Within
reasonable amounts of computing time we are able to accurately compute the
combinational dependency matrix as well as an approximation of our sequential
dependency matrix with a small number of iterations. We further show the
robustness of our proposed algorithm compared to a previous state-of-the-art
algorithm that times out or suffers from memory explosion on complex functions.
Finally, we present a case study in which the dependency matrix is used as
a signature in reverse engineering to effectively reduce the search space and
improve the performance of the underlying application.
The rest of the paper is organized as follows. Section 2 presents the fundamen-
tals of the work. In Sect. 3 we introduce our SAT-based approach to compute
the dependency matrix of combinational circuits, and extend it in Sect. 4 to
sequential circuits. The experimental results are presented in Sect. 5 and Sect. 6
concludes the work.
 SAT-Based Combinational and Sequential Dependency Computation 3

2 Background
2.1 Functional Dependencies

A Boolean function f (x1 , . . . , xn ) is functionally dependent in xi if fx̄i = fxi

where the co-factors fxi or fx̄i are obtained by setting xi to 1 or 0 in f , respec-
tively. We call fxi the positive co-factor and fx̄i the negative co-factor. The
function f is said to be positive unate in xi , if

fx̄i ≤ fxi (1)

and negative unate in xi , if

fx̄i ≥ fxi , (2)
where the comparisons ‘≤’ and ‘≥’ are applied to the binary strings that rep-
resent the truth tables of fx̄i and fxi . f is said to be unate in xi if it is either
positive or negative unate in xi . Clearly, a function f is both positive and neg-
ative unate in xi , if f does not depend on xi . Hence, we call f strictly positive
(negative) unate in xi , if f is positive (negative) unate in xi and depends on xi .
If f is neither positive nor negative unate in xi , we say that f is binate in xi .

Example 1. The functions x1 ∧ x2 and x1 ∨ x2 are positive unate in both x1 and

x2 . The function x1 → x2 is negative unate in x1 and positive unate in x2 . The
function x1 ⊕ x2 is binate in both variables.

Let f : Bn → Bm be a multi-output Boolean function where each output is

represented by a Boolean function fj (x1 , . . . , xn ). The dependency matrix D(f )
is an m × n matrix with entries dj,i where
⎧
⎪
⎪p if fj is strictly positive unate in xi ,
⎪
⎨n if fj is strictly negative unate in xi ,
dj,i = (3)
⎪
⎪ d if fj depends on, but is not unate in xi ,
⎪
⎩
• otherwise.

Example 2. Let f : B5 → B3 with f1 = x1 ∧ x2 , f2 = x3 → x5 , and f3 =

x1 ⊕ x2 ⊕ x5 . Then ⎡ ⎤
pp•••
D(f ) = ⎣• • n • p⎦ .
dd••d

2.2 Boolean Satisfiability

In our algorithm we translate decision problems into instances of the SAT prob-
lem [3]. SAT is the problem of deciding whether a function f , has an assignment
x for which f (x) = 1. Such an assignment is called a satisfying assignment. If f
has a satisfying assignment it is said to be satisfiable. Otherwise, f is said to be
unsatisfiable.
4 M. Soeken et al.

In general, SAT is NP-complete [6,17]. SAT solvers are algorithms that can
solve SAT problems and, while worst-case exponential, are nonetheless very effi-
cient for many practical problems. SAT solvers also return a satisfying assign-
ment if the instance is satisfiable. Most of the state-of-the-art SAT solvers are
conflict-driven and employ clause-learning techniques [10]. In incremental SAT
one asks whether f is satisfiable under the assumption of some variable assign-
ments. These assignments are only temporarily assumed, making it possible to
reuse the SAT instance and learned information when solving a sequence of sim-
ilar SAT problems. In the remainder of the paper, we refer to instances of SAT
as if they were calls to an incremental SAT solver. SAT?(f, α) is true if f is sat-
isfiable under the assumptions α, and UNSAT?(f, α) is true if f is unsatisfiable
under the assumptions α.

3 SAT-Based Dependency Computation

This section presents the SAT-based algorithm to compute the functional depen-
dencies of a function. We first describe the encoding into SAT, then an imple-
mentation of the algorithm, and finally possible optimizations.

3.1 SAT Encoding

We encode the test for functional dependence and unateness as an instance of
the SAT problem using the following theorem.

Theorem 1. Let f (x1 , . . . , xn ) be a Boolean function. Then

1. f is functionally dependent in xi , if and only if fx̄i ⊕ fxi is satisfiable,
2. f is positive unate in xi , if and only if fx̄i ∧ f¯xi is unsatisfiable, and
3. f is negative unate in xi , if and only if fxi ∧ f¯x̄i is unsatisfiable.

Proof. We only show the direction of “if”; the “only if” direction follows imme-
diately from the definition of functional dependency and unateness.
1. Let x be a satisfying assignment to fx̄i ⊕ fxi . Then, we have fx̄i (x) = fxi (x).
2. Assume the function was satisfiable and let x be a satisfying assignment.
Then fx̄i (x) = 1 while fxi (x) = 0 which contradicts Eq. (1).
3. Analogously to (2). 

In the implementation, we make use of the following immediate consequence of

the theorem.
Corollary 1. f is functionally independent in xi , if and only if fx̄i ⊕ fxi is
unsatisfiable.
In the following we consider multi-output functions f : Bn → Bm , where each
output is a function fj . In order to compute the full dependency matrix which
contains the dependency for each input-output pair, we transform the problem
to a sequence of SAT instances as illustrated by the generic miter in Fig. 1. The
 SAT-Based Combinational and Sequential Dependency Computation 5

e1
..
.
en
p1
x1 1
.. .. d1
. fx̄i .
xn m n1

x1 1 pm
.. ..
. fxi . dm
xn m
nm

Fig. 1. Generic miter to encode functional dependency as SAT instance

two boxes represent two copies of f . The upper copy, with inputs x1 , . . . , xn ,
is used as the negative co-factor, while the lower copy, with inputs x1 , . . . , xn ,
is used as the positive co-factor of f . The groups of three gates on the lower
right hand side realize the XOR operation which connect the outputs of the two
copies. The signals of the AND gates are exposed as outputs and will be used
to encode the unateness problems. The XNOR gates in the upper right of the
figure are used to force all but one of the inputs, to have equal values.
Let Π(fj ) be the characteristic Boolean function which is obtained by encod-
ing the miter in Fig. 1 for the function fj using encodings such as Tseytin [32]
or EMS [9]. Also let Ei = {xi = 0, xi = 1} ∪ {ek = 1 | k = i} be assignments
that lead to a correct interpretation of the miter for input xi , i.e., xi is set to 0,
xi is set to 1 and all the other inputs need to have the same value. We define
three problems on top of the incremental SAT interface:

DEP(fj , xi ) = SAT?(Π(fj ), Ei ∪ {dj = 1}) (4)

POS UNATE(fj , xi ) = UNSAT?(Π(fj ), Ei ∪ {pj = 1}) (5)
NEG UNATE(fj , xi ) = UNSAT?(Π(fj ), Ei ∪ {nj = 1}) (6)

Then the problems described in Theorem 1 and Corollary 1 can be solved

as follows. The function fj functionally depends on xi , if DEP(fj , xi ) holds.
And the function fj is positive (negative) unate in xi , if POS UNATE(fj , xi )
(NEG UNATE(fj , xi )) holds.

