See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/220728126

Sparse matrices in MATLAB*P:Design and implementation

Conference Paper · December 2004

DOI: 10.1007/978-3-540-30474-6_20 · Source: DBLP

CITATIONS READS

53 359

2 authors, including:

Viral B. Shah
Julia Computing
35 PUBLICATIONS   3,129 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Julia - A fresh approach to numerical computing View project

All content following this page was uploaded by Viral B. Shah on 03 June 2014.

The user has requested enhancement of the downloaded file.

 Sparse Matrices in Matlab*P:
Design and Implementation

Viral Shah and John R. Gilbert

{viral,gilbert}@cs.ucsb.edu?
Department of Computer Science
University of California, Santa Barbara

Abstract. Matlab*P is a flexible interactive system that enables com-

putational scientists and engineers to use a high-level language to pro-
gram cluster computers. The Matlab*P user writes code in the Mat-
lab language. Parallelism is available via data-parallel operations on
distributed objects and via task-parallel operations on multiple objects.
Matlab*P can store distributed matrices in either full or sparse for-
mat. As in Matlab, most matrix operations apply equally to full or
sparse operands. Here, we describe the design and implementation of
Matlab*P’s sparse matrix support, and an application to a problem in
computational fluid dynamics.

Introduction

Matlab is a widely used tool in scientific computing. It began in the 1970s as an

interactive interface to EISPACK, and LINPACK. Today, Matlab encompasses
several modern numerical libraries such as ATLAS, and FFTW, rich graphics
capabilities for visualization, and several toolboxes for such domains as control
theory, finance, and computational biology.
Almost all of today’s supercomputers are based on parallel architectures.
Companies such as IBM, Cray, SGI sell supercomputers with proprietary inter-
connects. Commodity clusters are omnipresent in research labs today. However,
the tools used to program them are still predominantly Fortran and C with MPI
or OpenMP.
Matlab*P brings interactivity to supercomputing. There have been sev-
eral efforts in the past to parallelize Matlab. The parallel Matlab survey [6]
discusses most of these projects. Perhaps the most notable project that pro-
vides a large scale integration of parallel libraries with a Matlab interface is
NetSolve [2]. NetSolve provides an interfaces by invoking RPC calls through a
special Matlab function, as opposed to Matlab*P which takes a unique ap-
proach to parallelization. The Matlab*P language is a superset of the Matlab
?
This material is based on research sponsored by Air Force Research Laboratories under
agreement number AFRL F30602-02-1-0181. The U.S. Government is authorized to
reproduce and distribute reprints for governmental purposes not withstanding any
copyright notation thereon.
language, and parallelism is propagated through programs using the the dlayout
object – p. In the case where these systems use the same underlying packages,
we believe that they can all achieve similar performance. However, we believe
that the systems have different design goals otherwise and it is unfair to compare
them.
Sparse matrices may have dimensions that are often in millions and enough
non–zeros that they cannot fit on one workstation. Sometimes, the sparse ma-
trices are themselves not too large, but due to the fill–in caused by intermediate
operations (for eg. LU factorization), it becomes necessary to distribute the fac-
tors over several processors. Iterative methods maybe a better way to solve such
large sparse systems. The goal of sparse matrix support in Matlab*P is to allow
the user perform operations on sparse matrices in the same way as in Matlab.

