X-Authentication-Warning: delorie.com: mailnull set sender to djgpp-workers-bounces using -f
Date: Sun, 03 Mar 2002 18:20:00 +0200
From: "Eli Zaretskii" <eliz AT is DOT elta DOT co DOT il>
Sender: halo1 AT zahav DOT net DOT il
To: djgpp-workers AT delorie DOT com
Message-Id: <5137-Sun03Mar2002182000+0200-eliz@is.elta.co.il>
X-Mailer: emacs 21.2.50 (via feedmail 8 I) and Blat ver 1.8.9
In-reply-to: <200203031431.g23EV4e07176@mother.ludd.luth.se> (message from
	Martin Str|mberg on Sun, 3 Mar 2002 15:31:04 +0100 (MET))
Subject: Re: cygnus libm test
References:  <200203031431 DOT g23EV4e07176 AT mother DOT ludd DOT luth DOT se>
Reply-To: djgpp-workers AT delorie DOT com
Errors-To: nobody AT delorie DOT com
X-Mailing-List: djgpp-workers AT delorie DOT com
X-Unsubscribes-To: listserv AT delorie DOT com
Precedence: bulk

> From: Martin Str|mberg <ams AT ludd DOT luth DOT se>
> Date: Sun, 3 Mar 2002 15:31:04 +0100 (MET)
> 
> asinf has a lot of inaccurate answers in bit 33.
> 
> atan2f has one inaccurate answer (ia, my abbr.) in bit 33.
> 
> gamma has two ias in bits 58 and 61.
> 
> gammaf has three ias, two in bit 33 and one in bit 31.
> 
> j0 has one ia in bit 61.
> 
> j0f has a lot of ias; mostly in bit 33, one in bit 31.
> 
> j1 has two ias, one bit 60 and one in 61.
> 
> j1f has a lot of ias in bits 32 and 33.
> 
> jn has a lot of ias in bits 59-61.
> 
> jnf has a lot of ias in bits 29-33.
> 
> log1p has six ias in bit 61.

These are not serious problems, but what I miss is the differences
against standard.results.  Does the new compile only add new
inaccuracies, or does it add some and remove others?

> log2 has two in bit 12 and one in bit 0!?

What results are those?  Are they finite or Inf/NaN?

> log2 also has errno wrong-

What value of errno?  If log2 is documented to produce that value,
it's okay.

> sqrt has two in bit 0!?

That's a sign bit?  A NaN, perhaps?  (Our NaN has its sign bit set.)

> y0f has a lot of ias in bit 33.
> 
> y1 has two in bit 61 and one in 59.
> 
> y1f has a lot of ias in bits 29-33.
> 
> yn has a lot of ias in bits 59-61.
> 
> ynf has a lot of ias in bits 29-33.

These are okay, unless the new compile adds much more new
inaccuracies.

> fpsetroundtoi/fpgetroundtoi has one ia: 00000000 should be 00000001.

This is a known limitation of emulations of these functiosn I threw
together to get the test suite to run.  I think you will see the same
in standard.results.