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.
35406 Commits
32 Branches
821MB
Tree: c3e15f7b39
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from 'c3e15f7b39'
${ noResults }
FFmpeg/libavcodec/x86/simple_idct.c

1170 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. #include "libavcodec/dsputil.h"

  23. #include "libavcodec/simple_idct.h"

  24. #include "libavutil/internal.h"

  25. #include "libavutil/mem.h"

  26. #include "dsputil_mmx.h"

  27. 

  28. #if HAVE_INLINE_ASM

  29. 

  30. /*

  31. 23170.475006

  32. 22725.260826

  33. 21406.727617

  34. 19265.545870

  35. 16384.000000

  36. 12872.826198

  37. 8866.956905

  38. 4520.335430

  39. */

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

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

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

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

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

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

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

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

  48. 

  49. #define ROW_SHIFT 11

  50. #define COL_SHIFT 20 // 6

  51. 

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

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

  54. 

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

  56.         1<<(ROW_SHIFT-1), 0, 1<<(ROW_SHIFT-1), 0,

  57. //        1<<(COL_SHIFT-1), 0, 1<<(COL_SHIFT-1), 0,

  58. //        0, 1<<(COL_SHIFT-1-16), 0, 1<<(COL_SHIFT-1-16),

  59.         1<<(ROW_SHIFT-1), 1, 1<<(ROW_SHIFT-1), 0,

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

  61. //        0, 0, 0, 0,

  62. //        0, 0, 0, 0,

  63. 

  64.  C4,  C4,  C4,  C4,

  65.  C4, -C4,  C4, -C4,

  66. 

  67.  C2,  C6,  C2,  C6,

  68.  C6, -C2,  C6, -C2,

  69. 

  70.  C1,  C3,  C1,  C3,

  71.  C5,  C7,  C5,  C7,

  72. 

  73.  C3, -C7,  C3, -C7,

  74. -C1, -C5, -C1, -C5,

  75. 

  76.  C5, -C1,  C5, -C1,

  77.  C7,  C3,  C7,  C3,

  78. 

  79.  C7, -C5,  C7, -C5,

  80.  C3, -C1,  C3, -C1

  81. };

  82. 

  83. static inline void idct(int16_t *block)

  84. {

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

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

  87. 

  88.         __asm__ volatile(

  89. #if 0 //Alternative, simpler variant

  90. 

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  158. 

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  232. 

  233. 

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  311.         "1:                             \n\t"\

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

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

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

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

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

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

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

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

  320.         "2:                             \n\t"

  321. 

  322. 

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

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

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

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

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

  328. 

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

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

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

  332. 

  333. 

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

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

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

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

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

  339. 

  340. #else

  341. 

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  419.         "1:                             \n\t"\

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

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

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

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

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

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

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

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

  428.         "2:                             \n\t"

  429. 

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  505. 

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  573. 

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

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

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

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

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

  579. 

  580. #undef IDCT

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  654. 

  655. 

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

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

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

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

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

  661.         "jmp 9f                         \n\t"

  662. 

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

  664.         "4:                             \n\t"

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

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

  667. 

  668. #undef IDCT

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  730. 

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

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

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

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

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

  736.         "jmp 9f                         \n\t"

  737. 

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

  739.         "6:                             \n\t"

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

  741. 

  742. #undef IDCT

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  795. 

  796. 

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

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

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

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

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

  802.         "jmp 9f                         \n\t"

  803. 

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

  805.         "2:                             \n\t"

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

  807. 

  808. #undef IDCT

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  873. 

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

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

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

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

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

  879.         "jmp 9f                         \n\t"

  880. 

