 
Summary: 12TH INTL. CONFERENCE IN ELECTRICAL IMPEDANCE TOMOGRAPHY (EIT 2011), 46 MAY 2011, UNIV. OF BATH 1
Scaling the EIT Problem
Alistair Boyle, Andy Adler, Andrea Borsic
Abstract
There are a number of interesting problems that could be tackled if the computing capacity of
current EIT systems can be improved. We examined the performance of two such systems and found
that a noticeable portion of the compute time is spent in finding the solution of sparse matrices. We
developed and used a new sparse matrix testbench, MeagreCrowd, to evaluate a selection of these
sparse matrix solvers and found that there are definite performance gains available.
I. INTRODUCTION
THE procedure used to solve inverse problems in areas such as Electrical Impedance
Tomography (EIT) involves numerous steps to determine a solution. Software such as
EIDORS [1] implement these algorithms using the numerical linear algebra tools in MatLab
to manipulate sparse matrices that result from a Finite Element Method (FEM) formulation.
Computing power has increased according to Moore's law for 40 years, approximately
doubling every 18 months [2]. Until recently, this has been achieved by increasing processor
clock frequencies and a corresponding improvement in instructionspersecond throughput and
memory bandwidth. Taking advantage of these improvements has involved little software cost.
Purchasing a new processor gains one the advantages of the new hardware when the software
is recompiled. In recent years, the benefits of increasing clock frequency have been limited
