[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

RE: Implementor support for the binary interchange formats

PS:  In COBOL terms, with regard to "conversion functions", when a numeric item is used as an arithmetic operand, the conversion is implicit.  The user does not have to code for them.  We might decide to provide in the standard for the ability to call those functions directly, but the COBOL user wouldn't be expected to use such calls explicitly as part of the arithmetic operation, and the COBOL programmer's reaction to seeing such a call in that enviroment would most likely be "What's THIS here for???".       
    -Chuck Stevens

From: charles.stevens@xxxxxxxx
To: marius.cornea@xxxxxxxxx
CC: stds-754@xxxxxxxxxxxxxxxxx
Subject: RE: Implementor support for the binary interchange formats
Date: Mon, 28 Feb 2011 10:26:58 -0700

I'm not talking about the binary ENCODING (e.g., "decimal128 using binary encoding"); already understood there were implementations providing that.
What I'm talking about is "binary128 format" ITSELF, specifically as distinct from decimal128 using EITHER encoding. 
    -Chuck Stevens. 

From: marius.cornea@xxxxxxxxx
To: charles.stevens@xxxxxxxx
CC: stds-754@xxxxxxxxxxxxxxxxx
Date: Mon, 28 Feb 2011 09:17:23 -0800
Subject: RE: Implementor support for the binary interchange formats

All the cases I mentioned are based on the Intel® Decimal Floating-Point Math Library (http://www.netlib.org/misc/intel/).

The library uses the binary encoding format, but provides conversion functions to/from the decimal encoding format.

To operate on data encoded using the decimal format would lead to a significant performance degradation, compared to operating directly on the native encoding format (binary).

It may also require changes to the source code, to insert calls to the conversion functions.

Thank you,

Marius Cornea


P.S.: Minor correction – DFP support in GCC for Intel Architecture started with GCC 4.3, not 4.4


From: Charles Stevens [mailto:charles.stevens@xxxxxxxx]
Sent: Monday, February 28, 2011 9:07 AM
To: Cornea, Marius
Cc: IEEE 754
Subject: RE: Implementor support for the binary interchange formats


Well, then, the question becomes this.  Is there an implementation (planned or actual) that supports the binary128 format directly in arithmetic and DOES NOT ALSO SUPPORT the decimal128 format in arithmetic?  You seem to indicate that the Intel architecture supports either one without prejudice. 
    -Chuck Stevens


754 | revision | FAQ | references | list archive