falkTX
/
FFmpeg
mirror of https://github.com/falkTX/FFmpeg.git
1
0
Fork 0
Code Issues 0 Releases 338 Wiki Activity
You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
43415 Commits
32 Branches
821MB
Tree: 82b7525173
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from '82b7525173'
${ noResults }
FFmpeg/libavcodec/x86/simple_idct.c

1174 lines
66KB
Raw Blame History 

  1. /*

  2.  * Simple IDCT MMX

  3.  *

  4.  * Copyright (c) 2001, 2002 Michael Niedermayer <michaelni@gmx.at>

  5.  *

  6.  * This file is part of Libav.

  7.  *

  8.  * Libav is free software; you can redistribute it and/or

  9.  * modify it under the terms of the GNU Lesser General Public

  10.  * License as published by the Free Software Foundation; either

  11.  * version 2.1 of the License, or (at your option) any later version.

  12.  *

  13.  * Libav is distributed in the hope that it will be useful,

  14.  * but WITHOUT ANY WARRANTY; without even the implied warranty of

  15.  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU

  16.  * Lesser General Public License for more details.

  17.  *

  18.  * You should have received a copy of the GNU Lesser General Public

  19.  * License along with Libav; if not, write to the Free Software

  20.  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA

  21.  */

  22. 

  23. #include "libavutil/internal.h"

  24. #include "libavutil/mem.h"

  25. #include "libavutil/x86/asm.h"

  26. 

  27. #include "libavcodec/idctdsp.h"

  28. 

  29. #include "idctdsp.h"

  30. #include "simple_idct.h"

  31. 

  32. #if HAVE_INLINE_ASM

  33. 

  34. /*

  35. 23170.475006

  36. 22725.260826

  37. 21406.727617

  38. 19265.545870

  39. 16384.000000

  40. 12872.826198

  41. 8866.956905

  42. 4520.335430

  43. */

  44. #define C0 23170 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5

  45. #define C1 22725 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5

  46. #define C2 21407 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5

  47. #define C3 19266 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5

  48. #define C4 16383 //cos(i*M_PI/16)*sqrt(2)*(1<<14) - 0.5

  49. #define C5 12873 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5

  50. #define C6 8867  //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5

  51. #define C7 4520  //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5

  52. 

  53. #define ROW_SHIFT 11

  54. #define COL_SHIFT 20 // 6

  55. 

  56. DECLARE_ASM_CONST(8, uint64_t, wm1010)= 0xFFFF0000FFFF0000ULL;

  57. DECLARE_ASM_CONST(8, uint64_t, d40000)= 0x0000000000040000ULL;

  58. 

  59. DECLARE_ALIGNED(8, static const int16_t, coeffs)[]= {

  60.         1<<(ROW_SHIFT-1), 0, 1<<(ROW_SHIFT-1), 0,

  61. //        1<<(COL_SHIFT-1), 0, 1<<(COL_SHIFT-1), 0,

  62. //        0, 1<<(COL_SHIFT-1-16), 0, 1<<(COL_SHIFT-1-16),

  63.         1<<(ROW_SHIFT-1), 1, 1<<(ROW_SHIFT-1), 0,

  64.         // the 1 = ((1<<(COL_SHIFT-1))/C4)<<ROW_SHIFT :)

  65. //        0, 0, 0, 0,

  66. //        0, 0, 0, 0,

  67. 

  68.  C4,  C4,  C4,  C4,

  69.  C4, -C4,  C4, -C4,

  70. 

  71.  C2,  C6,  C2,  C6,

  72.  C6, -C2,  C6, -C2,

  73. 

  74.  C1,  C3,  C1,  C3,

  75.  C5,  C7,  C5,  C7,

  76. 

  77.  C3, -C7,  C3, -C7,

  78. -C1, -C5, -C1, -C5,

  79. 

  80.  C5, -C1,  C5, -C1,

  81.  C7,  C3,  C7,  C3,

  82. 

  83.  C7, -C5,  C7, -C5,

  84.  C3, -C1,  C3, -C1

  85. };

  86. 

  87. static inline void idct(int16_t *block)

  88. {

  89.         DECLARE_ALIGNED(8, int64_t, align_tmp)[16];

  90.         int16_t * const temp= (int16_t*)align_tmp;

  91. 

  92.         __asm__ volatile(

  93. #if 0 //Alternative, simpler variant

  94. 

  95. #define ROW_IDCT(src0, src4, src1, src5, dst, rounder, shift) \

  96.         "movq " #src0 ", %%mm0          \n\t" /* R4     R0      r4      r0 */\

  97.         "movq " #src4 ", %%mm1          \n\t" /* R6     R2      r6      r2 */\

  98.         "movq " #src1 ", %%mm2          \n\t" /* R3     R1      r3      r1 */\

  99.         "movq " #src5 ", %%mm3          \n\t" /* R7     R5      r7      r5 */\

  100.         "movq 16(%2), %%mm4             \n\t" /* C4     C4      C4      C4 */\

  101.         "pmaddwd %%mm0, %%mm4           \n\t" /* C4R4+C4R0      C4r4+C4r0 */\

  102.         "movq 24(%2), %%mm5             \n\t" /* -C4    C4      -C4     C4 */\

  103.         "pmaddwd %%mm5, %%mm0           \n\t" /* -C4R4+C4R0     -C4r4+C4r0 */\

  104.         "movq 32(%2), %%mm5             \n\t" /* C6     C2      C6      C2 */\

  105.         "pmaddwd %%mm1, %%mm5           \n\t" /* C6R6+C2R2      C6r6+C2r2 */\

  106.         "movq 40(%2), %%mm6             \n\t" /* -C2    C6      -C2     C6 */\

  107.         "pmaddwd %%mm6, %%mm1           \n\t" /* -C2R6+C6R2     -C2r6+C6r2 */\

  108.         "movq 48(%2), %%mm7             \n\t" /* C3     C1      C3      C1 */\

  109.         "pmaddwd %%mm2, %%mm7           \n\t" /* C3R3+C1R1      C3r3+C1r1 */\

  110.         #rounder ", %%mm4               \n\t"\

  111.         "movq %%mm4, %%mm6              \n\t" /* C4R4+C4R0      C4r4+C4r0 */\

  112.         "paddd %%mm5, %%mm4             \n\t" /* A0             a0 */\

  113.         "psubd %%mm5, %%mm6             \n\t" /* A3             a3 */\

  114.         "movq 56(%2), %%mm5             \n\t" /* C7     C5      C7      C5 */\

  115.         "pmaddwd %%mm3, %%mm5           \n\t" /* C7R7+C5R5      C7r7+C5r5 */\

  116.         #rounder ", %%mm0               \n\t"\

  117.         "paddd %%mm0, %%mm1             \n\t" /* A1             a1 */\

  118.         "paddd %%mm0, %%mm0             \n\t" \

  119.         "psubd %%mm1, %%mm0             \n\t" /* A2             a2 */\

  120.         "pmaddwd 64(%2), %%mm2          \n\t" /* -C7R3+C3R1     -C7r3+C3r1 */\

  121.         "paddd %%mm5, %%mm7             \n\t" /* B0             b0 */\

  122.         "movq 72(%2), %%mm5             \n\t" /* -C5    -C1     -C5     -C1 */\

  123.         "pmaddwd %%mm3, %%mm5           \n\t" /* -C5R7-C1R5     -C5r7-C1r5 */\

  124.         "paddd %%mm4, %%mm7             \n\t" /* A0+B0          a0+b0 */\

  125.         "paddd %%mm4, %%mm4             \n\t" /* 2A0            2a0 */\

  126.         "psubd %%mm7, %%mm4             \n\t" /* A0-B0          a0-b0 */\

  127.         "paddd %%mm2, %%mm5             \n\t" /* B1             b1 */\

  128.         "psrad $" #shift ", %%mm7       \n\t"\

  129.         "psrad $" #shift ", %%mm4       \n\t"\

  130.         "movq %%mm1, %%mm2              \n\t" /* A1             a1 */\

  131.         "paddd %%mm5, %%mm1             \n\t" /* A1+B1          a1+b1 */\

  132.         "psubd %%mm5, %%mm2             \n\t" /* A1-B1          a1-b1 */\

  133.         "psrad $" #shift ", %%mm1       \n\t"\

  134.         "psrad $" #shift ", %%mm2       \n\t"\

  135.         "packssdw %%mm1, %%mm7          \n\t" /* A1+B1  a1+b1   A0+B0   a0+b0 */\

  136.         "packssdw %%mm4, %%mm2          \n\t" /* A0-B0  a0-b0   A1-B1   a1-b1 */\

  137.         "movq %%mm7, " #dst "           \n\t"\

  138.         "movq " #src1 ", %%mm1          \n\t" /* R3     R1      r3      r1 */\

  139.         "movq 80(%2), %%mm4             \n\t" /* -C1    C5      -C1     C5 */\

  140.         "movq %%mm2, 24+" #dst "        \n\t"\

  141.         "pmaddwd %%mm1, %%mm4           \n\t" /* -C1R3+C5R1     -C1r3+C5r1 */\

  142.         "movq 88(%2), %%mm7             \n\t" /* C3     C7      C3      C7 */\

  143.         "pmaddwd 96(%2), %%mm1          \n\t" /* -C5R3+C7R1     -C5r3+C7r1 */\

  144.         "pmaddwd %%mm3, %%mm7           \n\t" /* C3R7+C7R5      C3r7+C7r5 */\

  145.         "movq %%mm0, %%mm2              \n\t" /* A2             a2 */\

  146.         "pmaddwd 104(%2), %%mm3         \n\t" /* -C1R7+C3R5     -C1r7+C3r5 */\

  147.         "paddd %%mm7, %%mm4             \n\t" /* B2             b2 */\

  148.         "paddd %%mm4, %%mm2             \n\t" /* A2+B2          a2+b2 */\

  149.         "psubd %%mm4, %%mm0             \n\t" /* a2-B2          a2-b2 */\

  150.         "psrad $" #shift ", %%mm2       \n\t"\

  151.         "psrad $" #shift ", %%mm0       \n\t"\

  152.         "movq %%mm6, %%mm4              \n\t" /* A3             a3 */\

  153.         "paddd %%mm1, %%mm3             \n\t" /* B3             b3 */\

  154.         "paddd %%mm3, %%mm6             \n\t" /* A3+B3          a3+b3 */\

  155.         "psubd %%mm3, %%mm4             \n\t" /* a3-B3          a3-b3 */\

  156.         "psrad $" #shift ", %%mm6       \n\t"\

  157.         "packssdw %%mm6, %%mm2          \n\t" /* A3+B3  a3+b3   A2+B2   a2+b2 */\

  158.         "movq %%mm2, 8+" #dst "         \n\t"\

  159.         "psrad $" #shift ", %%mm4       \n\t"\

  160.         "packssdw %%mm0, %%mm4          \n\t" /* A2-B2  a2-b2   A3-B3   a3-b3 */\

  161.         "movq %%mm4, 16+" #dst "        \n\t"\

  162. 

