idct_sse2_xvid.c

ffmpeg移植到symbian的全部源代码

/*
 * XVID MPEG-4 VIDEO CODEC
 * - SSE2 inverse discrete cosine transform -
 *
 * Copyright(C) 2003 Pascal Massimino <[e-mail]>
 *
 * Conversion to gcc syntax with modifications
 * by Alexander Strange <[e-mail]>
 *
 * Originally from dct/x86_asm/fdct_sse2_skal.asm in Xvid.
 *
 * This file is part of FFmpeg.
 *
 * Vertical pass is an implementation of the scheme:
 *  Loeffler C., Ligtenberg A., and Moschytz C.S.:
 *  Practical Fast 1D DCT Algorithm with Eleven Multiplications,
 *  Proc. ICASSP 1989, 988-991.
 *
 * Horizontal pass is a double 4x4 vector/matrix multiplication,
 * (see also Intel's Application Note 922:
 *  http://developer.intel.com/vtune/cbts/strmsimd/922down.htm
 *  Copyright (C) 1999 Intel Corporation)
 *
 * More details at http://skal.planet-d.net/coding/dct.html
 *
 * 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
 */

#include "libavcodec/dsputil.h"
#include "libavcodec/i386/idct_xvid.h"

/*!
 * @file idct_sse2_xvid.c
 * @brief SSE2 idct compatible with xvidmmx
 */

#define X8(x)     x,x,x,x,x,x,x,x

#define ROW_SHIFT 11
#define COL_SHIFT 6

DECLARE_ASM_CONST(16, int16_t, tan1[]) = {X8(13036)}; // tan( pi/16)
DECLARE_ASM_CONST(16, int16_t, tan2[]) = {X8(27146)}; // tan(2pi/16) = sqrt(2)-1
DECLARE_ASM_CONST(16, int16_t, tan3[]) = {X8(43790)}; // tan(3pi/16)-1
DECLARE_ASM_CONST(16, int16_t, sqrt2[])= {X8(23170)}; // 0.5/sqrt(2)
DECLARE_ASM_CONST(8,  uint8_t, m127[]) = {X8(127)};

DECLARE_ASM_CONST(16, int16_t, iTab1[]) = {
 0x4000, 0x539f, 0xc000, 0xac61, 0x4000, 0xdd5d, 0x4000, 0xdd5d,
 0x4000, 0x22a3, 0x4000, 0x22a3, 0xc000, 0x539f, 0x4000, 0xac61,
 0x3249, 0x11a8, 0x4b42, 0xee58, 0x11a8, 0x4b42, 0x11a8, 0xcdb7,
 0x58c5, 0x4b42, 0xa73b, 0xcdb7, 0x3249, 0xa73b, 0x4b42, 0xa73b
};

DECLARE_ASM_CONST(16, int16_t, iTab2[]) = {
 0x58c5, 0x73fc, 0xa73b, 0x8c04, 0x58c5, 0xcff5, 0x58c5, 0xcff5,
 0x58c5, 0x300b, 0x58c5, 0x300b, 0xa73b, 0x73fc, 0x58c5, 0x8c04,
 0x45bf, 0x187e, 0x6862, 0xe782, 0x187e, 0x6862, 0x187e, 0xba41,
 0x7b21, 0x6862, 0x84df, 0xba41, 0x45bf, 0x84df, 0x6862, 0x84df
};

DECLARE_ASM_CONST(16, int16_t, iTab3[]) = {
 0x539f, 0x6d41, 0xac61, 0x92bf, 0x539f, 0xd2bf, 0x539f, 0xd2bf,
 0x539f, 0x2d41, 0x539f, 0x2d41, 0xac61, 0x6d41, 0x539f, 0x92bf,
 0x41b3, 0x1712, 0x6254, 0xe8ee, 0x1712, 0x6254, 0x1712, 0xbe4d,
 0x73fc, 0x6254, 0x8c04, 0xbe4d, 0x41b3, 0x8c04, 0x6254, 0x8c04
};

DECLARE_ASM_CONST(16, int16_t, iTab4[]) = {
 0x4b42, 0x6254, 0xb4be, 0x9dac, 0x4b42, 0xd746, 0x4b42, 0xd746,
 0x4b42, 0x28ba, 0x4b42, 0x28ba, 0xb4be, 0x6254, 0x4b42, 0x9dac,
 0x3b21, 0x14c3, 0x587e, 0xeb3d, 0x14c3, 0x587e, 0x14c3, 0xc4df,
 0x6862, 0x587e, 0x979e, 0xc4df, 0x3b21, 0x979e, 0x587e, 0x979e
};

