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

RE: Implementor support for the binary interchange formats



True, this needs to be qualified (was done actually in a subsequent email):

  Binary interchange formats are used whenever the IEEE 754-2008 Decimal Floating-Point arithmetic is used on Intel Architecture or equivalent processors, using the Intel® Decimal Floating-Point Math Library.

Thanks,

Marius Cornea

 

From: Mike Cowlishaw [mailto:mfc@xxxxxxxxxxxxxxx]
Sent: Monday, February 28, 2011 9:35 AM
To: 'IEEE 754'
Cc: Cornea, Marius
Subject: RE: Implementor support for the binary interchange formats

 

Marius wrote:

>> With all this discussion about DECIMAL interchange formats, I haven't seen much mention of the BINARY interchange formats. 

Hello Chuck,

Binary interchange formats are used whenever the IEEE 754-2008 Decimal Floating-Point arithmetic is used on Intel Architecture or equivalent processors. 

 

That's simply  not true.  I use that arithmetic on Intel processors every day, and always use the decimal encoding (except for some testing a few years ago) because it's faster overall.  There are at least four implementations I know of that use or provide the decimal encoding on Intel processors, and only one that uses the binary encoding.

 

In particular, this encoding is supported – among others - as the main encoding for decimal FP values in GCC (starting with version 4.4 and for systems mentioned above), in the Intel C/C++ Compiler (part of various Intel software products), and one HP C/C++ Compiler. 

 

In GCC the decimal and binary encodings are both supported; the decimal encoding was supported first, I believe.   Good to hear that Intel and HP compilers now support decimal FP, however!   :-)

 

Mike 

 


754 | revision | FAQ | references | list archive