  163. #define COL_IDCT(src0, src4, src1, src5, dst, shift) \

  164.         "movq " #src0 ", %%mm0          \n\t" /* R4     R0      r4      r0 */\

  165.         "movq " #src4 ", %%mm1          \n\t" /* R6     R2      r6      r2 */\

  166.         "movq " #src1 ", %%mm2          \n\t" /* R3     R1      r3      r1 */\

  167.         "movq " #src5 ", %%mm3          \n\t" /* R7     R5      r7      r5 */\

  168.         "movq 16(%2), %%mm4             \n\t" /* C4     C4      C4      C4 */\

  169.         "pmaddwd %%mm0, %%mm4           \n\t" /* C4R4+C4R0      C4r4+C4r0 */\

  170.         "movq 24(%2), %%mm5             \n\t" /* -C4    C4      -C4     C4 */\

  171.         "pmaddwd %%mm5, %%mm0           \n\t" /* -C4R4+C4R0     -C4r4+C4r0 */\

  172.         "movq 32(%2), %%mm5             \n\t" /* C6     C2      C6      C2 */\

  173.         "pmaddwd %%mm1, %%mm5           \n\t" /* C6R6+C2R2      C6r6+C2r2 */\

  174.         "movq 40(%2), %%mm6             \n\t" /* -C2    C6      -C2     C6 */\

  175.         "pmaddwd %%mm6, %%mm1           \n\t" /* -C2R6+C6R2     -C2r6+C6r2 */\

  176.         "movq %%mm4, %%mm6              \n\t" /* C4R4+C4R0      C4r4+C4r0 */\

  177.         "movq 48(%2), %%mm7             \n\t" /* C3     C1      C3      C1 */\

  178.         "pmaddwd %%mm2, %%mm7           \n\t" /* C3R3+C1R1      C3r3+C1r1 */\

  179.         "paddd %%mm5, %%mm4             \n\t" /* A0             a0 */\

  180.         "psubd %%mm5, %%mm6             \n\t" /* A3             a3 */\

  181.         "movq %%mm0, %%mm5              \n\t" /* -C4R4+C4R0     -C4r4+C4r0 */\

  182.         "paddd %%mm1, %%mm0             \n\t" /* A1             a1 */\

  183.         "psubd %%mm1, %%mm5             \n\t" /* A2             a2 */\

  184.         "movq 56(%2), %%mm1             \n\t" /* C7     C5      C7      C5 */\

  185.         "pmaddwd %%mm3, %%mm1           \n\t" /* C7R7+C5R5      C7r7+C5r5 */\

  186.         "pmaddwd 64(%2), %%mm2          \n\t" /* -C7R3+C3R1     -C7r3+C3r1 */\

  187.         "paddd %%mm1, %%mm7             \n\t" /* B0             b0 */\

  188.         "movq 72(%2), %%mm1             \n\t" /* -C5    -C1     -C5     -C1 */\

  189.         "pmaddwd %%mm3, %%mm1           \n\t" /* -C5R7-C1R5     -C5r7-C1r5 */\

  190.         "paddd %%mm4, %%mm7             \n\t" /* A0+B0          a0+b0 */\

  191.         "paddd %%mm4, %%mm4             \n\t" /* 2A0            2a0 */\

  192.         "psubd %%mm7, %%mm4             \n\t" /* A0-B0          a0-b0 */\

  193.         "paddd %%mm2, %%mm1             \n\t" /* B1             b1 */\

  194.         "psrad $" #shift ", %%mm7       \n\t"\

  195.         "psrad $" #shift ", %%mm4       \n\t"\

  196.         "movq %%mm0, %%mm2              \n\t" /* A1             a1 */\

  197.         "paddd %%mm1, %%mm0             \n\t" /* A1+B1          a1+b1 */\

  198.         "psubd %%mm1, %%mm2             \n\t" /* A1-B1          a1-b1 */\

  199.         "psrad $" #shift ", %%mm0       \n\t"\

  200.         "psrad $" #shift ", %%mm2       \n\t"\

  201.         "packssdw %%mm7, %%mm7          \n\t" /* A0+B0  a0+b0 */\

  202.         "movd %%mm7, " #dst "           \n\t"\

  203.         "packssdw %%mm0, %%mm0          \n\t" /* A1+B1  a1+b1 */\

  204.         "movd %%mm0, 16+" #dst "        \n\t"\

  205.         "packssdw %%mm2, %%mm2          \n\t" /* A1-B1  a1-b1 */\

  206.         "movd %%mm2, 96+" #dst "        \n\t"\

  207.         "packssdw %%mm4, %%mm4          \n\t" /* A0-B0  a0-b0 */\

  208.         "movd %%mm4, 112+" #dst "       \n\t"\

  209.         "movq " #src1 ", %%mm0          \n\t" /* R3     R1      r3      r1 */\

  210.         "movq 80(%2), %%mm4             \n\t" /* -C1    C5      -C1     C5 */\

  211.         "pmaddwd %%mm0, %%mm4           \n\t" /* -C1R3+C5R1     -C1r3+C5r1 */\

  212.         "movq 88(%2), %%mm7             \n\t" /* C3     C7      C3      C7 */\

  213.         "pmaddwd 96(%2), %%mm0          \n\t" /* -C5R3+C7R1     -C5r3+C7r1 */\

  214.         "pmaddwd %%mm3, %%mm7           \n\t" /* C3R7+C7R5      C3r7+C7r5 */\

  215.         "movq %%mm5, %%mm2              \n\t" /* A2             a2 */\

  216.         "pmaddwd 104(%2), %%mm3         \n\t" /* -C1R7+C3R5     -C1r7+C3r5 */\

  217.         "paddd %%mm7, %%mm4             \n\t" /* B2             b2 */\

  218.         "paddd %%mm4, %%mm2             \n\t" /* A2+B2          a2+b2 */\

  219.         "psubd %%mm4, %%mm5             \n\t" /* a2-B2          a2-b2 */\

  220.         "psrad $" #shift ", %%mm2       \n\t"\

  221.         "psrad $" #shift ", %%mm5       \n\t"\

  222.         "movq %%mm6, %%mm4              \n\t" /* A3             a3 */\

  223.         "paddd %%mm0, %%mm3             \n\t" /* B3             b3 */\

  224.         "paddd %%mm3, %%mm6             \n\t" /* A3+B3          a3+b3 */\

  225.         "psubd %%mm3, %%mm4             \n\t" /* a3-B3          a3-b3 */\

  226.         "psrad $" #shift ", %%mm6       \n\t"\

  227.         "psrad $" #shift ", %%mm4       \n\t"\

  228.         "packssdw %%mm2, %%mm2          \n\t" /* A2+B2  a2+b2 */\

  229.         "packssdw %%mm6, %%mm6          \n\t" /* A3+B3  a3+b3 */\

  230.         "movd %%mm2, 32+" #dst "        \n\t"\

  231.         "packssdw %%mm4, %%mm4          \n\t" /* A3-B3  a3-b3 */\

  232.         "packssdw %%mm5, %%mm5          \n\t" /* A2-B2  a2-b2 */\

  233.         "movd %%mm6, 48+" #dst "        \n\t"\

  234.         "movd %%mm4, 64+" #dst "        \n\t"\

  235.         "movd %%mm5, 80+" #dst "        \n\t"\

  236. 

  237. 

  238. #define DC_COND_ROW_IDCT(src0, src4, src1, src5, dst, rounder, shift) \

  239.         "movq " #src0 ", %%mm0          \n\t" /* R4     R0      r4      r0 */\

  240.         "movq " #src4 ", %%mm1          \n\t" /* R6     R2      r6      r2 */\

  241.         "movq " #src1 ", %%mm2          \n\t" /* R3     R1      r3      r1 */\

  242.         "movq " #src5 ", %%mm3          \n\t" /* R7     R5      r7      r5 */\

  243.         "movq "MANGLE(wm1010)", %%mm4   \n\t"\

  244.         "pand %%mm0, %%mm4              \n\t"\

  245.         "por %%mm1, %%mm4               \n\t"\

  246.         "por %%mm2, %%mm4               \n\t"\

  247.         "por %%mm3, %%mm4               \n\t"\

  248.         "packssdw %%mm4,%%mm4           \n\t"\

  249.         "movd %%mm4, %%eax              \n\t"\

  250.         "orl %%eax, %%eax               \n\t"\

  251.         "jz 1f                          \n\t"\

  252.         "movq 16(%2), %%mm4             \n\t" /* C4     C4      C4      C4 */\

  253.         "pmaddwd %%mm0, %%mm4           \n\t" /* C4R4+C4R0      C4r4+C4r0 */\

  254.         "movq 24(%2), %%mm5             \n\t" /* -C4    C4      -C4     C4 */\

  255.         "pmaddwd %%mm5, %%mm0           \n\t" /* -C4R4+C4R0     -C4r4+C4r0 */\

  256.         "movq 32(%2), %%mm5             \n\t" /* C6     C2      C6      C2 */\

  257.         "pmaddwd %%mm1, %%mm5           \n\t" /* C6R6+C2R2      C6r6+C2r2 */\

  258.         "movq 40(%2), %%mm6             \n\t" /* -C2    C6      -C2     C6 */\

  259.         "pmaddwd %%mm6, %%mm1           \n\t" /* -C2R6+C6R2     -C2r6+C6r2 */\

  260.         "movq 48(%2), %%mm7             \n\t" /* C3     C1      C3      C1 */\

  261.         "pmaddwd %%mm2, %%mm7           \n\t" /* C3R3+C1R1      C3r3+C1r1 */\

  262.         #rounder ", %%mm4               \n\t"\

  263.         "movq %%mm4, %%mm6              \n\t" /* C4R4+C4R0      C4r4+C4r0 */\

  264.         "paddd %%mm5, %%mm4             \n\t" /* A0             a0 */\

  265.         "psubd %%mm5, %%mm6             \n\t" /* A3             a3 */\

  266.         "movq 56(%2), %%mm5             \n\t" /* C7     C5      C7      C5 */\

  267.         "pmaddwd %%mm3, %%mm5           \n\t" /* C7R7+C5R5      C7r7+C5r5 */\

  268.         #rounder ", %%mm0               \n\t"\

  269.         "paddd %%mm0, %%mm1             \n\t" /* A1             a1 */\

  270.         "paddd %%mm0, %%mm0             \n\t" \

  271.         "psubd %%mm1, %%mm0             \n\t" /* A2             a2 */\

  272.         "pmaddwd 64(%2), %%mm2          \n\t" /* -C7R3+C3R1     -C7r3+C3r1 */\

  273.         "paddd %%mm5, %%mm7             \n\t" /* B0             b0 */\

  274.         "movq 72(%2), %%mm5             \n\t" /* -C5    -C1     -C5     -C1 */\

  275.         "pmaddwd %%mm3, %%mm5           \n\t" /* -C5R7-C1R5     -C5r7-C1r5 */\

  276.         "paddd %%mm4, %%mm7             \n\t" /* A0+B0          a0+b0 */\

  277.         "paddd %%mm4, %%mm4             \n\t" /* 2A0            2a0 */\

  278.         "psubd %%mm7, %%mm4             \n\t" /* A0-B0          a0-b0 */\

  279.         "paddd %%mm2, %%mm5             \n\t" /* B1             b1 */\

  280.         "psrad $" #shift ", %%mm7       \n\t"\

  281.         "psrad $" #shift ", %%mm4       \n\t"\

  282.         "movq %%mm1, %%mm2              \n\t" /* A1             a1 */\

  283.         "paddd %%mm5, %%mm1             \n\t" /* A1+B1          a1+b1 */\

  284.         "psubd %%mm5, %%mm2             \n\t" /* A1-B1          a1-b1 */\

  285.         "psrad $" #shift ", %%mm1       \n\t"\

  286.         "psrad $" #shift ", %%mm2       \n\t"\

  287.         "packssdw %%mm1, %%mm7          \n\t" /* A1+B1  a1+b1   A0+B0   a0+b0 */\

  288.         "packssdw %%mm4, %%mm2          \n\t" /* A0-B0  a0-b0   A1-B1   a1-b1 */\

  289.         "movq %%mm7, " #dst "           \n\t"\

  290.         "movq " #src1 ", %%mm1          \n\t" /* R3     R1      r3      r1 */\

  291.         "movq 80(%2), %%mm4             \n\t" /* -C1    C5      -C1     C5 */\

  292.         "movq %%mm2, 24+" #dst "        \n\t"\

  293.         "pmaddwd %%mm1, %%mm4           \n\t" /* -C1R3+C5R1     -C1r3+C5r1 */\

  294.         "movq 88(%2), %%mm7             \n\t" /* C3     C7      C3      C7 */\

  295.         "pmaddwd 96(%2), %%mm1          \n\t" /* -C5R3+C7R1     -C5r3+C7r1 */\

  296.         "pmaddwd %%mm3, %%mm7           \n\t" /* C3R7+C7R5      C3r7+C7r5 */\

  297.         "movq %%mm0, %%mm2              \n\t" /* A2             a2 */\

  298.         "pmaddwd 104(%2), %%mm3         \n\t" /* -C1R7+C3R5     -C1r7+C3r5 */\

  299.         "paddd %%mm7, %%mm4             \n\t" /* B2             b2 */\

  300.         "paddd %%mm4, %%mm2             \n\t" /* A2+B2          a2+b2 */\

  301.         "psubd %%mm4, %%mm0             \n\t" /* a2-B2          a2-b2 */\

  302.         "psrad $" #shift ", %%mm2       \n\t"\

  303.         "psrad $" #shift ", %%mm0       \n\t"\

  304.         "movq %%mm6, %%mm4              \n\t" /* A3             a3 */\

  305.         "paddd %%mm1, %%mm3             \n\t" /* B3             b3 */\

  306.         "paddd %%mm3, %%mm6             \n\t" /* A3+B3          a3+b3 */\

  307.         "psubd %%mm3, %%mm4             \n\t" /* a3-B3          a3-b3 */\

  308.         "psrad $" #shift ", %%mm6       \n\t"\

  309.         "packssdw %%mm6, %%mm2          \n\t" /* A3+B3  a3+b3   A2+B2   a2+b2 */\

  310.         "movq %%mm2, 8+" #dst "         \n\t"\

  311.         "psrad $" #shift ", %%mm4       \n\t"\

  312.         "packssdw %%mm0, %%mm4          \n\t" /* A2-B2  a2-b2   A3-B3   a3-b3 */\

  313.         "movq %%mm4, 16+" #dst "        \n\t"\

  314.         "jmp 2f                         \n\t"\

  315.         "1:                             \n\t"\

  316.         "pslld $16, %%mm0               \n\t"\

  317.         "#paddd "MANGLE(d40000)", %%mm0 \n\t"\

  318.         "psrad $13, %%mm0               \n\t"\

  319.         "packssdw %%mm0, %%mm0          \n\t"\

  320.         "movq %%mm0, " #dst "           \n\t"\

  321.         "movq %%mm0, 8+" #dst "         \n\t"\

  322.         "movq %%mm0, 16+" #dst "        \n\t"\

  323.         "movq %%mm0, 24+" #dst "        \n\t"\

  324.         "2:                             \n\t"

  325. 