3.2 Algorithm

Figure 2 displays the general flow of the algorithm. For each pair of an input
x = xi and an output y = fj the algorithm starts with a simple structural
dependency check. If x is outside of y’s structural cone of influence, it can be
concluded that y is independent of x. This is a very efficient check. Otherwise,
Visit https://textbookfull.com
now to explore a rich
collection of eBooks, textbook
and enjoy exciting offers!
6 M. Soeken et al.

Fig. 2. Functional dependency and unateness computation flow

the algorithm proceeds with a functional dependency check DEP(y, x) as defined

in Eq. (4). We omit the arguments from the boxes.
If the instance is unsatisfiable, y is independent from x as no assignment exists
that results in different logic values for y under the side constraint of yx̄ ⊕ yx .
In case the instance is satisfiable, x and y are at least functionally dependent.
Additionally, the SAT solver returns a satisfying assignment which is analyzed
for the logic value of y. In case yx̄ is 1 (and therefore yx is 0), y cannot be
positive unate in x as a counter example for Eq. (1) is found. Likewise, negative
unateness can be falsified if yx̄ is 0. Note that one of the two cases must hold as
the original instance requires a difference between yx̄ and yx .
In a last step, the algorithm specifically checks for unateness with an addi-
tional call to the SAT solver, unless it has been ruled out previously. If this SAT
call is unsatisfiable, unateness can be concluded, otherwise the algorithm returns
functional dependence.

3.3 Optimizations

As discussed above, we use incremental SAT solving because many of the calls to
the SAT solver are very similar. Hence, instead of encoding a miter-like structure
as illustrated in Fig. 1 for each input-output pair in an individual instance, we
encode the complete output cone of a target input xi in a single instance to profit
from incremental SAT solving. We enforce the co-factors of xi as unit clauses in
this instance. As we target combinational circuits, the direction of the co-factors
does not influence the satisfiability of the instance. Hence, we can restrict the
search space by enforcing xi to logic 1 and xi to logic 0 without loss of generality.
Furthermore, XOR gates are encoded for output pairs to enforce them to differ
using assumptions in the SAT solver.
 SAT-Based Combinational and Sequential Dependency Computation 7

On the output side, we iteratively run through each output in x’s cone of
influence and enforce a difference between fx̄i and fxi using an assumption. If
the resulting instance is UNSAT we can conclude independence. Otherwise, the
input-output pair is at least functionally dependent. By observing the direction
of the difference at the output, we consider the pair either as a candidate for
positive or negative unateness and run the respective check as described earlier.1
Additionally, we perform a forward looking logic analysis of each satisfiable
SAT instance to extract incidental information for future solver calls. In our
experiments we found quite often, that the difference not only propagates to
the target output, but also to multiple other outputs. Hence, we check the logic
values of all following outputs as well. Additionally, an output may incidentally
show differences in both directions and hence unateness can be ruled out without
additional SAT calls.
The described SAT instances are very similar to detecting a stuck-at-1 fault
at the input x. Hence, we employ encoding based speed-up techniques that are
known from solving such ATPG problems. By adding D-chains [31] that add
redundant information to the instance, the SAT solver can propagate the dif-
ferences more easily. Additionally, we tuned the SAT solver’s internal settings
towards the characteristics of the circuit-based SAT instances which are domi-
nated by a large number of rather simple SAT calls. For instance, we do not use
preprocessing techniques on the SAT instances.

4 Sequential Functional Dependency

While the prior definitions and algorithms are specified for combinational cir-
cuits, we also investigate the definition of dependency in sequential circuits.
To translate the properties from Sect. 2.1 to sequential circuits, we use a
similar approach as used in (un)bounded model checking: An output yj is called
sequential functionally dependent in an input xi if and only if there exists a num-
(k) (k)
ber k ∈ N, such that fj is functionally dependent in xi , where fj represents
the Boolean function of the output modelled over k steps.
(1) (k)
For fj the sequential circuit can be treated as a combinational one. For fj
with k > 1, the definition of sequential dependence follows the combinational one
if the sequential circuit is considered as an iterative logic array with an unlimited
number of time frames. Hence, such a definition allows to extend combinational
dependency in a natural way to sequential dependency.
In contrast to the complexity of the combinational dependency computa-
tion of a single input-output pair (which is NP-complete since it is an ATPG
problem), sequential dependency computation is identical to invariant check-
ing which can be expressed by an unbounded model checking approach and is
PSPACE-complete.
1
We also performed a structural check for each input-output pair if there potentially
exists an inverting path between them. If this is not the case, the additionally SAT-
call to check for unateness may be skipped. However, the performance impact was
insignificant and hence we did not employ this optimization.
8 M. Soeken et al.

A C
B

Fig. 3. Example circuit

Sequential independence is defined as the contrary of sequential dependence.

An output yj is called (strictly) sequential positive/negative unate in xi , if there
(k)
exists a k0 , such that for every number k ∈ N with k > k0 , fj is (strictly)
positive/negative unate in xi .

Example 3. Let f (k) be the Boolean function corresponding to the flip flop C in
the circuit in Fig. 3 in the k th step. Then C is alternating strictly positive and
negative unate in A. Thus, C is neither sequential positive nor negative unate
in A. However, C is sequential dependent in A.

4.1 Approximative Algorithm

We use the methods from Sect. 3 to compute the combinational dependency
matrix D(f ) for f (0) and then initialize the sequential dependency Matrix Ds (f )
as D(f ). For clarity, to refer to an entry of the dependency matrix with output
yj and input xi we write dyj ,xi instead of dj,i . Respective entries of the sequential
dependency matrix are denoted as dsj,i .
For each output y = fj and input x we check, if there exist xk and yl , such
that
– xk and yl correspond to the same flip flop ϕ,
– dsy,xk = • and
– dsyl ,x = •.
The path-dependence of y in x over ϕ is defined with the equation
⎧
⎪
⎨p if dy,xk , dyl ,x unate and dy,xk = dyl ,x ,
pdϕ (y, x) = n if dy,xk , dyl ,x unate and dy,xk = dyl ,x , (7)
⎪
⎩
d otherwise.

If pdϕ (y, x) = dsy,x , we may need to update the dependence value of the
sequential dependency matrix dsy,x :

pdϕ (y, x) if pdϕ (y, x) = dsy,x ∨ dsy,x = •,

dsy,x ← (8)
d otherwise.

Now we choose different yl , xk , y and/or x and start from the top until
we reach a fixed point. Our algorithm focuses on positive unateness (p) and
negative unateness (n), in contrast to strict positive unateness (p) and strict
negative unateness (n).
 SAT-Based Combinational and Sequential Dependency Computation 9

y2 x2
B

x1 x3 y6
A C E

y4 x4
D

Fig. 4. Example circuit (2)

According to the definitions in the previous section, all dependencies marked

as seq. positive unate (p), seq. negative unate (n) or seq. independent (•) by our
approximation are correctly classified as we will show in Sect. 4.2.
However, the dependencies marked as seq. functionally dependent (d) may
be inaccurate as dsj,i = d is an over-approximation. Hence, the algorithm allows
an accurate classification for three dependency conditions, while avoiding the
computational complexity of a completely accurate algorithm (that still can be
applied if completeness is needed).
To see that dsj,i = d does not generally imply sequential dependence, see
Fig. 4, where ds6,2 = d, ds2,1 = p, ds6,4 = d and ds4,1 = p. Therefore ds6,1 = d, but
because of the partly inverse reconvergence, y6 is sequentially independent in x1 .
If the XNOR-Gate in Fig. 4 was replaced by an XOR-Gate, y6 would be
sequentially dependent in x1 , while no values of its combinational dependency
matrix would differ from the combinational dependency matrix of the original
circuit. Since these two circuits have the same combinational dependency matrix,
but different sequential dependency matrices, it is not possible to build an exact
algorithm for sequential dependency, solely based on the combinational depen-
dency matrix.