1 User’s View
In addition to Matlab’s sparse and dense matrices, Matlab*P provides sup-
port for distributed sparse (dsparse) and distributed dense (ddense) matrices.
The system design of Matlab*P and operations on ddense matrices are de-
scribed elsewhere [12, 7].
The p operator provides for parallelism in Matlab*P. For example, a random
parallel dense matrix (ddense) distributed by rows across processors is created
as follows:
>> A = rand (100000*p, 100000)
Similarly, a random parallel sparse matrix (dsparse) also distributed across pro-
cessors by rows is created as follows: (An extra argument is required to specify
the density of non-zeros.)
>> S = sprand (1000000*p, 1000000, 0.001)
We use the overloading facilities in Matlab to define a dsparse object. The
Matlab*P language requires that all operations that can be performed in Mat-
lab be possible with Matlab*P. Our current implementation provides a work-
ing basis, but is not quite a drop–in replacement for existing Matlab programs.
Matlab*P achieves parallelism through polymorphism. Operations on ddense
matrices produce ddense matrices. But once initiated, sparsity propagates. Oper-
ations on dsparse matrices produce dsparse matrices. An operation on a mixture
of dsparse and ddense matrices produces a dsparse matrix unless the operator
destroys sparsity. The user can explicitly convert a ddense matrix to a dsparse
matrix using sparse(A). Similarly a dsparse matrix can be converted to a ddense
matrix using f ull(S). A dsparse matrix can also be converted into a Matlab
sparse matrix using S(:,:) or p2matlab(S). In addition to the data–parallel
SIMD view of distributed data, Matlab*P also provides a task–parallel SPMD
view through the so–called “MM–mode”.
Matlab*P currently also offers some preliminary graphics capabilities to
help users visualize dsparse matrices. This is based upon the parallel rendering
 Fig. 1. Matlab and Matlab*P Spy plots of a web crawl dsparse matrix

for ddense matrices [5]. Again, this demonstrates the philosophy that Matlab*P
should feel like Matlab. Figure 1 shows the spy plots (showing the non–zeros
of a matrix) of a web crawl matrix in Matlab*P and in Matlab.

2 Data Structures and Storage

Matlab stores sparse matrices on a single processor in a Compressed Sparse

Column (CSC) data structure [10]. The Matlab*P language allows for matrices
to be distributed by block rows or block columns. This is already the case for
ddense matrices [12, 7]. The current implementation supports only one distri-
bution for dsparse matrices – by block rows. This is a design choice to prevent
the combinatorial explosion of argument types. Block layout by rows makes the
Compressed Sparse Row data structure a logical choice to store the sparse ma-
trix slice on each processor. The choice to use a block row layout is not arbitrary,
but based on the following observations:

– The iterative methods community largely uses row based storage. Since we
believe that iterative methods will be the methods of choice for large sparse
matrices, we want to ensure maximum compatibility with existing code.
– A row based data structure also allows efficient implementation of matvec
(sparse matrix dense vector product) which is the workhorse of several itera-
tive methods such as Conjugate Gradient and Generalized Minimal Residual.

By default, a dsparse matrix in Matlab*P has the block row layout which
would be obtained by ScaLAPACK [3] for a ddense matrix of the same dimen-
sions. This allows for roughly the same number of rows on each processor. The
user can override this block row layout in a couple of ways. The Matlab sparse
function takes arguments specifying a vector of row indices i, a vector of column
indices j, a vector of non–zero values v, the number of rows m and the number
of columns n as follows:
>> S = sparse (i, j, v, m, n)
By using a vector layout which specifies the number of rows on each processor
instead of the scalar m which is simply the number of rows, the user can create
a dsparse matrix with the desired layout:
>> S = sparse (i, j, v, layout, n)
The block row layout of a dsparse matrix can also be changed after creation
with:
>> changelayout (S, newlayout)
The CSR data structure stores whole rows contiguously in a single array on each
processor. If a processor has nnz non–zeros, CSR uses an array of length nnz to
store the non–zeros and another array of length nnz to store column indices, as
shown in Figure 2. Row boundaries are specified by an array of length m + 1,
where m is the number of rows on that processor.

           


 
 

 
 


           


 

        

 
 
 
 
 

Fig. 2. Compressed Sparse Row (CSR) data structure

Assuming a 32–bit architecture and using double precision floating point

values for the non–zeros, an m × n real sparse matrix with nnz non-zeros uses
up 12nnz + 4m bytes of memory. Support for complex sparse matrices will be
available very soon in Matlab*P.
It would be simple to modify this data structure to allow some slack in each
row so that element–wise insertion, for example, could be efficient. However, the
current implementation uses the simplest possible data–structure for robustness
and efficiency in matrix and vector operations.

