virtualenv/FFmpeg
1
0
Fork 0
mirror of https://github.com/FFmpeg/FFmpeg.git synced 2024-11-26 19:01:44 +02:00
Code Issues Releases Activity
f79d847400
FFmpeg/libavutil/intreadwrite.h
Martin Storsjö f79d847400 intreadwrite: Use the __unaligned keyword on MSVC for ARM and x86_64 
AV_WN64 is meant for unaligned data, but the existing av_alias* unions
(without a definition for the av_alias attribute - we don't have one
for MSVC) indicate to the compiler that they would have sufficient
alignment for normal access, i.e. the compiler is free to assume
8 byte alignment.

On ARM, this makes sure that AV_WN64 (or two consecutive AV_WN32) is
done with two str instructions instead of one strd.

Signed-off-by: Martin Storsjö <martin@martin.st>
2016-08-02 11:33:23 +03:00

565 lines
16 KiB
C
Raw Blame History

/*
* This file is part of Libav.
*
* Libav 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.
*
* Libav 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 Libav; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
#ifndef AVUTIL_INTREADWRITE_H
#define AVUTIL_INTREADWRITE_H
#include <stdint.h>
#include "libavutil/avconfig.h"
#include "attributes.h"
#include "bswap.h"
typedef union {
uint64_t u64;
uint32_t u32[2];
uint16_t u16[4];
uint8_t u8 [8];
double f64;
float f32[2];
} av_alias av_alias64;
typedef union {
uint32_t u32;
uint16_t u16[2];
uint8_t u8 [4];
float f32;
} av_alias av_alias32;
typedef union {
uint16_t u16;
uint8_t u8 [2];
} av_alias av_alias16;
/*
* Arch-specific headers can provide any combination of
* AV_[RW][BLN](16|24|32|64) and AV_(COPY|SWAP|ZERO)(64|128) macros.
* Preprocessor symbols must be defined, even if these are implemented
* as inline functions.
*/
#ifdef HAVE_AV_CONFIG_H
#include "config.h"
#if ARCH_ARM
# include "arm/intreadwrite.h"
#elif ARCH_AVR32
# include "avr32/intreadwrite.h"
#elif ARCH_MIPS
# include "mips/intreadwrite.h"
#elif ARCH_PPC
# include "ppc/intreadwrite.h"
#elif ARCH_TOMI
# include "tomi/intreadwrite.h"
#elif ARCH_X86
# include "x86/intreadwrite.h"
#endif
#endif /* HAVE_AV_CONFIG_H */
/*
* Map AV_RNXX <-> AV_R[BL]XX for all variants provided by per-arch headers.
*/
#if AV_HAVE_BIGENDIAN
# if defined(AV_RN16) && !defined(AV_RB16)
# define AV_RB16(p) AV_RN16(p)
# elif !defined(AV_RN16) && defined(AV_RB16)
# define AV_RN16(p) AV_RB16(p)
# endif
# if defined(AV_WN16) && !defined(AV_WB16)
# define AV_WB16(p, v) AV_WN16(p, v)
# elif !defined(AV_WN16) && defined(AV_WB16)
# define AV_WN16(p, v) AV_WB16(p, v)
# endif
# if defined(AV_RN24) && !defined(AV_RB24)
# define AV_RB24(p) AV_RN24(p)
# elif !defined(AV_RN24) && defined(AV_RB24)
# define AV_RN24(p) AV_RB24(p)
# endif
# if defined(AV_WN24) && !defined(AV_WB24)
# define AV_WB24(p, v) AV_WN24(p, v)
# elif !defined(AV_WN24) && defined(AV_WB24)
# define