4.2 Proof of Correctness for p, n and •

The correctness of the classification of an input-output pair as either p, n or •

can be shown as follows:

p: Proof by contradiction: For the correctness of the return value p, let the
algorithm return p for output (or flip flop) y and input (or flip flop) x, but
y is not sequential positive unate in x. Then there exists an (arbitrary high)
k ∈ N, such that f (k) , the Boolean function of y, is not positive unate in
x. Following from the definition of unateness (cf. Sect. 2.1), there exists an
(k) (k)
input sequence x̂, such that fx (x̂) = 1 and fx (x̂) = 0. For clarity, we use
the abbreviations x[x] = 0, y [x] = 1 and x[x] = 1, y [x] = 0 where [x] and
[x] indicate the logic value for the respective case. There must exist a path
from x to y, where the path follows x = p0 , p1 , . . . , pm−1 , pm = y, all pi with
[x] [x]
0 < i < m represent flip flops and ∀i ≤ m : pi = pi .
For any i < m, pi+1 combinationally depends on pi , therefore the entry in
the combinational dependency matrix for pi+1 on pi (dpi+1 ,pi ) is not •, thus
10 M. Soeken et al.

d, p or n. As seen in Eq. 8, no dependency value gets overwritten by •, which

leads to dspi+1 ,pi ∈ {d, n, p} for all i. If dspi+1 ,pi in any calculation step was
d, dspi+1 ,pi would be d in the sequential dependency matrix, as d can not get
overwritten. Then, by Eq. 7, pdϕ (x, y) would be step-wise calculated as d,
which would result in dsy,x = d in contradiction to the algorithm returning p.
Thus, for any i < m, it holds that dspi+1 ,pi ∈ {n, p}.
[x] [x] [x] [x]
Let ISame = {i < m : pi = pi+1 } and IDiff = {i < m : pi = pi+1 },
[x] [x]
then IDiff contains an odd number of elements, because p0 = pm . For any
i ∈ ISame , it holds that dspi+1 ,pi = n resp. dspi+1 ,pi = p in every calculation
step and similarly for any i ∈ IDiff , always dspi+1 ,pi = n. The calculated
dependency pd for the given path along p0 , . . . , pm will then be calculated
based on an odd number of n and otherwise only p, which will by Eq. 7 result
in path dependence n. Therefore, by Eq. 8, the algorithm does not return p,
a contradiction.
n: The proof of the correctness of the return value n is analogous to the proof
of the correctness of p. The major difference is that IDiff contains an even
number of elements. This will force a path calculation to result in p, making
impossible, that the algorithm returns n.
•: Proof by contradiction: For the correctness of the return value •, let the algo-
rithm return • for output y and input x, but y is not sequential independent
in x, i.e. sequential dependent in x. Following from a similar argument as for
p, there must exist a path, which follows x = p0 , p1 , . . . , pm−1 , pm = y, all pi
[x] [x]
with 0 < i < m represent flip flops and ∀i ≤ m : pi = pi . By Eq. 8, every
dpi+1 ,pi in any calculation step is not •. Then pdϕ (x, y) would by Eq. 7 be
s

step-wise calculated not as •, which would by Eq. 8 result in dspi+1 ,pi = •

in the sequential dependency matrix, in contradiction to the algorithm
returning •. 

5 Experimental Results
We implemented the proposed approach in C++ on top of the ATPG framework
PHAETON [26] and the SAT solver antom [28]. All experiments were carried out
on a single core of an Intel Xeon machine running at 3.3 GHz, 64 GB main mem-
ory running Linux 3.13. For the evaluations, we used combinational arithmetic
benchmarks from EPFL2 as well as sequential benchmarks from the ITC’99 [7]
benchmark suite and industrial benchmarks provided by NXP (starting with ‘b’
and ‘p’ followed by a number, respectively). Finally, we applied the method to
the OpenCore benchmarks from the IWLS’05 [2] family. In order to keep the
section compact, we skipped the benchmarks that had either negligible runtime
or that could not be solved within a timeout of 2 h.

2
lsi.epfl.ch/benchmarks.
Random documents with unrelated
content Scribd suggests to you:
land, was a stream of water running from the mountains, that would answer
his purpose for manufacturing sugar. In a part of it that possessed all local
advantages, he had made a dam and collected part of the materials for a
mill, to be constructed in the simplest manner, but which, for the want of
mechanical aid, lay in almost the same rude state in which he had purchased
them. He anticipated, however, with great confidence, that by the time the
cane was ready for cutting, the mill would be prepared, when the hopes by
which he had first been stimulated, could scarcely fail to be realized. The
sugar cane grows wild upon the Sandwich Islands, and Mr. Wilkinson's
fields were from the native growth, which he had planted in prepared land,
and at our departure, the canes were upwards of six feet high.

Although several strangers, familiar with the arts of civilized countries,

have settled and lived, for many years, amongst the Sandwich Islanders, Mr.
Wilkinson was the first individual who ever attempted to put them in
practice, upon a scale, sufficiently extensive to improve, materially, the
agricultural condition of the islands, and thereby prove to the inhabitants
the wealth they possess, in a rich soil, and one of the finest climates on the
face of the globe. Should Mr. Wilkinson be successful, the result of his
experiment will probably do more towards the civilization of the natives,
and their ultimate advancement in knowledge and lasting prosperity, than
has yet been effected by all the white men that have lived amongst them.
This, I am aware, will be considered a rash assertion by the good people of
our country, who, from the best of motives, feel deeply engaged in foreign
missions, and, in expressing this opinion, I will also bear testimony to the
useful services of the missionaries. From the best information that I could
obtain upon the subject, I am satisfied, that they have improved the morals
of the people, and if they devote themselves to the schools that they have
established, and to introducing the arts of life and civilization, the time may
not be distant, when the natives will be prepared to receive Christianity. But
in their present condition, would it not be better, and more in the true spirit
of that benevolence and philanthropy which is inspired by our religion, to
teach them how to cultivate their land, to introduce grain and fruits,
congenial to the climate, and to plant and reap as we do, rather than imbue
their minds with a mysterious doctrine, which, being beyond their
comprehension, must resolve itself into a dark and intolerant superstition.
 The inhabitants of the Sandwich Islands, are mild, amiable, and
intelligent, and susceptible of the highest degree of moral and intellectual
improvement. As a people, it may be said, that the stamp of civilization is
scarcely perceptible yet upon them; and it is now inevitable, that they must
bear the impress of those, whom our philanthropists have sent to convert
them to Christianity; and, whether they become ignorant zealots, or
intelligent Christians, will depend upon their teachers. The situation of
these missionaries is, therefore, one of high responsibility, and ought not, in
Christian charity, to be occupied by the narrow-minded fanatic, or the
ignorant zealot. It is of the first importance that they, who dispose of the
vast means of the missionary society, should select only such men for
civilizing and Christianizing the Sandwich Islanders, as are distinguished
for their knowledge, and love of the useful arts of life, as well as for
practical piety.

