08 June 2009
Comparing apples
Gauging the power of a processor has always been an interesting process for engineers. For the longest time, it was the manufacturer – who also tended to be the supplier of the computer in which the chip was housed – who provided the information.
And it came as no surprise to find out that manufacturers did their very best to make their chips perform as well as possible. They had a number of ways of doing this – including optimising code to within an inch of its life – but none of these generally had any reflection on the way in which the processor would perform when in the hands of customers.
The accepted way of presenting the information was in terms of MIPS. Although the acronym was officially an abbreviation for millions of instructions per second, the general interpretation of the term was a 'meaningless indication of processor speed'.
MIPS, as a metric, was fairly useless because it wasn't an 'apples with apples' approach: you couldn't use the approach to compare risc and cisc processors. And there were VAX MIPS, Whetstone MIPS and the amusing pun of Dhrystone MIPS; all in different versions.
Things have, of course, changed and the Embedded Microprocessor Benchmark Consortium (EEMBC) has done a lot to engineer that change.
Now EEMBC has launched CoreMark and the benchmark is the first one from the organisation to be openly available. The benchmark – which works with most kinds of processor – exercises the processor using common operations, delivering, in EEMBC's opinion, a realistic mixture of operations.
CoreMark will help engineers to compare those apples; will it ever replace the hype of the MIPS figure?
Author
Graham Pitcher
Supporting Information
Websites
http://www.coremark.org
http://www.eembc.org
This material is protected by Findlay Media copyright
See Terms
and Conditions.
One-off usage is permitted but bulk copying is not.
For multiple copies contact the
sales team.