mirror of
https://github.com/AuxXxilium/linux_dsm_epyc7002.git
synced 2024-12-22 00:20:15 +07:00
db57f29d50
The code in _sp_maddf (formerly ieee754sp_madd) appears to have been
copied verbatim from ieee754sp_add, and although it's adding the
unpacked "r" & "z" floats it kept using macros that operate on "x" &
"y". This led to the addition being carried out incorrectly on some
mismash of the product, accumulator & multiplicand fields. Typically
this would lead to the assertions "ze == re" & "ze <= SP_EMAX" failing
since ze & re hadn't been operated upon.
Signed-off-by: Paul Burton <paul.burton@imgtec.com>
Fixes: e24c3bec3e
("MIPS: math-emu: Add support for the MIPS R6 MADDF FPU instruction")
Cc: Adam Buchbinder <adam.buchbinder@gmail.com>
Cc: Maciej W. Rozycki <macro@imgtec.com>
Cc: linux-mips@linux-mips.org
Cc: linux-kernel@vger.kernel.org
Patchwork: https://patchwork.linux-mips.org/patch/13159/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
209 lines
4.8 KiB
C
209 lines
4.8 KiB
C
/* IEEE754 floating point arithmetic
|
|
* single precision
|
|
*/
|
|
/*
|
|
* MIPS floating point support
|
|
* Copyright (C) 1994-2000 Algorithmics Ltd.
|
|
*
|
|
* 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; if not, write to the Free Software Foundation, Inc.,
|
|
* 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
|
|
*/
|
|
|
|
#include <linux/compiler.h>
|
|
|
|
#include "ieee754sp.h"
|
|
|
|
int ieee754sp_class(union ieee754sp x)
|
|
{
|
|
COMPXSP;
|
|
EXPLODEXSP;
|
|
return xc;
|
|
}
|
|
|
|
static inline int ieee754sp_isnan(union ieee754sp x)
|
|
{
|
|
return ieee754_class_nan(ieee754sp_class(x));
|
|
}
|
|
|
|
static inline int ieee754sp_issnan(union ieee754sp x)
|
|
{
|
|
int qbit;
|
|
|
|
assert(ieee754sp_isnan(x));
|
|
qbit = (SPMANT(x) & SP_MBIT(SP_FBITS - 1)) == SP_MBIT(SP_FBITS - 1);
|
|
return ieee754_csr.nan2008 ^ qbit;
|
|
}
|
|
|
|
|
|
/*
|
|
* Raise the Invalid Operation IEEE 754 exception
|
|
* and convert the signaling NaN supplied to a quiet NaN.
|
|
*/
|
|
union ieee754sp __cold ieee754sp_nanxcpt(union ieee754sp r)
|
|
{
|
|
assert(ieee754sp_issnan(r));
|
|
|
|
ieee754_setcx(IEEE754_INVALID_OPERATION);
|
|
if (ieee754_csr.nan2008) {
|
|
SPMANT(r) |= SP_MBIT(SP_FBITS - 1);
|
|
} else {
|
|
SPMANT(r) &= ~SP_MBIT(SP_FBITS - 1);
|
|
if (!ieee754sp_isnan(r))
|
|
SPMANT(r) |= SP_MBIT(SP_FBITS - 2);
|
|
}
|
|
|
|
return r;
|
|
}
|
|
|
|
static unsigned ieee754sp_get_rounding(int sn, unsigned xm)
|
|
{
|
|
/* inexact must round of 3 bits
|
|
*/
|
|
if (xm & (SP_MBIT(3) - 1)) {
|
|
switch (ieee754_csr.rm) {
|
|
case FPU_CSR_RZ:
|
|
break;
|
|
case FPU_CSR_RN:
|
|
xm += 0x3 + ((xm >> 3) & 1);
|
|
/* xm += (xm&0x8)?0x4:0x3 */
|
|
break;
|
|
case FPU_CSR_RU: /* toward +Infinity */
|
|
if (!sn) /* ?? */
|
|
xm += 0x8;
|
|
break;
|
|
case FPU_CSR_RD: /* toward -Infinity */
|
|
if (sn) /* ?? */
|
|
xm += 0x8;
|
|
break;
|
|
}
|
|
}
|
|
return xm;
|
|
}
|
|
|
|
|
|
/* generate a normal/denormal number with over,under handling
|
|
* sn is sign
|
|
* xe is an unbiased exponent
|
|
* xm is 3bit extended precision value.