3 Operations and Implementation

In this section, we describe the implementation of several sparse matrix oper-
ations in Matlab*P. All experiments are performed on a cluster of 2.6GHz
Pentium IV Xeon processors with 3GB RAM and a gigabit ethernet intercon-
nect.

3.1 Parallel Sorting

80
 !
 !
#
1e7 nnz(4/row)
70 1e6 nnz(32/row)


60

Time (Secs)
" !
#
#

50
# !

40

 #
30

 !
20
10

 #

0
2 4 6 8 10 12 14 16
Processors

Fig. 4. Scalability of sparse (Altix)

Fig. 3. Starching

Sorting of ddense vectors is an important building block for several paral-

lel sparse matrix operations in Matlab*P. Sorting is an extremely important
primitive for parallel irregular data structures in general. Several parallel sorting
algorithms have been surveyed in the literature [4, 8]. Cluster computers have
distributed memory, and the interconnect typically has high latency and low
bandwidth compared to shared memory computers. As a result, it is extremely
important to minimize the communication while sorting. We are experimenting
with SampleSort with different sampling techniques, and other median–based
sorting ideas. Although we have an efficient parallel sorting algorithm already
in Matlab*P, we are still trying to improve it, given the importance of having
fast sorting. We will describe results from our efforts in a future paper.
Several sorting algorithms produce a data distribution that is different from
the input distribution. Hence a reshuffling of the sorted data is often required.
We refer to this process as Starching, as shown in Figure 3.1. The distribution
after sorting is on the left, whereas the desired distribution is on the right in
the bipartite graph. The weights on the edges of the bipartite graph show the
communication required between pairs of processors during starching. This step
is required to ensure consistency of Matlab*P’s internal data structures.

3.2 Constructors

There are several ways to construct parallel sparse matrices in Matlab*P:

1. matlab2pp converts a sequential Matlab matrix to a distributed Mat-

lab*P matrix. If the input is a sparse matrix, the result is a dsparse matrix.
function s = sparse (i, j, v)

[j, perm] = sort(j);

i = i(perm); v = v(perm);

[i, perm] = sort(i);

j = j(perm); v = v(perm);

starch (i, j, v);

s = assemble (i, j, v);

Fig. 5. Implementation of sparse

2. sparse – sparse works with [i, j, v] triples, which specify the row value,
the column value and the non–zero value respectively. If i, j, v are ddense
vectors with nnz non–zeros, then sparse assembles a sparse matrix with
nnz non–zeros. If there are duplicate [i, j] indices, the corresponding values
are summed. The pseudocode for sparse is shown in Figure 3.1. However, in
our implementation, we implement this by sorting the vectors simultaneously
using row numbers as the primary key, and column numbers as the secondary
key.
The starch phase here is similar to the starching used in the parallel sort,
except that it redistributes the vectors so that row boundaries do not over-
lap among processors and the required block row distribution for the sparse
matrix is achieved. The assemble phase actually constructs a dsparse ma-
trix and fills it with the non–zero values. Figure 4 shows the scalability of
sparse on an SGI Altix 350. Although performance for commodity clusters
cannot be as good as that of an Altix, our initial experiments do indicate
good scalability on commodity clusters too. We will report more detailed
performance comparisons in a future paper.
3. spones, speye, spdiag, sprand etc. – Some basic functions implicitly
construct dsparse matrices.

3.3 Matrix Arithmetic

One of the goals in designing a sparse matrix data structure is that, wherever
possible, it should support matrix operations in time proportional to flops. As a
result, arithmetic on dsparse matrices is performed using a sparse accumulator
(SPA). Gilbert, Moler and Schreiber [10] discuss the design of the SPA in detail.
Matlab*P uses a separate SPA for each processor.

