This is the mail archive of the newlib@sourceware.org mailing list for the newlib 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: [PATCH] fmin/fmax behaviour on NaNs

From: Jeff Johnston <jjohnstn at redhat dot com>
To: Stephen Huw CLARKE <stephen dot clarke at st dot com>
Cc: newlib at sources dot redhat dot com
Date: Wed, 10 Aug 2005 17:03:28 -0400
Subject: Re: [PATCH] fmin/fmax behaviour on NaNs
References: <001901c59c36$99c21e50$840e81a4@bri0199>

Patch checked in. Thanks.

-- Jeff J.

Stephen Huw CLARKE wrote:

The implementations of fmax and fmin in newlib/libm/common
appear not to conform to section F.9.9.2/3 of the C99 spec.
That spec. states:
"If just one argument is a NaN, the fmax functions return the
other argument (if both arguments are NaNs, the functions
return a NaN)."
The following patch fixes this.

Steve.

--- newlib/libm/common/orig-sf_fmax.c 2002-06-07 22:59:56.000000000 +0100 +++ newlib/libm/common/sf_fmax.c 2005-08-08 17:13:08.000000000 +0100 @@ -15,9 +15,9 @@ #endif { if (__fpclassifyf(x) == FP_NAN) - return x; - if (__fpclassifyf(y) == FP_NAN) return y; + if (__fpclassifyf(y) == FP_NAN) + return x; return x > y ? x : y; } --- newlib/libm/common/orig-s_fmax.c 2002-06-07 22:59:56.000000000 +0100 +++ newlib/libm/common/s_fmax.c 2005-08-08 17:07:04.000000000 +0100 @@ -17,9 +17,9 @@ #endif { if (__fpclassifyd(x) == FP_NAN) - return x; - if (__fpclassifyd(y) == FP_NAN) return y; + if (__fpclassifyd(y) == FP_NAN) + return x; return x > y ? x : y; } --- newlib/libm/common/orig-sf_fmin.c 2002-06-07 22:59:56.000000000 +0100 +++ newlib/libm/common/sf_fmin.c 2005-08-08 17:24:57.000000000 +0100 @@ -15,9 +15,9 @@ #endif { if (__fpclassifyf(x) == FP_NAN) - return x; - if (__fpclassifyf(y) == FP_NAN) return y; + if (__fpclassifyf(y) == FP_NAN) + return x; return x < y ? x : y; } --- newlib/libm/common/orig-s_fmin.c 2002-06-07 22:59:56.000000000 +0100 +++ newlib/libm/common/s_fmin.c 2005-08-08 17:07:16.000000000 +0100 @@ -17,9 +17,9 @@ #endif { if (__fpclassifyd(x) == FP_NAN) - return x; - if (__fpclassifyd(y) == FP_NAN) return y; + if (__fpclassifyd(y) == FP_NAN) + return x; return x < y ? x : y; }

References:
- [PATCH] fmin/fmax behaviour on NaNs
  - From: Stephen Huw CLARKE

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