The harbour of Onavoora, is a place of general rendezvous for the whale

ships, cruising on the coast of Japan, and in the vicinity of the islands. The
months of January, February, and March, being the least favourable for their
business, they then leave their cruising ground, and go in search of
refreshment. No place is so convenient for them as the Sandwich Islands,
and the port of Onavoora, being the most commodious of any in the group,
and affording an abundant supply of vegetables, hogs, &c.,—they all
congregate there in the months above mentioned.

The average number that visit the island in the course of the year, is
upwards of fifty. More than twenty were there together, at several different
periods of our stay, some of which remained a few days, some weeks, and a
few one or two months, according to their several necessities. At such times
the seamen, from having been long confined to the narrow precinct of their
vessels, become very insubordinate, and frequently give way to the most
licentious indulgences of their passions, regardless of every obligation of
obedience due to their officers. We witnessed frequent instances of this
kind, and had the satisfaction of being constantly useful to the captains of
ships, and the whaling interest, by restraining the violent, and coercing
them to a proper sense of duty.
 A most unpleasant occurrence took place about six weeks after our
arrival at Onavoora, arising from this disposition of the seamen, and other
causes, which was afterwards greatly misrepresented in this country. Some
of the seamen of the Dolphin, who were on liberty, got into a frolic, and,
associating themselves with many others belonging to the whale ships,
determined to go to the houses of the high chiefs and missionaries, and
demand the repeal of a restriction that deprived them of the society of
females. They produced a riot that gave rise to a considerable degree of
excitement for a few minutes; but Captain Percival, with some of the
officers of the Dolphin, and captains of whale-ships, promptly suppressed
it, and prevented any serious outrage. It was afterwards unjustly and most
ungenerously ascribed to the officers of the Dolphin.

The white population of Onavoora is of a varied character, from the

agent of our North West traders, to the most abandoned members of society.
There are from fifty to a hundred, permanently settled at Onavoora; the
least respectable of whom, maintain themselves by keeping tippling shops
Tor sailors, and practising such chicaneries as are suggested by
opportunities and the absence of law. The season for the whalers to visit the
Island is the time of their harvest, when, besides their gains from
entertaining the seamen, they frequently prevail upon them to desert for the
sake of the reward for their apprehension, or to strip them of what little
money or clothes they may be possessed of. Some of them have married
women of the Island, and live much in the same way as the natives. Of all
the permanent settlers at Onavoora, the most remarkable was a Spaniard by
the name of Meninne, who had been in the Islands upwards of thirty years.
Of his manner of getting there, a variety of stories were told. His own
account was, that being invited on board of a vessel that visited the coast of
California, he fell asleep in the evening, and, when he awoke, found himself
at sea, and the vessel running, with a fair wind, for the Sandwich Islands.
All his entreaties to be returned to land were in vain, and he arrived at
Owhyhee, where, poor and friendless, he was landed amongst the natives.
He wandered from island to island, in a forlorn and wretched condition,
until he was taken under the patronage of Tamahamaha. For a long time, he
followed the fortunes of the native warrior, and at length fixed his residence
at Oahoo, from whence he occasionally embarked as interpreter on board of
merchant vessels, bound on smuggling voyages to the coast of America.
After passing through a variety of fortune, being several times made
prisoner and threatened with death, he finally quitted his vagrant life, and
remained permanently at Oahoo, where, by great industry and economy, he
has acquired an amount of property which, for that place, is a princely
fortune. Besides having money in the United States and England, he owns
nearly all the cattle on the Island of Oahoo, a number of horses, flocks of
goats, sheep, &c. He has extensive possessions in land, which he holds
from the chiefs in consideration of his services, and a great many houses.
He has taken some pains to introduce exotics in the Islands, and besides the
lemon, orange, and other valuable fruits, has a vineyard from which he
makes a considerable quantity of wine annually. He is said to be extremely
selfish, and so jealous of preserving to himself a monopoly of these
valuable fruits, that he has been accused of destroying the young plants of
others. He is considered as ranking amongst the chiefs in the enjoyment of
certain privileges; and, speaking the language of the Islands well, is called
upon as interpreter for the government upon all important occasions. He has
had thirty-seven children by various wives, and is yet in the prime of life.
Like most men who have felt the cold hand of poverty, and afterwards
acquired wealth, he is extremely penurious, and thinks of nothing so much
as adding to his fortune without knowing how to enjoy it.

On the 3d of April, the young king and all the high chiefs, were invited
on board to spend the day with us. In the morning, we dressed the schooner
in all the flags we could muster, and made the best preparation our limited
accommodation would admit of to receive our distinguished guests. No
indication of their appearance was seen for some time after every thing was
in readiness, and our boats on shore in waiting, and we were kept in
suspense until we had almost despaired of seeing them. All at once the
whole town was in an uproar, and the people were running and hallooing in
every direction. The young king in advance, walked arm in arm with one of
the officers of the Dolphin. Next came Boque and his spouse, with other
high dignitaries, and in the rear a multitude of people of both sexes and all
ages. When they had embarked, the eyes of the people were turned upon
another object, not less interesting to them than the king. Crimacu, or Billy
Pitt, too ill to walk, was gravely making his way to the beach in a hand cart,
where he got into a boat and came on board soon after the king. We
received them with manned yards, and a salute of twenty-one guns. The
king, who was a boy of ten or twelve years old, was dressed in a military
uniform that was sent to him from the king of England, and upon his
shoulders he wore a pair of golden epaulets, with crowns on them. He
appeared to be sensible of his rank, and, upon two or three occasions,
addressed his attendants in a way that indicated a wish to make a display of
his authority. The chiefs conducted themselves towards him with a
becoming respect, but without bestowing upon him any attention that might
interfere with their own enjoyment or convenience. Instead of partaking of
what we had provided for our guests, they, with one or two exceptions,
unceremoniously uncovered their poye pots, which had been brought with
them by an attendant, and ate heartily of their favourite food, using their
fingers as is the native custom. Boque again displayed his major general's
uniform, and was amongst the most polished of our guests. After passing a
number of hours on board, they returned to the shore delighted with their
visit. The same compliment was paid them as at their reception. It was a
day of great enjoyment with the common people. They received their chiefs
on landing with loud huzzas, and followed them in crowds to their
respective habitations.

We passed all the winter months at Onavoora, during which we had

frequent heavy rains. Some of the storms were attended with violent gales,
and disagreeably cold weather. About the 3d of May, an influenza made its
appearance amongst the inhabitants of Oahoo, and in two days not a well
native was any where to be seen. The market, from being well attended,
was deserted. In a week, the distress was so general and so great, that it was
feared the poor people would perish with hunger. I visited several families,
not a member of which was able to help himself or others, and all were
totally destitute of food. A great many of the people died, and amongst the
rest two chiefs, one of whom was George Tamauri, a native of the Island of
Atooi, who was educated in this country. The other was Cahaliha, next in
affinity to the king. Scarcely any of the white people were affected. Not
even those living on shore. In about ten days, the people from the country
began to make their appearance in the market, the distress was greatly
alleviated, and soon afterwards the general good health restored. At such
times of suffering, the condition of these people is truly wretched. They
have no floor to their huts but the hard-beaten ground from which their
naked bodies are separated only by two or three thin mats, and during the
rainy season the earth becomes perfectly saturated with water. Their huts
are generally built in a very frail manner, and in a driving storm are not
proof against the rain. This, alone, is sufficient to account for the thinly
populated state of the islands, without charging the natives with the acts of
inhumanity that have been ascribed to them by some people.