3.4 Indexing, Assignment and Concatenation

The syntax of matrix indexing in Matlab*P is the same as in Matlab. It is

of the form A(p, q). p and q can each be either a range (1 : n), or a permutation
vector or scalars. Depending on the context, however, this can mean different
things.
>> B = A(p,q)
In this case, the indexing is done on the right side of “=” which specifies that
B is a submatrix of A. This is the subsref operation in Matlab.
>> B(p,q) = A
On the other hand, indexing on the left side of “=” specifies that A should be
stored in a submatrix of B. This is the subsasgn operation in Matlab.
If p and q are both integers, A(p, q) directly accesses the dsparse data struc-
ture. If p or q are vectors or a range, A(p, q) calls find and sparse. find is the
reverse of sparse – it converts the matrix from CSR to [i, j, v] format. In this
format, it is very easy to find [i, j] pairs which satisfy the indexing criteria. The
resulting submatrix is then assembled by simply calling sparse .
Matlab also supports horizontal and vertical concatenation of matrices. The
following code, for example, concatenates A and B horizontally, C and D hori-
zontally, and finally concatenates the results of these two operations vertically.
>> S = [ A B; C D ]
The basic primitives, find and sparse are used to provide support for concate-
nation operations in Matlab*P.

3.5 Matvec
The matvec operation multiplies a dsparse matrix with a ddense column vector,
producing a ddense column vector as a result. Matvec is the kernel for many
iterative methods.
For the matvec, y = Ax, we have A and x distributed across processors by
rows. The submatrix of A at each processor will need a piece of x depending upon
its sparsity structure. When matvec is invoked for the first time on a dsparse
matrix A, Matlab*P computes a communication schedule for A and caches it.
When more matvecs are performed using A, this communication schedule does
not need to be recomputed, which saves some computing and communication
overhead, at the cost of extra space required to save the schedule. Matlab*P
also overlaps the communication and computation during matvec. This way,
each processor starts computing the result of the matvec whenever it receives a
piece of the vector from any other processor. Figure 6 also shows how matvec
scales in Matlab*P, since it forms the main computational kernel for conjugate
gradient.
Communication in matvec can be reduced by performing graph partitioning
of the graph of the sparse matrix. If fewer edges cross processors, lesser commu-
nication is required during matvec. Matlab*P can use several of the available
tools for graph partitioning. However, by default, Matlab*P does not perform
graph partitioning during matvec. The philosophy behind this decision is sim-
ilar to that in Matlab, that reorganizing data to make later operations more
efficient should be possible, but not automatic.
 0.4

0.35

Time per iteration of CG (secs)

0.3

0.25

0.2

0.15

0.1
Matlab
0.05
Matlab*P
0
40 50 60 70 80
Problem Size (grid3d)

Fig. 6. Time per iteration of CG, scalability of matvec

3.6 Solutions of Linear Systems

Matlab solves the linear system Ax = b with the matrix division operator,
x = A\b. In sequential Matlab, A\b is implemented as a polyalgorithm [10],
where every test in the polyalgorithm is cheaper than the next one.

1. If A is not square, solve the least squares problem.

2. Otherwise, if A is triangular, perform a triangular solve.
3. Otherwise, test whether A is a permutation of a triangular matrix (a “morally
triangular” matrix), permute it, and solve it if so.
4. Otherwise, if A is Hermitian and has positive real diagonal elements, find
a symmetric minimum degree ordering p of A, and perform the cholesky
factorization of A(p, p). If successful, finish with two sparse triangular solves.
5. Otherwise, find a column minimum degree order p, and perform the LU
factorization of A(:, p). Finish with two sparse triangular solves.

Different issues arise in parallel polyalgorithms. For example, morally trian-

gular matrices and symmetric matrices are harder to detect in parallel. One also
expects to be able to use iterative methods. Design for the right polyalgorithm
for \ in parallel is an active research problem. For now, Matlab*P uses a par-
allel general direct sparse solver for \, which is SuperLU DIST [14] by default,
although a user can choose to use MUMPS [1] too.
The open question at this point is, should Matlab*P use preconditioned
iterative methods to solve sparse linear systems instead of direct methods. Cur-
rently, iterative methods are not usable as a black box, and not yet suitable for
Matlab*P.