|
|
*/
|
|
union ieee754sp ieee754sp_format(int sn, int xe, unsigned xm)
|
|
{
|
|
assert(xm); /* we don't gen exact zeros (probably should) */
|
|
|
|
assert((xm >> (SP_FBITS + 1 + 3)) == 0); /* no excess */
|
|
assert(xm & (SP_HIDDEN_BIT << 3));
|
|
|
|
if (xe < SP_EMIN) {
|
|
/* strip lower bits */
|
|
int es = SP_EMIN - xe;
|
|
|
|
if (ieee754_csr.nod) {
|
|
ieee754_setcx(IEEE754_UNDERFLOW);
|
|
ieee754_setcx(IEEE754_INEXACT);
|
|
|
|
switch(ieee754_csr.rm) {
|
|
case FPU_CSR_RN:
|
|
case FPU_CSR_RZ:
|
|
return ieee754sp_zero(sn);
|
|
case FPU_CSR_RU: /* toward +Infinity */
|
|
if (sn == 0)
|
|
return ieee754sp_min(0);
|
|
else
|
|
return ieee754sp_zero(1);
|
|
case FPU_CSR_RD: /* toward -Infinity */
|
|
if (sn == 0)
|
|
return ieee754sp_zero(0);
|
|
else
|
|
return ieee754sp_min(1);
|
|
}
|
|
}
|
|
|
|
if (xe == SP_EMIN - 1 &&
|
|
ieee754sp_get_rounding(sn, xm) >> (SP_FBITS + 1 + 3))
|
|
{
|
|
/* Not tiny after rounding */
|
|
ieee754_setcx(IEEE754_INEXACT);
|
|
xm = ieee754sp_get_rounding(sn, xm);
|
|
xm >>= 1;
|
|
/* Clear grs bits */
|
|
xm &= ~(SP_MBIT(3) - 1);
|
|
xe++;
|
|
} else {
|
|
/* sticky right shift es bits
|
|
*/
|
|
xm = XSPSRS(xm, es);
|
|
xe += es;
|
|
assert((xm & (SP_HIDDEN_BIT << 3)) == 0);
|
|
assert(xe == SP_EMIN);
|
|
}
|
|
}
|
|
if (xm & (SP_MBIT(3) - 1)) {
|
|
ieee754_setcx(IEEE754_INEXACT);
|
|
if ((xm & (SP_HIDDEN_BIT << 3)) == 0) {
|
|
ieee754_setcx(IEEE754_UNDERFLOW);
|
|
}
|
|
|
|
/* inexact must round of 3 bits
|
|
*/
|
|
xm = ieee754sp_get_rounding(sn, xm);
|
|
/* adjust exponent for rounding add overflowing
|
|
*/
|
|
if (xm >> (SP_FBITS + 1 + 3)) {
|
|
/* add causes mantissa overflow */
|
|
xm >>= 1;
|
|
xe++;
|
|
}
|
|
}
|
|
/* strip grs bits */
|
|
xm >>= 3;
|
|
|
|
assert((xm >> (SP_FBITS + 1)) == 0); /* no excess */
|
|
assert(xe >= SP_EMIN);
|
|
|
|
if (xe > SP_EMAX) {
|
|
ieee754_setcx(IEEE754_OVERFLOW);
|
|
ieee754_setcx(IEEE754_INEXACT);
|
|
/* -O can be table indexed by (rm,sn) */
|
|
switch (ieee754_csr.rm) {
|
|
case FPU_CSR_RN:
|
|
return ieee754sp_inf(sn);
|
|
case FPU_CSR_RZ:
|
|
return ieee754sp_max(sn);
|
|
case FPU_CSR_RU: /* toward +Infinity */
|
|
if (sn == 0)
|
|
return ieee754sp_inf(0);
|
|
else
|
|
return ieee754sp_max(1);
|
|
case FPU_CSR_RD: /* toward -Infinity */
|
|
if (sn == 0)
|
|
return ieee754sp_max(0);
|
|
else
|
|
return ieee754sp_inf(1);
|
|
}
|
|
}
|
|
/* gen norm/denorm/zero */
|
|
|
|
if ((xm & SP_HIDDEN_BIT) == 0) {
|
|
/* we underflow (tiny/zero) */
|
|
assert(xe == SP_EMIN);
|
|
if (ieee754_csr.mx & IEEE754_UNDERFLOW)
|
|
ieee754_setcx(IEEE754_UNDERFLOW);
|
|
return buildsp(sn, SP_EMIN - 1 + SP_EBIAS, xm);
|
|
} else {
|
|
assert((xm >> (SP_FBITS + 1)) == 0); /* no excess */
|
|
assert(xm & SP_HIDDEN_BIT);
|
|
|
|
return buildsp(sn, xe + SP_EBIAS, xm & ~SP_HIDDEN_BIT);
|
|
}
|
|
}
|