  326. 

  327. //IDCT(      src0,   src4,   src1,   src5,    dst,    rounder, shift)

  328. ROW_IDCT(    (%0),  8(%0), 16(%0), 24(%0),  0(%1),paddd 8(%2), 11)

  329. /*ROW_IDCT(  32(%0), 40(%0), 48(%0), 56(%0), 32(%1), paddd (%2), 11)

  330. ROW_IDCT(  64(%0), 72(%0), 80(%0), 88(%0), 64(%1), paddd (%2), 11)

  331. ROW_IDCT(  96(%0),104(%0),112(%0),120(%0), 96(%1), paddd (%2), 11)*/

  332. 

  333. DC_COND_ROW_IDCT(  32(%0), 40(%0), 48(%0), 56(%0), 32(%1),paddd (%2), 11)

  334. DC_COND_ROW_IDCT(  64(%0), 72(%0), 80(%0), 88(%0), 64(%1),paddd (%2), 11)

  335. DC_COND_ROW_IDCT(  96(%0),104(%0),112(%0),120(%0), 96(%1),paddd (%2), 11)

  336. 

  337. 

  338. //IDCT(      src0,   src4,   src1,    src5,    dst, shift)

  339. COL_IDCT(    (%1), 64(%1), 32(%1),  96(%1),  0(%0), 20)

  340. COL_IDCT(   8(%1), 72(%1), 40(%1), 104(%1),  4(%0), 20)

  341. COL_IDCT(  16(%1), 80(%1), 48(%1), 112(%1),  8(%0), 20)

  342. COL_IDCT(  24(%1), 88(%1), 56(%1), 120(%1), 12(%0), 20)

  343. 

  344. #else

  345. 

  346. #define DC_COND_IDCT(src0, src4, src1, src5, dst, rounder, shift) \

  347.         "movq " #src0 ", %%mm0          \n\t" /* R4     R0      r4      r0 */\

  348.         "movq " #src4 ", %%mm1          \n\t" /* R6     R2      r6      r2 */\

  349.         "movq " #src1 ", %%mm2          \n\t" /* R3     R1      r3      r1 */\

  350.         "movq " #src5 ", %%mm3          \n\t" /* R7     R5      r7      r5 */\

  351.         "movq "MANGLE(wm1010)", %%mm4   \n\t"\

  352.         "pand %%mm0, %%mm4              \n\t"\

  353.         "por %%mm1, %%mm4               \n\t"\

  354.         "por %%mm2, %%mm4               \n\t"\

  355.         "por %%mm3, %%mm4               \n\t"\

  356.         "packssdw %%mm4,%%mm4           \n\t"\

  357.         "movd %%mm4, %%eax              \n\t"\

  358.         "orl %%eax, %%eax               \n\t"\

  359.         "jz 1f                          \n\t"\

  360.         "movq 16(%2), %%mm4             \n\t" /* C4     C4      C4      C4 */\

  361.         "pmaddwd %%mm0, %%mm4           \n\t" /* C4R4+C4R0      C4r4+C4r0 */\

  362.         "movq 24(%2), %%mm5             \n\t" /* -C4    C4      -C4     C4 */\

  363.         "pmaddwd %%mm5, %%mm0           \n\t" /* -C4R4+C4R0     -C4r4+C4r0 */\

  364.         "movq 32(%2), %%mm5             \n\t" /* C6     C2      C6      C2 */\

  365.         "pmaddwd %%mm1, %%mm5           \n\t" /* C6R6+C2R2      C6r6+C2r2 */\

  366.         "movq 40(%2), %%mm6             \n\t" /* -C2    C6      -C2     C6 */\

  367.         "pmaddwd %%mm6, %%mm1           \n\t" /* -C2R6+C6R2     -C2r6+C6r2 */\

  368.         "movq 48(%2), %%mm7             \n\t" /* C3     C1      C3      C1 */\

  369.         "pmaddwd %%mm2, %%mm7           \n\t" /* C3R3+C1R1      C3r3+C1r1 */\

  370.         #rounder ", %%mm4               \n\t"\

  371.         "movq %%mm4, %%mm6              \n\t" /* C4R4+C4R0      C4r4+C4r0 */\

  372.         "paddd %%mm5, %%mm4             \n\t" /* A0             a0 */\

  373.         "psubd %%mm5, %%mm6             \n\t" /* A3             a3 */\

  374.         "movq 56(%2), %%mm5             \n\t" /* C7     C5      C7      C5 */\

  375.         "pmaddwd %%mm3, %%mm5           \n\t" /* C7R7+C5R5      C7r7+C5r5 */\

  376.         #rounder ", %%mm0               \n\t"\

  377.         "paddd %%mm0, %%mm1             \n\t" /* A1             a1 */\

  378.         "paddd %%mm0, %%mm0             \n\t" \

  379.         "psubd %%mm1, %%mm0             \n\t" /* A2             a2 */\

  380.         "pmaddwd 64(%2), %%mm2          \n\t" /* -C7R3+C3R1     -C7r3+C3r1 */\

  381.         "paddd %%mm5, %%mm7             \n\t" /* B0             b0 */\

  382.         "movq 72(%2), %%mm5             \n\t" /* -C5    -C1     -C5     -C1 */\

  383.         "pmaddwd %%mm3, %%mm5           \n\t" /* -C5R7-C1R5     -C5r7-C1r5 */\

  384.         "paddd %%mm4, %%mm7             \n\t" /* A0+B0          a0+b0 */\

  385.         "paddd %%mm4, %%mm4             \n\t" /* 2A0            2a0 */\

  386.         "psubd %%mm7, %%mm4             \n\t" /* A0-B0          a0-b0 */\

  387.         "paddd %%mm2, %%mm5             \n\t" /* B1             b1 */\

  388.         "psrad $" #shift ", %%mm7       \n\t"\

  389.         "psrad $" #shift ", %%mm4       \n\t"\

  390.         "movq %%mm1, %%mm2              \n\t" /* A1             a1 */\

  391.         "paddd %%mm5, %%mm1             \n\t" /* A1+B1          a1+b1 */\

  392.         "psubd %%mm5, %%mm2             \n\t" /* A1-B1          a1-b1 */\

  393.         "psrad $" #shift ", %%mm1       \n\t"\

  394.         "psrad $" #shift ", %%mm2       \n\t"\

  395.         "packssdw %%mm1, %%mm7          \n\t" /* A1+B1  a1+b1   A0+B0   a0+b0 */\

  396.         "packssdw %%mm4, %%mm2          \n\t" /* A0-B0  a0-b0   A1-B1   a1-b1 */\

  397.         "movq %%mm7, " #dst "           \n\t"\

  398.         "movq " #src1 ", %%mm1          \n\t" /* R3     R1      r3      r1 */\

  399.         "movq 80(%2), %%mm4             \n\t" /* -C1    C5      -C1     C5 */\

  400.         "movq %%mm2, 24+" #dst "        \n\t"\

  401.         "pmaddwd %%mm1, %%mm4           \n\t" /* -C1R3+C5R1     -C1r3+C5r1 */\

  402.         "movq 88(%2), %%mm7             \n\t" /* C3     C7      C3      C7 */\

  403.         "pmaddwd 96(%2), %%mm1          \n\t" /* -C5R3+C7R1     -C5r3+C7r1 */\

  404.         "pmaddwd %%mm3, %%mm7           \n\t" /* C3R7+C7R5      C3r7+C7r5 */\

  405.         "movq %%mm0, %%mm2              \n\t" /* A2             a2 */\

  406.         "pmaddwd 104(%2), %%mm3         \n\t" /* -C1R7+C3R5     -C1r7+C3r5 */\

  407.         "paddd %%mm7, %%mm4             \n\t" /* B2             b2 */\

  408.         "paddd %%mm4, %%mm2             \n\t" /* A2+B2          a2+b2 */\

  409.         "psubd %%mm4, %%mm0             \n\t" /* a2-B2          a2-b2 */\

  410.         "psrad $" #shift ", %%mm2       \n\t"\

  411.         "psrad $" #shift ", %%mm0       \n\t"\

  412.         "movq %%mm6, %%mm4              \n\t" /* A3             a3 */\

  413.         "paddd %%mm1, %%mm3             \n\t" /* B3             b3 */\

  414.         "paddd %%mm3, %%mm6             \n\t" /* A3+B3          a3+b3 */\

  415.         "psubd %%mm3, %%mm4             \n\t" /* a3-B3          a3-b3 */\

  416.         "psrad $" #shift ", %%mm6       \n\t"\

  417.         "packssdw %%mm6, %%mm2          \n\t" /* A3+B3  a3+b3   A2+B2   a2+b2 */\

  418.         "movq %%mm2, 8+" #dst "         \n\t"\

  419.         "psrad $" #shift ", %%mm4       \n\t"\

  420.         "packssdw %%mm0, %%mm4          \n\t" /* A2-B2  a2-b2   A3-B3   a3-b3 */\

  421.         "movq %%mm4, 16+" #dst "        \n\t"\

  422.         "jmp 2f                         \n\t"\

  423.         "1:                             \n\t"\

  424.         "pslld $16, %%mm0               \n\t"\

  425.         "paddd "MANGLE(d40000)", %%mm0  \n\t"\

  426.         "psrad $13, %%mm0               \n\t"\

  427.         "packssdw %%mm0, %%mm0          \n\t"\

  428.         "movq %%mm0, " #dst "           \n\t"\

  429.         "movq %%mm0, 8+" #dst "         \n\t"\

  430.         "movq %%mm0, 16+" #dst "        \n\t"\

  431.         "movq %%mm0, 24+" #dst "        \n\t"\

  432.         "2:                             \n\t"

  433. 

  434. #define Z_COND_IDCT(src0, src4, src1, src5, dst, rounder, shift, bt) \

  435.         "movq " #src0 ", %%mm0          \n\t" /* R4     R0      r4      r0 */\

  436.         "movq " #src4 ", %%mm1          \n\t" /* R6     R2      r6      r2 */\

  437.         "movq " #src1 ", %%mm2          \n\t" /* R3     R1      r3      r1 */\

  438.         "movq " #src5 ", %%mm3          \n\t" /* R7     R5      r7      r5 */\

  439.         "movq %%mm0, %%mm4              \n\t"\

  440.         "por %%mm1, %%mm4               \n\t"\

  441.         "por %%mm2, %%mm4               \n\t"\

  442.         "por %%mm3, %%mm4               \n\t"\

  443.         "packssdw %%mm4,%%mm4           \n\t"\

  444.         "movd %%mm4, %%eax              \n\t"\

  445.         "orl %%eax, %%eax               \n\t"\

  446.         "jz " #bt "                     \n\t"\

  447.         "movq 16(%2), %%mm4             \n\t" /* C4     C4      C4      C4 */\

  448.         "pmaddwd %%mm0, %%mm4           \n\t" /* C4R4+C4R0      C4r4+C4r0 */\

  449.         "movq 24(%2), %%mm5             \n\t" /* -C4    C4      -C4     C4 */\

  450.         "pmaddwd %%mm5, %%mm0           \n\t" /* -C4R4+C4R0     -C4r4+C4r0 */\

  451.         "movq 32(%2), %%mm5             \n\t" /* C6     C2      C6      C2 */\

  452.         "pmaddwd %%mm1, %%mm5           \n\t" /* C6R6+C2R2      C6r6+C2r2 */\

  453.         "movq 40(%2), %%mm6             \n\t" /* -C2    C6      -C2     C6 */\

  454.         "pmaddwd %%mm6, %%mm1           \n\t" /* -C2R6+C6R2     -C2r6+C6r2 */\

  455.         "movq 48(%2), %%mm7             \n\t" /* C3     C1      C3      C1 */\