AV_WN24(p, v) AV_WB24(p, v)
# endif
# if defined(AV_RN32) && !defined(AV_RB32)
# define AV_RB32(p) AV_RN32(p)
# elif !defined(AV_RN32) && defined(AV_RB32)
# define AV_RN32(p) AV_RB32(p)
# endif
# if defined(AV_WN32) && !defined(AV_WB32)
# define AV_WB32(p, v) AV_WN32(p, v)
# elif !defined(AV_WN32) && defined(AV_WB32)
# define AV_WN32(p, v) AV_WB32(p, v)
# endif
# if defined(AV_RN64) && !defined(AV_RB64)
# define AV_RB64(p) AV_RN64(p)
# elif !defined(AV_RN64) && defined(AV_RB64)
# define AV_RN64(p) AV_RB64(p)
# endif
# if defined(AV_WN64) && !defined(AV_WB64)
# define AV_WB64(p, v) AV_WN64(p, v)
# elif !defined(AV_WN64) && defined(AV_WB64)
# define AV_WN64(p, v) AV_WB64(p, v)
# endif
#else /* AV_HAVE_BIGENDIAN */
# if defined(AV_RN16) && !defined(AV_RL16)
# define AV_RL16(p) AV_RN16(p)
# elif !defined(AV_RN16) && defined(AV_RL16)
# define AV_RN16(p) AV_RL16(p)
# endif
# if defined(AV_WN16) && !defined(AV_WL16)
# define AV_WL16(p, v) AV_WN16(p, v)
# elif !defined(AV_WN16) && defined(AV_WL16)
# define AV_WN16(p, v) AV_WL16(p, v)
# endif
# if defined(AV_RN24) && !defined(AV_RL24)
# define AV_RL24(p) AV_RN24(p)
# elif !defined(AV_RN24) && defined(AV_RL24)
# define AV_RN24(p) AV_RL24(p)
# endif
# if defined(AV_WN24) && !defined(AV_WL24)
# define AV_WL24(p, v) AV_WN24(p, v)
# elif !defined(AV_WN24) && defined(AV_WL24)
# define AV_WN24(p, v) AV_WL24(p, v)
# endif
# if defined(AV_RN32) && !defined(AV_RL32)
# define AV_RL32(p) AV_RN32(p)
# elif !defined(AV_RN32) && defined(AV_RL32)
# define AV_RN32(p) AV_RL32(p)
# endif
# if defined(AV_WN32) && !defined(AV_WL32)
# define AV_WL32(p, v) AV_WN32(p, v)
# elif !defined(AV_WN32) && defined(AV_WL32)
# define AV_WN32(p, v) AV_WL32(p, v)
# endif
# if defined(AV_RN64) && !defined(AV_RL64)
# define AV_RL64(p) AV_RN64(p)
# elif !defined(AV_RN64) && defined(AV_RL64)
# define AV_RN64(p) AV_RL64(p)
# endif
# if defined(AV_WN64) && !defined(AV_WL64)
# define AV_WL64(p, v) AV_WN64(p, v)
# elif !defined(AV_WN64) && defined(AV_WL64)
# define AV_WN64(p, v) AV_WL64(p, v)
# endif
#endif /* !AV_HAVE_BIGENDIAN */
/*
* Define AV_[RW]N helper macros to simplify definitions not provided
* by per-arch headers.
*/
#if defined(__GNUC__) && !defined(__TI_COMPILER_VERSION__)
union unaligned_64 { uint64_t l; } __attribute__((packed)) av_alias;
union unaligned_32 { uint32_t l; } __attribute__((packed)) av_alias;
union unaligned_16 { uint16_t l; } __attribute__((packed)) av_alias;
# define AV_RN(s, p) (((const union unaligned_##s *) (p))->l)
# define AV_WN(s, p, v) ((((union unaligned_##s *) (p))->l) = (v))
#elif defined(__DECC)
# define AV_RN(s, p) (*((const __unaligned uint##s##_t*)(p)))
# define AV_WN(s, p, v) (*((__unaligned uint##s##_t*)(p)) = (v))
#elif defined(_MSC_VER) && (defined(_M_ARM) || defined(_M_X64)) && AV_HAVE_FAST_UNALIGNED
