From mboxrd@z Thu Jan  1 00:00:00 1970
X-Spam-Checker-Version: SpamAssassin 3.4.4 (2020-01-24) on inbox.vuxu.org
X-Spam-Level: 
X-Spam-Status: No, score=0.0 required=5.0 tests=none autolearn=ham
	autolearn_force=no version=3.4.4
Received: (qmail 30224 invoked from network); 22 Jan 2021 04:56:27 -0000
Received: from 1ess.inri.net (216.126.196.35)
  by inbox.vuxu.org with ESMTPUTF8; 22 Jan 2021 04:56:27 -0000
Received: from vultr.musolino.id.au ([45.76.123.158]) by 1ess; Thu Jan 21 23:28:08 -0500 2021
Received: from 58.170.131.8 ([58.170.131.8]) by vultr; Fri Jan 22 15:27:31 +1100 2021
Message-ID: <DF333D35FBF66DF7B8B2BF77E3A53C45@musolino.id.au>
From: Alex Musolino <alex@musolino.id.au>
Date: Fri, 22 Jan 2021 14:57:29 +1030
To: 9front@9front.org
In-Reply-To: <ABFD8ED6-5269-49D5-85A6-8693BF646CC9@gmail.com>
MIME-Version: 1.0
Content-Type: text/plain; charset="US-ASCII"
Content-Transfer-Encoding: 7bit
List-ID: <9front.9front.org>
List-Help: <http://lists.9front.org>
X-Glyph: ➈
X-Bullshit: stable dependency GPU reduce/map layer 
Subject: Re: [9front] ape modf infinite recursion
Reply-To: 9front@9front.org
Precedence: bulk

We might need an isNaN check in there but it looks like we're also
generating the wrong code for the <0 comparison.  6c emits UCOMISD
followed by JCC.  The UCOMISD will set all three ZF, PF, CF flags to 1
when either operand is NaN to indicate there is no ordering, but the
JCC only looks at CF and incorrectly determines that NaN < 0.