Doc No:           WG14/N1369

Date:               2009-02-27

Project:            Decimal floating-point TR 24732

Reply to:         Edison Kwok

ekwok@ca.ibm.com

 

Subject:           Problems with FLOAT_CONST_DECIMAL64 pragma  

 

This paper proposes to clarify the intended behavior of the FLOAT_CONST_DECIMAL64 pragma, and possibly to rename it to a more meaningful name.

 

Background

 

The current wording in section 7.1.1 of the TR implies that the FLOAT_CONST_DECIMAL64 pragma applies to unsuffixed hexadecimal floating-point constants as well, but clearly the intent of the TR is not to support hexadecimal floating-constants for decimal types (see constraints violation in section 7.)

 

Also the name of the pragma may be misleading. In N1295 where the pragma was originally proposed, the more verbose name UNSUFFIXED_FLOAT_CONSTANT_IS_DECIMAL64 was suggested. The committee thought it was too long and decided to abbreviate it to FLOAT_CONST_DECIMAL64. At least one committee member thinks that a name such as DOUBLE_CONST_DECIMAL64 may be a better name for the intended purpose of the pragma.

 

Suggested changes to TR

 

In 7.1.1 of the TR, change "unsuffixed floating-point constants" to "unsuffixed decimal floating-point constants".

 

Change the name of the pragma to DOUBLE_CONST_DECIMAL64 (or some other appropriate name.)