Open Software Solutions

Source: Microsoft Support

RAPID PUBLISHING ARTICLES PROVIDE INFORMATION DIRECTLY FROM WITHIN THE MICROSOFT SUPPORT ORGANIZATION. THE INFORMATION CONTAINED HEREIN IS CREATED IN RESPONSE TO EMERGING OR UNIQUE TOPICS, OR IS INTENDED SUPPLEMENT OTHER KNOWLEDGE BASE INFORMATION.

Write a VC++ program that does many mathematical calculations using floating-point values. Compile it with both Visual Studio 2003 (VC++ 7.1) & 2005  (VC++ 8.0) compilers, and run both executables independently.

The VC++ 7.1-compiled executable takes less time to execute compared to the VC++ 8.0 executable.

By default in Visual Studio 2005, the /fp:precise floating-point model is used, which converts floats to doubles, and then does double-precision math. In Visual Studio 2003, this conversion does not happen, and hence it takes less time.

If performance is to be given more priority than precision, compile the VC++ 8.0 (Visual Studio 2005) code with /fp:fast option. However, if the user code is sensitive to the precision with which the commutation is done, /fp:fast may cause unexpected results. If speed is the primary goal, then /fp:fast should be tried, but making sure to verify that the program is giving acceptable results in calculations.

In VS2003, /Op- is the default floating model used. Whereas in VS2005, the new floating-point models (fast, precise, and strict) were added, with the default being /fp:precise. In many ways, /Op- is more like /fp:fast than /fp:precise, so to get the default behavior from VS2003, /fp:fast should be used. But /fp:fast in VC++ 8.0 will not always exactly match /Op- from VC++ 7.1.

MICROSOFT AND/OR ITS SUPPLIERS MAKE NO REPRESENTATIONS OR WARRANTIES ABOUT THE SUITABILITY, RELIABILITY OR ACCURACY OF THE INFORMATION CONTAINED IN THE DOCUMENTS AND RELATED GRAPHICS PUBLISHED ON THIS WEBSITE (THE “MATERIALS”) FOR ANY PURPOSE. THE MATERIALS MAY INCLUDE TECHNICAL INACCURACIES OR TYPOGRAPHICAL ERRORS AND MAY BE REVISED AT ANY TIME WITHOUT NOTICE.

TO THE MAXIMUM EXTENT PERMITTED BY APPLICABLE LAW, MICROSOFT AND/OR ITS SUPPLIERS DISCLAIM AND EXCLUDE ALL REPRESENTATIONS, WARRANTIES, AND CONDITIONS WHETHER EXPRESS, IMPLIED OR STATUTORY, INCLUDING BUT NOT LIMITED TO REPRESENTATIONS, WARRANTIES, OR CONDITIONS OF TITLE, NON INFRINGEMENT, SATISFACTORY CONDITION OR QUALITY, MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE, WITH RESPECT TO THE MATERIALS.