# define AV_RN(s, p) (*((const __unaligned uint##s##_t*)(p)))
# define AV_WN(s, p, v) (*((__unaligned uint##s##_t*)(p)) = (v))
#elif AV_HAVE_FAST_UNALIGNED
# define AV_RN(s, p) (((const av_alias##s*)(p))->u##s)
# define AV_WN(s, p, v) (((av_alias##s*)(p))->u##s = (v))
#else
#ifndef AV_RB16
# define AV_RB16(x) \
((((const uint8_t*)(x))[0] << 8) | \
((const uint8_t*)(x))[1])
#endif
#ifndef AV_WB16
# define AV_WB16(p, val) do { \
uint16_t d = val; \
((uint8_t*)(p))[1] = (d); \
((uint8_t*)(p))[0] = (d)>>8; \
} while(0)
#endif
#ifndef AV_RL16
# define AV_RL16(x) \
((((const uint8_t*)(x))[1] << 8) | \
((const uint8_t*)(x))[0])
#endif
#ifndef AV_WL16
# define AV_WL16(p, val) do { \
uint16_t d = val; \
((uint8_t*)(p))[0] = (d); \
((uint8_t*)(p))[1] = (d)>>8; \
} while(0)
#endif
#ifndef AV_RB32
# define AV_RB32(x) \
(((uint32_t)((const uint8_t*)(x))[0] << 24) | \
(((const uint8_t*)(x))[1] << 16) | \
(((const uint8_t*)(x))[2] << 8) | \
((const uint8_t*)(x))[3])
#endif
#ifndef AV_WB32
# define AV_WB32(p, val) do { \
uint32_t d = val; \
((uint8_t*)(p))[3] = (d); \
((uint8_t*)(p))[2] = (d)>>8; \
((uint8_t*)(p))[1] = (d)>>16; \
((uint8_t*)(p))[0] = (d)>>24; \
} while(0)
#endif
#ifndef AV_RL32
# define AV_RL32(x) \
(((uint32_t)((const uint8_t*)(x))[3] << 24) | \
(((const uint8_t*)(x))[2] << 16) | \
(((const uint8_t*)(x))[1] << 8) | \
((const uint8_t*)(x))[0])
#endif
#ifndef AV_WL32
# define AV_WL32(p, val) do { \
uint32_t d = val; \
((uint8_t*)(p))[0] = (d); \
((uint8_t*)(p))[1] = (d)>>8; \
((uint8_t*)(p))[2] = (d)>>16; \
((uint8_t*)(p))[3] = (d)>>24; \
} while(0)
#endif
#ifndef AV_RB64
# define AV_RB64(x) \
(((uint64_t)((const uint8_t*)(x))[0] << 56) | \
((uint64_t)((const uint8_t*)(x))[1] << 48) | \
((uint64_t)((const uint8_t*)(x))[2] << 40) | \
((uint64_t)((const uint8_t*)(x))[3] << 32) | \
((uint64_t)((const uint8_t*)(x))[4] << 24) | \
((uint64_t)((const uint8_t*)(x))[5] << 16) | \
((uint64_t)((const uint8_t*)(x))[6] << 8) | \
(uint64_t)((const uint8_t*)(x))[7])
#endif
#ifndef AV_WB64
# define AV_WB64(p, val) do { \
uint64_t d = val; \
((uint8_t*)(p))[7] = (d); \
((uint8_t*)(p))[6] = (d)>>8; \
((uint8_t*)(p))[5] = (d)>>16; \
((uint8_t*)(p))[4] = (d)>>24; \
((uint8_t*)(p))[3] = (d)>>32; \
((uint8_t*)(p))[2] = (d)>>40; \
((uint8_t*)(p))[1] = (d)>>48; \
((uint8_t*)(p))[0] = (d)>>56; \
} while(0)
#endif
#ifndef AV_RL64
# define AV_RL64(x) \
(((uint64_t)((const uint8_t*)(x))[7] << 56) | \
((uint64_t)((const uint8_t*)(x))[6] << 48) | \
((uint64_t)((const uint8_t*)(x))[5] << 40) | \
((uint64_t)((const uint8_t*)(x))[4] << 32) | \
((uint64_t)((const uint8_t*)(x))[3] << 24) | \
((uint64_t)((const uint8_t*)(x))[2] << 16) | \
((uint64_t)((const uint8_t*)(x))[1] << 8) | \
(uint64_t)((const uint8_t*)(x))[0])
#endif
#ifndef AV_WL64
# define AV_WL64(p, val) do { \
uint64_t d = val; \
