mirror of
https://github.com/FFmpeg/FFmpeg.git
synced 2024-12-23 12:43:46 +02:00
cb8ad005bb
IEEE-754 differentiates two different kind of NaNs. Quiet and Signaling ones. They are differentiated by the MSB of the mantissa. For whatever reason, actual hardware conversion of half to single always sets the signaling bit to 1 if the mantissa is != 0, and to 0 if it's 0. So our code has to follow suite or fate-testing hardware float16 will be impossible.
80 lines
2.5 KiB
C
80 lines
2.5 KiB
C
/*
|
|
* This file is part of FFmpeg.
|
|
*
|
|
* FFmpeg is free software; you can redistribute it and/or
|
|
* modify it under the terms of the GNU Lesser General Public
|
|
* License as published by the Free Software Foundation; either
|
|
* version 2.1 of the License, or (at your option) any later version.
|
|
*
|
|
* FFmpeg is distributed in the hope that it will be useful,
|
|
* but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
|
* Lesser General Public License for more details.
|
|
*
|
|
* You should have received a copy of the GNU Lesser General Public
|
|
* License along with FFmpeg; if not, write to the Free Software
|
|
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
|
|
*/
|
|
|
|
#ifndef AVUTIL_HALF2FLOAT_H
|
|
#define AVUTIL_HALF2FLOAT_H
|
|
|
|
#include <stdint.h>
|
|
|
|
static uint32_t convertmantissa(uint32_t i)
|
|
{
|
|
int32_t m = i << 13; // Zero pad mantissa bits
|
|
int32_t e = 0; // Zero exponent
|
|
|
|
while (!(m & 0x00800000)) { // While not normalized
|
|
e -= 0x00800000; // Decrement exponent (1<<23)
|
|
m <<= 1; // Shift mantissa
|
|
}
|
|
|
|
m &= ~0x00800000; // Clear leading 1 bit
|
|
e += 0x38800000; // Adjust bias ((127-14)<<23)
|
|
|
|
return m | e; // Return combined number
|
|
}
|
|
|
|
static void half2float_table(uint32_t *mantissatable, uint32_t *exponenttable,
|
|
uint16_t *offsettable)
|
|
{
|
|
mantissatable[0] = 0;
|
|
for (int i = 1; i < 1024; i++)
|
|
mantissatable[i] = convertmantissa(i);
|
|
for (int i = 1024; i < 2048; i++)
|
|
mantissatable[i] = 0x38000000UL + ((i - 1024) << 13UL);
|
|
for (int i = 2048; i < 3072; i++)
|
|
mantissatable[i] = mantissatable[i - 1024] | 0x400000UL;
|
|
mantissatable[2048] = mantissatable[1024];
|
|
|
|
exponenttable[0] = 0;
|
|
for (int i = 1; i < 31; i++)
|
|
exponenttable[i] = i << 23;
|
|
for (int i = 33; i < 63; i++)
|
|
exponenttable[i] = 0x80000000UL + ((i - 32) << 23UL);
|
|
exponenttable[31]= 0x47800000UL;
|
|
exponenttable[32]= 0x80000000UL;
|
|
exponenttable[63]= 0xC7800000UL;
|
|
|
|
offsettable[0] = 0;
|
|
for (int i = 1; i < 64; i++)
|
|
offsettable[i] = 1024;
|
|
offsettable[31] = 2048;
|
|
offsettable[32] = 0;
|
|
offsettable[63] = 2048;
|
|
}
|
|
|
|
static uint32_t half2float(uint16_t h, const uint32_t *mantissatable, const uint32_t *exponenttable,
|
|
const uint16_t *offsettable)
|
|
{
|
|
uint32_t f;
|
|
|
|
f = mantissatable[offsettable[h >> 10] + (h & 0x3ff)] + exponenttable[h >> 10];
|
|
|
|
return f;
|
|
}
|
|
|
|
#endif /* AVUTIL_HALF2FLOAT_H */
|