2010-01-14 20:58:12 +01:00
|
|
|
/*
|
|
|
|
* Copyright (c) 2010 Mans Rullgard <mans@mansr.com>
|
|
|
|
*
|
|
|
|
* 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_INTMATH_H
|
|
|
|
#define AVUTIL_INTMATH_H
|
|
|
|
|
2012-10-18 19:15:38 +02:00
|
|
|
#include <stdint.h>
|
|
|
|
|
2010-01-14 20:58:12 +01:00
|
|
|
#include "config.h"
|
2010-03-09 02:19:28 +01:00
|
|
|
#include "attributes.h"
|
2010-01-14 20:58:12 +01:00
|
|
|
|
2012-10-20 16:29:32 +02:00
|
|
|
#if ARCH_ARM
|
|
|
|
# include "arm/intmath.h"
|
|
|
|
#endif
|
2015-02-25 23:52:33 +01:00
|
|
|
#if ARCH_X86
|
|
|
|
# include "x86/intmath.h"
|
|
|
|
#endif
|
2012-10-20 16:29:32 +02:00
|
|
|
|
2014-10-26 15:43:15 +01:00
|
|
|
#if HAVE_FAST_CLZ
|
2015-10-16 08:00:13 +02:00
|
|
|
#if AV_GCC_VERSION_AT_LEAST(3,4)
|
2014-10-26 15:43:15 +01:00
|
|
|
#ifndef ff_log2
|
2015-07-18 21:52:42 +02:00
|
|
|
# define ff_log2(x) (31 - __builtin_clz((x)|1))
|
2014-10-26 15:43:15 +01:00
|
|
|
# ifndef ff_log2_16bit
|
|
|
|
# define ff_log2_16bit av_log2
|
|
|
|
# endif
|
|
|
|
#endif /* ff_log2 */
|
2010-01-14 20:58:12 +01:00
|
|
|
#endif /* AV_GCC_VERSION_AT_LEAST(3,4) */
|
2015-07-18 21:52:42 +02:00
|
|
|
#endif
|
2010-01-14 20:58:12 +01:00
|
|
|
|
2012-10-18 19:15:38 +02:00
|
|
|
extern const uint8_t ff_log2_tab[256];
|
|
|
|
|
|
|
|
#ifndef ff_log2
|
|
|
|
#define ff_log2 ff_log2_c
|
|
|
|
static av_always_inline av_const int ff_log2_c(unsigned int v)
|
|
|
|
{
|
|
|
|
int n = 0;
|
|
|
|
if (v & 0xffff0000) {
|
|
|
|
v >>= 16;
|
|
|
|
n += 16;
|
|
|
|
}
|
|
|
|
if (v & 0xff00) {
|
|
|
|
v >>= 8;
|
|
|
|
n += 8;
|
|
|
|
}
|
|
|
|
n += ff_log2_tab[v];
|
|
|
|
|
|
|
|
return n;
|
|
|
|
}
|
|
|
|
#endif
|
|
|
|
|
|
|
|
#ifndef ff_log2_16bit
|
|
|
|
#define ff_log2_16bit ff_log2_16bit_c
|
|
|
|
static av_always_inline av_const int ff_log2_16bit_c(unsigned int v)
|
|
|
|
{
|
|
|
|
int n = 0;
|
|
|
|
if (v & 0xff00) {
|
|
|
|
v >>= 8;
|
|
|
|
n += 8;
|
|
|
|
}
|
|
|
|
n += ff_log2_tab[v];
|
|
|
|
|
|
|
|
return n;
|
|
|
|
}
|
|
|
|
#endif
|
|
|
|
|
|
|
|
#define av_log2 ff_log2
|
|
|
|
#define av_log2_16bit ff_log2_16bit
|
|
|
|
|
2012-10-26 20:48:40 +02:00
|
|
|
/**
|
|
|
|
* @addtogroup lavu_math
|
|
|
|
* @{
|
|
|
|
*/
|
|
|
|
|
2014-10-26 15:43:15 +01:00
|
|
|
#if HAVE_FAST_CLZ
|
2015-10-16 08:00:13 +02:00
|
|
|
#if AV_GCC_VERSION_AT_LEAST(3,4)
|
2014-10-26 15:43:15 +01:00
|
|
|
#ifndef ff_ctz
|
2015-07-18 21:52:42 +02:00
|
|
|
#define ff_ctz(v) __builtin_ctz(v)
|
2014-10-26 15:43:15 +01:00
|
|
|
#endif
|
avutil/mathematics: speed up av_gcd by using Stein's binary GCD algorithm
This uses Stein's binary GCD algorithm:
https://en.wikipedia.org/wiki/Binary_GCD_algorithm
to get a roughly 4x speedup over Euclidean GCD on standard architectures
with a compiler intrinsic for ctzll, and a roughly 2x speedup otherwise.