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

  882.         "3:                             \n\t"

  883. #undef IDCT

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  902.         "movq %%mm0, %%mm1              \n\t" /* A1             a1 */\

  903.         "paddd %%mm3, %%mm0             \n\t" /* A1+B1          a1+b1 */\

  904.         "psubd %%mm3, %%mm1             \n\t" /* A1-B1          a1-b1 */\

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

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

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

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

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

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

  911.         "packssdw %%mm1, %%mm1          \n\t" /* A1-B1  a1-b1 */\

  912.         "movd %%mm1, 96+" #dst "        \n\t"\

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

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

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

  916.         "pmaddwd %%mm2, %%mm4           \n\t" /* -C1R3+C5R1     -C1r3+C5r1 */\

  917.         "pmaddwd 96(%2), %%mm2          \n\t" /* -C5R3+C7R1     -C5r3+C7r1 */\

  918.         "movq %%mm5, %%mm1              \n\t" /* A2             a2 */\

  919.         "paddd %%mm4, %%mm1             \n\t" /* A2+B2          a2+b2 */\

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

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

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

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

  924.         "paddd %%mm2, %%mm6             \n\t" /* A3+B3          a3+b3 */\

  925.         "psubd %%mm2, %%mm4             \n\t" /* a3-B3          a3-b3 */\

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

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

  928.         "packssdw %%mm1, %%mm1          \n\t" /* A2+B2  a2+b2 */\

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

  930.         "movd %%mm1, 32+" #dst "        \n\t"\

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

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

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

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

  935.         "movd %%mm5, 80+" #dst "        \n\t"

  936. 

  937. 

  938. //IDCT(  src0,   src4,   src1,    src5,    dst, shift)

  939. IDCT(    (%1), 64(%1), 32(%1),  96(%1),  0(%0), 20)

  940. IDCT(   8(%1), 72(%1), 40(%1), 104(%1),  4(%0), 20)

  941. IDCT(  16(%1), 80(%1), 48(%1), 112(%1),  8(%0), 20)

  942. IDCT(  24(%1), 88(%1), 56(%1), 120(%1), 12(%0), 20)

  943.         "jmp 9f                         \n\t"

  944. 

  945.         "# .p2align 4                   \n\t"\

  946.         "5:                             \n\t"

  947. #undef IDCT

  948. #define IDCT(src0, src4, src1, src5, dst, shift) \

  949.         "movq " #src0 ", %%mm0          \n\t" /* R4     R0      r4      r0 */\

  950.         "movq " #src4 ", %%mm1          \n\t" /* R6     R2      r6      r2 */\

  951.         "movq 16(%2), %%mm4             \n\t" /* C4     C4      C4      C4 */\

  952.         "pmaddwd %%mm0, %%mm4           \n\t" /* C4R4+C4R0      C4r4+C4r0 */\

  953.         "movq 24(%2), %%mm5             \n\t" /* -C4    C4      -C4     C4 */\

  954.         "pmaddwd %%mm5, %%mm0           \n\t" /* -C4R4+C4R0     -C4r4+C4r0 */\

  955.         "movq 32(%2), %%mm5             \n\t" /* C6     C2      C6      C2 */\

  956.         "pmaddwd %%mm1, %%mm5           \n\t" /* C6R6+C2R2      C6r6+C2r2 */\

  957.         "movq 40(%2), %%mm6             \n\t" /* -C2    C6      -C2     C6 */\

  958.         "pmaddwd %%mm6, %%mm1           \n\t" /* -C2R6+C6R2     -C2r6+C6r2 */\

  959.         "movq %%mm4, %%mm6              \n\t" /* C4R4+C4R0      C4r4+C4r0 */\

  960.         "paddd %%mm5, %%mm4             \n\t" /* A0             a0 */\

  961.         "psubd %%mm5, %%mm6             \n\t" /* A3             a3 */\

  962.         "movq %%mm0, %%mm5              \n\t" /* -C4R4+C4R0     -C4r4+C4r0 */\

  963.         "paddd %%mm1, %%mm0             \n\t" /* A1             a1 */\

  964.         "psubd %%mm1, %%mm5             \n\t" /* A2             a2 */\

  965.         "movq 8+" #src0 ", %%mm2        \n\t" /* R4     R0      r4      r0 */\

