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.
44689 Commits
32 Branches
283MB
Tree: c5714b51aa
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from 'c5714b51aa'
${ noResults }
FFmpeg/libavcodec/x86/simple_idct.c

919 lines
51KB
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. #define DC_COND_IDCT(src0, src4, src1, src5, dst, rounder, shift) \

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  170.         "1:                             \n\t"\

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

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

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

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

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

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

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

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

  179.         "2:                             \n\t"

  180. 

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  256. 

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  324. 

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

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

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

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

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

  330. 

  331. #undef IDCT

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  405. 

  406. 

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

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

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

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

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

  412.         "jmp 9f                         \n\t"

  413. 

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

  415.         "4:                             \n\t"

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

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

  418. 

  419. #undef IDCT

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  481. 

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

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

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

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

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

  487.         "jmp 9f                         \n\t"

  488. 

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

  490.         "6:                             \n\t"

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

  492. 

  493. #undef IDCT

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  546. 

  547. 

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

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

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

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

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

  553.         "jmp 9f                         \n\t"

  554. 

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

  556.         "2:                             \n\t"

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

  558. 

  559. #undef IDCT

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  624. 

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

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

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

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

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

  630.         "jmp 9f                         \n\t"

  631. 

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

  633.         "3:                             \n\t"

  634. #undef IDCT

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  662.         "packssdw %%mm1, %%mm1          \n\t" /* A1-B1  a1-b1 */\

  663.         "movd %%mm1, 96+" #dst "        \n\t"\

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

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

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

  667.         "pmaddwd %%mm2, %%mm4           \n\t" /* -C1R3+C5R1     -C1r3+C5r1 */\

  668.         "pmaddwd 96(%2), %%mm2          \n\t" /* -C5R3+C7R1     -C5r3+C7r1 */\

  669.         "movq %%mm5, %%mm1              \n\t" /* A2             a2 */\

  670.         "paddd %%mm4, %%mm1             \n\t" /* A2+B2          a2+b2 */\

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

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

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

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

  675.         "paddd %%mm2, %%mm6             \n\t" /* A3+B3          a3+b3 */\

  676.         "psubd %%mm2, %%mm4             \n\t" /* a3-B3          a3-b3 */\

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

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

  679.         "packssdw %%mm1, %%mm1          \n\t" /* A2+B2  a2+b2 */\

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

  681.         "movd %%mm1, 32+" #dst "        \n\t"\

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

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

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

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

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

  687. 

  688. 

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

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

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

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

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

  694.         "jmp 9f                         \n\t"

  695. 

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

  697.         "5:                             \n\t"

  698. #undef IDCT

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  716.         "movq 8+" #src0 ", %%mm2        \n\t" /* R4     R0      r4      r0 */\

  717.         "movq 8+" #src4 ", %%mm3        \n\t" /* R6     R2      r6      r2 */\

  718.         "movq 16(%2), %%mm1             \n\t" /* C4     C4      C4      C4 */\

  719.         "pmaddwd %%mm2, %%mm1           \n\t" /* C4R4+C4R0      C4r4+C4r0 */\

  720.         "movq 24(%2), %%mm7             \n\t" /* -C4    C4      -C4     C4 */\

  721.         "pmaddwd %%mm7, %%mm2           \n\t" /* -C4R4+C4R0     -C4r4+C4r0 */\

  722.         "movq 32(%2), %%mm7             \n\t" /* C6     C2      C6      C2 */\

  723.         "pmaddwd %%mm3, %%mm7           \n\t" /* C6R6+C2R2      C6r6+C2r2 */\

  724.         "pmaddwd 40(%2), %%mm3          \n\t" /* -C2R6+C6R2     -C2r6+C6r2 */\

  725.         "paddd %%mm1, %%mm7             \n\t" /* A0             a0 */\

  726.         "paddd %%mm1, %%mm1             \n\t" /* 2C0            2c0 */\

  727.         "psubd %%mm7, %%mm1             \n\t" /* A3             a3 */\

  728.         "paddd %%mm2, %%mm3             \n\t" /* A1             a1 */\

  729.         "paddd %%mm2, %%mm2             \n\t" /* 2C1            2c1 */\

  730.         "psubd %%mm3, %%mm2             \n\t" /* A2             a2 */\

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

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

  733.         "psrad $" #shift ", %%mm3       \n\t"\

  734.         "packssdw %%mm7, %%mm4          \n\t" /* A0     a0 */\

  735.         "movq %%mm4, " #dst "           \n\t"\

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

  737.         "packssdw %%mm3, %%mm0          \n\t" /* A1     a1 */\

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

  739.         "movq %%mm0, 96+" #dst "        \n\t"\

  740.         "movq %%mm4, 112+" #dst "       \n\t"\

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

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

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

  744.         "packssdw %%mm2, %%mm5          \n\t" /* A2-B2  a2-b2 */\

  745.         "movq %%mm5, 32+" #dst "        \n\t"\

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

  747.         "packssdw %%mm1, %%mm6          \n\t" /* A3+B3  a3+b3 */\

  748.         "movq %%mm6, 48+" #dst "        \n\t"\

  749.         "movq %%mm6, 64+" #dst "        \n\t"\

  750.         "movq %%mm5, 80+" #dst "        \n\t"

  751. 

