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.
6808 Commits
32 Branches
390MB
Tree: 33592a633f
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from '33592a633f'
${ noResults }
FFmpeg/libavcodec/ppc/h264_altivec.c

437 lines
18KB
Raw Blame History 

  1. /*

  2.  * Copyright (c) 2004 Romain Dolbeau <romain@dolbeau.org>

  3.  *

  4.  * This library is free software; you can redistribute it and/or

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

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

  7.  * version 2 of the License, or (at your option) any later version.

  8.  *

  9.  * This library is distributed in the hope that it will be useful,

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

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

  12.  * Lesser General Public License for more details.

  13.  *

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

  15.  * License along with this library; if not, write to the Free Software

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

  17.  */

  18. 

  19. #include "../dsputil.h"

  20. 

  21. #include "gcc_fixes.h"

  22. 

  23. #include "dsputil_altivec.h"

  24. 

  25. #define PUT_OP_U8_ALTIVEC(d, s, dst) d = s

  26. #define AVG_OP_U8_ALTIVEC(d, s, dst) d = vec_avg(dst, s)

  27. 

  28. #define OP_U8_ALTIVEC                          PUT_OP_U8_ALTIVEC

  29. #define PREFIX_h264_chroma_mc8_altivec         put_h264_chroma_mc8_altivec

  30. #define PREFIX_h264_chroma_mc8_num             altivec_put_h264_chroma_mc8_num

  31. #define PREFIX_h264_qpel16_h_lowpass_altivec   put_h264_qpel16_h_lowpass_altivec

  32. #define PREFIX_h264_qpel16_h_lowpass_num       altivec_put_h264_qpel16_h_lowpass_num

  33. #define PREFIX_h264_qpel16_v_lowpass_altivec   put_h264_qpel16_v_lowpass_altivec

  34. #define PREFIX_h264_qpel16_v_lowpass_num       altivec_put_h264_qpel16_v_lowpass_num

  35. #define PREFIX_h264_qpel16_hv_lowpass_altivec  put_h264_qpel16_hv_lowpass_altivec

  36. #define PREFIX_h264_qpel16_hv_lowpass_num      altivec_put_h264_qpel16_hv_lowpass_num

  37. #include "h264_template_altivec.c"

  38. #undef OP_U8_ALTIVEC

  39. #undef PREFIX_h264_chroma_mc8_altivec

  40. #undef PREFIX_h264_chroma_mc8_num

  41. #undef PREFIX_h264_qpel16_h_lowpass_altivec

  42. #undef PREFIX_h264_qpel16_h_lowpass_num

  43. #undef PREFIX_h264_qpel16_v_lowpass_altivec

  44. #undef PREFIX_h264_qpel16_v_lowpass_num

  45. #undef PREFIX_h264_qpel16_hv_lowpass_altivec

  46. #undef PREFIX_h264_qpel16_hv_lowpass_num

  47. 

  48. #define OP_U8_ALTIVEC                          AVG_OP_U8_ALTIVEC

  49. #define PREFIX_h264_chroma_mc8_altivec         avg_h264_chroma_mc8_altivec

  50. #define PREFIX_h264_chroma_mc8_num             altivec_avg_h264_chroma_mc8_num

  51. #define PREFIX_h264_qpel16_h_lowpass_altivec   avg_h264_qpel16_h_lowpass_altivec

  52. #define PREFIX_h264_qpel16_h_lowpass_num       altivec_avg_h264_qpel16_h_lowpass_num

  53. #define PREFIX_h264_qpel16_v_lowpass_altivec   avg_h264_qpel16_v_lowpass_altivec

  54. #define PREFIX_h264_qpel16_v_lowpass_num       altivec_avg_h264_qpel16_v_lowpass_num

  55. #define PREFIX_h264_qpel16_hv_lowpass_altivec  avg_h264_qpel16_hv_lowpass_altivec

  56. #define PREFIX_h264_qpel16_hv_lowpass_num      altivec_avg_h264_qpel16_hv_lowpass_num

  57. #include "h264_template_altivec.c"

  58. #undef OP_U8_ALTIVEC

  59. #undef PREFIX_h264_chroma_mc8_altivec

  60. #undef PREFIX_h264_chroma_mc8_num

  61. #undef PREFIX_h264_qpel16_h_lowpass_altivec

  62. #undef PREFIX_h264_qpel16_h_lowpass_num

  63. #undef PREFIX_h264_qpel16_v_lowpass_altivec

  64. #undef PREFIX_h264_qpel16_v_lowpass_num

  65. #undef PREFIX_h264_qpel16_hv_lowpass_altivec

  66. #undef PREFIX_h264_qpel16_hv_lowpass_num

  67. 