  966.         "movq 8+" #src4 ", %%mm3        \n\t" /* R6     R2      r6      r2 */\

  967.         "movq 16(%2), %%mm1             \n\t" /* C4     C4      C4      C4 */\

  968.         "pmaddwd %%mm2, %%mm1           \n\t" /* C4R4+C4R0      C4r4+C4r0 */\

  969.         "movq 24(%2), %%mm7             \n\t" /* -C4    C4      -C4     C4 */\

  970.         "pmaddwd %%mm7, %%mm2           \n\t" /* -C4R4+C4R0     -C4r4+C4r0 */\

  971.         "movq 32(%2), %%mm7             \n\t" /* C6     C2      C6      C2 */\

  972.         "pmaddwd %%mm3, %%mm7           \n\t" /* C6R6+C2R2      C6r6+C2r2 */\

  973.         "pmaddwd 40(%2), %%mm3          \n\t" /* -C2R6+C6R2     -C2r6+C6r2 */\

  974.         "paddd %%mm1, %%mm7             \n\t" /* A0             a0 */\

  975.         "paddd %%mm1, %%mm1             \n\t" /* 2C0            2c0 */\

  976.         "psubd %%mm7, %%mm1             \n\t" /* A3             a3 */\

  977.         "paddd %%mm2, %%mm3             \n\t" /* A1             a1 */\

  978.         "paddd %%mm2, %%mm2             \n\t" /* 2C1            2c1 */\

  979.         "psubd %%mm3, %%mm2             \n\t" /* A2             a2 */\

  980.         "psrad $" #shift ", %%mm4       \n\t"\

  981.         "psrad $" #shift ", %%mm7       \n\t"\

  982.         "psrad $" #shift ", %%mm3       \n\t"\

  983.         "packssdw %%mm7, %%mm4          \n\t" /* A0     a0 */\

  984.         "movq %%mm4, " #dst "           \n\t"\

  985.         "psrad $" #shift ", %%mm0       \n\t"\

  986.         "packssdw %%mm3, %%mm0          \n\t" /* A1     a1 */\

  987.         "movq %%mm0, 16+" #dst "        \n\t"\

  988.         "movq %%mm0, 96+" #dst "        \n\t"\

  989.         "movq %%mm4, 112+" #dst "       \n\t"\

  990.         "psrad $" #shift ", %%mm5       \n\t"\

  991.         "psrad $" #shift ", %%mm6       \n\t"\

  992.         "psrad $" #shift ", %%mm2       \n\t"\

  993.         "packssdw %%mm2, %%mm5          \n\t" /* A2-B2  a2-b2 */\

  994.         "movq %%mm5, 32+" #dst "        \n\t"\

  995.         "psrad $" #shift ", %%mm1       \n\t"\

  996.         "packssdw %%mm1, %%mm6          \n\t" /* A3+B3  a3+b3 */\

  997.         "movq %%mm6, 48+" #dst "        \n\t"\

  998.         "movq %%mm6, 64+" #dst "        \n\t"\

  999.         "movq %%mm5, 80+" #dst "        \n\t"

  1000. 

  1001. 

  1002. //IDCT(  src0,   src4,   src1,    src5,    dst, shift)

  1003. IDCT(    0(%1), 64(%1), 32(%1),  96(%1),  0(%0), 20)

  1004. //IDCT(   8(%1), 72(%1), 40(%1), 104(%1),  4(%0), 20)

  1005. IDCT(  16(%1), 80(%1), 48(%1), 112(%1),  8(%0), 20)

  1006. //IDCT(  24(%1), 88(%1), 56(%1), 120(%1), 12(%0), 20)

  1007.         "jmp 9f                         \n\t"

  1008. 

  1009. 