At the moment, the compiler intrinsic is used on GCC and Clang due to
its easy availability.
Quick note regarding overflow: yes, subtractions on int64_t can, but the
llabs takes care of that. The llabs is also guaranteed to be safe, with
no annoying INT64_MIN business since INT64_MIN being a power of 2, is
shifted down before being sent to llabs.
The binary GCD needs ff_ctzll, an extension of ff_ctz for long long (int64_t). On
GCC, this is provided by a built-in. On Microsoft, there is a
BitScanForward64 analog of BitScanForward that should work; but I can't confirm.
Apparently it is not available on 32 bit builds; so this may or may not
work correctly. On Intel, per the documentation there is only an
intrinsic for _bit_scan_forward and people have posted on forums
regarding _bit_scan_forward64, but often their documentation is
woeful. Again, I don't have it, so I can't test.
As such, to be safe, for now only the GCC/Clang intrinsic is added, the rest
use a compiled version based on the De-Bruijn method of Leiserson et al:
http://supertech.csail.mit.edu/papers/debruijn.pdf.
Tested with FATE, sample benchmark (x86-64, GCC 5.2.0, Haswell)
with a START_TIMER and STOP_TIMER in libavutil/rationsl.c, followed by a
make fate.
aac-am00_88.err:
builtin:
714 decicycles in av_gcd, 4095 runs, 1 skips
de-bruijn:
1440 decicycles in av_gcd, 4096 runs, 0 skips
previous:
2889 decicycles in av_gcd, 4096 runs, 0 skips
Signed-off-by: Ganesh Ajjanagadde <gajjanagadde@gmail.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
2015-10-11 03:58:47 +02:00
|
|
|
#ifndef ff_ctzll
|
|
|
|
#define ff_ctzll(v) __builtin_ctzll(v)
|
|
|
|
#endif
|
2015-12-16 19:28:39 +01:00
|
|
|
#ifndef ff_clz
|
|
|
|
#define ff_clz(v) __builtin_clz(v)
|
|
|
|
#endif
|
2014-10-26 15:43:15 +01:00
|
|
|
#endif
|
2012-10-26 20:48:40 +02:00
|
|
|
#endif
|
|
|
|
|
|
|
|
#ifndef ff_ctz
|
|
|
|
#define ff_ctz ff_ctz_c
|
2015-10-11 23:43:29 +02:00
|
|
|
/**
|
|
|
|
* Trailing zero bit count.
|
|
|
|
*
|
|
|
|
* @param v input value. If v is 0, the result is undefined.
|
|
|
|
* @return the number of trailing 0-bits
|
|
|
|
*/
|
2015-10-14 16:26:59 +02:00
|
|
|
/* We use the De-Bruijn method outlined in:
|
|
|
|
* http://supertech.csail.mit.edu/papers/debruijn.pdf. */
|
2012-10-26 20:48:40 +02:00
|
|
|
static av_always_inline av_const int ff_ctz_c(int v)
|
|
|
|
{
|
2015-10-14 16:26:59 +02:00
|
|
|
static const uint8_t debruijn_ctz32[32] = {
|
|
|
|
0, 1, 28, 2, 29, 14, 24, 3, 30, 22, 20, 15, 25, 17, 4, 8,
|
|
|
|
31, 27, 13, 23, 21, 19, 16, 7, 26, 12, 18, 6, 11, 5, 10, 9
|
|
|
|
};
|
|
|
|
return debruijn_ctz32[(uint32_t)((v & -v) * 0x077CB531U) >> 27];
|
2012-10-26 20:48:40 +02:00
|
|
|
}
|
|
|
|
#endif
|
|
|
|
|
avutil/mathematics: speed up av_gcd by using Stein's binary GCD algorithm
This uses Stein's binary GCD algorithm:
https://en.wikipedia.org/wiki/Binary_GCD_algorithm
to get a roughly 4x speedup over Euclidean GCD on standard architectures
with a compiler intrinsic for ctzll, and a roughly 2x speedup otherwise.