  456.         "pmaddwd %%mm2, %%mm7           \n\t" /* C3R3+C1R1      C3r3+C1r1 */\

  457.         #rounder ", %%mm4               \n\t"\

  458.         "movq %%mm4, %%mm6              \n\t" /* C4R4+C4R0      C4r4+C4r0 */\

  459.         "paddd %%mm5, %%mm4             \n\t" /* A0             a0 */\

  460.         "psubd %%mm5, %%mm6             \n\t" /* A3             a3 */\

  461.         "movq 56(%2), %%mm5             \n\t" /* C7     C5      C7      C5 */\

  462.         "pmaddwd %%mm3, %%mm5           \n\t" /* C7R7+C5R5      C7r7+C5r5 */\

  463.         #rounder ", %%mm0               \n\t"\

  464.         "paddd %%mm0, %%mm1             \n\t" /* A1             a1 */\

  465.         "paddd %%mm0, %%mm0             \n\t" \

  466.         "psubd %%mm1, %%mm0             \n\t" /* A2             a2 */\

  467.         "pmaddwd 64(%2), %%mm2          \n\t" /* -C7R3+C3R1     -C7r3+C3r1 */\

  468.         "paddd %%mm5, %%mm7             \n\t" /* B0             b0 */\

  469.         "movq 72(%2), %%mm5             \n\t" /* -C5    -C1     -C5     -C1 */\

  470.         "pmaddwd %%mm3, %%mm5           \n\t" /* -C5R7-C1R5     -C5r7-C1r5 */\

  471.         "paddd %%mm4, %%mm7             \n\t" /* A0+B0          a0+b0 */\

  472.         "paddd %%mm4, %%mm4             \n\t" /* 2A0            2a0 */\

  473.         "psubd %%mm7, %%mm4             \n\t" /* A0-B0          a0-b0 */\

  474.         "paddd %%mm2, %%mm5             \n\t" /* B1             b1 */\

  475.         "psrad $" #shift ", %%mm7       \n\t"\

  476.         "psrad $" #shift ", %%mm4       \n\t"\

  477.         "movq %%mm1, %%mm2              \n\t" /* A1             a1 */\

  478.         "paddd %%mm5, %%mm1             \n\t" /* A1+B1          a1+b1 */\

  479.         "psubd %%mm5, %%mm2             \n\t" /* A1-B1          a1-b1 */\

  480.         "psrad $" #shift ", %%mm1       \n\t"\

  481.         "psrad $" #shift ", %%mm2       \n\t"\

  482.         "packssdw %%mm1, %%mm7          \n\t" /* A1+B1  a1+b1   A0+B0   a0+b0 */\

  483.         "packssdw %%mm4, %%mm2          \n\t" /* A0-B0  a0-b0   A1-B1   a1-b1 */\

  484.         "movq %%mm7, " #dst "           \n\t"\

  485.         "movq " #src1 ", %%mm1          \n\t" /* R3     R1      r3      r1 */\

  486.         "movq 80(%2), %%mm4             \n\t" /* -C1    C5      -C1     C5 */\

  487.         "movq %%mm2, 24+" #dst "        \n\t"\

  488.         "pmaddwd %%mm1, %%mm4           \n\t" /* -C1R3+C5R1     -C1r3+C5r1 */\

  489.         "movq 88(%2), %%mm7             \n\t" /* C3     C7      C3      C7 */\

  490.         "pmaddwd 96(%2), %%mm1          \n\t" /* -C5R3+C7R1     -C5r3+C7r1 */\

  491.         "pmaddwd %%mm3, %%mm7           \n\t" /* C3R7+C7R5      C3r7+C7r5 */\

  492.         "movq %%mm0, %%mm2              \n\t" /* A2             a2 */\

  493.         "pmaddwd 104(%2), %%mm3         \n\t" /* -C1R7+C3R5     -C1r7+C3r5 */\

  494.         "paddd %%mm7, %%mm4             \n\t" /* B2             b2 */\

  495.         "paddd %%mm4, %%mm2             \n\t" /* A2+B2          a2+b2 */\

  496.         "psubd %%mm4, %%mm0             \n\t" /* a2-B2          a2-b2 */\

  497.         "psrad $" #shift ", %%mm2       \n\t"\

  498.         "psrad $" #shift ", %%mm0       \n\t"\

  499.         "movq %%mm6, %%mm4              \n\t" /* A3             a3 */\

  500.         "paddd %%mm1, %%mm3             \n\t" /* B3             b3 */\

  501.         "paddd %%mm3, %%mm6             \n\t" /* A3+B3          a3+b3 */\

  502.         "psubd %%mm3, %%mm4             \n\t" /* a3-B3          a3-b3 */\

  503.         "psrad $" #shift ", %%mm6       \n\t"\

  504.         "packssdw %%mm6, %%mm2          \n\t" /* A3+B3  a3+b3   A2+B2   a2+b2 */\

  505.         "movq %%mm2, 8+" #dst "         \n\t"\

  506.         "psrad $" #shift ", %%mm4       \n\t"\

  507.         "packssdw %%mm0, %%mm4          \n\t" /* A2-B2  a2-b2   A3-B3   a3-b3 */\

  508.         "movq %%mm4, 16+" #dst "        \n\t"\

  509. 

  510. #define ROW_IDCT(src0, src4, src1, src5, dst, rounder, shift) \

  511.         "movq " #src0 ", %%mm0          \n\t" /* R4     R0      r4      r0 */\

  512.         "movq " #src4 ", %%mm1          \n\t" /* R6     R2      r6      r2 */\

  513.         "movq " #src1 ", %%mm2          \n\t" /* R3     R1      r3      r1 */\

  514.         "movq " #src5 ", %%mm3          \n\t" /* R7     R5      r7      r5 */\

  515.         "movq 16(%2), %%mm4             \n\t" /* C4     C4      C4      C4 */\

  516.         "pmaddwd %%mm0, %%mm4           \n\t" /* C4R4+C4R0      C4r4+C4r0 */\

  517.         "movq 24(%2), %%mm5             \n\t" /* -C4    C4      -C4     C4 */\

  518.         "pmaddwd %%mm5, %%mm0           \n\t" /* -C4R4+C4R0     -C4r4+C4r0 */\

  519.         "movq 32(%2), %%mm5             \n\t" /* C6     C2      C6      C2 */\

  520.         "pmaddwd %%mm1, %%mm5           \n\t" /* C6R6+C2R2      C6r6+C2r2 */\

  521.         "movq 40(%2), %%mm6             \n\t" /* -C2    C6      -C2     C6 */\

  522.         "pmaddwd %%mm6, %%mm1           \n\t" /* -C2R6+C6R2     -C2r6+C6r2 */\

  523.         "movq 48(%2), %%mm7             \n\t" /* C3     C1      C3      C1 */\

  524.         "pmaddwd %%mm2, %%mm7           \n\t" /* C3R3+C1R1      C3r3+C1r1 */\

  525.         #rounder ", %%mm4               \n\t"\

  526.         "movq %%mm4, %%mm6              \n\t" /* C4R4+C4R0      C4r4+C4r0 */\

  527.         "paddd %%mm5, %%mm4             \n\t" /* A0             a0 */\

  528.         "psubd %%mm5, %%mm6             \n\t" /* A3             a3 */\

  529.         "movq 56(%2), %%mm5             \n\t" /* C7     C5      C7      C5 */\

  530.         "pmaddwd %%mm3, %%mm5           \n\t" /* C7R7+C5R5      C7r7+C5r5 */\

  531.         #rounder ", %%mm0               \n\t"\

  532.         "paddd %%mm0, %%mm1             \n\t" /* A1             a1 */\

  533.         "paddd %%mm0, %%mm0             \n\t" \

  534.         "psubd %%mm1, %%mm0             \n\t" /* A2             a2 */\

  535.         "pmaddwd 64(%2), %%mm2          \n\t" /* -C7R3+C3R1     -C7r3+C3r1 */\

  536.         "paddd %%mm5, %%mm7             \n\t" /* B0             b0 */\

  537.         "movq 72(%2), %%mm5             \n\t" /* -C5    -C1     -C5     -C1 */\

  538.         "pmaddwd %%mm3, %%mm5           \n\t" /* -C5R7-C1R5     -C5r7-C1r5 */\

  539.         "paddd %%mm4, %%mm7             \n\t" /* A0+B0          a0+b0 */\

  540.         "paddd %%mm4, %%mm4             \n\t" /* 2A0            2a0 */\

  541.         "psubd %%mm7, %%mm4             \n\t" /* A0-B0          a0-b0 */\

  542.         "paddd %%mm2, %%mm5             \n\t" /* B1             b1 */\

  543.         "psrad $" #shift ", %%mm7       \n\t"\

  544.         "psrad $" #shift ", %%mm4       \n\t"\

  545.         "movq %%mm1, %%mm2              \n\t" /* A1             a1 */\

  546.         "paddd %%mm5, %%mm1             \n\t" /* A1+B1          a1+b1 */\

  547.         "psubd %%mm5, %%mm2             \n\t" /* A1-B1          a1-b1 */\

  548.         "psrad $" #shift ", %%mm1       \n\t"\

  549.         "psrad $" #shift ", %%mm2       \n\t"\

  550.         "packssdw %%mm1, %%mm7          \n\t" /* A1+B1  a1+b1   A0+B0   a0+b0 */\

  551.         "packssdw %%mm4, %%mm2          \n\t" /* A0-B0  a0-b0   A1-B1   a1-b1 */\

  552.         "movq %%mm7, " #dst "           \n\t"\

  553.         "movq " #src1 ", %%mm1          \n\t" /* R3     R1      r3      r1 */\

  554.         "movq 80(%2), %%mm4             \n\t" /* -C1    C5      -C1     C5 */\

  555.         "movq %%mm2, 24+" #dst "        \n\t"\

  556.         "pmaddwd %%mm1, %%mm4           \n\t" /* -C1R3+C5R1     -C1r3+C5r1 */\

  557.         "movq 88(%2), %%mm7             \n\t" /* C3     C7      C3      C7 */\

  558.         "pmaddwd 96(%2), %%mm1          \n\t" /* -C5R3+C7R1     -C5r3+C7r1 */\

  559.         "pmaddwd %%mm3, %%mm7           \n\t" /* C3R7+C7R5      C3r7+C7r5 */\

  560.         "movq %%mm0, %%mm2              \n\t" /* A2             a2 */\

  561.         "pmaddwd 104(%2), %%mm3         \n\t" /* -C1R7+C3R5     -C1r7+C3r5 */\

  562.         "paddd %%mm7, %%mm4             \n\t" /* B2             b2 */\

  563.         "paddd %%mm4, %%mm2             \n\t" /* A2+B2          a2+b2 */\

  564.         "psubd %%mm4, %%mm0             \n\t" /* a2-B2          a2-b2 */\

  565.         "psrad $" #shift ", %%mm2       \n\t"\

  566.         "psrad $" #shift ", %%mm0       \n\t"\

  567.         "movq %%mm6, %%mm4              \n\t" /* A3             a3 */\

  568.         "paddd %%mm1, %%mm3             \n\t" /* B3             b3 */\

  569.         "paddd %%mm3, %%mm6             \n\t" /* A3+B3          a3+b3 */\

  570.         "psubd %%mm3, %%mm4             \n\t" /* a3-B3          a3-b3 */\

  571.         "psrad $" #shift ", %%mm6       \n\t"\

  572.         "packssdw %%mm6, %%mm2          \n\t" /* A3+B3  a3+b3   A2+B2   a2+b2 */\

  573.         "movq %%mm2, 8+" #dst "         \n\t"\

  574.         "psrad $" #shift ", %%mm4       \n\t"\

  575.         "packssdw %%mm0, %%mm4          \n\t" /* A2-B2  a2-b2   A3-B3   a3-b3 */\

  576.         "movq %%mm4, 16+" #dst "        \n\t"\

  577. 

  578. //IDCT(         src0,   src4,   src1,   src5,    dst,   rounder, shift)

  579. DC_COND_IDCT(  0(%0),  8(%0), 16(%0), 24(%0),  0(%1),paddd 8(%2), 11)

  580. Z_COND_IDCT(  32(%0), 40(%0), 48(%0), 56(%0), 32(%1),paddd (%2), 11, 4f)

  581. Z_COND_IDCT(  64(%0), 72(%0), 80(%0), 88(%0), 64(%1),paddd (%2), 11, 2f)

  582. Z_COND_IDCT(  96(%0),104(%0),112(%0),120(%0), 96(%1),paddd (%2), 11, 1f)

  583. 