3.7 Iterative methods - Conjugate Gradient

Conjugate Gradient is an iterative method used to solve a symmetric, positive
definite system of equations. The same code is used for Matlab*P and matlab,
except that the input is dsparse in the Matlab*P case. In Fig 6, grid3d(k) is
a routine used from the meshpart [9] toolbox, which returns a k 3 × k 3 symmetric
positive definite matrix A with the structure of the k × k × k 7–point grid.

4 An Application in Computational Fluid Dynamics

g
y

1
ρ1 , µ 1
x
c =1

z 2
ρ , µ2
c =0 2

Fig. 7. Geometry of the Hele-Shaw cell (left). The heavier fluid is placed above the
lighter one. Either one of the fluids can be the more viscous one. Mesh point distribution
in the computational domain (right). A Chebyshev grid is employed in the y–direction,
and compact finite differences in the z–direction.

We are using a prototype version of Matlab*P in collaboration with a num-

ber of domain scientists for applications in computational science and engineer-
ing. We describe an application here.
Goyal and Meiburg [11] are studying the influence of viscosity variations on
the density–driven instability of two miscible fluids. The two fluids, of different
density and viscosity are in a vertical Hele–Shaw cell as shown in figure 7. This
problem is used to model porous media flows and finds applications in enhanced
oil recovery, fixed bed regeneration and groundwater flows.
Fig. 7 shows the discretization of the problem, which yields an algebraic
system of the form Aφ = σBφ. The eigenvalue σ represents the growth rate of
the perturbations, while the eigenvector φ reflects the shape of the perturbations.
A positive (negative) eigenvalue indicates unstable (stable) behavior. The system
has a 5×5 block structure reflecting the 5 variables at each mesh point (3 velocity
components u, v and w, relative concentration of the heavier fluid c, and pressure
p).
A discretization of 165 × 25 points turns out to be sufficient for this prob-
lem. Since we solve for 5 variables at each grid point, the matrix A is of the
 Fig. 8. Spy plots of the matrices A and B

size 20, 625 × 20, 625. The number of non–zeros is 3, 812, 450. The matrix is un-
symmetric, both in values and nonzero structure, as shown in the spy plots in
Figure 4. In order to calculate the largest eigenvalue, we use the power method
with shift and invert in Matlab*P.

function lambda = peigs (A, B, sigma, iter)

[m n] = size (A);
C = A - sigma * B;
y = rand (n*p, 1);

for k=1:iter
q = y ./ norm (y);
v = B * q;
y = C \ v;

theta = dot (q, y);

res = norm (y - theta * q);
if res <= 0.0001, break; end
end

lambda = 1 / theta + sigma;

Fig. 9. Matlab*P code for power method with shift and invert

The original non–Matlab*P code used LAPACK with ARPACK [13], while
the Matlab*P code is using SuperLU DIST with the power method as shown
in figure 4. We use a guess of 0.1 to initialize the power method and it converges
to 0.0194 which is enough precision for linear stability analysis. We use a cluster
with 16 processors to solve the generalized eigenvalue problem. Each node has
a 2.6GHz Pentium Xeon CPU, 3GB of RAM and a gigabit ethernet connection.
Results are presented in Table 1.

Table 1. Time to solve the generalized eigenvalue problem

No. of processors Time (seconds)

4 90
8 39
16 33

As a next step we want to incorporate a variable viscosity net flow through

the Hele–Shaw cell to incorporate the potentially destabilizing effects of vis-
cous fingering into play, so that the possibility of complex interactions between
density- and viscosity-driven instabilities arises. The existence of a more com-
plex flow field necessitates a finer grid and a larger domain size for the linear
stability calculations as compared to the previous case discussed where the two
fluids were essentially at rest with respect to each other. We expect that we will
require about 10 times more computing resources (CPU and memory) to tackle
these challenges .