The harbour of Onavoora is formed by coral reefs, that extend upwards

of a mile from the shore. The entrance to it is narrow and somewhat
difficult. It affords water enough for a sloop of war. A pilot in and out of the
harbour, is always employed. It is very secure, being quite land locked. The
holding ground is good, and vessels may lay in safety close along side of
the beach. The water is perfectly smooth in good weather, and the bottom
and shores being nearly every where of soft mud, a vessel might be driven
upon them in a gale without sustaining any material injury. The visits of the
numerous whale ships has made Onavoora a place of considerable trade.
All of them spend more or less money for their necessary refreshments, and
when out of repair, their disbursements are frequently very considerable.
For the supplies afforded by the natives, thousands of dollars are annually
received by them, which they give in return for silk, cotton, calicoes, cloth,
&c. Two or three stores, well stocked with a great variety of goods, are
supported in Onavoora by this interchange of commodities; and, from the
way in which the trade is conducted, there is but little doubt that the
proprietors are rewarded with handsome profits. The sandal wood has
become scarce upon the islands, from the large quantities formerly taken
away by our traders, and does not now form a very lucrative or extensive
article of commerce. Upon this article, some of our merchants are said to
have made large sums of money in exchange for whole cargoes of goods,
with the chiefs of the Sandwich Islands. It is related of Tamahamaha, that,
after making a purchase of this kind, which he paid for in ship loads of
sandal wood, brought from the mountains with great labour to himself and
people, he has been known to load a number of canoes secretly at night,
with fine broadcloths, and take them out to sea, where he sunk them with
stones. His only reason for it was, that the possession of them would make
his chiefs and people too luxurious and idle, and bringing sandal wood from
the mountains to pay for more, would give them employment. A doubtful
policy, certainly, if the statement be true. Be this as it may, himself and
successor have entailed upon their descendants a national debt of several
hundred thousands of dollars, now due to our merchants, which they will
not soon be able to liquidate.

On the 11th of May, after long anticipating our much wished for
departure, we got underway, and saluting the fort as we passed it with
twenty-one guns, stood out to sea and shaped our course for Chili. Nothing
material transpired from the time of our sailing until the 7th of June, when
standing along with a fresh trade, and the night dark and squally, at 10 P.M.,
the lookout-ahead reported land close aboard. We tacked and lay to for the
night, to survey our newly discovered island on the following morning. At
daylight, it bore S.S.W., about six miles from us, and appeared in three
small hummocs, covered thickly with trees and bushes, every where
bounded by a coral reef and heavy surf. We hesitated for some time whether
we should land, apprehending that it would be attended with too much risk,
merely for the gratification of curiosity; but this feeling operated so
powerfully upon us, that there was no resisting the desire to land where no
one had ever been before. Accordingly, two boats were sent off, and
watching a favourable opportunity passed through the surf in safety, and
landed on the coral bank where they were left high and dry by the receding
wave. It was on the lee side of the island, and a coral reef stretched off
about fifty or a hundred yards from the shore, full of holes, and almost dry
at low water. In the holes we sought for fish, as at Caroline Island, but
found very few. On traversing the island, we could find no fruit or vegetable
of any description except bup. In most places it was covered with trees and
bushes of a thick growth, almost impenetrable. In the bushes we found a
great many tropical birds setting, so tame that we could take them off of
their nests with our hands, and in getting upon the weather side of the island
where there was a clear space, we found a species of small gull, so
numerous, that when they rose from the ground at our approach, they
appeared to form almost a compact mass. The sand was literally covered
with their eggs, which, upon examination, proved to be unfit for use, with
few exceptions. The birds flew and hovered so near to us that we caught
several of them with our hands. After amusing ourselves a little while with
the novel spectacle of such numbers of birds so very tame, we collected all
the old eggs within a small space, and in less than twenty minutes after it
had been thus cleared, it was again covered with fresh eggs by the birds that
were constantly lighting. We might, in a few hours, have loaded our boats
with them. When we returned to our boats to go on board, the tide had risen,
and with it the surf had increased to an alarming degree. At first, we held a
consultation whether we should attempt to pass through it or remain until it
should again subside with the falling of the tide; but the day was far
advanced, and the appearance of the weather such, as in all probability
would render our situation extremely painful, there being every prospect of
an increase of wind. We therefore determined to put our fortune to the test,
and were not a little discouraged when on one of the boats attempting the
surf was thrown back by the second roller that she encountered with the
utmost violence, upsetting her, and scattering the people in different
directions, some of them escaping with their lives with the greatest
difficulty. Notwithstanding the ill success of our first attempt, we
determined on a second, rather than risk the consequences of longer delay.
Embracing a favourable opportunity where the surf appeared least violent,
we put off, and passed through it in safety without further accident.

The island was little more than a mile long, and from a hundred yards to
a quarter of a mile wide. It was every where very low. By a meridian
observation, we placed the north-west end of it in latitude south 21 degrees
48 minutes, and longitude by chronometer 154 degrees 54 minutes west. In
compliment to the commander of our squadron in the Pacific Ocean, we
called it Hull's Island. It may be comprehended within the group of Society
Islands.