  584. #undef IDCT

  585. #define IDCT(src0, src4, src1, src5, dst, shift) \

  586.         "movq " #src0 ", %%mm0          \n\t" /* R4     R0      r4      r0 */\

  587.         "movq " #src4 ", %%mm1          \n\t" /* R6     R2      r6      r2 */\

  588.         "movq " #src1 ", %%mm2          \n\t" /* R3     R1      r3      r1 */\

  589.         "movq " #src5 ", %%mm3          \n\t" /* R7     R5      r7      r5 */\

  590.         "movq 16(%2), %%mm4             \n\t" /* C4     C4      C4      C4 */\

  591.         "pmaddwd %%mm0, %%mm4           \n\t" /* C4R4+C4R0      C4r4+C4r0 */\

  592.         "movq 24(%2), %%mm5             \n\t" /* -C4    C4      -C4     C4 */\

  593.         "pmaddwd %%mm5, %%mm0           \n\t" /* -C4R4+C4R0     -C4r4+C4r0 */\

  594.         "movq 32(%2), %%mm5             \n\t" /* C6     C2      C6      C2 */\

  595.         "pmaddwd %%mm1, %%mm5           \n\t" /* C6R6+C2R2      C6r6+C2r2 */\

  596.         "movq 40(%2), %%mm6             \n\t" /* -C2    C6      -C2     C6 */\

  597.         "pmaddwd %%mm6, %%mm1           \n\t" /* -C2R6+C6R2     -C2r6+C6r2 */\

  598.         "movq %%mm4, %%mm6              \n\t" /* C4R4+C4R0      C4r4+C4r0 */\

  599.         "movq 48(%2), %%mm7             \n\t" /* C3     C1      C3      C1 */\

  600.         "pmaddwd %%mm2, %%mm7           \n\t" /* C3R3+C1R1      C3r3+C1r1 */\

  601.         "paddd %%mm5, %%mm4             \n\t" /* A0             a0 */\

  602.         "psubd %%mm5, %%mm6             \n\t" /* A3             a3 */\

  603.         "movq %%mm0, %%mm5              \n\t" /* -C4R4+C4R0     -C4r4+C4r0 */\

  604.         "paddd %%mm1, %%mm0             \n\t" /* A1             a1 */\

  605.         "psubd %%mm1, %%mm5             \n\t" /* A2             a2 */\

  606.         "movq 56(%2), %%mm1             \n\t" /* C7     C5      C7      C5 */\

  607.         "pmaddwd %%mm3, %%mm1           \n\t" /* C7R7+C5R5      C7r7+C5r5 */\

  608.         "pmaddwd 64(%2), %%mm2          \n\t" /* -C7R3+C3R1     -C7r3+C3r1 */\

  609.         "paddd %%mm1, %%mm7             \n\t" /* B0             b0 */\

  610.         "movq 72(%2), %%mm1             \n\t" /* -C5    -C1     -C5     -C1 */\

  611.         "pmaddwd %%mm3, %%mm1           \n\t" /* -C5R7-C1R5     -C5r7-C1r5 */\

  612.         "paddd %%mm4, %%mm7             \n\t" /* A0+B0          a0+b0 */\

  613.         "paddd %%mm4, %%mm4             \n\t" /* 2A0            2a0 */\

  614.         "psubd %%mm7, %%mm4             \n\t" /* A0-B0          a0-b0 */\

  615.         "paddd %%mm2, %%mm1             \n\t" /* B1             b1 */\

  616.         "psrad $" #shift ", %%mm7       \n\t"\

  617.         "psrad $" #shift ", %%mm4       \n\t"\

  618.         "movq %%mm0, %%mm2              \n\t" /* A1             a1 */\

  619.         "paddd %%mm1, %%mm0             \n\t" /* A1+B1          a1+b1 */\

  620.         "psubd %%mm1, %%mm2             \n\t" /* A1-B1          a1-b1 */\

  621.         "psrad $" #shift ", %%mm0       \n\t"\

  622.         "psrad $" #shift ", %%mm2       \n\t"\

  623.         "packssdw %%mm7, %%mm7          \n\t" /* A0+B0  a0+b0 */\

  624.         "movd %%mm7, " #dst "           \n\t"\

  625.         "packssdw %%mm0, %%mm0          \n\t" /* A1+B1  a1+b1 */\

  626.         "movd %%mm0, 16+" #dst "        \n\t"\

  627.         "packssdw %%mm2, %%mm2          \n\t" /* A1-B1  a1-b1 */\

  628.         "movd %%mm2, 96+" #dst "        \n\t"\

  629.         "packssdw %%mm4, %%mm4          \n\t" /* A0-B0  a0-b0 */\

  630.         "movd %%mm4, 112+" #dst "       \n\t"\

  631.         "movq " #src1 ", %%mm0          \n\t" /* R3     R1      r3      r1 */\

  632.         "movq 80(%2), %%mm4             \n\t" /* -C1    C5      -C1     C5 */\

  633.         "pmaddwd %%mm0, %%mm4           \n\t" /* -C1R3+C5R1     -C1r3+C5r1 */\

  634.         "movq 88(%2), %%mm7             \n\t" /* C3     C7      C3      C7 */\

  635.         "pmaddwd 96(%2), %%mm0          \n\t" /* -C5R3+C7R1     -C5r3+C7r1 */\

  636.         "pmaddwd %%mm3, %%mm7           \n\t" /* C3R7+C7R5      C3r7+C7r5 */\

  637.         "movq %%mm5, %%mm2              \n\t" /* A2             a2 */\

  638.         "pmaddwd 104(%2), %%mm3         \n\t" /* -C1R7+C3R5     -C1r7+C3r5 */\

  639.         "paddd %%mm7, %%mm4             \n\t" /* B2             b2 */\

  640.         "paddd %%mm4, %%mm2             \n\t" /* A2+B2          a2+b2 */\

  641.         "psubd %%mm4, %%mm5             \n\t" /* a2-B2          a2-b2 */\

  642.         "psrad $" #shift ", %%mm2       \n\t"\

  643.         "psrad $" #shift ", %%mm5       \n\t"\

  644.         "movq %%mm6, %%mm4              \n\t" /* A3             a3 */\

  645.         "paddd %%mm0, %%mm3             \n\t" /* B3             b3 */\

  646.         "paddd %%mm3, %%mm6             \n\t" /* A3+B3          a3+b3 */\

  647.         "psubd %%mm3, %%mm4             \n\t" /* a3-B3          a3-b3 */\

  648.         "psrad $" #shift ", %%mm6       \n\t"\

  649.         "psrad $" #shift ", %%mm4       \n\t"\

  650.         "packssdw %%mm2, %%mm2          \n\t" /* A2+B2  a2+b2 */\

  651.         "packssdw %%mm6, %%mm6          \n\t" /* A3+B3  a3+b3 */\

  652.         "movd %%mm2, 32+" #dst "        \n\t"\

  653.         "packssdw %%mm4, %%mm4          \n\t" /* A3-B3  a3-b3 */\

  654.         "packssdw %%mm5, %%mm5          \n\t" /* A2-B2  a2-b2 */\

  655.         "movd %%mm6, 48+" #dst "        \n\t"\

  656.         "movd %%mm4, 64+" #dst "        \n\t"\

  657.         "movd %%mm5, 80+" #dst "        \n\t"

  658. 

  659. 

  660. //IDCT(  src0,   src4,   src1,    src5,    dst, shift)

  661. IDCT(    (%1), 64(%1), 32(%1),  96(%1),  0(%0), 20)

  662. IDCT(   8(%1), 72(%1), 40(%1), 104(%1),  4(%0), 20)

  663. IDCT(  16(%1), 80(%1), 48(%1), 112(%1),  8(%0), 20)

  664. IDCT(  24(%1), 88(%1), 56(%1), 120(%1), 12(%0), 20)

  665.         "jmp 9f                         \n\t"

  666. 

  667.         "# .p2align 4                   \n\t"\

  668.         "4:                             \n\t"

  669. Z_COND_IDCT(  64(%0), 72(%0), 80(%0), 88(%0), 64(%1),paddd (%2), 11, 6f)

  670. Z_COND_IDCT(  96(%0),104(%0),112(%0),120(%0), 96(%1),paddd (%2), 11, 5f)

  671. 

  672. #undef IDCT

  673. #define IDCT(src0, src4, src1, src5, dst, shift) \

  674.         "movq " #src0 ", %%mm0          \n\t" /* R4     R0      r4      r0 */\

  675.         "movq " #src4 ", %%mm1          \n\t" /* R6     R2      r6      r2 */\

  676.         "movq " #src5 ", %%mm3          \n\t" /* R7     R5      r7      r5 */\

  677.         "movq 16(%2), %%mm4             \n\t" /* C4     C4      C4      C4 */\

  678.         "pmaddwd %%mm0, %%mm4           \n\t" /* C4R4+C4R0      C4r4+C4r0 */\

  679.         "movq 24(%2), %%mm5             \n\t" /* -C4    C4      -C4     C4 */\

  680.         "pmaddwd %%mm5, %%mm0           \n\t" /* -C4R4+C4R0     -C4r4+C4r0 */\

  681.         "movq 32(%2), %%mm5             \n\t" /* C6     C2      C6      C2 */\

  682.         "pmaddwd %%mm1, %%mm5           \n\t" /* C6R6+C2R2      C6r6+C2r2 */\

  683.         "movq 40(%2), %%mm6             \n\t" /* -C2    C6      -C2     C6 */\

  684.         "pmaddwd %%mm6, %%mm1           \n\t" /* -C2R6+C6R2     -C2r6+C6r2 */\

  685.         "movq %%mm4, %%mm6              \n\t" /* C4R4+C4R0      C4r4+C4r0 */\

  686.         "paddd %%mm5, %%mm4             \n\t" /* A0             a0 */\

  687.         "psubd %%mm5, %%mm6             \n\t" /* A3             a3 */\

  688.         "movq %%mm0, %%mm5              \n\t" /* -C4R4+C4R0     -C4r4+C4r0 */\

  689.         "paddd %%mm1, %%mm0             \n\t" /* A1             a1 */\

  690.         "psubd %%mm1, %%mm5             \n\t" /* A2             a2 */\

  691.         "movq 56(%2), %%mm1             \n\t" /* C7     C5      C7      C5 */\

  692.         "pmaddwd %%mm3, %%mm1           \n\t" /* C7R7+C5R5      C7r7+C5r5 */\

  693.         "movq 72(%2), %%mm7             \n\t" /* -C5    -C1     -C5     -C1 */\

  694.         "pmaddwd %%mm3, %%mm7           \n\t" /* -C5R7-C1R5     -C5r7-C1r5 */\

  695.         "paddd %%mm4, %%mm1             \n\t" /* A0+B0          a0+b0 */\

  696.         "paddd %%mm4, %%mm4             \n\t" /* 2A0            2a0 */\

  697.         "psubd %%mm1, %%mm4             \n\t" /* A0-B0          a0-b0 */\

  698.         "psrad $" #shift ", %%mm1       \n\t"\

  699.         "psrad $" #shift ", %%mm4       \n\t"\

  700.         "movq %%mm0, %%mm2              \n\t" /* A1             a1 */\

  701.         "paddd %%mm7, %%mm0             \n\t" /* A1+B1          a1+b1 */\

  702.         "psubd %%mm7, %%mm2             \n\t" /* A1-B1          a1-b1 */\

  703.         "psrad $" #shift ", %%mm0       \n\t"\

  704.         "psrad $" #shift ", %%mm2       \n\t"\

  705.         "packssdw %%mm1, %%mm1          \n\t" /* A0+B0  a0+b0 */\

  706.         "movd %%mm1, " #dst "           \n\t"\

  707.         "packssdw %%mm0, %%mm0          \n\t" /* A1+B1  a1+b1 */\

  708.         "movd %%mm0, 16+" #dst "        \n\t"\

  709.         "packssdw %%mm2, %%mm2          \n\t" /* A1-B1  a1-b1 */\

  710.         "movd %%mm2, 96+" #dst "        \n\t"\

  711.         "packssdw %%mm4, %%mm4          \n\t" /* A0-B0  a0-b0 */\

  712.         "movd %%mm4, 112+" #dst "       \n\t"\

  713.         "movq 88(%2), %%mm1             \n\t" /* C3     C7      C3      C7 */\

  714.         "pmaddwd %%mm3, %%mm1           \n\t" /* C3R7+C7R5      C3r7+C7r5 */\

  715.         "movq %%mm5, %%mm2              \n\t" /* A2             a2 */\

  716.         "pmaddwd 104(%2), %%mm3         \n\t" /* -C1R7+C3R5     -C1r7+C3r5 */\

  717.         "paddd %%mm1, %%mm2             \n\t" /* A2+B2          a2+b2 */\

  718.         "psubd %%mm1, %%mm5             \n\t" /* a2-B2          a2-b2 */\

  719.         "psrad $" #shift ", %%mm2       \n\t"\

  720.         "psrad $" #shift ", %%mm5       \n\t"\

  721.         "movq %%mm6, %%mm1              \n\t" /* A3             a3 */\

  722.         "paddd %%mm3, %%mm6             \n\t" /* A3+B3          a3+b3 */\

  723.         "psubd %%mm3, %%mm1             \n\t" /* a3-B3          a3-b3 */\

  724.         "psrad $" #shift ", %%mm6       \n\t"\

  725.         "psrad $" #shift ", %%mm1       \n\t"\

  726.         "packssdw %%mm2, %%mm2          \n\t" /* A2+B2  a2+b2 */\

  727.         "packssdw %%mm6, %%mm6          \n\t" /* A3+B3  a3+b3 */\

  728.         "movd %%mm2, 32+" #dst "        \n\t"\

  729.         "packssdw %%mm1, %%mm1          \n\t" /* A3-B3  a3-b3 */\

  730.         "packssdw %%mm5, %%mm5          \n\t" /* A2-B2  a2-b2 */\

  731.         "movd %%mm6, 48+" #dst "        \n\t"\

  732.         "movd %%mm1, 64+" #dst "        \n\t"\

  733.         "movd %%mm5, 80+" #dst "        \n\t"

  734. 