((uint8_t*)(p))[0] = (d); \
((uint8_t*)(p))[1] = (d)>>8; \
((uint8_t*)(p))[2] = (d)>>16; \
((uint8_t*)(p))[3] = (d)>>24; \
((uint8_t*)(p))[4] = (d)>>32; \
((uint8_t*)(p))[5] = (d)>>40; \
((uint8_t*)(p))[6] = (d)>>48; \
((uint8_t*)(p))[7] = (d)>>56; \
} while(0)
#endif
#if AV_HAVE_BIGENDIAN
# define AV_RN(s, p) AV_RB##s(p)
# define AV_WN(s, p, v) AV_WB##s(p, v)
#else
# define AV_RN(s, p) AV_RL##s(p)
# define AV_WN(s, p, v) AV_WL##s(p, v)
#endif
#endif /* HAVE_FAST_UNALIGNED */
#ifndef AV_RN16
# define AV_RN16(p) AV_RN(16, p)
#endif
#ifndef AV_RN32
# define AV_RN32(p) AV_RN(32, p)
#endif
#ifndef AV_RN64
# define AV_RN64(p) AV_RN(64, p)
#endif
#ifndef AV_WN16
# define AV_WN16(p, v) AV_WN(16, p, v)
#endif
#ifndef AV_WN32
# define AV_WN32(p, v) AV_WN(32, p, v)
#endif
#ifndef AV_WN64
# define AV_WN64(p, v) AV_WN(64, p, v)
#endif
#if AV_HAVE_BIGENDIAN
# define AV_RB(s, p) AV_RN##s(p)
# define AV_WB(s, p, v) AV_WN##s(p, v)
# define AV_RL(s, p) av_bswap##s(AV_RN##s(p))
# define AV_WL(s, p, v) AV_WN##s(p, av_bswap##s(v))
#else
# define AV_RB(s, p) av_bswap##s(AV_RN##s(p))
# define AV_WB(s, p, v) AV_WN##s(p, av_bswap##s(v))
# define AV_RL(s, p) AV_RN##s(p)
# define AV_WL(s, p, v) AV_WN##s(p, v)
#endif
#define AV_RB8(x) (((const uint8_t*)(x))[0])
#define AV_WB8(p, d) do { ((uint8_t*)(p))[0] = (d); } while(0)
#define AV_RL8(x) AV_RB8(x)
#define AV_WL8(p, d) AV_WB8(p, d)
#ifndef AV_RB16
# define AV_RB16(p) AV_RB(16, p)
#endif
#ifndef AV_WB16
# define AV_WB16(p, v) AV_WB(16, p, v)
#endif
#ifndef AV_RL16
# define AV_RL16(p) AV_RL(16, p)
#endif
#ifndef AV_WL16
# define AV_WL16(p, v) AV_WL(16, p, v)
#endif
#ifndef AV_RB32
# define AV_RB32(p) AV_RB(32, p)
#endif
#ifndef AV_WB32
# define AV_WB32(p, v) AV_WB(32, p, v)
#endif
#ifndef AV_RL32
# define AV_RL32(p) AV_RL(32, p)
#endif
#ifndef AV_WL32
# define AV_WL32(p, v) AV_WL(32, p, v)
#endif
#ifndef AV_RB64
# define AV_RB64(p) AV_RB(64, p)
#endif
#ifndef AV_WB64
# define AV_WB64(p, v) AV_WB(64, p, v)
#endif
#ifndef AV_RL64
# define AV_RL64(p) AV_RL(64, p)
#endif
#ifndef AV_WL64
# define AV_WL64(p, v) AV_WL(64, p, v)
#endif
#ifndef AV_RB24
# define AV_RB24(x) \
((((const uint8_t*)(x))[0] << 16) | \
(((const uint8_t*)(x))[1] << 8) | \
((const uint8_t*)(x))[2])
#endif
#ifndef AV_WB24
# define AV_WB24(p, d) do { \
((uint8_t*)(p))[2] = (d); \
((uint8_t*)(p))[1] = (d)>>8; \
((uint8_t*)(p))[0] = (d)>>16; \
} while(0)
#endif
#ifndef AV_RL24
# define AV_RL24(x) \
((((const uint8_t*)(x))[2] << 16) | \
(((const uint8_t*)(x))[1] << 8) | \
((const uint8_t*)(x))[0])
#endif
#ifndef AV_WL24
# define AV_WL24(p, d) do { \
((uint8_t*)(p))[0] = (d); \
((uint8_t*)(p))[1] = (d)>>8; \
((uint8_t*)(p))[2] = (d)>>16; \
} while(0)
#endif
/*
* The AV_[RW]NA macros access naturally aligned data
* in a type-safe way.