  1010.         "# .p2align 4                   \n\t"\

  1011.         "1:                             \n\t"

  1012. #undef IDCT

  1013. #define IDCT(src0, src4, src1, src5, dst, shift) \

  1014.         "movq " #src0 ", %%mm0          \n\t" /* R4     R0      r4      r0 */\

  1015.         "movq " #src4 ", %%mm1          \n\t" /* R6     R2      r6      r2 */\

  1016.         "movq " #src1 ", %%mm2          \n\t" /* R3     R1      r3      r1 */\

  1017.         "movq 16(%2), %%mm4             \n\t" /* C4     C4      C4      C4 */\

  1018.         "pmaddwd %%mm0, %%mm4           \n\t" /* C4R4+C4R0      C4r4+C4r0 */\

  1019.         "movq 24(%2), %%mm5             \n\t" /* -C4    C4      -C4     C4 */\

  1020.         "pmaddwd %%mm5, %%mm0           \n\t" /* -C4R4+C4R0     -C4r4+C4r0 */\

  1021.         "movq 32(%2), %%mm5             \n\t" /* C6     C2      C6      C2 */\

  1022.         "pmaddwd %%mm1, %%mm5           \n\t" /* C6R6+C2R2      C6r6+C2r2 */\

  1023.         "movq 40(%2), %%mm6             \n\t" /* -C2    C6      -C2     C6 */\

  1024.         "pmaddwd %%mm6, %%mm1           \n\t" /* -C2R6+C6R2     -C2r6+C6r2 */\

  1025.         "movq %%mm4, %%mm6              \n\t" /* C4R4+C4R0      C4r4+C4r0 */\

  1026.         "movq 48(%2), %%mm7             \n\t" /* C3     C1      C3      C1 */\

  1027.         "pmaddwd %%mm2, %%mm7           \n\t" /* C3R3+C1R1      C3r3+C1r1 */\

  1028.         "paddd %%mm5, %%mm4             \n\t" /* A0             a0 */\

  1029.         "psubd %%mm5, %%mm6             \n\t" /* A3             a3 */\

  1030.         "movq %%mm0, %%mm5              \n\t" /* -C4R4+C4R0     -C4r4+C4r0 */\

  1031.         "paddd %%mm1, %%mm0             \n\t" /* A1             a1 */\

  1032.         "psubd %%mm1, %%mm5             \n\t" /* A2             a2 */\

  1033.         "movq 64(%2), %%mm1             \n\t"\

  1034.         "pmaddwd %%mm2, %%mm1           \n\t" /* -C7R3+C3R1     -C7r3+C3r1 */\

  1035.         "paddd %%mm4, %%mm7             \n\t" /* A0+B0          a0+b0 */\

  1036.         "paddd %%mm4, %%mm4             \n\t" /* 2A0            2a0 */\

  1037.         "psubd %%mm7, %%mm4             \n\t" /* A0-B0          a0-b0 */\

  1038.         "psrad $" #shift ", %%mm7       \n\t"\

  1039.         "psrad $" #shift ", %%mm4       \n\t"\

  1040.         "movq %%mm0, %%mm3              \n\t" /* A1             a1 */\

  1041.         "paddd %%mm1, %%mm0             \n\t" /* A1+B1          a1+b1 */\

  1042.         "psubd %%mm1, %%mm3             \n\t" /* A1-B1          a1-b1 */\

  1043.         "psrad $" #shift ", %%mm0       \n\t"\

  1044.         "psrad $" #shift ", %%mm3       \n\t"\

  1045.         "packssdw %%mm7, %%mm7          \n\t" /* A0+B0  a0+b0 */\

  1046.         "movd %%mm7, " #dst "           \n\t"\

  1047.         "packssdw %%mm0, %%mm0          \n\t" /* A1+B1  a1+b1 */\

  1048.         "movd %%mm0, 16+" #dst "        \n\t"\

  1049.         "packssdw %%mm3, %%mm3          \n\t" /* A1-B1  a1-b1 */\

  1050.         "movd %%mm3, 96+" #dst "        \n\t"\

  1051.         "packssdw %%mm4, %%mm4          \n\t" /* A0-B0  a0-b0 */\

  1052.         "movd %%mm4, 112+" #dst "       \n\t"\

  1053.         "movq 80(%2), %%mm4             \n\t" /* -C1    C5      -C1     C5 */\