  735. //IDCT(  src0,   src4,   src1,    src5,    dst, shift)

  736. IDCT(    (%1), 64(%1), 32(%1),  96(%1),  0(%0), 20)

  737. IDCT(   8(%1), 72(%1), 40(%1), 104(%1),  4(%0), 20)

  738. IDCT(  16(%1), 80(%1), 48(%1), 112(%1),  8(%0), 20)

  739. IDCT(  24(%1), 88(%1), 56(%1), 120(%1), 12(%0), 20)

  740.         "jmp 9f                         \n\t"

  741. 

  742.         "# .p2align 4                   \n\t"\

  743.         "6:                             \n\t"

  744. Z_COND_IDCT(  96(%0),104(%0),112(%0),120(%0), 96(%1),paddd (%2), 11, 7f)

  745. 

  746. #undef IDCT

  747. #define IDCT(src0, src4, src1, src5, dst, shift) \

  748.         "movq " #src0 ", %%mm0          \n\t" /* R4     R0      r4      r0 */\

  749.         "movq " #src5 ", %%mm3          \n\t" /* R7     R5      r7      r5 */\

  750.         "movq 16(%2), %%mm4             \n\t" /* C4     C4      C4      C4 */\

  751.         "pmaddwd %%mm0, %%mm4           \n\t" /* C4R4+C4R0      C4r4+C4r0 */\

  752.         "movq 24(%2), %%mm5             \n\t" /* -C4    C4      -C4     C4 */\

  753.         "pmaddwd %%mm5, %%mm0           \n\t" /* -C4R4+C4R0     -C4r4+C4r0 */\

  754.         "movq %%mm4, %%mm6              \n\t" /* C4R4+C4R0      C4r4+C4r0 */\

  755.         "movq %%mm0, %%mm5              \n\t" /* -C4R4+C4R0     -C4r4+C4r0 */\

  756.         "movq 56(%2), %%mm1             \n\t" /* C7     C5      C7      C5 */\

  757.         "pmaddwd %%mm3, %%mm1           \n\t" /* C7R7+C5R5      C7r7+C5r5 */\

  758.         "movq 72(%2), %%mm7             \n\t" /* -C5    -C1     -C5     -C1 */\

  759.         "pmaddwd %%mm3, %%mm7           \n\t" /* -C5R7-C1R5     -C5r7-C1r5 */\

  760.         "paddd %%mm4, %%mm1             \n\t" /* A0+B0          a0+b0 */\

  761.         "paddd %%mm4, %%mm4             \n\t" /* 2A0            2a0 */\

  762.         "psubd %%mm1, %%mm4             \n\t" /* A0-B0          a0-b0 */\

  763.         "psrad $" #shift ", %%mm1       \n\t"\

  764.         "psrad $" #shift ", %%mm4       \n\t"\

  765.         "movq %%mm0, %%mm2              \n\t" /* A1             a1 */\

  766.         "paddd %%mm7, %%mm0             \n\t" /* A1+B1          a1+b1 */\

  767.         "psubd %%mm7, %%mm2             \n\t" /* A1-B1          a1-b1 */\

  768.         "psrad $" #shift ", %%mm0       \n\t"\

  769.         "psrad $" #shift ", %%mm2       \n\t"\

  770.         "packssdw %%mm1, %%mm1          \n\t" /* A0+B0  a0+b0 */\

  771.         "movd %%mm1, " #dst "           \n\t"\

  772.         "packssdw %%mm0, %%mm0          \n\t" /* A1+B1  a1+b1 */\

  773.         "movd %%mm0, 16+" #dst "        \n\t"\

  774.         "packssdw %%mm2, %%mm2          \n\t" /* A1-B1  a1-b1 */\

  775.         "movd %%mm2, 96+" #dst "        \n\t"\

  776.         "packssdw %%mm4, %%mm4          \n\t" /* A0-B0  a0-b0 */\

  777.         "movd %%mm4, 112+" #dst "       \n\t"\

  778.         "movq 88(%2), %%mm1             \n\t" /* C3     C7      C3      C7 */\

  779.         "pmaddwd %%mm3, %%mm1           \n\t" /* C3R7+C7R5      C3r7+C7r5 */\

  780.         "movq %%mm5, %%mm2              \n\t" /* A2             a2 */\

  781.         "pmaddwd 104(%2), %%mm3         \n\t" /* -C1R7+C3R5     -C1r7+C3r5 */\

  782.         "paddd %%mm1, %%mm2             \n\t" /* A2+B2          a2+b2 */\

  783.         "psubd %%mm1, %%mm5             \n\t" /* a2-B2          a2-b2 */\

  784.         "psrad $" #shift ", %%mm2       \n\t"\

  785.         "psrad $" #shift ", %%mm5       \n\t"\

  786.         "movq %%mm6, %%mm1              \n\t" /* A3             a3 */\

  787.         "paddd %%mm3, %%mm6             \n\t" /* A3+B3          a3+b3 */\

  788.         "psubd %%mm3, %%mm1             \n\t" /* a3-B3          a3-b3 */\

  789.         "psrad $" #shift ", %%mm6       \n\t"\

  790.         "psrad $" #shift ", %%mm1       \n\t"\

  791.         "packssdw %%mm2, %%mm2          \n\t" /* A2+B2  a2+b2 */\

  792.         "packssdw %%mm6, %%mm6          \n\t" /* A3+B3  a3+b3 */\

  793.         "movd %%mm2, 32+" #dst "        \n\t"\

  794.         "packssdw %%mm1, %%mm1          \n\t" /* A3-B3  a3-b3 */\

  795.         "packssdw %%mm5, %%mm5          \n\t" /* A2-B2  a2-b2 */\

  796.         "movd %%mm6, 48+" #dst "        \n\t"\

  797.         "movd %%mm1, 64+" #dst "        \n\t"\

  798.         "movd %%mm5, 80+" #dst "        \n\t"

  799. 

  800. 

  801. //IDCT(  src0,   src4,   src1,    src5,    dst, shift)

  802. IDCT(    (%1), 64(%1), 32(%1),  96(%1),  0(%0), 20)

  803. IDCT(   8(%1), 72(%1), 40(%1), 104(%1),  4(%0), 20)

  804. IDCT(  16(%1), 80(%1), 48(%1), 112(%1),  8(%0), 20)

  805. IDCT(  24(%1), 88(%1), 56(%1), 120(%1), 12(%0), 20)

  806.         "jmp 9f                         \n\t"

  807. 

  808.         "# .p2align 4                   \n\t"\

  809.         "2:                             \n\t"

  810. Z_COND_IDCT(  96(%0),104(%0),112(%0),120(%0), 96(%1),paddd (%2), 11, 3f)

  811. 

  812. #undef IDCT

  813. #define IDCT(src0, src4, src1, src5, dst, shift) \

  814.         "movq " #src0 ", %%mm0          \n\t" /* R4     R0      r4      r0 */\

  815.         "movq " #src1 ", %%mm2          \n\t" /* R3     R1      r3      r1 */\

  816.         "movq " #src5 ", %%mm3          \n\t" /* R7     R5      r7      r5 */\

  817.         "movq 16(%2), %%mm4             \n\t" /* C4     C4      C4      C4 */\

  818.         "pmaddwd %%mm0, %%mm4           \n\t" /* C4R4+C4R0      C4r4+C4r0 */\

  819.         "movq 24(%2), %%mm5             \n\t" /* -C4    C4      -C4     C4 */\

  820.         "pmaddwd %%mm5, %%mm0           \n\t" /* -C4R4+C4R0     -C4r4+C4r0 */\

  821.         "movq %%mm4, %%mm6              \n\t" /* C4R4+C4R0      C4r4+C4r0 */\

  822.         "movq 48(%2), %%mm7             \n\t" /* C3     C1      C3      C1 */\

  823.         "pmaddwd %%mm2, %%mm7           \n\t" /* C3R3+C1R1      C3r3+C1r1 */\

  824.         "movq %%mm0, %%mm5              \n\t" /* -C4R4+C4R0     -C4r4+C4r0 */\

  825.         "movq 56(%2), %%mm1             \n\t" /* C7     C5      C7      C5 */\

  826.         "pmaddwd %%mm3, %%mm1           \n\t" /* C7R7+C5R5      C7r7+C5r5 */\

  827.         "pmaddwd 64(%2), %%mm2          \n\t" /* -C7R3+C3R1     -C7r3+C3r1 */\

  828.         "paddd %%mm1, %%mm7             \n\t" /* B0             b0 */\

  829.         "movq 72(%2), %%mm1             \n\t" /* -C5    -C1     -C5     -C1 */\

  830.         "pmaddwd %%mm3, %%mm1           \n\t" /* -C5R7-C1R5     -C5r7-C1r5 */\

  831.         "paddd %%mm4, %%mm7             \n\t" /* A0+B0          a0+b0 */\

  832.         "paddd %%mm4, %%mm4             \n\t" /* 2A0            2a0 */\

  833.         "psubd %%mm7, %%mm4             \n\t" /* A0-B0          a0-b0 */\

  834.         "paddd %%mm2, %%mm1             \n\t" /* B1             b1 */\

  835.         "psrad $" #shift ", %%mm7       \n\t"\

  836.         "psrad $" #shift ", %%mm4       \n\t"\

  837.         "movq %%mm0, %%mm2              \n\t" /* A1             a1 */\

  838.         "paddd %%mm1, %%mm0             \n\t" /* A1+B1          a1+b1 */\

  839.         "psubd %%mm1, %%mm2             \n\t" /* A1-B1          a1-b1 */\

  840.         "psrad $" #shift ", %%mm0       \n\t"\

  841.         "psrad $" #shift ", %%mm2       \n\t"\

  842.         "packssdw %%mm7, %%mm7          \n\t" /* A0+B0  a0+b0 */\

  843.         "movd %%mm7, " #dst "           \n\t"\

  844.         "packssdw %%mm0, %%mm0          \n\t" /* A1+B1  a1+b1 */\

  845.         "movd %%mm0, 16+" #dst "        \n\t"\

  846.         "packssdw %%mm2, %%mm2          \n\t" /* A1-B1  a1-b1 */\

  847.         "movd %%mm2, 96+" #dst "        \n\t"\

  848.         "packssdw %%mm4, %%mm4          \n\t" /* A0-B0  a0-b0 */\

  849.         "movd %%mm4, 112+" #dst "       \n\t"\

  850.         "movq " #src1 ", %%mm0          \n\t" /* R3     R1      r3      r1 */\

  851.         "movq 80(%2), %%mm4             \n\t" /* -C1    C5      -C1     C5 */\

  852.         "pmaddwd %%mm0, %%mm4           \n\t" /* -C1R3+C5R1     -C1r3+C5r1 */\

  853.         "movq 88(%2), %%mm7             \n\t" /* C3     C7      C3      C7 */\

  854.         "pmaddwd 96(%2), %%mm0          \n\t" /* -C5R3+C7R1     -C5r3+C7r1 */\

  855.         "pmaddwd %%mm3, %%mm7           \n\t" /* C3R7+C7R5      C3r7+C7r5 */\

  856.         "movq %%mm5, %%mm2              \n\t" /* A2             a2 */\

  857.         "pmaddwd 104(%2), %%mm3         \n\t" /* -C1R7+C3R5     -C1r7+C3r5 */\

  858.         "paddd %%mm7, %%mm4             \n\t" /* B2             b2 */\

  859.         "paddd %%mm4, %%mm2             \n\t" /* A2+B2          a2+b2 */\

  860.         "psubd %%mm4, %%mm5             \n\t" /* a2-B2          a2-b2 */\

  861.         "psrad $" #shift ", %%mm2       \n\t"\

  862.         "psrad $" #shift ", %%mm5       \n\t"\

  863.         "movq %%mm6, %%mm4              \n\t" /* A3             a3 */\

  864.         "paddd %%mm0, %%mm3             \n\t" /* B3             b3 */\

  865.         "paddd %%mm3, %%mm6             \n\t" /* A3+B3          a3+b3 */\

  866.         "psubd %%mm3, %%mm4             \n\t" /* a3-B3          a3-b3 */\

  867.         "psrad $" #shift ", %%mm6       \n\t"\

  868.         "psrad $" #shift ", %%mm4       \n\t"\

  869.         "packssdw %%mm2, %%mm2          \n\t" /* A2+B2  a2+b2 */\

  870.         "packssdw %%mm6, %%mm6          \n\t" /* A3+B3  a3+b3 */\

  871.         "movd %%mm2, 32+" #dst "        \n\t"\

  872.         "packssdw %%mm4, %%mm4          \n\t" /* A3-B3  a3-b3 */\

  873.         "packssdw %%mm5, %%mm5          \n\t" /* A2-B2  a2-b2 */\

  874.         "movd %%mm6, 48+" #dst "        \n\t"\

  875.         "movd %%mm4, 64+" #dst "        \n\t"\

  876.         "movd %%mm5, 80+" #dst "        \n\t"

  877. 