In the afternoon, June the 4th, we made sail; and on the following day at
half-past 6 o'clock, A.M. discovered the island of Ramitaria, on the lee bow,
about eight leagues from us. This island was not laid down in any of our
charts, having been discovered only three or four years before; but we had
seen a gentlemen who had stopped there, and it was included in our list of
islands. On approaching it we were pleased to find that it differed from
many of the islands we had visited, being of a moderate elevation. It is
about three miles long and one or two wide. We ran nearly round it before
we found a place to land, the surf breaking high every where, and the shores
bounded by large rocks of coral. At last, we came to the principal
settlement, which was situated close to the shore, where a large white-
washed house indicated the former visits and influences of the missionaries.
The people, to the number of one or two hundred, were assembled on the
beach inviting us to land. Here, also, was a considerable surf and some
coral rocks, which made the landing not altogether free from difficulty.
When the boat came near the shore and while she was yet shooting rapidly
through the surf, the natives, who had already advanced to meet us, laid
hold as many as could get round her, and with loud shouting, carried us
high upon the beach. At this somewhat unexpected reception, the boat's
crew instinctively seized their pistols, thinking at first, that the natives were
hostile in their disposition towards us. It was but a momentery panic which
passed away with the kind salutations we immediately afterwards received.
When I enquired for the chief, a young man was pointed out to me in the
crowd, distinguished from the rest by an old hat on his head, that he had
obtained from some former visiter. He seemed not to be treated with the
least respect by the people, who jostled him in the crowd with the most
perfect carelessness. Scarcely had I addressed him, when a stout native
came up with an air of some importance, and saluting me, told me in the
language of the Society Islands, that he was the missionary. Upon his
approach, the chief immediately shrunk back into the crowd. He called to
him a Malay, who was not far off, and bade him ask what we wanted. The
Malay, whose name was Manoo, spoke English very well, which was a
source of great satisfaction to us, as we could thereby communicate our
wants freely, and it introduced us at once to each others' acquaintance. I
explained to Manoo that we were in want of water and such refreshments as
the island might afford. He immediately proceeded in company with the
missionary to show me where water was to be obtained. We passed through
a forest of very large trees over a plain that extended more than half a mile,
when we came to a marsh of reeds and rank grass, where there was from
one to two feet of water, covering an area of two or three acres. This would
not answer our purpose as the water was not very good, and its distance
from the place of embarkation rendered it too laborious an undertaking to
water the vessel from it. When I had remarked this to Manoo and the
missionary, they replied that there was better water, but it was still more
distant. I suffered them to conduct me to it, and we took a footpath over
rising ground, and through another delightful forest of bread-fruit and other
wide-spreading trees, passing many fine tarrow patches, and at the distance
of half a mile from the marsh, we came to a spring of excellent water.
Manoo and the missionary both expressed a great deal of disappointment
when I spoke in terms of disapprobation of this also as a watering-place—it
being altogether too far from the shore. On our return from the spring, we
took a different footpath from that by which we came, and ascended to a
more elevated part of the island to see the work of the missionary, as it was
called by Manoo. This consisted of two or three enclosures by means of
stakes, in the midst of the forest, where the trees had been cut down for
several acres, which was cleared and planted with sweet potatoes and
tobacco. The enclosures were made, and all the labour of clearing the forest
was done, as Manoo remarked, by such of the wicked and disobedient, as
had resisted the authority and ordinances of the white missionary during his
residence on the island some months previous. Since then, the white
missionary had returned to Otaheite, and sent this native missionary, who
belonged to that island, to represent him in his absence. Leaving the
enclosures of tobacco and sweet potatoes, we came into a thicket where the
trees were overrun with the vines of the yam, growing wild and covered
with beautiful blue flowers that gave a picturesque appearance to the forest,
and filled it with their fragrance. We passed five or six stone columns that
had been sixty or seventy feet high, and twenty or thirty in circumference.
They were in a dilapidated state, having in part been thrown down by order
of the missionaries. Manoo told me that they were monuments erected in
honour of the Indian god. Several of these ruins were standing by the side
of an old burial-place in the edge of the woods near the shore. Each of the
graves were neatly enclosed with a wall of stone. When we returned to the
village, I was taken to a large frame building called the missionary house,
where the missionary had prepared a roasted pig and some tarrow, for our
dinner. He designed to entertain me after the manner of the whites, and with
this view, had placed our repast on a large coarse table that stood in the
middle of the room with benches round it. When we were seated, he
unlocked a chest and took from it a plate for each of us, and a knife and
fork, all of which were extremely dirty, and the knives and forks quite
covered with rust. This, however, he did not seem to remark, although he
evidently wished me to think that he knew how to be polite, after the
fashion of my country people. He acquitted himself pretty well, to his own
satisfaction, until he attempted to use the knife and fork—but that was
altogether too much for him. After making several trials in vain to cut his
meat, he asked me to assist him; and finally, before he had half finished his
dinner, laid down his clumsy instruments and used his fingers. The chief,
and as many of the natives as could get into the room came round us, but
none of them were invited to partake with the missionary and myself,
except Manoo, whilst the hungry crowd stood looking wistfully at us.

Soon after we had finished our dinner, the captain and several of the
officers landed, and we exchanged several articles with the natives for their
pigs, yams, &c. The following day was their Sabbath and our Saturday, and
they insisted that we should remain until Monday, before any exchange of
commodities took place. When, however, we declared our determination to
depart that evening, they began collecting whatever might be acceptable to
us. Hogs of various sizes, were brought to the beach in great numbers.
Needles, jack-knives, and old clothes, were our articles of traffic, than
which we could have offered them nothing more valuable. By sundown, we
had collected from thirty to forty hogs, and a good supply of yams. In a few
hours more, we might have obtained twice as many upon the same terms.

Towards the close of the day, when the missionary felt assured that it
was our determination to depart, he asked if I was a doctor, or had any skill
in medicine; and, although I replied in the negative, insisted upon my going
to his house to see, and prescribe for his wife, who, he told me, was
extremely ill. Upon entering his hut, we found her laying on a mat on the
floor; and notwithstanding the weather was oppressively warm, she was
covered over with a great many pieces of the tappa cloth, head and all, and
perspiring most profusely. The missionary, with great gravity, but most
unceremoniously, removed all the covering, and pointed out to me her
infirmity, which was nothing more than a common bile, with which she
seemed to be suffering considerable pain. I declined prescribing, although
repeatedly requested to do so; and at sun-down, we embarked and made
sail. Had it been convenient for us to remain two days longer at Ramitarias,
we could have obtained an abundant supply of whatever the island
produced, for a very trifling consideration; but although there was
anchorage, it was unsheltered, and too near the shore for us to ride in safety.

At six, P.M. on the 10th of June, we took our departure from Ramitarias;
and at day-light, on the 13th, made the Island of Toubouai, bearing to the
northward and eastward, about eight leagues from us. In getting in with the
S.W. part of the island, we found an extensive reef, upon which the surf was
breaking with great violence. We hauled round to the east side, passing two
small uninhabited islands, but there was not the slightest appearance of a
landing-place, the surf breaking heavily as far as we could see. At four,
P.M. we anchored on the north side, in seven fathoms water, and sent boats
in search of the harbour, along the west shore. In the evening, they returned,
having found it, and on the following morning, we got underway, and beat
up for it. When we had advanced near the opening, through the coral reefs
that extend from the shore several miles, a Mr. Strong, an American, came
on board, and piloted us in through a difficult passage. The channel was
narrow, and very crooked; but we had not less than three and a half fathoms
water. Our anchorage was within a coral reef, about a mile from the shore,
in four and a half fathoms. On the day previous to entering the harbour, we
discovered that the head of our mainmast was decayed, and badly sprung.
In this situation, we congratulated ourselves in having found a secure
harbour, where the water was tolerably smooth, which was very essential in
fixing the mast securely.

Nearly all the inhabitants of Toubouai, consisting of about two hundred,

were living on the side of the island where we were anchored, in two
different settlements. One of them was the residence of two Otaheite
missionaries, who governed in spiritual and temporal affairs; and at the
other, was the residence of the king, who, since the coming of the
missionaries, retained but a small share of the power that formerly
appertained to his station. He lived on apparently good terms with the
missionaries, fearing their influence with the people, but secretly declaring
his aversion to them. He was the more afraid of offending them, as there
was a living example of their displeasure constantly before him, in the
person of one who had formerly exercised the regal authority in the
missionary village, of which they had divested him for disobedience, and
compelled him to live in the condition of a common private person. Soon
after the arrival of the missionaries, the people became divided into two
parties, one of which advocated matrimony, and the other the unrestrained
indulgence which is practised in a state of nature. The first, called
themselves the missionary party; the latter, the party of the Tutiori. After a
long struggle for the ascendency, the Tutiories took up arms and retired
from the habitable part of the island, declaring themselves independent of
the missionaries. They remained for several days undisturbed in their
disobedience, when a party was sent to bring them to terms. The Tutiories
at first retreated, but finally made a stand, and after a slight show of
resistance, submitted, promising to go home and live in the observance of
the missionary precepts. The dethroned king, was at the head of the
Tutiories.

At the lower village, as it was called by us, where the king lived, was a
party of our countrymen, who had been there for a number of months
building a vessel. They had completed the frame and commenced planking,
when, unfortunately, a quarrel arose between them and the people of the
missionary village, which terminated in open hostility, and the loss of
several lives. One of the white men only, was killed. Scarcely had they
made peace with the natives, when they quarrelled among themselves, and
nearly half of their number (four or five) discontinued their work, and
waited only for an opportunity to leave the island. This mutinous
disposition of a few, paralized the efforts of the whole party, and it was
probable from appearances, that the labour they had bestowed with so much
effect, would be entirely lost to themselves and their employer, Captain
Dana of Massachusetts, whom we had seen at Oahoo.