  1054.         "pmaddwd %%mm2, %%mm4           \n\t" /* -C1R3+C5R1     -C1r3+C5r1 */\

  1055.         "pmaddwd 96(%2), %%mm2          \n\t" /* -C5R3+C7R1     -C5r3+C7r1 */\

  1056.         "movq %%mm5, %%mm3              \n\t" /* A2             a2 */\

  1057.         "paddd %%mm4, %%mm3             \n\t" /* A2+B2          a2+b2 */\

  1058.         "psubd %%mm4, %%mm5             \n\t" /* a2-B2          a2-b2 */\

  1059.         "psrad $" #shift ", %%mm3       \n\t"\

  1060.         "psrad $" #shift ", %%mm5       \n\t"\

  1061.         "movq %%mm6, %%mm4              \n\t" /* A3             a3 */\

  1062.         "paddd %%mm2, %%mm6             \n\t" /* A3+B3          a3+b3 */\

  1063.         "psubd %%mm2, %%mm4             \n\t" /* a3-B3          a3-b3 */\

  1064.         "psrad $" #shift ", %%mm6       \n\t"\

  1065.         "packssdw %%mm3, %%mm3          \n\t" /* A2+B2  a2+b2 */\

  1066.         "movd %%mm3, 32+" #dst "        \n\t"\

  1067.         "psrad $" #shift ", %%mm4       \n\t"\

  1068.         "packssdw %%mm6, %%mm6          \n\t" /* A3+B3  a3+b3 */\

  1069.         "movd %%mm6, 48+" #dst "        \n\t"\

  1070.         "packssdw %%mm4, %%mm4          \n\t" /* A3-B3  a3-b3 */\

  1071.         "packssdw %%mm5, %%mm5          \n\t" /* A2-B2  a2-b2 */\

  1072.         "movd %%mm4, 64+" #dst "        \n\t"\

  1073.         "movd %%mm5, 80+" #dst "        \n\t"

  1074. 

  1075. 

  1076. //IDCT(  src0,   src4,   src1,    src5,    dst, shift)

  1077. IDCT(    (%1), 64(%1), 32(%1),  96(%1),  0(%0), 20)

  1078. IDCT(   8(%1), 72(%1), 40(%1), 104(%1),  4(%0), 20)

  1079. IDCT(  16(%1), 80(%1), 48(%1), 112(%1),  8(%0), 20)

  1080. IDCT(  24(%1), 88(%1), 56(%1), 120(%1), 12(%0), 20)

  1081.         "jmp 9f                         \n\t"

  1082. 

  1083. 

  1084.         "# .p2align 4                   \n\t"

  1085.         "7:                             \n\t"

  1086. #undef IDCT

  1087. #define IDCT(src0, src4, src1, src5, dst, shift) \

  1088.         "movq " #src0 ", %%mm0          \n\t" /* R4     R0      r4      r0 */\

  1089.         "movq 16(%2), %%mm4             \n\t" /* C4     C4      C4      C4 */\

  1090.         "pmaddwd %%mm0, %%mm4           \n\t" /* C4R4+C4R0      C4r4+C4r0 */\

  1091.         "movq 24(%2), %%mm5             \n\t" /* -C4    C4      -C4     C4 */\

  1092.         "pmaddwd %%mm5, %%mm0           \n\t" /* -C4R4+C4R0     -C4r4+C4r0 */\

  1093.         "psrad $" #shift ", %%mm4       \n\t"\

  1094.         "psrad $" #shift ", %%mm0       \n\t"\

  1095.         "movq 8+" #src0 ", %%mm2        \n\t" /* R4     R0      r4      r0 */\

  1096.         "movq 16(%2), %%mm1             \n\t" /* C4     C4      C4      C4 */\