  68. #define H264_MC(OPNAME, SIZE, CODETYPE) \

  69. static void OPNAME ## h264_qpel ## SIZE ## _mc00_ ## CODETYPE (uint8_t *dst, uint8_t *src, int stride){\

  70.     OPNAME ## pixels ## SIZE ## _ ## CODETYPE(dst, src, stride, SIZE);\

  71. }\

  72. \

  73. static void OPNAME ## h264_qpel ## SIZE ## _mc10_ ## CODETYPE(uint8_t *dst, uint8_t *src, int stride){ \

  74.     DECLARE_ALIGNED_16(uint8_t, half[SIZE*SIZE]);\

  75.     put_h264_qpel ## SIZE ## _h_lowpass_ ## CODETYPE(half, src, SIZE, stride);\

  76.     OPNAME ## pixels ## SIZE ## _l2_ ## CODETYPE(dst, src, half, stride, stride, SIZE);\

  77. }\

  78. \

  79. static void OPNAME ## h264_qpel ## SIZE ## _mc20_ ## CODETYPE(uint8_t *dst, uint8_t *src, int stride){\

  80.     OPNAME ## h264_qpel ## SIZE ## _h_lowpass_ ## CODETYPE(dst, src, stride, stride);\

  81. }\

  82. \

  83. static void OPNAME ## h264_qpel ## SIZE ## _mc30_ ## CODETYPE(uint8_t *dst, uint8_t *src, int stride){\

  84.     DECLARE_ALIGNED_16(uint8_t, half[SIZE*SIZE]);\

  85.     put_h264_qpel ## SIZE ## _h_lowpass_ ## CODETYPE(half, src, SIZE, stride);\

  86.     OPNAME ## pixels ## SIZE ## _l2_ ## CODETYPE(dst, src+1, half, stride, stride, SIZE);\

  87. }\

  88. \

  89. static void OPNAME ## h264_qpel ## SIZE ## _mc01_ ## CODETYPE(uint8_t *dst, uint8_t *src, int stride){\

  90.     DECLARE_ALIGNED_16(uint8_t, half[SIZE*SIZE]);\

  91.     put_h264_qpel ## SIZE ## _v_lowpass_ ## CODETYPE(half, src, SIZE, stride);\

  92.     OPNAME ## pixels ## SIZE ## _l2_ ## CODETYPE(dst, src, half, stride, stride, SIZE);\

  93. }\

  94. \

  95. static void OPNAME ## h264_qpel ## SIZE ## _mc02_ ## CODETYPE(uint8_t *dst, uint8_t *src, int stride){\

  96.     OPNAME ## h264_qpel ## SIZE ## _v_lowpass_ ## CODETYPE(dst, src, stride, stride);\

  97. }\

  98. \

  99. static void OPNAME ## h264_qpel ## SIZE ## _mc03_ ## CODETYPE(uint8_t *dst, uint8_t *src, int stride){\

  100.     DECLARE_ALIGNED_16(uint8_t, half[SIZE*SIZE]);\

  101.     put_h264_qpel ## SIZE ## _v_lowpass_ ## CODETYPE(half, src, SIZE, stride);\

  102.     OPNAME ## pixels ## SIZE ## _l2_ ## CODETYPE(dst, src+stride, half, stride, stride, SIZE);\

  103. }\

  104. \

  105. static void OPNAME ## h264_qpel ## SIZE ## _mc11_ ## CODETYPE(uint8_t *dst, uint8_t *src, int stride){\

  106.     DECLARE_ALIGNED_16(uint8_t, halfH[SIZE*SIZE]);\

  107.     DECLARE_ALIGNED_16(uint8_t, halfV[SIZE*SIZE]);\

  108.     put_h264_qpel ## SIZE ## _h_lowpass_ ## CODETYPE(halfH, src, SIZE, stride);\

  109.     put_h264_qpel ## SIZE ## _v_lowpass_ ## CODETYPE(halfV, src, SIZE, stride);\

