Wolfram Research, Inc. announced that it believes that its software is unaffected by the recently discovered “flag erratum” bug in the Pentium Pro and Pentium II microprocessors.
“Accuracy is paramount,” said Wolfram President/CEO Stephen Wolfram. “Many, many people are depending on us, directly and indirectly, to get things right. We work closely with the hardware vendors to help find and analyze any problems that may affect the accuracy of our products.”
Roger Germundsson, Director of Research and Development at Wolfram Research, led the company’s effort to determine the bug’s possible impact on Mathematica. “We have been in contact with Intel and have used the information provided to assess our code,” Germundsson said. “We believe this bug will not affect Mathematica.”
Microprocessors are designed to store numbers and manipulate numbers internally using various fixed-size formats, each of which can handle numbers only up to a given size. Mathematica supports arbitrary-precision arithmetic without these constraints because it treats numbers in a manner independent of the details of the microprocessor hardware. For this reason, Mathematica is widely used by groups that design and test floating-point hardware.