You are on page 1of 17

Parallel Computing in

CFD
L Sugumar
OE12S010
Department of Ocean Engineering
IIT Madras

Contents

Introduction
Computer Memory Architectures
Parallel Programming Models
CFD Computationally intensive task
Conclusion
References

Introduction

Traditionally software has been written for serial


computation.

CPU executes the set of instruction one by one in


serial computation

Parallel computing is the simultaneous use of


multiple compute resources to solve a
computational problem.

Computer Memory Architectures

Single Processor and memory(RAM)

Multiple Processor share same memory.

Simple distributed memory architecture.

Hybrid distributed memory architecture.

Parallel Programming Models


In software development considering the
characteristics and capabilities of the hardware for
which the software is being build is vital.

Shared memory architecture OpenMP


Distributed memory architecture MPI
Hybrid distributed memory architecture(OpenMP + MPI)

Serial code computation of 2 D steady state heat equation

Parallel code computation of Laplace equation


in shared memory architecture.

Parallel code computation of Laplace equation


in distributed memory architecture

At the end of parallel computation one gets the same


temperature distribution contour as that of the serial
computation with substantial computational time reduction.

What basically drives the CFD community to go behind


parallel computing?

Calculation of Lift and Drag of an Airfoil section in high speed


compressible flow( say transonic flow)

Computational domain discretisation Grid generation

Big computational domain with fine grid

Far view of the grid

Near view of the grid

Can you guess many grid points are there in the above?
58500 grid points.
At Each grid points one has to solve
Mass continuity equation
X Momentum equation
Y- Momentum equation
Energy equation
Turbulence Equation, based on the turbulence model it can be
either one or two more equations.

So CFD is computationally intensive task and it needs a great


deal of memory, processing speed and storage capacity.

Conclusion
High computational time and memory issue are the two
prime factors which basically drives the CFD community to
go behind parallel computing.
Parallel computing is very fast, it saves time and money of
an organization.
Parallel computing is the future of the computing.

References

Internet Parallel Computing Archive, wotug.ukc.ac.uk/parallel

Introduction to Parallel Computing,


www.llnl.gov/computing/tutorials/parallel_comp/#Whatis

Parallel Programming in C with MPI and OpenMP, Michael J.


Quinn, McGraw Hill Higher Education, 2003

The New Turing Omnibus, A. K. Dewdney, Henry Holt and


Company, 1993

Thank you

You might also like