At the moment, the compiler intrinsic is used on GCC and Clang due to
its easy availability.
Quick note regarding overflow: yes, subtractions on int64_t can, but the
llabs takes care of that. The llabs is also guaranteed to be safe, with
no annoying INT64_MIN business since INT64_MIN being a power of 2, is
shifted down before being sent to llabs.
The binary GCD needs ff_ctzll, an extension of ff_ctz for long long (int64_t). On
GCC, this is provided by a built-in. On Microsoft, there is a
BitScanForward64 analog of BitScanForward that should work; but I can't confirm.
Apparently it is not available on 32 bit builds; so this may or may not
work correctly. On Intel, per the documentation there is only an
intrinsic for _bit_scan_forward and people have posted on forums
regarding _bit_scan_forward64, but often their documentation is
woeful. Again, I don't have it, so I can't test.
As such, to be safe, for now only the GCC/Clang intrinsic is added, the rest
use a compiled version based on the De-Bruijn method of Leiserson et al:
http://supertech.csail.mit.edu/papers/debruijn.pdf.
Tested with FATE, sample benchmark (x86-64, GCC 5.2.0, Haswell)
with a START_TIMER and STOP_TIMER in libavutil/rationsl.c, followed by a
make fate.
aac-am00_88.err:
builtin:
714 decicycles in av_gcd, 4095 runs, 1 skips
de-bruijn:
1440 decicycles in av_gcd, 4096 runs, 0 skips
previous:
2889 decicycles in av_gcd, 4096 runs, 0 skips
Signed-off-by: Ganesh Ajjanagadde <gajjanagadde@gmail.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
2015-10-11 03:58:47 +02:00
|
|
|
#ifndef ff_ctzll
|
|
|
|
#define ff_ctzll ff_ctzll_c
|
|
|
|
/* We use the De-Bruijn method outlined in:
|
|
|
|
* http://supertech.csail.mit.edu/papers/debruijn.pdf. */
|
|
|
|
static av_always_inline av_const int ff_ctzll_c(long long v)
|
|
|
|
{
|
2015-10-11 04:21:01 +02:00
|
|
|
static const uint8_t debruijn_ctz64[64] = {
|
avutil/mathematics: speed up av_gcd by using Stein's binary GCD algorithm
This uses Stein's binary GCD algorithm:
https://en.wikipedia.org/wiki/Binary_GCD_algorithm
to get a roughly 4x speedup over Euclidean GCD on standard architectures
with a compiler intrinsic for ctzll, and a roughly 2x speedup otherwise.
At the moment, the compiler intrinsic is used on GCC and Clang due to
its easy availability.
Quick note regarding overflow: yes, subtractions on int64_t can, but the
llabs takes care of that. The llabs is also guaranteed to be safe, with
no annoying INT64_MIN business since INT64_MIN being a power of 2, is
shifted down before being sent to llabs.
The binary GCD needs ff_ctzll, an extension of ff_ctz for long long (int64_t). On
GCC, this is provided by a built-in. On Microsoft, there is a
BitScanForward64 analog of BitScanForward that should work; but I can't confirm.
Apparently it is not available on 32 bit builds; so this may or may not
work correctly. On Intel, per the documentation there is only an
intrinsic for _bit_scan_forward and people have posted on forums
regarding _bit_scan_forward64, but often their documentation is
woeful. Again, I don't have it, so I can't test.
As such, to be safe, for now only the GCC/Clang intrinsic is added, the rest
use a compiled version based on the De-Bruijn method of Leiserson et al:
http://supertech.csail.mit.edu/papers/debruijn.pdf.
