Virtualization
/
berkeley-softfloat-3
mirror of https://github.com/ucb-bar/berkeley-softfloat-3.git
1
0
Fork 0
Code Issues Projects Releases Wiki Activity

Adding support for BF16 / BF16 <-> conversions in 8086-SSE

This commit is contained in:
Nicolas Brunie 2023-09-19 12:54:43 -07:00
parent aac867e874
commit 1172fb3e0a
4 changed files with 138 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

7
build/Linux-x86_64-GCC/Makefile
View File

@ -94,6 +94,8 @@ OBJS_SPECIALIZE = \
s_f16UIToCommonNaN$(OBJ) \
s_commonNaNToF16UI$(OBJ) \
s_propagateNaNF16UI$(OBJ) \
s_bf16UIToCommonNaN$(OBJ) \
s_commonNaNToBF16UI$(OBJ) \
s_f32UIToCommonNaN$(OBJ) \
s_commonNaNToF32UI$(OBJ) \
s_propagateNaNF32UI$(OBJ) \
@ -114,6 +116,8 @@ OBJS_OTHERS = \
s_roundToUI64$(OBJ) \
s_roundToI32$(OBJ) \
s_roundToI64$(OBJ) \
s_normSubnormalBF16Sig$(OBJ) \
s_roundPackToBF16$(OBJ) \
s_normSubnormalF16Sig$(OBJ) \
s_roundPackToF16$(OBJ) \
s_normRoundPackToF16$(OBJ) \
@ -172,6 +176,8 @@ OBJS_OTHERS = \
i64_to_extF80M$(OBJ) \
i64_to_f128$(OBJ) \
i64_to_f128M$(OBJ) \
bf16_isSignalingNaN$(OBJ) \
bf16_to_f32$(OBJ) \
f16_to_ui32$(OBJ) \
f16_to_ui64$(OBJ) \
f16_to_i32$(OBJ) \
@ -209,6 +215,7 @@ OBJS_OTHERS = \
f32_to_ui64_r_minMag$(OBJ) \
f32_to_i32_r_minMag$(OBJ) \
f32_to_i64_r_minMag$(OBJ) \
f32_to_bf16$(OBJ) \
f32_to_f16$(OBJ) \
f32_to_f64$(OBJ) \
f32_to_extF80$(OBJ) \

59
source/8086-SSE/s_bf16UIToCommonNaN.c Normal file
View File

@ -0,0 +1,59 @@
/*============================================================================
This C source file is part of the SoftFloat IEEE Floating-Point Arithmetic
Package, Release 3e, by John R. Hauser.
Copyright 2011, 2012, 2013, 2014 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 <stdint.h>
#include "platform.h"
#include "specialize.h"
#include "softfloat.h"
/*----------------------------------------------------------------------------
| Assuming `uiA' has the bit pattern of a 32-bit floating-point NaN, converts
| this NaN to the common NaN form, and stores the resulting common NaN at the
| location pointed to by `zPtr'. If the NaN is a signaling NaN, the invalid
| exception is raised.
*----------------------------------------------------------------------------*/
void softfloat_bf16UIToCommonNaN( uint_fast16_t uiA, struct commonNaN *zPtr )
{
if ( softfloat_isSigNaNBF16UI( uiA ) ) {
softfloat_raiseFlags( softfloat_flag_invalid );
}
zPtr->sign = uiA>>15;
zPtr->v64 = (uint_fast64_t) uiA<<56;
zPtr->v0 = 0;
}

51
source/8086-SSE/s_commonNaNToBF16UI.c Normal file
View File

@ -0,0 +1,51 @@
/*============================================================================
This C source file is part of the SoftFloat IEEE Floating-Point Arithmetic
Package, Release 3e, by John R. Hauser.
Copyright 2011, 2012, 2013, 2014 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 <stdint.h>
#include "platform.h"
#include "specialize.h"
/*----------------------------------------------------------------------------
| Converts the common NaN pointed to by `aPtr' into a 16-bit floating-point
| NaN, and returns the bit pattern of this value as an unsigned integer.
*----------------------------------------------------------------------------*/
uint_fast16_t softfloat_commonNaNToBF16UI( const struct commonNaN *aPtr )
{
return (uint_fast16_t) aPtr->sign<<15 | 0x7FC0 | aPtr->v64>>56;
}

21
source/8086-SSE/specialize.h
View File

@ -117,6 +117,27 @@ uint_fast16_t softfloat_commonNaNToF16UI( const struct commonNaN *aPtr );
uint_fast16_t
softfloat_propagateNaNF16UI( uint_fast16_t uiA, uint_fast16_t uiB );
/*----------------------------------------------------------------------------
| Returns true when 16-bit unsigned integer 'uiA' has the bit pattern of a
| 16-bit brain floating-point (BF16) signaling NaN.
| Note: This macro evaluates its argument more than once.
*----------------------------------------------------------------------------*/
#define softfloat_isSigNaNBF16UI( uiA ) ((((uiA) & 0x7FC0) == 0x7F80) && ((uiA) & 0x003F))
/*----------------------------------------------------------------------------
| Assuming 'uiA' has the bit pattern of a 16-bit BF16 floating-point NaN, converts
| this NaN to the common NaN form, and stores the resulting common NaN at the
| location pointed to by 'zPtr'. If the NaN is a signaling NaN, the invalid
| exception is raised.
*----------------------------------------------------------------------------*/
void softfloat_bf16UIToCommonNaN( uint_fast16_t uiA, struct commonNaN *zPtr );
/*----------------------------------------------------------------------------
| Converts the common NaN pointed to by 'aPtr' into a 16-bit floating-point
| NaN, and returns the bit pattern of this value as an unsigned integer.
*----------------------------------------------------------------------------*/
uint_fast16_t softfloat_commonNaNToBF16UI( const struct commonNaN *aPtr );
/*----------------------------------------------------------------------------
| The bit pattern for a default generated 32-bit floating-point NaN.
*----------------------------------------------------------------------------*/