The king and myself, became high carnies, (or great friends) the day
after our arrival; and from that time until our departure, he did everything in
his power to merit my regard. My mess was constantly supplied with bread-
fruit, cocoa-nuts, tarrow, and bananas; and whenever I was on shore, he
waited upon me everywhere with the most friendly attention. Our friendship
commenced in the following manner, simple enough, it is true, but I believe
it was not the less sincere on that account. Seeing me with a jack-knife in
my hand, he expressed a wish to look at it, when I told him that he might
have it. He received it, and after observing me for a moment, put his hand
upon my arm and remarked, now you and I will be high carnies. I sincerely
reciprocated his kindness, so strongly recommended by the disinterested
simplicity with which it was proffered. Two or three days after our arrival, I
landed with a view of traversing the island. I found king Dick, as I used to
call him, upon a bank of coral that stretched out from his house, with a long
wooden spear in his hand, looking about in the holes for fish. He begged me
to wait until he had taken one for his dinner, and he would accompany me.
Accordingly, having succeeded in a few minutes afterwards, we set off
together.
 The island was two or three miles wide, and we had not proceeded far,
when we came to an extensive marsh that runs through the middle of it.
King Dick stopped, and insisted upon carrying me over on his back. Feeling
that it was too menial an office to be performed by a king, although an
untutored native, I remonstrated with him, and positively refused to be
carried; but after resisting his importunity for some time, he took my
musket in his hand, and I mounted on his shoulders. The marsh was several
hundred yards wide, and king Dick found his burthen very heavy before he
landed me on the opposite side. We there entered upon a romantic and
extensive plain, covered with cocoa-nut, and bread-fruit trees, plantain
walks, papayas, &c.; interspersed through which, were numerous huts that
appeared to have been long deserted. We traversed this plain for miles, with
little variation; the same pleasing prospect everywhere presenting itself,
when suddenly king Dick stopped, and made the whole forest ring with the
shrill notes of his voice. After repeating it for several times, he was
answered at a distance, and soon afterwards, we saw a native making his
approaches towards us. King Dick said something to him, and he ascended
a cocoa-nut tree, after the manner of the people of the Marquesas, and threw
down a sufficient quantity of nuts to quench our thirst. He afterwards
conducted us to a house where there were two women and several children.
They were delighted to see us, and hospitably proffered whatever they had
of refreshment. They gave us a preparation of dried bread-fruit, of which
king Dick ate very heartily, but to me it was scarcely palatable. Near the
house there were several citron and lemon trees, the only ones that I saw on
the island.

In our route back, we met with large quantities of sugar cane in a wild
state. We passed the burial-place of the village, where, beside two or three
newly made graves, were rudely carved images placed there in conformity
to a native custom, that had not yet been abolished by the influence of the
missionaries. When we arrived at king Dick's house, we found dinner
waiting for us. It consisted of the fish he had taken in the morning before
our departure, roasted fowls, bread-fruit, tarrow, &c. The whole was
wrapped up in plantain leaves, and placed on a mat upon the floor, around
which, we all seated ourselves, there being a number of visiters present.
During our repast, king Dick went frequently to a large chest, where he kept
a bottle of rum I had given him, and, after proffering me the bottle, he
would help himself, and lock it up in the chest again, without taking the
least notice of the rest of his visiters. No miser ever guarded his treasure
more penuriously than the king did his bottle of rum; not only on this, but
all other occasions. His house was large, and a lounging place for all the
idle people of his village; and, although there were always a number present
when I made my visits, and he never failed to offer me a portion of his
favourite beverage, the wistful lookers-on received not the slightest notice. I
once ventured to propose to him to give some to his friends, but he said no,
it was too good for them. When, after a week's stay at Toubouai, our
departure was spoken of, king Dick expressed the liveliest regret, and
proposed going with us. We told him that he would never be able to get
back again; to which he replied, that he should not care about returning. The
day before we left, he sent me, on board, an abundant sea-stock, of fruits
and vegetables, and a good-sized hog. As a parting gift, I dressed him up in
an old uniform of mine, that seemed to give him infinite pleasure. To his
presents, his wife added several pieces of the tappa cloth, some of which,
were as fine specimens of the native manufacture, as I ever met with
anywhere.

At Toubouai, we added considerably to our collection of curiosities. The

most ingeniously wrought article, was a lash, used by the natives for
brushing the flies off of their backs. The handles were carved to represent a
man's face, or some animal familiar to them. The lash itself, was, in several
strands, finely braided from twine of the cocoa-nut husk. The natives were,
in general, naked, except a wrapper round their waists, and had a sickly and
feeble look. The island is very fertile, producing, in many parts, large
quantities of cocoa-nut and bread-fruit, and is capable, without further
improvement, of sustaining some thousands of people. The population has
greatly diminished within the recollection of recent visiters. It must have
been much greater when visited by the mutineers of the Bounty, who built a
fort on the island, but afterwards became dissatisfied, and left it in
consequence of the treacherous and hostile character of the natives. The
only survivor of those guilty men, who is now the patriarch of Pitcairn's
Island, states, that they had war with the natives previous to leaving
Toubouai, and that they killed a great many of them. His estimate of the
inhabitants at that time, is fifteen thousand. King Dick told me, that the huts
we saw on the side of the island opposite to where we anchored, were once
inhabited by people who were all dead. We obtained wood and water of a
good quality, without much labour. Hogs were scarce, and we got none but
what were presented to us. The natives gave us a few domestic fowls, in
exchange for writing paper. The paper was used by the women for making
paste-board for a bonnet, that had just been introduced amongst them by the
wives of the Otaheite missionaries.

On the island of Toubouai, there is a tree, the bark of which, makes

excellent cordage. It is produced in great abundance, and grows to the size
of six or eight inches in diameter. After the ship builders were interrupted in
their work, by the discontent of some of the party, the remainder made a
rope-walk, and devoted themselves to making rope of this bark. They had
several small cables, and a considerable quantity of rope. We purchased
some, and found it to wear nearly, or quite, as well as hemp. Arrow root,
grows spontaneously upon the island, and has been taken away by several
vessels, as an article of commerce. We were informed that the scarcity of
hogs was in consequence of a great many having been carried off recently,
by a vessel belonging to the missionaries. Also, that they lay all the small
islands under contribution, and annually send their small vessels to collect
the hogs, which they barrel, and send to Port Jackson. This may be true or
not. It was positively asserted.