*/
#define AV_RNA(s, p) (((const av_alias##s*)(p))->u##s)
#define AV_WNA(s, p, v) (((av_alias##s*)(p))->u##s = (v))
#ifndef AV_RN16A
# define AV_RN16A(p) AV_RNA(16, p)
#endif
#ifndef AV_RN32A
# define AV_RN32A(p) AV_RNA(32, p)
#endif
#ifndef AV_RN64A
# define AV_RN64A(p) AV_RNA(64, p)
#endif
#ifndef AV_WN16A
# define AV_WN16A(p, v) AV_WNA(16, p, v)
#endif
#ifndef AV_WN32A
# define AV_WN32A(p, v) AV_WNA(32, p, v)
#endif
#ifndef AV_WN64A
# define AV_WN64A(p, v) AV_WNA(64, p, v)
#endif
/*
* The AV_COPYxxU macros are suitable for copying data to/from unaligned
* memory locations.
*/
#define AV_COPYU(n, d, s) \
do { \
uint##n##_t val = AV_RN##n(s); \
AV_WN##n(d, val); \
} while (0)
#ifndef AV_COPY16U
# define AV_COPY16U(d, s) AV_COPYU(16, d, s)
#endif
#ifndef AV_COPY32U
# define AV_COPY32U(d, s) AV_COPYU(32, d, s)
#endif
#ifndef AV_COPY64U
# define AV_COPY64U(d, s) AV_COPYU(64, d, s)
#endif
#ifndef AV_COPY128U
# define AV_COPY128U(d, s) \
do { \
AV_COPY64U(d, s); \
AV_COPY64U((char *)(d) + 8, (const char *)(s) + 8); \
} while(0)
#endif
/* Parameters for AV_COPY*, AV_SWAP*, AV_ZERO* must be
* naturally aligned. They may be implemented using MMX,
* so emms_c() must be called before using any float code
* afterwards.
*/
#define AV_COPY(n, d, s) \
(((av_alias##n*)(d))->u##n = ((const av_alias##n*)(s))->u##n)
#ifndef AV_COPY16
# define AV_COPY16(d, s) AV_COPY(16, d, s)
#endif
#ifndef AV_COPY32
# define AV_COPY32(d, s) AV_COPY(32, d, s)
#endif
#ifndef AV_COPY64
# define AV_COPY64(d, s) AV_COPY(64, d, s)
#endif
#ifndef AV_COPY128
# define AV_COPY128(d, s) \
do { \
AV_COPY64(d, s); \
AV_COPY64((char*)(d)+8, (char*)(s)+8); \
} while(0)
#endif
#define AV_SWAP(n, a, b) FFSWAP(av_alias##n, *(av_alias##n*)(a), *(av_alias##n*)(b))
#ifndef AV_SWAP64
# define AV_SWAP64(a, b) AV_SWAP(64, a, b)
#endif
#define AV_ZERO(n, d) (((av_alias##n*)(d))->u##n = 0)
#ifndef AV_ZERO16
# define AV_ZERO16(d) AV_ZERO(16, d)
#endif
#ifndef AV_ZERO32
# define AV_ZERO32(d) AV_ZERO(32, d)
#endif
#ifndef AV_ZERO64
# define AV_ZERO64(d) AV_ZERO(64, d)
#endif
#ifndef AV_ZERO128
# define AV_ZERO128(d) \
do { \
AV_ZERO64(d); \
AV_ZERO64((char*)(d)+8); \
} while(0)
#endif
#endif /* AVUTIL_INTREADWRITE_H */
Reference in New Issue View Git Blame Copy Permalink