  878. //IDCT(  src0,   src4,   src1,    src5,    dst, shift)

  879. IDCT(    (%1), 64(%1), 32(%1),  96(%1),  0(%0), 20)

  880. IDCT(   8(%1), 72(%1), 40(%1), 104(%1),  4(%0), 20)

  881. IDCT(  16(%1), 80(%1), 48(%1), 112(%1),  8(%0), 20)

  882. IDCT(  24(%1), 88(%1), 56(%1), 120(%1), 12(%0), 20)

  883.         "jmp 9f                         \n\t"

  884. 

  885.         "# .p2align 4                   \n\t"\

  886.         "3:                             \n\t"

  887. #undef IDCT

  888. #define IDCT(src0, src4, src1, src5, dst, shift) \

  889.         "movq " #src0 ", %%mm0          \n\t" /* R4     R0      r4      r0 */\

  890.         "movq " #src1 ", %%mm2          \n\t" /* R3     R1      r3      r1 */\

  891.         "movq 16(%2), %%mm4             \n\t" /* C4     C4      C4      C4 */\

  892.         "pmaddwd %%mm0, %%mm4           \n\t" /* C4R4+C4R0      C4r4+C4r0 */\

  893.         "movq 24(%2), %%mm5             \n\t" /* -C4    C4      -C4     C4 */\

  894.         "pmaddwd %%mm5, %%mm0           \n\t" /* -C4R4+C4R0     -C4r4+C4r0 */\

  895.         "movq %%mm4, %%mm6              \n\t" /* C4R4+C4R0      C4r4+C4r0 */\

  896.         "movq 48(%2), %%mm7             \n\t" /* C3     C1      C3      C1 */\

  897.         "pmaddwd %%mm2, %%mm7           \n\t" /* C3R3+C1R1      C3r3+C1r1 */\

  898.         "movq %%mm0, %%mm5              \n\t" /* -C4R4+C4R0     -C4r4+C4r0 */\

  899.         "movq 64(%2), %%mm3             \n\t"\

  900.         "pmaddwd %%mm2, %%mm3           \n\t" /* -C7R3+C3R1     -C7r3+C3r1 */\

  901.         "paddd %%mm4, %%mm7             \n\t" /* A0+B0          a0+b0 */\

  902.         "paddd %%mm4, %%mm4             \n\t" /* 2A0            2a0 */\

  903.         "psubd %%mm7, %%mm4             \n\t" /* A0-B0          a0-b0 */\

  904.         "psrad $" #shift ", %%mm7       \n\t"\

  905.         "psrad $" #shift ", %%mm4       \n\t"\

  906.         "movq %%mm0, %%mm1              \n\t" /* A1             a1 */\

  907.         "paddd %%mm3, %%mm0             \n\t" /* A1+B1          a1+b1 */\

  908.         "psubd %%mm3, %%mm1             \n\t" /* A1-B1          a1-b1 */\

  909.         "psrad $" #shift ", %%mm0       \n\t"\

  910.         "psrad $" #shift ", %%mm1       \n\t"\

  911.         "packssdw %%mm7, %%mm7          \n\t" /* A0+B0  a0+b0 */\

  912.         "movd %%mm7, " #dst "           \n\t"\

  913.         "packssdw %%mm0, %%mm0          \n\t" /* A1+B1  a1+b1 */\

  914.         "movd %%mm0, 16+" #dst "        \n\t"\

  915.         "packssdw %%mm1, %%mm1          \n\t" /* A1-B1  a1-b1 */\

  916.         "movd %%mm1, 96+" #dst "        \n\t"\

  917.         "packssdw %%mm4, %%mm4          \n\t" /* A0-B0  a0-b0 */\

  918.         "movd %%mm4, 112+" #dst "       \n\t"\

  919.         "movq 80(%2), %%mm4             \n\t" /* -C1    C5      -C1     C5 */\

  920.         "pmaddwd %%mm2, %%mm4           \n\t" /* -C1R3+C5R1     -C1r3+C5r1 */\

  921.         "pmaddwd 96(%2), %%mm2          \n\t" /* -C5R3+C7R1     -C5r3+C7r1 */\

  922.         "movq %%mm5, %%mm1              \n\t" /* A2             a2 */\

  923.         "paddd %%mm4, %%mm1             \n\t" /* A2+B2          a2+b2 */\

  924.         "psubd %%mm4, %%mm5             \n\t" /* a2-B2          a2-b2 */\

  925.         "psrad $" #shift ", %%mm1       \n\t"\

  926.         "psrad $" #shift ", %%mm5       \n\t"\

  927.         "movq %%mm6, %%mm4              \n\t" /* A3             a3 */\

  928.         "paddd %%mm2, %%mm6             \n\t" /* A3+B3          a3+b3 */\

  929.         "psubd %%mm2, %%mm4             \n\t" /* a3-B3          a3-b3 */\

  930.         "psrad $" #shift ", %%mm6       \n\t"\

  931.         "psrad $" #shift ", %%mm4       \n\t"\

  932.         "packssdw %%mm1, %%mm1          \n\t" /* A2+B2  a2+b2 */\

  933.         "packssdw %%mm6, %%mm6          \n\t" /* A3+B3  a3+b3 */\

  934.         "movd %%mm1, 32+" #dst "        \n\t"\

  935.         "packssdw %%mm4, %%mm4          \n\t" /* A3-B3  a3-b3 */\

  936.         "packssdw %%mm5, %%mm5          \n\t" /* A2-B2  a2-b2 */\

  937.         "movd %%mm6, 48+" #dst "        \n\t"\

  938.         "movd %%mm4, 64+" #dst "        \n\t"\

  939.         "movd %%mm5, 80+" #dst "        \n\t"

  940. 

  941. 

  942. //IDCT(  src0,   src4,   src1,    src5,    dst, shift)

  943. IDCT(    (%1), 64(%1), 32(%1),  96(%1),  0(%0), 20)

  944. IDCT(   8(%1), 72(%1), 40(%1), 104(%1),  4(%0), 20)

  945. IDCT(  16(%1), 80(%1), 48(%1), 112(%1),  8(%0), 20)

  946. IDCT(  24(%1), 88(%1), 56(%1), 120(%1), 12(%0), 20)

  947.         "jmp 9f                         \n\t"

  948. 

  949.         "# .p2align 4                   \n\t"\

  950.         "5:                             \n\t"

  951. #undef IDCT

  952. #define IDCT(src0, src4, src1, src5, dst, shift) \

  953.         "movq " #src0 ", %%mm0          \n\t" /* R4     R0      r4      r0 */\

  954.         "movq " #src4 ", %%mm1          \n\t" /* R6     R2      r6      r2 */\

  955.         "movq 16(%2), %%mm4             \n\t" /* C4     C4      C4      C4 */\

  956.         "pmaddwd %%mm0, %%mm4           \n\t" /* C4R4+C4R0      C4r4+C4r0 */\

  957.         "movq 24(%2), %%mm5             \n\t" /* -C4    C4      -C4     C4 */\

  958.         "pmaddwd %%mm5, %%mm0           \n\t" /* -C4R4+C4R0     -C4r4+C4r0 */\

  959.         "movq 32(%2), %%mm5             \n\t" /* C6     C2      C6      C2 */\

  960.         "pmaddwd %%mm1, %%mm5           \n\t" /* C6R6+C2R2      C6r6+C2r2 */\

  961.         "movq 40(%2), %%mm6             \n\t" /* -C2    C6      -C2     C6 */\

  962.         "pmaddwd %%mm6, %%mm1           \n\t" /* -C2R6+C6R2     -C2r6+C6r2 */\

  963.         "movq %%mm4, %%mm6              \n\t" /* C4R4+C4R0      C4r4+C4r0 */\

  964.         "paddd %%mm5, %%mm4             \n\t" /* A0             a0 */\

  965.         "psubd %%mm5, %%mm6             \n\t" /* A3             a3 */\

  966.         "movq %%mm0, %%mm5              \n\t" /* -C4R4+C4R0     -C4r4+C4r0 */\

  967.         "paddd %%mm1, %%mm0             \n\t" /* A1             a1 */\

  968.         "psubd %%mm1, %%mm5             \n\t" /* A2             a2 */\

  969.         "movq 8+" #src0 ", %%mm2        \n\t" /* R4     R0      r4      r0 */\

  970.         "movq 8+" #src4 ", %%mm3        \n\t" /* R6     R2      r6      r2 */\

  971.         "movq 16(%2), %%mm1             \n\t" /* C4     C4      C4      C4 */\

  972.         "pmaddwd %%mm2, %%mm1           \n\t" /* C4R4+C4R0      C4r4+C4r0 */\

  973.         "movq 24(%2), %%mm7             \n\t" /* -C4    C4      -C4     C4 */\

  974.         "pmaddwd %%mm7, %%mm2           \n\t" /* -C4R4+C4R0     -C4r4+C4r0 */\

  975.         "movq 32(%2), %%mm7             \n\t" /* C6     C2      C6      C2 */\

  976.         "pmaddwd %%mm3, %%mm7           \n\t" /* C6R6+C2R2      C6r6+C2r2 */\

  977.         "pmaddwd 40(%2), %%mm3          \n\t" /* -C2R6+C6R2     -C2r6+C6r2 */\

  978.         "paddd %%mm1, %%mm7             \n\t" /* A0             a0 */\

  979.         "paddd %%mm1, %%mm1             \n\t" /* 2C0            2c0 */\

  980.         "psubd %%mm7, %%mm1             \n\t" /* A3             a3 */\

  981.         "paddd %%mm2, %%mm3             \n\t" /* A1             a1 */\

  982.         "paddd %%mm2, %%mm2             \n\t" /* 2C1            2c1 */\

  983.         "psubd %%mm3, %%mm2             \n\t" /* A2             a2 */\

  984.         "psrad $" #shift ", %%mm4       \n\t"\

  985.         "psrad $" #shift ", %%mm7       \n\t"\

  986.         "psrad $" #shift ", %%mm3       \n\t"\

  987.         "packssdw %%mm7, %%mm4          \n\t" /* A0     a0 */\

  988.         "movq %%mm4, " #dst "           \n\t"\

  989.         "psrad $" #shift ", %%mm0       \n\t"\

  990.         "packssdw %%mm3, %%mm0          \n\t" /* A1     a1 */\

  991.         "movq %%mm0, 16+" #dst "        \n\t"\

  992.         "movq %%mm0, 96+" #dst "        \n\t"\

  993.         "movq %%mm4, 112+" #dst "       \n\t"\

  994.         "psrad $" #shift ", %%mm5       \n\t"\

  995.         "psrad $" #shift ", %%mm6       \n\t"\

  996.         "psrad $" #shift ", %%mm2       \n\t"\

  997.         "packssdw %%mm2, %%mm5          \n\t" /* A2-B2  a2-b2 */\

  998.         "movq %%mm5, 32+" #dst "        \n\t"\

  999.         "psrad $" #shift ", %%mm1       \n\t"\

  1000.         "packssdw %%mm1, %%mm6          \n\t" /* A3+B3  a3+b3 */\

  1001.         "movq %%mm6, 48+" #dst "        \n\t"\

  1002.         "movq %%mm6, 64+" #dst "        \n\t"\

  1003.         "movq %%mm5, 80+" #dst "        \n\t"

  1004. 

