Research information of Coding
efficiency projects
Research Information from the Internet
Prepared by:
Divya Markend Joshi
Date: October 2, 2004
Global Computers, Inc. is a global
company that strives to achieve customer satisfaction in every
aspect of services delivered to the customers. However, despite
being in the computer or rather in
the Information Technology (IT)
industry, the company’s IT department is in a worrying state as it
is facing problems regarding the computing power needed for their
supercomputers and mainframes.
The department has narrowed down the
core of the problem to being their programs not being coded for
efficiency. This has therefore caused Global Computers, Inc.
thousands and millions of dollars annually to enhance its computing
power. That is not all, the number of supercomputers and mainframes
increase every year because the utilization of the systems are not
being maximized, hence the situation.
Several companies and corporations that
have formally implemented the Coding
Performance Improvement project have
clearly benefited from it. I researched on two sites that
stated the success of implementing the
Coding Performance Improvement project.
A team of researchers from NASA carried
out the Coding Performance Improvement project to measure the
performance of the baseline AMR Navier-Stokes code. A description of
the problem
was presented including the input as
well as the techniques used to measure wall-clock times
along with serial and parallel results
of the baseline on the code ‘halem’ machine at GSFC. The
approximate peak memory of usage for two
benchmark problems in serial was also measured.
After implementing the new codes, the
researches were able to present a timing measurement
that displayed an improvement in
performance after various code optimizations. Through
vigorous review on the original baseline
codes, errors were encountered, and as such
necessary changes were made to the
codes.
The primary outcome of the improvised
codes was the reduced wall-clock time by factors ranging
from 1.76 to 7.893. These factors varied
depending on the complexity of the problem and the
number of processors. Another outcome of
executing this project was that they were able to
reduce the amount of memory that was
needed by approximately 5% and 20%.
Another successful story of implementing
the Coding Performance Improvement project is that of
an aeronautical company. Lockheed Martin
had also experienced major problems in the
company’s management system because the
system lagged way behind time and it was almost impossible to get
any work done. Furthermore, being a company that deals with
real-time systems, the company had to act fast in order to stay in
the industry. Lockheed Martin applied Six Sigma
principles in the Coding Performance
Improvement project. The company reworked on the codes
for the management system. The outcome
was an obvious cost reduction in maintaining the
system. Apart from that, the amount of
time wasted just waiting for the system to load also
saw a significant improvement.
Initially, Lockheed Martin used two
approaches to deal with the problem. The first approach is
known as Diminishing Manufacturing
Sources (DMS). The second approach procures a
lifetime quantity while still in
production or a periodic major programmed retrofit is
conducted. However, both approaches were
expensive and either risky or wasteful.
A programmed retrofit typically involved
the redesign of the computer hardware components and
the recoding of the software. This is
further followed by an extensive recertification testing,
particularly when there are flight
critical functions. So, in order to curb this problem, the
Lockheed Martin team developed a
life-cycle technology management approach. The
life-cycle technology management
approach focuses on true OSA and "evolutionary
technology refreshment" that achieves
software portability (independence from hardware), both
within an avionics box and throughout
the entire aircraft. Standardized, non-proprietary
interfaces provide true "plug-and-play"
capability, much like the mix-and-match capability of
hardware peripherals and software in
today’s commercial personal computers.
The new approach allows boards or
modules, incorporating new technology, to be changed out
as preferred spares on an attrition
basis, with no impact to form, fit or function. Large
spares inventories are not required. The
newer-generation technology also offers performance
growth potential, lower cost and higher
reliability.