  752. 

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

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

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

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

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

  758.         "jmp 9f                         \n\t"

  759. 

  760. 

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

  762.         "1:                             \n\t"

  763. #undef IDCT

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  784.         "movq 64(%2), %%mm1             \n\t"\

  785.         "pmaddwd %%mm2, %%mm1           \n\t" /* -C7R3+C3R1     -C7r3+C3r1 */\

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

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

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

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

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

  791.         "movq %%mm0, %%mm3              \n\t" /* A1             a1 */\

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

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

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

  795.         "psrad $" #shift ", %%mm3       \n\t"\

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

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

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

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

  800.         "packssdw %%mm3, %%mm3          \n\t" /* A1-B1  a1-b1 */\

  801.         "movd %%mm3, 96+" #dst "        \n\t"\

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

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

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

  805.         "pmaddwd %%mm2, %%mm4           \n\t" /* -C1R3+C5R1     -C1r3+C5r1 */\

  806.         "pmaddwd 96(%2), %%mm2          \n\t" /* -C5R3+C7R1     -C5r3+C7r1 */\

  807.         "movq %%mm5, %%mm3              \n\t" /* A2             a2 */\

  808.         "paddd %%mm4, %%mm3             \n\t" /* A2+B2          a2+b2 */\

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

  810.         "psrad $" #shift ", %%mm3       \n\t"\

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

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

  813.         "paddd %%mm2, %%mm6             \n\t" /* A3+B3          a3+b3 */\

  814.         "psubd %%mm2, %%mm4             \n\t" /* a3-B3          a3-b3 */\

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

  816.         "packssdw %%mm3, %%mm3          \n\t" /* A2+B2  a2+b2 */\

  817.         "movd %%mm3, 32+" #dst "        \n\t"\

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

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

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

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

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

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

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

  825. 

  826. 

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

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

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

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

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

  832.         "jmp 9f                         \n\t"

  833. 

  834. 

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

  836.         "7:                             \n\t"

  837. #undef IDCT

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

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

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

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

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

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

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

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

  846.         "movq 8+" #src0 ", %%mm2        \n\t" /* R4     R0      r4      r0 */\

  847.         "movq 16(%2), %%mm1             \n\t" /* C4     C4      C4      C4 */\

  848.         "pmaddwd %%mm2, %%mm1           \n\t" /* C4R4+C4R0      C4r4+C4r0 */\

  849.         "movq 24(%2), %%mm7             \n\t" /* -C4    C4      -C4     C4 */\

  850.         "pmaddwd %%mm7, %%mm2           \n\t" /* -C4R4+C4R0     -C4r4+C4r0 */\

  851.         "movq 32(%2), %%mm7             \n\t" /* C6     C2      C6      C2 */\

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

  853.         "packssdw %%mm1, %%mm4          \n\t" /* A0     a0 */\

  854.         "movq %%mm4, " #dst "           \n\t"\

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

  856.         "packssdw %%mm2, %%mm0          \n\t" /* A1     a1 */\

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

  858.         "movq %%mm0, 96+" #dst "        \n\t"\

  859.         "movq %%mm4, 112+" #dst "       \n\t"\

  860.         "movq %%mm0, 32+" #dst "        \n\t"\

  861.         "movq %%mm4, 48+" #dst "        \n\t"\

  862.         "movq %%mm4, 64+" #dst "        \n\t"\

  863.         "movq %%mm0, 80+" #dst "        \n\t"

  864. 

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

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

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

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

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

  870. 

  871. 

  872. /*

  873. Input

  874.  00 40 04 44 20 60 24 64

  875.  10 30 14 34 50 70 54 74

  876.  01 41 03 43 21 61 23 63

  877.  11 31 13 33 51 71 53 73

  878.  02 42 06 46 22 62 26 66

  879.  12 32 16 36 52 72 56 76

  880.  05 45 07 47 25 65 27 67

  881.  15 35 17 37 55 75 57 77

  882. 

  883. Temp

  884.  00 04 10 14 20 24 30 34

  885.  40 44 50 54 60 64 70 74

  886.  01 03 11 13 21 23 31 33

  887.  41 43 51 53 61 63 71 73

  888.  02 06 12 16 22 26 32 36

  889.  42 46 52 56 62 66 72 76

  890.  05 07 15 17 25 27 35 37

  891.  45 47 55 57 65 67 75 77

  892. */

  893. 

  894. "9: \n\t"

  895.                 :: "r" (block), "r" (temp), "r" (coeffs)

  896.                 : "%eax"

  897.         );

  898. }

  899. 

  900. void ff_simple_idct_mmx(int16_t *block)

  901. {

  902.     idct(block);

  903. }

  904. 

  905. //FIXME merge add/put into the idct

  906. 

  907. void ff_simple_idct_put_mmx(uint8_t *dest, ptrdiff_t line_size, int16_t *block)

  908. {

  909.     idct(block);

  910.     ff_put_pixels_clamped(block, dest, line_size);

  911. }

  912. void ff_simple_idct_add_mmx(uint8_t *dest, ptrdiff_t line_size, int16_t *block)

  913. {

  914.     idct(block);

  915.     ff_add_pixels_clamped(block, dest, line_size);

  916. }

  917. 

  918. #endif /* HAVE_INLINE_ASM */