Files
gem5/ext/softfloat/s_mulAddF16.c
Xuan Hu 461520d0ab ext: Update softfloat to 3d full version
* Add all softfloat source files without any change.
* Remove useless file softfloat.mk.in, since gem5 use Scons.
* Add `use_fast_int64` in SConscript to distinguish src of two strategies for data
  larger than 64 bits.
  * The SoftFloat library uses two strategies to handle data larger than 64bit. One is
    spliting data into `fast_int64`, and the other is using pointer. Two strategies
    are distinguished by macro `SOFTFLOAT_FAST_INT64`. But not all "*.c" files are
    guarded by this macro, which leads to including useless files in compiling progress
    and compiling error. `use_fast_int64` used in SConscript can exclude unnecessary
    files.

Change-Id: I7cec10412c00a35c247299cd92d83cdee9066410
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/66552
Maintainer: Jason Lowe-Power <power.jg@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Hoa Nguyen <hoanguyen@ucdavis.edu>
2023-03-28 10:47:18 +00:00

227 lines
8.1 KiB
C

/*============================================================================
This C source file is part of the SoftFloat IEEE Floating-Point Arithmetic
Package, Release 3d, by John R. Hauser.
Copyright 2011, 2012, 2013, 2014, 2015, 2016 The Regents of the University of
California. All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are met:
1. Redistributions of source code must retain the above copyright notice,
this list of conditions, and the following disclaimer.
2. Redistributions in binary form must reproduce the above copyright notice,
this list of conditions, and the following disclaimer in the documentation
and/or other materials provided with the distribution.
3. Neither the name of the University nor the names of its contributors may
be used to endorse or promote products derived from this software without
specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS "AS IS", AND ANY
EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE, ARE
DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE FOR ANY
DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
(INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
=============================================================================*/
#include <stdbool.h>
#include <stdint.h>
#include "platform.h"
#include "internals.h"
#include "specialize.h"
#include "softfloat.h"
float16_t
softfloat_mulAddF16(
uint_fast16_t uiA, uint_fast16_t uiB, uint_fast16_t uiC, uint_fast8_t op )
{
bool signA;
int_fast8_t expA;
uint_fast16_t sigA;
bool signB;
int_fast8_t expB;
uint_fast16_t sigB;
bool signC;
int_fast8_t expC;
uint_fast16_t sigC;
bool signProd;
uint_fast16_t magBits, uiZ;
struct exp8_sig16 normExpSig;
int_fast8_t expProd;
uint_fast32_t sigProd;
bool signZ;
int_fast8_t expZ;
uint_fast16_t sigZ;
int_fast8_t expDiff;
uint_fast32_t sig32Z, sig32C;
int_fast8_t shiftDist;
union ui16_f16 uZ;
/*------------------------------------------------------------------------
*------------------------------------------------------------------------*/
signA = signF16UI( uiA );
expA = expF16UI( uiA );
sigA = fracF16UI( uiA );
signB = signF16UI( uiB );
expB = expF16UI( uiB );
sigB = fracF16UI( uiB );
signC = signF16UI( uiC ) ^ (op == softfloat_mulAdd_subC);
expC = expF16UI( uiC );
sigC = fracF16UI( uiC );
signProd = signA ^ signB ^ (op == softfloat_mulAdd_subProd);
/*------------------------------------------------------------------------
*------------------------------------------------------------------------*/
if ( expA == 0x1F ) {
if ( sigA || ((expB == 0x1F) && sigB) ) goto propagateNaN_ABC;
magBits = expB | sigB;
goto infProdArg;
}
if ( expB == 0x1F ) {
if ( sigB ) goto propagateNaN_ABC;
magBits = expA | sigA;