  1005. 

  1006. //IDCT(  src0,   src4,   src1,    src5,    dst, shift)

  1007. IDCT(    0(%1), 64(%1), 32(%1),  96(%1),  0(%0), 20)

  1008. //IDCT(   8(%1), 72(%1), 40(%1), 104(%1),  4(%0), 20)

  1009. IDCT(  16(%1), 80(%1), 48(%1), 112(%1),  8(%0), 20)

  1010. //IDCT(  24(%1), 88(%1), 56(%1), 120(%1), 12(%0), 20)

  1011.         "jmp 9f                         \n\t"

  1012. 

  1013. 

  1014.         "# .p2align 4                   \n\t"\

  1015.         "1:                             \n\t"

  1016. #undef IDCT

  1017. #define IDCT(src0, src4, src1, src5, dst, shift) \

  1018.         "movq " #src0 ", %%mm0          \n\t" /* R4     R0      r4      r0 */\

  1019.         "movq " #src4 ", %%mm1          \n\t" /* R6     R2      r6      r2 */\

  1020.         "movq " #src1 ", %%mm2          \n\t" /* R3     R1      r3      r1 */\

  1021.         "movq 16(%2), %%mm4             \n\t" /* C4     C4      C4      C4 */\

  1022.         "pmaddwd %%mm0, %%mm4           \n\t" /* C4R4+C4R0      C4r4+C4r0 */\

  1023.         "movq 24(%2), %%mm5             \n\t" /* -C4    C4      -C4     C4 */\

  1024.         "pmaddwd %%mm5, %%mm0           \n\t" /* -C4R4+C4R0     -C4r4+C4r0 */\

  1025.         "movq 32(%2), %%mm5             \n\t" /* C6     C2      C6      C2 */\

  1026.         "pmaddwd %%mm1, %%mm5           \n\t" /* C6R6+C2R2      C6r6+C2r2 */\

  1027.         "movq 40(%2), %%mm6             \n\t" /* -C2    C6      -C2     C6 */\

  1028.         "pmaddwd %%mm6, %%mm1           \n\t" /* -C2R6+C6R2     -C2r6+C6r2 */\

  1029.         "movq %%mm4, %%mm6              \n\t" /* C4R4+C4R0      C4r4+C4r0 */\

  1030.         "movq 48(%2), %%mm7             \n\t" /* C3     C1      C3      C1 */\

  1031.         "pmaddwd %%mm2, %%mm7           \n\t" /* C3R3+C1R1      C3r3+C1r1 */\

  1032.         "paddd %%mm5, %%mm4             \n\t" /* A0             a0 */\

  1033.         "psubd %%mm5, %%mm6             \n\t" /* A3             a3 */\

  1034.         "movq %%mm0, %%mm5              \n\t" /* -C4R4+C4R0     -C4r4+C4r0 */\

  1035.         "paddd %%mm1, %%mm0             \n\t" /* A1             a1 */\

  1036.         "psubd %%mm1, %%mm5             \n\t" /* A2             a2 */\

  1037.         "movq 64(%2), %%mm1             \n\t"\

  1038.         "pmaddwd %%mm2, %%mm1           \n\t" /* -C7R3+C3R1     -C7r3+C3r1 */\

  1039.         "paddd %%mm4, %%mm7             \n\t" /* A0+B0          a0+b0 */\

  1040.         "paddd %%mm4, %%mm4             \n\t" /* 2A0            2a0 */\

  1041.         "psubd %%mm7, %%mm4             \n\t" /* A0-B0          a0-b0 */\

  1042.         "psrad $" #shift ", %%mm7       \n\t"\

  1043.         "psrad $" #shift ", %%mm4       \n\t"\

  1044.         "movq %%mm0, %%mm3              \n\t" /* A1             a1 */\

  1045.         "paddd %%mm1, %%mm0             \n\t" /* A1+B1          a1+b1 */\

  1046.         "psubd %%mm1, %%mm3             \n\t" /* A1-B1          a1-b1 */\

  1047.         "psrad $" #shift ", %%mm0       \n\t"\

  1048.         "psrad $" #shift ", %%mm3       \n\t"\

  1049.         "packssdw %%mm7, %%mm7          \n\t" /* A0+B0  a0+b0 */\

  1050.         "movd %%mm7, " #dst "           \n\t"\

  1051.         "packssdw %%mm0, %%mm0          \n\t" /* A1+B1  a1+b1 */\

  1052.         "movd %%mm0, 16+" #dst "        \n\t"\

  1053.         "packssdw %%mm3, %%mm3          \n\t" /* A1-B1  a1-b1 */\

  1054.         "movd %%mm3, 96+" #dst "        \n\t"\

  1055.         "packssdw %%mm4, %%mm4          \n\t" /* A0-B0  a0-b0 */\

  1056.         "movd %%mm4, 112+" #dst "       \n\t"\

  1057.         "movq 80(%2), %%mm4             \n\t" /* -C1    C5      -C1     C5 */\

  1058.         "pmaddwd %%mm2, %%mm4           \n\t" /* -C1R3+C5R1     -C1r3+C5r1 */\

  1059.         "pmaddwd 96(%2), %%mm2          \n\t" /* -C5R3+C7R1     -C5r3+C7r1 */\

  1060.         "movq %%mm5, %%mm3              \n\t" /* A2             a2 */\

  1061.         "paddd %%mm4, %%mm3             \n\t" /* A2+B2          a2+b2 */\

  1062.         "psubd %%mm4, %%mm5             \n\t" /* a2-B2          a2-b2 */\

  1063.         "psrad $" #shift ", %%mm3       \n\t"\

  1064.         "psrad $" #shift ", %%mm5       \n\t"\

  1065.         "movq %%mm6, %%mm4              \n\t" /* A3             a3 */\

  1066.         "paddd %%mm2, %%mm6             \n\t" /* A3+B3          a3+b3 */\

  1067.         "psubd %%mm2, %%mm4             \n\t" /* a3-B3          a3-b3 */\

  1068.         "psrad $" #shift ", %%mm6       \n\t"\

  1069.         "packssdw %%mm3, %%mm3          \n\t" /* A2+B2  a2+b2 */\

  1070.         "movd %%mm3, 32+" #dst "        \n\t"\

  1071.         "psrad $" #shift ", %%mm4       \n\t"\

  1072.         "packssdw %%mm6, %%mm6          \n\t" /* A3+B3  a3+b3 */\

  1073.         "movd %%mm6, 48+" #dst "        \n\t"\

  1074.         "packssdw %%mm4, %%mm4          \n\t" /* A3-B3  a3-b3 */\

  1075.         "packssdw %%mm5, %%mm5          \n\t" /* A2-B2  a2-b2 */\

  1076.         "movd %%mm4, 64+" #dst "        \n\t"\

  1077.         "movd %%mm5, 80+" #dst "        \n\t"

  1078. 

  1079. 

  1080. //IDCT(  src0,   src4,   src1,    src5,    dst, shift)

  1081. IDCT(    (%1), 64(%1), 32(%1),  96(%1),  0(%0), 20)

  1082. IDCT(   8(%1), 72(%1), 40(%1), 104(%1),  4(%0), 20)

  1083. IDCT(  16(%1), 80(%1), 48(%1), 112(%1),  8(%0), 20)

  1084. IDCT(  24(%1), 88(%1), 56(%1), 120(%1), 12(%0), 20)

  1085.         "jmp 9f                         \n\t"

  1086. 

  1087. 

  1088.         "# .p2align 4                   \n\t"

  1089.         "7:                             \n\t"

  1090. #undef IDCT

  1091. #define IDCT(src0, src4, src1, src5, dst, shift) \

  1092.         "movq " #src0 ", %%mm0          \n\t" /* R4     R0      r4      r0 */\

  1093.         "movq 16(%2), %%mm4             \n\t" /* C4     C4      C4      C4 */\

  1094.         "pmaddwd %%mm0, %%mm4           \n\t" /* C4R4+C4R0      C4r4+C4r0 */\

  1095.         "movq 24(%2), %%mm5             \n\t" /* -C4    C4      -C4     C4 */\

  1096.         "pmaddwd %%mm5, %%mm0           \n\t" /* -C4R4+C4R0     -C4r4+C4r0 */\

  1097.         "psrad $" #shift ", %%mm4       \n\t"\

  1098.         "psrad $" #shift ", %%mm0       \n\t"\

  1099.         "movq 8+" #src0 ", %%mm2        \n\t" /* R4     R0      r4      r0 */\

  1100.         "movq 16(%2), %%mm1             \n\t" /* C4     C4      C4      C4 */\

  1101.         "pmaddwd %%mm2, %%mm1           \n\t" /* C4R4+C4R0      C4r4+C4r0 */\

  1102.         "movq 24(%2), %%mm7             \n\t" /* -C4    C4      -C4     C4 */\

  1103.         "pmaddwd %%mm7, %%mm2           \n\t" /* -C4R4+C4R0     -C4r4+C4r0 */\

  1104.         "movq 32(%2), %%mm7             \n\t" /* C6     C2      C6      C2 */\

  1105.         "psrad $" #shift ", %%mm1       \n\t"\

  1106.         "packssdw %%mm1, %%mm4          \n\t" /* A0     a0 */\

  1107.         "movq %%mm4, " #dst "           \n\t"\

  1108.         "psrad $" #shift ", %%mm2       \n\t"\

  1109.         "packssdw %%mm2, %%mm0          \n\t" /* A1     a1 */\

  1110.         "movq %%mm0, 16+" #dst "        \n\t"\

  1111.         "movq %%mm0, 96+" #dst "        \n\t"\

  1112.         "movq %%mm4, 112+" #dst "       \n\t"\

  1113.         "movq %%mm0, 32+" #dst "        \n\t"\

  1114.         "movq %%mm4, 48+" #dst "        \n\t"\

  1115.         "movq %%mm4, 64+" #dst "        \n\t"\

  1116.         "movq %%mm0, 80+" #dst "        \n\t"

  1117. 

  1118. //IDCT(  src0,   src4,   src1,    src5,    dst, shift)

  1119. IDCT(   0(%1), 64(%1), 32(%1),  96(%1),  0(%0), 20)

  1120. //IDCT(   8(%1), 72(%1), 40(%1), 104(%1),  4(%0), 20)

  1121. IDCT(  16(%1), 80(%1), 48(%1), 112(%1),  8(%0), 20)

  1122. //IDCT(  24(%1), 88(%1), 56(%1), 120(%1), 12(%0), 20)

  1123. 

  1124. 

  1125. #endif

  1126. 

  1127. /*

  1128. Input

  1129.  00 40 04 44 20 60 24 64

  1130.  10 30 14 34 50 70 54 74

  1131.  01 41 03 43 21 61 23 63

  1132.  11 31 13 33 51 71 53 73

  1133.  02 42 06 46 22 62 26 66

  1134.  12 32 16 36 52 72 56 76

  1135.  05 45 07 47 25 65 27 67

  1136.  15 35 17 37 55 75 57 77

  1137. 

  1138. Temp

  1139.  00 04 10 14 20 24 30 34

  1140.  40 44 50 54 60 64 70 74

  1141.  01 03 11 13 21 23 31 33

  1142.  41 43 51 53 61 63 71 73

  1143.  02 06 12 16 22 26 32 36

  1144.  42 46 52 56 62 66 72 76

  1145.  05 07 15 17 25 27 35 37

  1146.  45 47 55 57 65 67 75 77

  1147. */

  1148. 

  1149. "9: \n\t"

  1150.                 :: "r" (block), "r" (temp), "r" (coeffs)

  1151.                 : "%eax"

  1152.         );

  1153. }

  1154. 

  1155. void ff_simple_idct_mmx(int16_t *block)

  1156. {

  1157.     idct(block);

  1158. }

  1159. 

  1160. //FIXME merge add/put into the idct

  1161. 

  1162. void ff_simple_idct_put_mmx(uint8_t *dest, int line_size, int16_t *block)

  1163. {

  1164.     idct(block);

  1165.     ff_put_pixels_clamped(block, dest, line_size);

  1166. }

  1167. void ff_simple_idct_add_mmx(uint8_t *dest, int line_size, int16_t *block)

  1168. {

  1169.     idct(block);

  1170.     ff_add_pixels_clamped(block, dest, line_size);

  1171. }

  1172. 

  1173. #endif /* HAVE_INLINE_ASM */