  110.     OPNAME ## pixels ## SIZE ## _l2_ ## CODETYPE(dst, halfH, halfV, stride, SIZE, SIZE);\

  111. }\

  112. \

  113. static void OPNAME ## h264_qpel ## SIZE ## _mc31_ ## CODETYPE(uint8_t *dst, uint8_t *src, int stride){\

  114.     DECLARE_ALIGNED_16(uint8_t, halfH[SIZE*SIZE]);\

  115.     DECLARE_ALIGNED_16(uint8_t, halfV[SIZE*SIZE]);\

  116.     put_h264_qpel ## SIZE ## _h_lowpass_ ## CODETYPE(halfH, src, SIZE, stride);\

  117.     put_h264_qpel ## SIZE ## _v_lowpass_ ## CODETYPE(halfV, src+1, SIZE, stride);\

  118.     OPNAME ## pixels ## SIZE ## _l2_ ## CODETYPE(dst, halfH, halfV, stride, SIZE, SIZE);\

  119. }\

  120. \

  121. static void OPNAME ## h264_qpel ## SIZE ## _mc13_ ## CODETYPE(uint8_t *dst, uint8_t *src, int stride){\

  122.     DECLARE_ALIGNED_16(uint8_t, halfH[SIZE*SIZE]);\

  123.     DECLARE_ALIGNED_16(uint8_t, halfV[SIZE*SIZE]);\

  124.     put_h264_qpel ## SIZE ## _h_lowpass_ ## CODETYPE(halfH, src + stride, SIZE, stride);\

  125.     put_h264_qpel ## SIZE ## _v_lowpass_ ## CODETYPE(halfV, src, SIZE, stride);\

  126.     OPNAME ## pixels ## SIZE ## _l2_ ## CODETYPE(dst, halfH, halfV, stride, SIZE, SIZE);\

  127. }\

  128. \

  129. static void OPNAME ## h264_qpel ## SIZE ## _mc33_ ## CODETYPE(uint8_t *dst, uint8_t *src, int stride){\

  130.     DECLARE_ALIGNED_16(uint8_t, halfH[SIZE*SIZE]);\

  131.     DECLARE_ALIGNED_16(uint8_t, halfV[SIZE*SIZE]);\

  132.     put_h264_qpel ## SIZE ## _h_lowpass_ ## CODETYPE(halfH, src + stride, SIZE, stride);\

  133.     put_h264_qpel ## SIZE ## _v_lowpass_ ## CODETYPE(halfV, src+1, SIZE, stride);\

  134.     OPNAME ## pixels ## SIZE ## _l2_ ## CODETYPE(dst, halfH, halfV, stride, SIZE, SIZE);\

  135. }\

  136. \

  137. static void OPNAME ## h264_qpel ## SIZE ## _mc22_ ## CODETYPE(uint8_t *dst, uint8_t *src, int stride){\

  138.     DECLARE_ALIGNED_16(int16_t, tmp[SIZE*(SIZE+8)]);\

  139.     OPNAME ## h264_qpel ## SIZE ## _hv_lowpass_ ## CODETYPE(dst, tmp, src, stride, SIZE, stride);\

  140. }\

  141. \

  142. static void OPNAME ## h264_qpel ## SIZE ## _mc21_ ## CODETYPE(uint8_t *dst, uint8_t *src, int stride){\

  143.     DECLARE_ALIGNED_16(uint8_t, halfH[SIZE*SIZE]);\

  144.     DECLARE_ALIGNED_16(uint8_t, halfHV[SIZE*SIZE]);\

  145.     DECLARE_ALIGNED_16(int16_t, tmp[SIZE*(SIZE+8)]);\

  146.     put_h264_qpel ## SIZE ## _h_lowpass_ ## CODETYPE(halfH, src, SIZE, stride);\

  147.     put_h264_qpel ## SIZE ## _hv_lowpass_ ## CODETYPE(halfHV, tmp, src, SIZE, SIZE, stride);\

  148.     OPNAME ## pixels ## SIZE ## _l2_ ## CODETYPE(dst, halfH, halfHV, stride, SIZE, SIZE);\

  149. }\

  150. \

  151. static void OPNAME ## h264_qpel ## SIZE ## _mc23_ ## CODETYPE(uint8_t *dst, uint8_t *src, int stride){\