5 Conclusion
The implementation of sparse matrices in Matlab*P is work in progress. Cur-
rent available functionality includes being able to construct sparse matrices,
perform element–wise arithmetic and indexing operations on them, multiply a
sparse matrix with a dense vector and solve linear systems. This level of func-
tionality allows us to implement several algorithms such as conjugate gradient
and the power method.
Much remains to be done. A complete implementation of sparse matrices
requires matrix–matrix multiplication and several factorizations (Cholesky, QR,
SVD etc). Improvements in the sorting code can lead to general improvements in
many parts of Matlab*P. It is also important to make existing graph partition-
ers available in Matlab*P – Meshpart and ParMetis. Several preconditioning
methods also need to be implemented for Matlab*P, since iterative methods
might possibly be the way to solve large linear systems.
The goal of sparse matrix support in Matlab*P is to provide an interactive
environment for users to perform operations on large sparse matrices in parallel,
while being compatible with Matlab. Our current implementation is ready to
be used for simple real life problems.

6 Acknowledgements
David Cheng made useful contributions to the sorting code. We had several useful
discussions with Parry Husbands, Per–Olof Persson, Ron Choy, Alan Edelman,
Nisheet Goyal and Eckart Meiburg.
 References
1. Patrick Amestoy, Iain S. Duff, and Jean-Yves L’Excellent. Multifrontal solvers
within the PARASOL environment. In PARA, pages 7–11, 1998.
2. D. Arnold, S. Agrawal, S. Blackford, J. Dongarra, M. Miller, K. Seymour, K. Sagi,
Z. Shi, and S. Vadhiyar. Users’ Guide to NetSolve V1.4.1. Innovative Computing
Dept. Technical Report ICL-UT-02-05, University of Tennessee, Knoxville, TN,
June 2002.
3. L. S. Blackford, J. Choi, A. Cleary, E. D’Azevedo, J. Demmel, I. Dhillon, J. Don-
garra, S. Hammarling, G. Henry, A. Petitet, K. Stanley, D. Walker, and R. C.
Whaley. ScaLAPACK Users’ Guide. SIAM, Philadelphia, PA, 1997.
4. Guy E. Blelloch, Charles E. Leiserson, Bruce M. Maggs, C. Greg Plaxton,
Stephen J. Smith, and Marco Zagha. A comparison of sorting algorithms for the
connection machine cm-2. In Proceedings of the third annual ACM symposium on
Parallel algorithms and architectures, pages 3–16. ACM Press, 1991.
5. Oskar Bruning, Jack Holloway, and Adnan Sulejmanpasic. Matlab *p visualization
package. 2002.
6. Long Yin Choy. Parallel Matlab survey. 2001. http://theory.csail.mit.edu/
∼cly/survey.html.
7. Long Yin Choy. MATLAB*P 2.0: Interactive supercomputing made practical. M.S.
Thesis, EECS, 2002.
8. D. E. Culler, A. Dusseau, R. Martin, and K. E. Schauser. Fast parallel sorting under
LogP: from theory to practice. In Proceedings of the Workshop on Portability and
Performance for Parallel Processing, Southampton, England, July 1993. Wiley.
9. John. R. Gilbert, Gary L. Miller, and Shang-Hua Teng. Geometric mesh partition-
ing: Implementation and experiments. SIAM Journal on Scientific Computing,
19(6):2091–2110, 1998.
10. John R. Gilbert, Cleve Moler, and Robert Schreiber. Sparse matrices in MATLAB:
Design and implementation. SIAM Journal on Matrix Analysis and Applications,
13(1):333–356, 1992.
11. Nisheet Goyal and Eckart Meiburg. Unstable density stratification of miscible fluids
in a vertical hele-shaw cell: Influence of variable viscosity on the linear stability.
Journal of Fluid Mechanics, (To appear), 2004.
12. P. Husbands and C. Isbell. MATLAB*P: A tool for interactive supercomputing.
The Ninth SIAM Conference on Parallel Processing for Scientific Computing, 1999.
13. R. B. Lehoucq, D. C. Sorensen, and C. Yang. ARPACK Users Guide: Solution of
Large Scale Eigenvalue Problems with Implicitly Restarted Arnoldi Methods. SIAM,
Philadelphia, 1998.
14. Xiaoye S. Li and James W. Demmel. Superlu dist: A scalable distributed–memory
sparse direct solver for unsymmetric linear systems. ACM Trans. Math. Softw.,
29(2):110–140, 2003.

View publication stats