linux-mips
[Top] [All Lists]

Re: Compiler problem in glibc

To: Hartvig Ekner <hartvige@mips.com>
Subject: Re: Compiler problem in glibc
From: "H . J . Lu" <hjl@lucon.org>
Date: Sun, 17 Mar 2002 10:41:24 -0800
Cc: user alias <linux-mips@oss.sgi.com>
In-reply-to: <200203171052.g2HAqGb27844@coplin09.mips.com>; from hartvige@mips.com on Sun, Mar 17, 2002 at 11:52:15AM +0100
References: <200203171052.g2HAqGb27844@coplin09.mips.com>
Sender: owner-linux-mips@oss.sgi.com
User-agent: Mutt/1.2.5i
On Sun, Mar 17, 2002 at 11:52:15AM +0100, Hartvig Ekner wrote:
> I have found a problem in glibc caused by the gcc-2.96-99.1 compiler
> from H.J's miniport.
> 
> in the exp() function (file w_expf.c), there is code like: 
> 
> #ifdef __STDC__
>         float __expf(float x)           /* wrapper expf */
> #else
>         float __expf(x)                 /* wrapper expf */
>         float x;
> #endif
> {
> #ifdef _IEEE_LIBM
>         return __ieee754_expf(x);
> #else
>         float z;
>         z = __ieee754_expf(x);
>         if(_LIB_VERSION == _IEEE_) return z;
> 
>         if(__finitef(x)) {
>             if(x>o_threshold)
> 
> 
> (IEEE_LIBM is not set). Note that there are two function calls (ieee754_expf
> and finitef()) followed by a FP if-statement (x>o_threshold). This 
> translates into:
> 

I believe it has been reported before and is fixed in gcc 3.1 at the
time.


H.J.

<Prev in Thread] Current Thread [Next in Thread>