  152.     DECLARE_ALIGNED_16(uint8_t, halfH[SIZE*SIZE]);\

  153.     DECLARE_ALIGNED_16(uint8_t, halfHV[SIZE*SIZE]);\

  154.     DECLARE_ALIGNED_16(int16_t, tmp[SIZE*(SIZE+8)]);\

  155.     put_h264_qpel ## SIZE ## _h_lowpass_ ## CODETYPE(halfH, src + stride, SIZE, stride);\

  156.     put_h264_qpel ## SIZE ## _hv_lowpass_ ## CODETYPE(halfHV, tmp, src, SIZE, SIZE, stride);\

  157.     OPNAME ## pixels ## SIZE ## _l2_ ## CODETYPE(dst, halfH, halfHV, stride, SIZE, SIZE);\

  158. }\

  159. \

  160. static void OPNAME ## h264_qpel ## SIZE ## _mc12_ ## CODETYPE(uint8_t *dst, uint8_t *src, int stride){\

  161.     DECLARE_ALIGNED_16(uint8_t, halfV[SIZE*SIZE]);\

  162.     DECLARE_ALIGNED_16(uint8_t, halfHV[SIZE*SIZE]);\

  163.     DECLARE_ALIGNED_16(int16_t, tmp[SIZE*(SIZE+8)]);\

  164.     put_h264_qpel ## SIZE ## _v_lowpass_ ## CODETYPE(halfV, src, SIZE, stride);\

  165.     put_h264_qpel ## SIZE ## _hv_lowpass_ ## CODETYPE(halfHV, tmp, src, SIZE, SIZE, stride);\

  166.     OPNAME ## pixels ## SIZE ## _l2_ ## CODETYPE(dst, halfV, halfHV, stride, SIZE, SIZE);\

  167. }\

  168. \

  169. static void OPNAME ## h264_qpel ## SIZE ## _mc32_ ## CODETYPE(uint8_t *dst, uint8_t *src, int stride){\

  170.     DECLARE_ALIGNED_16(uint8_t, halfV[SIZE*SIZE]);\

  171.     DECLARE_ALIGNED_16(uint8_t, halfHV[SIZE*SIZE]);\

  172.     DECLARE_ALIGNED_16(int16_t, tmp[SIZE*(SIZE+8)]);\

  173.     put_h264_qpel ## SIZE ## _v_lowpass_ ## CODETYPE(halfV, src+1, SIZE, stride);\

  174.     put_h264_qpel ## SIZE ## _hv_lowpass_ ## CODETYPE(halfHV, tmp, src, SIZE, SIZE, stride);\

  175.     OPNAME ## pixels ## SIZE ## _l2_ ## CODETYPE(dst, halfV, halfHV, stride, SIZE, SIZE);\

  176. }\

  177. 

  178. /* this code assume that stride % 16 == 0 */

  179. void put_no_rnd_h264_chroma_mc8_altivec(uint8_t * dst, uint8_t * src, int stride, int h, int x, int y) {

  180.     signed int ABCD[4] __attribute__((aligned(16))) =

  181.                         {((8 - x) * (8 - y)),

  182.                           ((x) * (8 - y)),

  183.                           ((8 - x) * (y)),

  184.                           ((x) * (y))};

  185.     register int i;

  186.     vector unsigned char fperm;

  187.     const vector signed int vABCD = vec_ld(0, ABCD);

  188.     const vector signed short vA = vec_splat((vector signed short)vABCD, 1);

  189.     const vector signed short vB = vec_splat((vector signed short)vABCD, 3);

  190.     const vector signed short vC = vec_splat((vector signed short)vABCD, 5);

  191.     const vector signed short vD = vec_splat((vector signed short)vABCD, 7);

  192.     const vector signed int vzero = vec_splat_s32(0);

  193.     const vector signed short v28ss = vec_sub(vec_sl(vec_splat_s16(1),vec_splat_u16(5)),vec_splat_s16(4));

  194.     const vector unsigned short v6us = vec_splat_u16(6);

  195.     register int loadSecond = (((unsigned long)src) % 16) <= 7 ? 0 : 1;

  196.     register int reallyBadAlign = (((unsigned long)src) % 16) == 15 ? 1 : 0;

  197. 

