This is the mail archive of the libstdc++@gcc.gnu.org mailing list for the libstdc++ project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

RE: Linux doubles precision bug?


Ian

Thank you so much for your reply.  I also received good advise from
Phillip Jordan.  He noted that the -ffloat-store option would solve my
problem but that there is an performance hit using this compiler option.

Betty 

-----Original Message-----
From: Ian Lance Taylor [mailto:iant@google.com] 
Sent: Tuesday, February 26, 2008 3:58 PM
To: Betty Evans
Cc: libstdc++@gcc.gnu.org
Subject: Re: Linux doubles precision bug?

"Betty Evans" <bevans@ittvis.com> writes:

> Running the same code gives me different results on Linux 32 bit and
64
> bit computers.  I'm not sure if it is a bug or not but other compilers
> give the same results when run on 32 and 64-bit systems.

This question would be better directed to the mailing list
gcc-help@gcc.gnu.org.  Please take any followups there.  Thanks.

You didn't mention the processor, but let's assume it's x86.  On x86,
32-bit floating point code uses the x87 registers with 80-bit
precision.  64-bit floating point code uses the SSE registers with
64-bit precision.  Different results are normal and expected.

You may find it interesting to look at http://gcc.gnu.org/PR323 .

Ian


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]