  1097.         "pmaddwd %%mm2, %%mm1           \n\t" /* C4R4+C4R0      C4r4+C4r0 */\

  1098.         "movq 24(%2), %%mm7             \n\t" /* -C4    C4      -C4     C4 */\

  1099.         "pmaddwd %%mm7, %%mm2           \n\t" /* -C4R4+C4R0     -C4r4+C4r0 */\

  1100.         "movq 32(%2), %%mm7             \n\t" /* C6     C2      C6      C2 */\

  1101.         "psrad $" #shift ", %%mm1       \n\t"\

  1102.         "packssdw %%mm1, %%mm4          \n\t" /* A0     a0 */\

  1103.         "movq %%mm4, " #dst "           \n\t"\

  1104.         "psrad $" #shift ", %%mm2       \n\t"\

  1105.         "packssdw %%mm2, %%mm0          \n\t" /* A1     a1 */\

  1106.         "movq %%mm0, 16+" #dst "        \n\t"\

  1107.         "movq %%mm0, 96+" #dst "        \n\t"\

  1108.         "movq %%mm4, 112+" #dst "       \n\t"\

  1109.         "movq %%mm0, 32+" #dst "        \n\t"\

  1110.         "movq %%mm4, 48+" #dst "        \n\t"\

  1111.         "movq %%mm4, 64+" #dst "        \n\t"\

  1112.         "movq %%mm0, 80+" #dst "        \n\t"

  1113. 

  1114. //IDCT(  src0,   src4,   src1,    src5,    dst, shift)

  1115. IDCT(   0(%1), 64(%1), 32(%1),  96(%1),  0(%0), 20)

  1116. //IDCT(   8(%1), 72(%1), 40(%1), 104(%1),  4(%0), 20)

  1117. IDCT(  16(%1), 80(%1), 48(%1), 112(%1),  8(%0), 20)

  1118. //IDCT(  24(%1), 88(%1), 56(%1), 120(%1), 12(%0), 20)

  1119. 

  1120. 

  1121. #endif

  1122. 

  1123. /*

  1124. Input

  1125.  00 40 04 44 20 60 24 64

  1126.  10 30 14 34 50 70 54 74

  1127.  01 41 03 43 21 61 23 63

  1128.  11 31 13 33 51 71 53 73

  1129.  02 42 06 46 22 62 26 66

  1130.  12 32 16 36 52 72 56 76

  1131.  05 45 07 47 25 65 27 67

  1132.  15 35 17 37 55 75 57 77

  1133. 

  1134. Temp

  1135.  00 04 10 14 20 24 30 34

  1136.  40 44 50 54 60 64 70 74

  1137.  01 03 11 13 21 23 31 33

  1138.  41 43 51 53 61 63 71 73

  1139.  02 06 12 16 22 26 32 36

  1140.  42 46 52 56 62 66 72 76

  1141.  05 07 15 17 25 27 35 37

  1142.  45 47 55 57 65 67 75 77

  1143. */

  1144. 

  1145. "9: \n\t"

  1146.                 :: "r" (block), "r" (temp), "r" (coeffs)

  1147.                 : "%eax"

  1148.         );

  1149. }

  1150. 

  1151. void ff_simple_idct_mmx(int16_t *block)

  1152. {

  1153.     idct(block);

  1154. }

  1155. 

  1156. //FIXME merge add/put into the idct

  1157. 

  1158. void ff_simple_idct_put_mmx(uint8_t *dest, int line_size, DCTELEM *block)

  1159. {

  1160.     idct(block);

  1161.     ff_put_pixels_clamped_mmx(block, dest, line_size);

  1162. }

  1163. void ff_simple_idct_add_mmx(uint8_t *dest, int line_size, DCTELEM *block)

  1164. {

  1165.     idct(block);

  1166.     ff_add_pixels_clamped_mmx(block, dest, line_size);

  1167. }

  1168. 

  1169. #endif /* HAVE_INLINE_ASM */