  198.     vector unsigned char vsrcAuc, vsrcBuc, vsrcperm0, vsrcperm1;

  199.     vector unsigned char vsrc0uc, vsrc1uc;

  200.     vector signed short vsrc0ssH, vsrc1ssH;

  201.     vector unsigned char vsrcCuc, vsrc2uc, vsrc3uc;

  202.     vector signed short vsrc2ssH, vsrc3ssH, psum;

  203.     vector unsigned char vdst, ppsum, vfdst, fsum;

  204. 

  205.     if (((unsigned long)dst) % 16 == 0) {

  206.       fperm = (vector unsigned char)AVV(0x10, 0x11, 0x12, 0x13,

  207.                                         0x14, 0x15, 0x16, 0x17,

  208.                                         0x08, 0x09, 0x0A, 0x0B,

  209.                                         0x0C, 0x0D, 0x0E, 0x0F);

  210.     } else {

  211.       fperm = (vector unsigned char)AVV(0x00, 0x01, 0x02, 0x03,

  212.                                         0x04, 0x05, 0x06, 0x07,

  213.                                         0x18, 0x19, 0x1A, 0x1B,

  214.                                         0x1C, 0x1D, 0x1E, 0x1F);

  215.     }

  216. 

  217.     vsrcAuc = vec_ld(0, src);

  218. 

  219.     if (loadSecond)

  220.       vsrcBuc = vec_ld(16, src);

  221.     vsrcperm0 = vec_lvsl(0, src);

  222.     vsrcperm1 = vec_lvsl(1, src);

  223. 

  224.     vsrc0uc = vec_perm(vsrcAuc, vsrcBuc, vsrcperm0);

  225.     if (reallyBadAlign)

  226.       vsrc1uc = vsrcBuc;

  227.     else

  228.       vsrc1uc = vec_perm(vsrcAuc, vsrcBuc, vsrcperm1);

  229. 

  230.     vsrc0ssH = (vector signed short)vec_mergeh((vector unsigned char)vzero,

  231.                                                (vector unsigned char)vsrc0uc);

  232.     vsrc1ssH = (vector signed short)vec_mergeh((vector unsigned char)vzero,

  233.                                                (vector unsigned char)vsrc1uc);

  234. 

  235.     if (!loadSecond) {// -> !reallyBadAlign

  236.       for (i = 0 ; i < h ; i++) {

  237. 

  238. 

  239.         vsrcCuc = vec_ld(stride + 0, src);

  240. 

  241.         vsrc2uc = vec_perm(vsrcCuc, vsrcCuc, vsrcperm0);

  242.         vsrc3uc = vec_perm(vsrcCuc, vsrcCuc, vsrcperm1);

  243. 

  244.         vsrc2ssH = (vector signed short)vec_mergeh((vector unsigned char)vzero,

  245.                                                 (vector unsigned char)vsrc2uc);

  246.         vsrc3ssH = (vector signed short)vec_mergeh((vector unsigned char)vzero,

  247.                                                 (vector unsigned char)vsrc3uc);

  248. 

  249.         psum = vec_mladd(vA, vsrc0ssH, vec_splat_s16(0));

  250.         psum = vec_mladd(vB, vsrc1ssH, psum);

  251.         psum = vec_mladd(vC, vsrc2ssH, psum);

  252.         psum = vec_mladd(vD, vsrc3ssH, psum);

  253.         psum = vec_add(v28ss, psum);

  254.         psum = vec_sra(psum, v6us);

  255. 

  256.         vdst = vec_ld(0, dst);

  257.         ppsum = (vector unsigned char)vec_packsu(psum, psum);

  258.         fsum = vec_perm(vdst, ppsum, fperm);

  259. 

  260.         vec_st(fsum, 0, dst);

  261. 

  262.         vsrc0ssH = vsrc2ssH;

  263.         vsrc1ssH = vsrc3ssH;

  264. 

  265.         dst += stride;

  266.         src += stride;

  267.       }

  268.     } else {

  269.         vector unsigned char vsrcDuc;

  270.       for (i = 0 ; i < h ; i++) {

  271.         vsrcCuc = vec_ld(stride + 0, src);

  272.         vsrcDuc = vec_ld(stride + 16, src);

  273. 

