From mboxrd@z Thu Jan  1 00:00:00 1970
Return-Path: <keithp@keithp.com>
Received: from elaine.keithp.com (home.keithp.com [63.227.221.253])
 by sourceware.org (Postfix) with ESMTPS id 0EA403857C54
 for <newlib@sourceware.org>; Sat,  8 Aug 2020 22:34:17 +0000 (GMT)
DMARC-Filter: OpenDMARC Filter v1.3.2 sourceware.org 0EA403857C54
Received: from localhost (localhost [127.0.0.1])
 by elaine.keithp.com (Postfix) with ESMTP id BDBE33F2D304
 for <newlib@sourceware.org>; Sat,  8 Aug 2020 15:34:15 -0700 (PDT)
X-Virus-Scanned: Debian amavisd-new at keithp.com
Received: from elaine.keithp.com ([127.0.0.1])
 by localhost (elaine.keithp.com [127.0.0.1]) (amavisd-new, port 10024)
 with LMTP id Aabu6HrFo-qZ for <newlib@sourceware.org>;
 Sat,  8 Aug 2020 15:34:15 -0700 (PDT)
Received: from keithp.com (koto.keithp.com [10.0.0.2])
 by elaine.keithp.com (Postfix) with ESMTPSA id E94CB3F2D301
 for <newlib@sourceware.org>; Sat,  8 Aug 2020 15:34:14 -0700 (PDT)
Received: by keithp.com (Postfix, from userid 1000)
 id ABA7715821C1; Sat,  8 Aug 2020 15:34:14 -0700 (PDT)
From: Keith Packard <keithp@keithp.com>
To: newlib@sourceware.org
Subject: [PATCH 0/3] ARM with only 32-bit floats do not have fast 64-bit FMA
Date: Sat,  8 Aug 2020 15:34:10 -0700
Message-Id: <20200808223413.4015633-1-keithp@keithp.com>
X-Mailer: git-send-email 2.28.0
MIME-Version: 1.0
Content-Transfer-Encoding: 8bit
X-Spam-Status: No, score=-5.6 required=5.0 tests=BAYES_00, DKIM_SIGNED,
 DKIM_VALID, DKIM_VALID_AU, DKIM_VALID_EF, SPF_HELO_NONE, SPF_PASS,
 TXREP autolearn=ham autolearn_force=no version=3.4.2
X-Spam-Checker-Version: SpamAssassin 3.4.2 (2018-09-13) on
 server2.sourceware.org
X-BeenThere: newlib@sourceware.org
X-Mailman-Version: 2.1.29
Precedence: list
List-Id: Newlib mailing list <newlib.sourceware.org>
List-Unsubscribe: <https://sourceware.org/mailman/options/newlib>,
 <mailto:newlib-request@sourceware.org?subject=unsubscribe>
List-Archive: <https://sourceware.org/pipermail/newlib/>
List-Post: <mailto:newlib@sourceware.org>
List-Help: <mailto:newlib-request@sourceware.org?subject=help>
List-Subscribe: <https://sourceware.org/mailman/listinfo/newlib>,
 <mailto:newlib-request@sourceware.org?subject=subscribe>
X-List-Received-Date: Sat, 08 Aug 2020 22:34:19 -0000

I added some new test configurations to my CI system for picolibc and
discovered that when the new math code was built on 32-bit ARM
processors with only single-precision floating hardware, several math
functions were returning imprecise results. I got the expected results
on processors with no FPU and on processors with both 32- and 64- bit
FPUs.

I discovered that the affected functions were using the 'fma' function
on this hardware, even though (lacking 64-bit HW support), that
function was being emulated without the required precision.

This all boiled down to math_config.h incorrectly detecting 64-bit FMA
support on ARM processors.

This patch series contains three changes:

 1. fix the fast FMA process so that 32-bit ARM processors without 64-bit FMA
    support don't use 'fma' for the new math functions

 2. Add detection of fast FMAF, which 32-bit ARM processors with only
    32-bit FPUs *do* support.

 3. Add ARM versions of fma and fmaf which are used when those
    instructions are available.