Our stock of bread was nearly exhausted, and what remained, was in a
damaged state. Having failed in all our attempts to obtain a supply of yams,
that would enable us to reach the coast of South America, an examination
was held upon the bread, to the end, that we might adopt such expedients as
should be deemed most proper. Valparaiso, was our most direct port; there
was but one inhabited island (Oparro,) in our way, and that at some distance
from a straight course. To avoid, however, so disagreeable a circumstance
as being without bread, we determined to touch at Oparro, and get whatever
we could for a substitute. Accordingly, we put to sea on the 22d of June,
and on the 25th, made Oparro, a little before day-light, five leagues from us.
When we came near, its appearance was rude and inhospitable in the
extreme, being a mass of rugged mountains, about a thousand feet high, and
eight or ten miles in circumference. On the north-west side, there were no
signs of inhabitants. The mountains rose almost perpendicularly from the
Ocean, and we sounded frequently, close to the shore, without getting
bottom. In rounding the island, to the northward and eastward, we opened a
valley, where there was a number of huts, and some patches of tarrow. A
boat was sent in shore, and two of the natives came on board, who informed
us that there was a harbour to the eastward. We continued on, and after
landing at one or two other small valleys, where there were huts and tarrow
patches, opened a beautiful deep bay, that had an appearance of great
fertility. Soon afterwards, an Englishman came on board, and offered to
pilot us in. He had not advanced far with the vessel, before she grounded on
a bank of coral, but fell off immediately, on putting the helm down and
throwing all aback. We were everywhere surrounded by shoals of coral, and
fearful of accidents, came to outside of the harbour, where we had ten
fathoms of water on one side, and five on the other. The boats were sent off
immediately, in search of tarrow. It was planted everywhere, in large
patches, where there was a small valley, through which a stream of water
found its way from the mountains to the sea, and at a distance, the hills
were green with another species we had not before seen, called mountain
tarrow. The latter is superior to the low land tarrow, and will keep longer at
sea.

When we landed in the bay, which was several miles deep, we found
two different missionary establishments, occupied by natives of Otaheite.
No one else, that we saw, seemed to have any authority. They permitted us
to dig as much tarrow as we pleased, without asking any return for it, and
by night, we had as much on board as we knew what to do with. One of the
missionary establishments was situated at the extreme depth of the bay. On
landing there, I was met by the missionaries themselves, and conducted to
their house, where their wives, who were also from Otaheite, received me,
dressed in their own island costume, with large straw bonnets on their
heads. After remaining with them for a few minutes, I repaired to the place
where the boats' crew were employed in collecting tarrow. A group of
natives, male and female, had assembled round, none of whom, could be
prevailed upon by the rewards we offered them, to assist our people in their
occupation. They had a sickly look, almost without an exception. Their
dress differed altogether from any we had before seen. It consisted of a
heavy mat of grass, weighing from ten to fifteen pounds, which was thrown
over their shoulders, and another light mat of the same material, for the
loins. Their deportment was modest and retiring, and they evinced a
disposition to have but little intercourse with us. A few of their houses were
scattered about upon the hills. They were extremely miserable, and might,
without disparagement, be compared to dog kennels. They were long, and
very narrow, and about three or four feet high, so that when one entered
them, it was necessary to get down upon the hands and knees. The Otaheite
missionaries were distinguished from the natives, by wearing the tappa
cloth, of their own island, and an old cloth jacket they had obtained from
white visiters. I added something to their stock of clothes, before I took
leave of them, for which, they gave me many thanks. When I returned on
board, I found the other two missionaries there, and several of the natives,
who had accompanied the captain. They spent the night with us, and the
next morning, showed us where to obtain a quantity of the mountain tarrow.
It is large, and very much resembles the West India yam. For a vessel long
at sea, and requiring vegetables, there is none more valuable.

The Englishman who came on board to act as our pilot, was residing at
the island, in charge of a party which had been left there by the English
Consul at the Sandwich Islands, to collect beach la mer, a valuable article of
commerce, at Canton. The collecting of sandal-wood, was also to be an
object of their attention; but they told us, that both were scarce, and difficult
to obtain—a statement, we considered, of doubtful veracity.

At 2, P.M., on the 27th of June, we got underway, and made sail for
Valparaiso, which is distant from Oparro, about three thousand five hundred
miles. This island, which is called by the discoverer, Oparro, is called by the
natives, Lapa. It is situated in latitude 27 degrees 34 minutes south, west
longitude 144 degrees. At 5, P.M., we saw the islands, called by Quiros, Los
Corones, and hauled up, until eight, to clear them.

For the first three days after we sailed, it blew a severe gale from the
westward, after which, it became more moderate, but nearly all our passage
to Valparaiso, was wet and boisterous. It was less disagreeable, however,
than we anticipated, as the wind was almost constantly fair. We had the
more reason to apprehend a severe trial of our fortitude, as it was in the
dead of winter that we were advancing into a high latitude, after having
been a long time within the tropics, and besides that, our sails and rigging
were very much worn, some of our articles of provisions were quite
exhausted, and others, of the first necessity, reduced to a small quantity. If,
therefore, we had encountered tedious gales ahead, we must have suffered
greatly in our shattered and ill-fitted condition.

At 4, P.M., on the 19th of July, we made the island of Mas a Fuera, on

the coast of Chili, seventy miles from us, and at midnight, passed close to it.
At day-light, Juan Fernandes, was in sight, fifty miles off. We passed it a
little after meridian. It is very mountainous, but well covered with trees.
The interesting fable of Robinson Crusoe's adventures, has given it a lasting
fame, and rendered it an object of curiosity to all who visit this part of the
Pacific Ocean. It is very fertile, and has been tolerably well cultivated. The
Spanish captain-general of Chili, formerly made it a place of banishment,
and after the revolution took place in that country, it was appropriated to the
same purpose, by the patriots and royalists, as they alternately came into
power. A considerable town was built by the exiles, who were sent there at
different times, and the finest fruits of Chili are produced in great
abundance. The cattle that have been left upon the island, are running wild
in large herds, and several persons have found it profitable to send parties
there to kill them for their hides. The island produces some sandal-wood,
but it is small, and has never been collected in large quantities.

Fish, that very much resemble our codfish, and a variety of other kinds
are taken in the greatest abundance around Juan Fernandez and Mas a
Fuera. It is believed that if a fishery were established there by some of our
enterprising countrymen, it would be found a source of great emolument.
The privileges that might be considered necessary for the prosperity of a
company formed with this object, could easily be obtained from the
government of Chili, and there is no apparent cause why the most
successful results should not be realized. It seems only necessary to call the
attention of our capitalists to this subject, to have all its advantages secured
to our country. It has a fine harbour for the prevailing winds of summer, but
in the winter season, when the winds set in from the northward, it is
exposed. It lies a little more than three hundred miles from the coast of
Chili, and in the summer months I have known open boats to pass between
it and Valparaiso.
 On the 23d of July, we anchored in the harbour of Valparaiso, a little
before day-light, to the gratification of our friends, who were becoming
very much alarmed for our safety, no information of us having been
received during the whole period of our absence. Thus, in a vessel of 180
tons burthen, poorly fitted, and having on board only about four months'
provisions, when we sailed from the coast of Peru, we performed a cruise of
upwards of eleven months in an unfrequented Ocean, rendering to our
countrymen, and many of the people whom we visited, important benefits,
besides realizing the most successful results in the primary object of our
cruise. Its beneficial effects will long be felt by our countrymen, who are
engaged in the whale-fishery; and, although we suffered many hardships,
privations, and dangers, we were happy in being the instruments, in the
hands of Providence and our government, of proving that crime cannot go
unpunished in the remotest part of the earth, and that no situation is so
perilous as to justify despair.

[1] A tree that resembles the locust. It bears a pod, like that of a bean, which
is given by the Peruvians to their horses.

[2] A retail grocery and tippling shop.

[3] Spirits distilled from the grape.

[4] Mattee, mattee—Very bad.

[5] Very good, or very well.

[6] Don't you want a wife?

[7] Com. Porter.

[8] A bird commonly called the sand-snipe. We afterwards saw them so

frequently at sea, where no land was known to exist, that their presence