  274.         vsrc2uc = vec_perm(vsrcCuc, vsrcDuc, vsrcperm0);

  275.         if (reallyBadAlign)

  276.           vsrc3uc = vsrcDuc;

  277.         else

  278.           vsrc3uc = vec_perm(vsrcCuc, vsrcDuc, vsrcperm1);

  279. 

  280.         vsrc2ssH = (vector signed short)vec_mergeh((vector unsigned char)vzero,

  281.                                                 (vector unsigned char)vsrc2uc);

  282.         vsrc3ssH = (vector signed short)vec_mergeh((vector unsigned char)vzero,

  283.                                                 (vector unsigned char)vsrc3uc);

  284. 

  285.         psum = vec_mladd(vA, vsrc0ssH, vec_splat_s16(0));

  286.         psum = vec_mladd(vB, vsrc1ssH, psum);

  287.         psum = vec_mladd(vC, vsrc2ssH, psum);

  288.         psum = vec_mladd(vD, vsrc3ssH, psum);

  289.         psum = vec_add(v28ss, psum);

  290.         psum = vec_sr(psum, v6us);

  291. 

  292.         vdst = vec_ld(0, dst);

  293.         ppsum = (vector unsigned char)vec_pack(psum, psum);

  294.         fsum = vec_perm(vdst, ppsum, fperm);

  295. 

  296.         vec_st(fsum, 0, dst);

  297. 

  298.         vsrc0ssH = vsrc2ssH;

  299.         vsrc1ssH = vsrc3ssH;

  300. 

  301.         dst += stride;

  302.         src += stride;

  303.       }

  304.     }

  305. }

  306. 

  307. static inline void put_pixels16_l2_altivec( uint8_t * dst, const uint8_t * src1,

  308.                                     const uint8_t * src2, int dst_stride,

  309.                                     int src_stride1, int h)

  310. {

  311.     int i;

  312.     vector unsigned char a, b, d, tmp1, tmp2, mask, mask_, edges, align;

  313. 

  314.     mask_ = vec_lvsl(0, src2);

  315. 

  316.     for (i = 0; i < h; i++) {

  317. 

  318.         tmp1 = vec_ld(i * src_stride1, src1);

  319.         mask = vec_lvsl(i * src_stride1, src1);

  320.         tmp2 = vec_ld(i * src_stride1 + 15, src1);

  321. 

  322.         a = vec_perm(tmp1, tmp2, mask);

  323. 

  324.         tmp1 = vec_ld(i * 16, src2);

  325.         tmp2 = vec_ld(i * 16 + 15, src2);

  326. 

  327.         b = vec_perm(tmp1, tmp2, mask_);

  328. 

  329.         tmp1 = vec_ld(0, dst);

  330.         mask = vec_lvsl(0, dst);

  331.         tmp2 = vec_ld(15, dst);

  332. 

  333.         d = vec_avg(a, b);

  334. 

  335.         edges = vec_perm(tmp2, tmp1, mask);

  336. 

  337.         align = vec_lvsr(0, dst);

  338. 

  339.         tmp2 = vec_perm(d, edges, align);

  340.         tmp1 = vec_perm(edges, d, align);

  341. 

  342.         vec_st(tmp2, 15, dst);

  343.         vec_st(tmp1, 0 , dst);

  344. 

  345.         dst += dst_stride;

  346.     }

  347. }

  348. 

  349. static inline void avg_pixels16_l2_altivec( uint8_t * dst, const uint8_t * src1,

  350.                                     const uint8_t * src2, int dst_stride,

  351.                                     int src_stride1, int h)

  352. {

  353.     int i;

  354.     vector unsigned char a, b, d, tmp1, tmp2, mask, mask_, edges, align;

  355. 

  356.     mask_ = vec_lvsl(0, src2);

  357. 

  358.     for (i = 0; i < h; i++) {

  359. 

  360.         tmp1 = vec_ld(i * src_stride1, src1);

  361.         mask = vec_lvsl(i * src_stride1, src1);

  362.         tmp2 = vec_ld(i * src_stride1 + 15, src1);

  363. 

  364.         a = vec_perm(tmp1, tmp2, mask);

  365. 

  366.         tmp1 = vec_ld(i * 16, src2);

  367.         tmp2 = vec_ld(i * 16 + 15, src2);

  368. 

  369.         b = vec_perm(tmp1, tmp2, mask_);

  370. 

