![Aleksandar Markovic](/assets/img/avatar_default.png)
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>
91 lines
2 KiB
C
91 lines
2 KiB
C
/* IEEE754 floating point arithmetic
|
|
* single precision
|
|
*/
|
|
/*
|
|
* 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 "ieee754sp.h"
|
|
|
|
union ieee754sp ieee754sp_rint(union ieee754sp x)
|
|
{
|
|
union ieee754sp ret;
|
|
u32 residue;
|
|
int sticky;
|
|
int round;
|
|
int odd;
|
|
|
|
COMPXDP; /* <-- DP needed for 64-bit mantissa tmp */
|
|
|
|
ieee754_clearcx();
|
|
|
|
EXPLODEXSP;
|
|
FLUSHXSP;
|
|
|
|
if (xc == IEEE754_CLASS_SNAN)
|
|
return ieee754sp_nanxcpt(x);
|
|
|
|
if ((xc == IEEE754_CLASS_QNAN) ||
|
|
(xc == IEEE754_CLASS_INF) ||
|
|
(xc == IEEE754_CLASS_ZERO))
|
|
return x;
|
|
|
|
if (xe >= SP_FBITS)
|
|
return x;
|
|
|
|
if (xe < -1) {
|
|
residue = xm;
|
|
round = 0;
|
|
sticky = residue != 0;
|
|
xm = 0;
|
|
} else {
|
|
residue = xm << (xe + 1);
|
|
residue <<= 31 - SP_FBITS;
|
|
round = (residue >> 31) != 0;
|
|
sticky = (residue << 1) != 0;
|
|
xm >>= SP_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 = ieee754sp_flong(xm);
|
|
SPSIGN(ret) = xs;
|
|
|
|
return ret;
|
|
}
|