linux/arch/mips/math-emu/dp_rint.c
Aleksandar Markovic 3ec404d88c MIPS: math-emu: RINT.<D|S>: Fix several problems by reimplementation
Reimplement RINT.<D|S> kernel emulation so that all RINT.<D|S>
specifications are met.

For the sake of simplicity, let's analyze RINT.S only. Prior to
this patch, RINT.S emulation was essentially implemented as (in
pseudocode) <output> = ieee754sp_flong(ieee754sp_tlong(<input>)),
where ieee754sp_tlong() and ieee754sp_flong() are functions
providing conversion from double to integer, and from integer
to double, respectively. On surface, this implementation looks
correct, but actually fails in many cases. Following problems
were detected:

1. NaN and infinity cases will not be handled properly. The
   function ieee754sp_flong() never returns NaN nor infinity.
2. For RINT.S, for all inputs larger than LONG_MAX, and smaller
   than FLT_MAX, the result will be wrong, and the overflow
   exception will be erroneously set. A similar problem for
   negative inputs exists as well.
3. For some rounding modes, for some negative inputs close to zero,
   the return value will be zero, and should be -zero. This is
   because ieee754sp_flong() never returns -zero.

This patch removes the problems above by implementing dedicated
functions for RINT.<D|S> emulation.

The core of the new function functionality is adapted version of
the core of the function ieee754sp_tlong(). However, there are many
details that are implemented to match RINT.<D|S> specification. It
should be said that the functionality of ieee754sp_tlong() actually
closely corresponds to CVT.L.S instruction, and it is used while
emulating CVT.L.S. However, RINT.S and CVT.L.S instructions differ
in many aspects. This patch fulfills missing support for RINT.<D|S>.

Signed-off-by: Miodrag Dinic <miodrag.dinic@imgtec.com>
Signed-off-by: Goran Ferenc <goran.ferenc@imgtec.com>
Signed-off-by: Aleksandar Markovic <aleksandar.markovic@imgtec.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: Douglas Leung <douglas.leung@imgtec.com>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Hans Verkuil <hans.verkuil@cisco.com>
Cc: James Hogan <james.hogan@imgtec.com>
Cc: Maciej W. Rozycki <macro@imgtec.com>
Cc: Masahiro Yamada <yamada.masahiro@socionext.com>
Cc: Mauro Carvalho Chehab <mchehab@kernel.org>
Cc: Paul Burton <paul.burton@imgtec.com>
Cc: Petar Jovanovic <petar.jovanovic@imgtec.com>
Cc: Raghu Gandham <raghu.gandham@imgtec.com>
Cc: Randy Dunlap <rdunlap@infradead.org>
Cc: linux-mips@linux-mips.org
Cc: linux-kernel@vger.kernel.org
Patchwork: https://patchwork.linux-mips.org/patch/17141/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2017-08-29 15:21:56 +02:00

90 lines
1.9 KiB
C

/* IEEE754 floating point arithmetic
* double precision: common utilities
*/
/*
* MIPS floating point support
* Copyright (C) 1994-2000 Algorithmics Ltd.
* Copyright (C) 2017 Imagination Technologies, Ltd.
* Author: Aleksandar Markovic <aleksandar.markovic@imgtec.com>
*
* This program is free software; you can distribute it and/or modify it
* under the terms of the GNU General Public License (Version 2) as
* published by the Free Software Foundation.
*
* This program is distributed in the hope it will be useful, but WITHOUT
* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
* FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
* for more details.
*
* You should have received a copy of the GNU General Public License along
* with this program.
*/
#include "ieee754dp.h"
union ieee754dp ieee754dp_rint(union ieee754dp x)
{
union ieee754dp ret;
u64 residue;
int sticky;
int round;
int odd;
COMPXDP;
ieee754_clearcx();
EXPLODEXDP;
FLUSHXDP;
if (xc == IEEE754_CLASS_SNAN)
return ieee754dp_nanxcpt(x);
if ((xc == IEEE754_CLASS_QNAN) ||
(xc == IEEE754_CLASS_INF) ||
(xc == IEEE754_CLASS_ZERO))
return x;
if (xe >= DP_FBITS)
return x;
if (xe < -1) {
residue = xm;
round = 0;
sticky = residue != 0;
xm = 0;
} else {
residue = xm << (64 - DP_FBITS + xe);
round = (residue >> 63) != 0;
sticky = (residue << 1) != 0;
xm >>= DP_FBITS - xe;
}
odd = (xm & 0x1) != 0x0;
switch (ieee754_csr.rm) {
case FPU_CSR_RN: /* toward nearest */
if (round && (sticky || odd))
xm++;
break;
case FPU_CSR_RZ: /* toward zero */
break;
case FPU_CSR_RU: /* toward +infinity */
if ((round || sticky) && !xs)
xm++;
break;
case FPU_CSR_RD: /* toward -infinity */
if ((round || sticky) && xs)
xm++;
break;
}
if (round || sticky)
ieee754_setcx(IEEE754_INEXACT);
ret = ieee754dp_flong(xm);
DPSIGN(ret) = xs;
return ret;
}