  371.         tmp1 = vec_ld(0, dst);

  372.         mask = vec_lvsl(0, dst);

  373.         tmp2 = vec_ld(15, dst);

  374. 

  375.         d = vec_avg(vec_perm(tmp1, tmp2, mask), vec_avg(a, b));

  376. 

  377.         edges = vec_perm(tmp2, tmp1, mask);

  378. 

  379.         align = vec_lvsr(0, dst);

  380. 

  381.         tmp2 = vec_perm(d, edges, align);

  382.         tmp1 = vec_perm(edges, d, align);

  383. 

  384.         vec_st(tmp2, 15, dst);

  385.         vec_st(tmp1, 0 , dst);

  386. 

  387.         dst += dst_stride;

  388.     }

  389. }

  390. 

  391. /* Implemented but could be faster

  392. #define put_pixels16_l2_altivec(d,s1,s2,ds,s1s,h) put_pixels16_l2(d,s1,s2,ds,s1s,16,h)

  393. #define avg_pixels16_l2_altivec(d,s1,s2,ds,s1s,h) avg_pixels16_l2(d,s1,s2,ds,s1s,16,h)

  394.  */

  395. 

  396.   H264_MC(put_, 16, altivec)

  397.   H264_MC(avg_, 16, altivec)

  398. 

  399. void dsputil_h264_init_ppc(DSPContext* c, AVCodecContext *avctx) {

  400. 

  401. #ifdef HAVE_ALTIVEC

  402.   if (has_altivec()) {

  403.     c->put_h264_chroma_pixels_tab[0] = put_h264_chroma_mc8_altivec;

  404.     c->put_no_rnd_h264_chroma_pixels_tab[0] = put_no_rnd_h264_chroma_mc8_altivec;

  405.     c->avg_h264_chroma_pixels_tab[0] = avg_h264_chroma_mc8_altivec;

  406. 

  407. #define dspfunc(PFX, IDX, NUM) \

  408.     c->PFX ## _pixels_tab[IDX][ 0] = PFX ## NUM ## _mc00_altivec; \

  409.     c->PFX ## _pixels_tab[IDX][ 1] = PFX ## NUM ## _mc10_altivec; \

  410.     c->PFX ## _pixels_tab[IDX][ 2] = PFX ## NUM ## _mc20_altivec; \

  411.     c->PFX ## _pixels_tab[IDX][ 3] = PFX ## NUM ## _mc30_altivec; \

  412.     c->PFX ## _pixels_tab[IDX][ 4] = PFX ## NUM ## _mc01_altivec; \

  413.     c->PFX ## _pixels_tab[IDX][ 5] = PFX ## NUM ## _mc11_altivec; \

  414.     c->PFX ## _pixels_tab[IDX][ 6] = PFX ## NUM ## _mc21_altivec; \

  415.     c->PFX ## _pixels_tab[IDX][ 7] = PFX ## NUM ## _mc31_altivec; \

  416.     c->PFX ## _pixels_tab[IDX][ 8] = PFX ## NUM ## _mc02_altivec; \

  417.     c->PFX ## _pixels_tab[IDX][ 9] = PFX ## NUM ## _mc12_altivec; \

  418.     c->PFX ## _pixels_tab[IDX][10] = PFX ## NUM ## _mc22_altivec; \

  419.     c->PFX ## _pixels_tab[IDX][11] = PFX ## NUM ## _mc32_altivec; \

  420.     c->PFX ## _pixels_tab[IDX][12] = PFX ## NUM ## _mc03_altivec; \

  421.     c->PFX ## _pixels_tab[IDX][13] = PFX ## NUM ## _mc13_altivec; \

  422.     c->PFX ## _pixels_tab[IDX][14] = PFX ## NUM ## _mc23_altivec; \

  423.     c->PFX ## _pixels_tab[IDX][15] = PFX ## NUM ## _mc33_altivec

  424. 

  425.     dspfunc(put_h264_qpel, 0, 16);

  426.     dspfunc(avg_h264_qpel, 0, 16);

  427. #undef dspfunc

  428. 

  429.   } else

  430. #endif /* HAVE_ALTIVEC */

  431.   {

  432.     // Non-AltiVec PPC optimisations

  433. 

  434.     // ... pending ...

  435.   }

  436. }