Article ID: 257496 - Last Review: August 7, 2007 - Revision: 1.2 FIX: Optimized Code Incorrectly Compares Floating Point NumbersThis article was previously published under Q257496 On This PageSYMPTOMS
When running an application as a compiled EXE and a comparison of numeric data types is evaluated, incorrect results may be returned when the comparison involves numbers of the data type Single or Double and the application was compiled with optimizations.
CAUSE
This bug is caused by a problem in the way the compiler creates optimized code for floating point comparisons. However, the problem does not always occur just by simply comparing one floating point value to another. It is affected by the additional code that surrounds the actual code that performs the calculation. For example, in the Steps to Reproduce Behavior in the "More Information" section of this article, it is the combination of the following that produces the incorrect results:
RESOLUTION
There are two possible means to solve this issue. The recommended way is to install the latest service pack for Visual Studio 6.0. An alternative way is to recompile the executable with no compiler optimizations.
STATUSMicrosoft has confirmed that this is a bug in the Microsoft products that are listed at the beginning of this article. This bug was corrected in the latest service pack for Visual Studio 6.0. For additional information about Visual Studio service packs, click the following article numbers to view the articles in the Microsoft Knowledge Base: 194022
(http://support.microsoft.com/kb/194022/EN-US/
)
INFO: Visual Studio 6.0 Service Packs, What, Where, Why To download the latest Visual Studio service pack, visit the following Microsoft Web site:194295 (http://support.microsoft.com/kb/194295/EN-US/ ) HOWTO: Tell That a Visual Studio Service Pack Is Installed http://msdn2.microsoft.com/en-us/vstudio/Aa718353.aspx
(http://msdn2.microsoft.com/en-us/vstudio/Aa718353.aspx)
MORE INFORMATIONSteps to Reproduce Behavior
| Article Translations
|
Back to the top