goto infProdArg;
}
if ( expC == 0x1F ) {
if ( sigC ) {
uiZ = 0;
goto propagateNaN_ZC;
}
uiZ = uiC;
goto uiZ;
}
/*------------------------------------------------------------------------
*------------------------------------------------------------------------*/
if ( ! expA ) {
if ( ! sigA ) goto zeroProd;
normExpSig = softfloat_normSubnormalF16Sig( sigA );
expA = normExpSig.exp;
sigA = normExpSig.sig;
}
if ( ! expB ) {
if ( ! sigB ) goto zeroProd;
normExpSig = softfloat_normSubnormalF16Sig( sigB );
expB = normExpSig.exp;
sigB = normExpSig.sig;
}
/*------------------------------------------------------------------------
*------------------------------------------------------------------------*/
expProd = expA + expB - 0xE;
sigA = (sigA | 0x0400)<<4;
sigB = (sigB | 0x0400)<<4;
sigProd = (uint_fast32_t) sigA * sigB;
if ( sigProd < 0x20000000 ) {
--expProd;
sigProd <<= 1;
}
signZ = signProd;
if ( ! expC ) {
if ( ! sigC ) {
expZ = expProd - 1;
sigZ = sigProd>>15 | ((sigProd & 0x7FFF) != 0);
goto roundPack;
}
normExpSig = softfloat_normSubnormalF16Sig( sigC );
expC = normExpSig.exp;
sigC = normExpSig.sig;
}
sigC = (sigC | 0x0400)<<3;
/*------------------------------------------------------------------------
*------------------------------------------------------------------------*/
expDiff = expProd - expC;
if ( signProd == signC ) {
/*--------------------------------------------------------------------
*--------------------------------------------------------------------*/
if ( expDiff <= 0 ) {
expZ = expC;
sigZ = sigC + softfloat_shiftRightJam32( sigProd, 16 - expDiff );
} else {
expZ = expProd;
sig32Z =
sigProd
+ softfloat_shiftRightJam32(
(uint_fast32_t) sigC<<16, expDiff );
sigZ = sig32Z>>16 | ((sig32Z & 0xFFFF) != 0 );
}
if ( sigZ < 0x4000 ) {
--expZ;
sigZ <<= 1;
}
} else {
/*--------------------------------------------------------------------
*--------------------------------------------------------------------*/
sig32C = (uint_fast32_t) sigC<<16;
if ( expDiff < 0 ) {
signZ = signC;
expZ = expC;
sig32Z = sig32C - softfloat_shiftRightJam32( sigProd, -expDiff );
} else if ( ! expDiff ) {
expZ = expProd;
sig32Z = sigProd - sig32C;
if ( ! sig32Z ) goto completeCancellation;
if ( sig32Z & 0x80000000 ) {
signZ = ! signZ;
sig32Z = -sig32Z;
}
} else {
expZ = expProd;
sig32Z = sigProd - softfloat_shiftRightJam32( sig32C, expDiff );
}
shiftDist = softfloat_countLeadingZeros32( sig32Z ) - 1;
expZ -= shiftDist;
shiftDist -= 16;
if ( shiftDist < 0 ) {
sigZ =
sig32Z>>(-shiftDist)
| ((uint32_t) (sig32Z<<(shiftDist & 31)) != 0);
} else {
sigZ = (uint_fast16_t) sig32Z<<shiftDist;
}
}
roundPack:
return softfloat_roundPackToF16( signZ, expZ, sigZ );
/*------------------------------------------------------------------------
*------------------------------------------------------------------------*/
propagateNaN_ABC:
uiZ = softfloat_propagateNaNF16UI( uiA, uiB );
goto propagateNaN_ZC;
/*------------------------------------------------------------------------
*------------------------------------------------------------------------*/
infProdArg:
if ( magBits ) {
uiZ = packToF16UI( signProd, 0x1F, 0 );
if ( expC != 0x1F ) goto uiZ;
if ( sigC ) goto propagateNaN_ZC;
if ( signProd == signC ) goto uiZ;
}
softfloat_raiseFlags( softfloat_flag_invalid );
uiZ = defaultNaNF16UI;
propagateNaN_ZC:
uiZ = softfloat_propagateNaNF16UI( uiZ, uiC );
goto uiZ;
/*------------------------------------------------------------------------
*------------------------------------------------------------------------*/
zeroProd:
uiZ = uiC;
if ( ! (expC | sigC) && (signProd != signC) ) {
completeCancellation:
uiZ =
packToF16UI(
(softfloat_roundingMode == softfloat_round_min), 0, 0 );
}
uiZ:
uZ.ui = uiZ;
return uZ.f;
}