DECLARE_ASM_CONST(16, int32_t, walkenIdctRounders[]) = {
 65536, 65536, 65536, 65536,
  3597,  3597,  3597,  3597,
  2260,  2260,  2260,  2260,
  1203,  1203,  1203,  1203,
   120,   120,   120,   120,
   512,   512,   512,   512
};

// Temporary storage before the column pass
#define ROW1 "%%xmm6"
#define ROW3 "%%xmm4"
#define ROW5 "%%xmm5"
#define ROW7 "%%xmm7"

#define CLEAR_ODD(r) "pxor  "r","r" \n\t"
#define PUT_ODD(dst) "pshufhw  $0x1B, %%xmm2, "dst"   \n\t"

#ifdef ARCH_X86_64

# define ROW0 "%%xmm8"
# define REG0 ROW0
# define ROW2 "%%xmm9"
# define REG2 ROW2
# define ROW4 "%%xmm10"
# define REG4 ROW4
# define ROW6 "%%xmm11"
# define REG6 ROW6
# define CLEAR_EVEN(r) CLEAR_ODD(r)
# define PUT_EVEN(dst) PUT_ODD(dst)
# define XMMS "%%xmm12"
# define MOV_32_ONLY "#"
# define SREG2 REG2
# define TAN3 "%%xmm13"
# define TAN1 "%%xmm14"

#else

# define ROW0 "(%0)"
# define REG0 "%%xmm4"
# define ROW2 "2*16(%0)"
# define REG2 "%%xmm4"
# define ROW4 "4*16(%0)"
# define REG4 "%%xmm6"
# define ROW6 "6*16(%0)"
# define REG6 "%%xmm6"
# define CLEAR_EVEN(r)
# define PUT_EVEN(dst) \
    "pshufhw  $0x1B, %%xmm2, %%xmm2   \n\t" \
    "movdqa          %%xmm2, "dst"    \n\t"
# define XMMS "%%xmm2"
# define MOV_32_ONLY "movdqa "
# define SREG2 "%%xmm7"
# define TAN3 "%%xmm0"
# define TAN1 "%%xmm2"

#endif

#define ROUND(x) "paddd   "MANGLE(x)

#define JZ(reg, to)                         \
    "testl     "reg","reg"            \n\t" \
    "jz        "to"                   \n\t"

#define JNZ(reg, to)                        \
    "testl     "reg","reg"            \n\t" \
    "jnz       "to"                   \n\t"

#define TEST_ONE_ROW(src, reg, clear)       \
    clear                                   \
    "movq     "src", %%mm1            \n\t" \
    "por    8+"src", %%mm1            \n\t" \
    "paddusb  %%mm0, %%mm1            \n\t" \
    "pmovmskb %%mm1, "reg"            \n\t"

#define TEST_TWO_ROWS(row1, row2, reg1, reg2, clear1, clear2) \
    clear1                                  \
    clear2                                  \
    "movq     "row1", %%mm1           \n\t" \
    "por    8+"row1", %%mm1           \n\t" \
    "movq     "row2", %%mm2           \n\t" \
    "por    8+"row2", %%mm2           \n\t" \
    "paddusb   %%mm0, %%mm1           \n\t" \
    "paddusb   %%mm0, %%mm2           \n\t" \
    "pmovmskb  %%mm1, "reg1"          \n\t" \
    "pmovmskb  %%mm2, "reg2"          \n\t"

///IDCT pass on rows.
#define iMTX_MULT(src, table, rounder, put) \
    "movdqa        "src", %%xmm3      \n\t" \
    "movdqa       %%xmm3, %%xmm0      \n\t" \
    "pshufd   $0x11, %%xmm3, %%xmm1   \n\t" /* 4602 */ \
    "punpcklqdq   %%xmm0, %%xmm0      \n\t" /* 0246 */ \
    "pmaddwd     "table", %%xmm0      \n\t" \
    "pmaddwd  16+"table", %%xmm1      \n\t" \
    "pshufd   $0xBB, %%xmm3, %%xmm2   \n\t" /* 5713 */ \
    "punpckhqdq   %%xmm3, %%xmm3      \n\t" /* 1357 */ \
    "pmaddwd  32+"table", %%xmm2      \n\t" \
    "pmaddwd  48+"table", %%xmm3      \n\t" \
    "paddd        %%xmm1, %%xmm0      \n\t" \
    "paddd        %%xmm3, %%xmm2      \n\t" \
    rounder",     %%xmm0              \n\t" \
    "movdqa       %%xmm2, %%xmm3      \n\t" \
    "paddd        %%xmm0, %%xmm2      \n\t" \
    "psubd        %%xmm3, %%xmm0      \n\t" \
    "psrad           $11, %%xmm2      \n\t" \
    "psrad           $11, %%xmm0      \n\t" \
    "packssdw     %%xmm0, %%xmm2      \n\t" \
    put                                     \
    "1:                               \n\t"