Tested with FATE, sample benchmark (x86-64, GCC 5.2.0, Haswell)
with a START_TIMER and STOP_TIMER in libavutil/rationsl.c, followed by a
make fate.
aac-am00_88.err:
builtin:
714 decicycles in av_gcd, 4095 runs, 1 skips
de-bruijn:
1440 decicycles in av_gcd, 4096 runs, 0 skips
previous:
2889 decicycles in av_gcd, 4096 runs, 0 skips
Signed-off-by: Ganesh Ajjanagadde <gajjanagadde@gmail.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
2015-10-11 03:58:47 +02:00
|
|
|
0, 1, 2, 53, 3, 7, 54, 27, 4, 38, 41, 8, 34, 55, 48, 28,
|
|
|
|
62, 5, 39, 46, 44, 42, 22, 9, 24, 35, 59, 56, 49, 18, 29, 11,
|
|
|
|
63, 52, 6, 26, 37, 40, 33, 47, 61, 45, 43, 21, 23, 58, 17, 10,
|
|
|
|
51, 25, 36, 32, 60, 20, 57, 16, 50, 31, 19, 15, 30, 14, 13, 12
|
|
|
|
};
|
2015-10-22 14:10:10 +02:00
|
|
|
return debruijn_ctz64[(uint64_t)((v & -v) * 0x022FDD63CC95386DU) >> 58];
|
avutil/mathematics: speed up av_gcd by using Stein's binary GCD algorithm
This uses Stein's binary GCD algorithm:
https://en.wikipedia.org/wiki/Binary_GCD_algorithm
to get a roughly 4x speedup over Euclidean GCD on standard architectures
with a compiler intrinsic for ctzll, and a roughly 2x speedup otherwise.
At the moment, the compiler intrinsic is used on GCC and Clang due to
its easy availability.
Quick note regarding overflow: yes, subtractions on int64_t can, but the
llabs takes care of that. The llabs is also guaranteed to be safe, with
no annoying INT64_MIN business since INT64_MIN being a power of 2, is
shifted down before being sent to llabs.
The binary GCD needs ff_ctzll, an extension of ff_ctz for long long (int64_t). On
GCC, this is provided by a built-in. On Microsoft, there is a
BitScanForward64 analog of BitScanForward that should work; but I can't confirm.
Apparently it is not available on 32 bit builds; so this may or may not
work correctly. On Intel, per the documentation there is only an
intrinsic for _bit_scan_forward and people have posted on forums
regarding _bit_scan_forward64, but often their documentation is
woeful. Again, I don't have it, so I can't test.
As such, to be safe, for now only the GCC/Clang intrinsic is added, the rest
use a compiled version based on the De-Bruijn method of Leiserson et al:
http://supertech.csail.mit.edu/papers/debruijn.pdf.
Tested with FATE, sample benchmark (x86-64, GCC 5.2.0, Haswell)
with a START_TIMER and STOP_TIMER in libavutil/rationsl.c, followed by a
make fate.
aac-am00_88.err:
builtin:
714 decicycles in av_gcd, 4095 runs, 1 skips
de-bruijn:
1440 decicycles in av_gcd, 4096 runs, 0 skips
previous:
2889 decicycles in av_gcd, 4096 runs, 0 skips
Signed-off-by: Ganesh Ajjanagadde <gajjanagadde@gmail.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
2015-10-11 03:58:47 +02:00
|
|
|
}
|
|
|
|
#endif
|
|
|
|
|
2015-12-16 19:28:39 +01:00
|
|
|
#ifndef ff_clz
|
|
|
|
#define ff_clz ff_clz_c
|
|
|
|
static av_always_inline av_const unsigned ff_clz_c(unsigned x)
|
|
|
|
{
|
|
|
|
unsigned i = sizeof(x) * 8;
|
|
|
|
|
|
|
|
while (x) {
|
|
|
|
x >>= 1;
|
|
|
|
i--;
|
|
|
|
}
|
|
|
|
|
|
|
|
return i;
|
|
|
|
}
|
|
|
|
#endif
|
|
|
|
|
2011-11-20 20:38:24 +01:00
|
|
|
/**
|
|
|
|
* @}
|
|
|
|
*/
|
2010-01-14 20:58:12 +01:00
|
|
|
#endif /* AVUTIL_INTMATH_H */
|