| 1 | /* |
| 2 | * This file is part of FFmpeg. |
| 3 | * |
| 4 | * FFmpeg 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 2.1 of the License, or (at your option) any later version. |
| 8 | * |
| 9 | * FFmpeg 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 FFmpeg; if not, write to the Free Software |
| 16 | * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA |
| 17 | */ |
| 18 | |
| 19 | #include "config.h" |
| 20 | #include <stdint.h> |
| 21 | #if HAVE_ALTIVEC_H |
| 22 | #include <altivec.h> |
| 23 | #endif |
| 24 | |
| 25 | #include "libavutil/attributes.h" |
| 26 | #include "libavutil/cpu.h" |
| 27 | #include "libavutil/ppc/cpu.h" |
| 28 | #include "libavutil/ppc/types_altivec.h" |
| 29 | #include "libavutil/ppc/util_altivec.h" |
| 30 | #include "libavcodec/mpegvideoencdsp.h" |
| 31 | |
| 32 | #if HAVE_ALTIVEC |
| 33 | |
| 34 | #if HAVE_VSX |
| 35 | static int pix_norm1_altivec(uint8_t *pix, int line_size) |
| 36 | { |
| 37 | int i, s = 0; |
| 38 | const vector unsigned int zero = |
| 39 | (const vector unsigned int) vec_splat_u32(0); |
| 40 | vector unsigned int sv = (vector unsigned int) vec_splat_u32(0); |
| 41 | vector signed int sum; |
| 42 | |
| 43 | for (i = 0; i < 16; i++) { |
| 44 | /* Read the potentially unaligned pixels. */ |
| 45 | //vector unsigned char pixl = vec_ld(0, pix); |
| 46 | //vector unsigned char pixr = vec_ld(15, pix); |
| 47 | //vector unsigned char pixv = vec_perm(pixl, pixr, perm); |
| 48 | vector unsigned char pixv = vec_vsx_ld(0, pix); |
| 49 | |
| 50 | /* Square the values, and add them to our sum. */ |
| 51 | sv = vec_msum(pixv, pixv, sv); |
| 52 | |
| 53 | pix += line_size; |
| 54 | } |
| 55 | /* Sum up the four partial sums, and put the result into s. */ |
| 56 | sum = vec_sums((vector signed int) sv, (vector signed int) zero); |
| 57 | sum = vec_splat(sum, 3); |
| 58 | vec_vsx_st(sum, 0, &s); |
| 59 | return s; |
| 60 | } |
| 61 | #else |
| 62 | static int pix_norm1_altivec(uint8_t *pix, int line_size) |
| 63 | { |
| 64 | int i, s = 0; |
| 65 | const vector unsigned int zero = |
| 66 | (const vector unsigned int) vec_splat_u32(0); |
| 67 | vector unsigned char perm = vec_lvsl(0, pix); |
| 68 | vector unsigned int sv = (vector unsigned int) vec_splat_u32(0); |
| 69 | vector signed int sum; |
| 70 | |
| 71 | for (i = 0; i < 16; i++) { |
| 72 | /* Read the potentially unaligned pixels. */ |
| 73 | vector unsigned char pixl = vec_ld(0, pix); |
| 74 | vector unsigned char pixr = vec_ld(15, pix); |
| 75 | vector unsigned char pixv = vec_perm(pixl, pixr, perm); |
| 76 | |
| 77 | /* Square the values, and add them to our sum. */ |
| 78 | sv = vec_msum(pixv, pixv, sv); |
| 79 | |
| 80 | pix += line_size; |
| 81 | } |
| 82 | /* Sum up the four partial sums, and put the result into s. */ |
| 83 | sum = vec_sums((vector signed int) sv, (vector signed int) zero); |
| 84 | sum = vec_splat(sum, 3); |
| 85 | vec_ste(sum, 0, &s); |
| 86 | |
| 87 | return s; |
| 88 | } |
| 89 | #endif /* HAVE_VSX */ |
| 90 | |
| 91 | #if HAVE_VSX |
| 92 | static int pix_sum_altivec(uint8_t *pix, int line_size) |
| 93 | { |
| 94 | int i, s; |
| 95 | const vector unsigned int zero = |
| 96 | (const vector unsigned int) vec_splat_u32(0); |
| 97 | vector unsigned int sad = (vector unsigned int) vec_splat_u32(0); |
| 98 | vector signed int sumdiffs; |
| 99 | |
| 100 | for (i = 0; i < 16; i++) { |
| 101 | /* Read the potentially unaligned 16 pixels into t1. */ |
| 102 | //vector unsigned char pixl = vec_ld(0, pix); |
| 103 | //vector unsigned char pixr = vec_ld(15, pix); |
| 104 | //vector unsigned char t1 = vec_perm(pixl, pixr, perm); |
| 105 | vector unsigned char t1 = vec_vsx_ld(0, pix); |
| 106 | |
| 107 | /* Add each 4 pixel group together and put 4 results into sad. */ |
| 108 | sad = vec_sum4s(t1, sad); |
| 109 | |
| 110 | pix += line_size; |
| 111 | } |
| 112 | |
| 113 | /* Sum up the four partial sums, and put the result into s. */ |
| 114 | sumdiffs = vec_sums((vector signed int) sad, (vector signed int) zero); |
| 115 | sumdiffs = vec_splat(sumdiffs, 3); |
| 116 | vec_vsx_st(sumdiffs, 0, &s); |
| 117 | return s; |
| 118 | } |
| 119 | #else |
| 120 | static int pix_sum_altivec(uint8_t *pix, int line_size) |
| 121 | { |
| 122 | int i, s; |
| 123 | const vector unsigned int zero = |
| 124 | (const vector unsigned int) vec_splat_u32(0); |
| 125 | vector unsigned char perm = vec_lvsl(0, pix); |
| 126 | vector unsigned int sad = (vector unsigned int) vec_splat_u32(0); |
| 127 | vector signed int sumdiffs; |
| 128 | |
| 129 | for (i = 0; i < 16; i++) { |
| 130 | /* Read the potentially unaligned 16 pixels into t1. */ |
| 131 | vector unsigned char pixl = vec_ld(0, pix); |
| 132 | vector unsigned char pixr = vec_ld(15, pix); |
| 133 | vector unsigned char t1 = vec_perm(pixl, pixr, perm); |
| 134 | |
| 135 | /* Add each 4 pixel group together and put 4 results into sad. */ |
| 136 | sad = vec_sum4s(t1, sad); |
| 137 | |
| 138 | pix += line_size; |
| 139 | } |
| 140 | |
| 141 | /* Sum up the four partial sums, and put the result into s. */ |
| 142 | sumdiffs = vec_sums((vector signed int) sad, (vector signed int) zero); |
| 143 | sumdiffs = vec_splat(sumdiffs, 3); |
| 144 | vec_ste(sumdiffs, 0, &s); |
| 145 | |
| 146 | return s; |
| 147 | } |
| 148 | |
| 149 | #endif /* HAVE_VSX */ |
| 150 | |
| 151 | #endif /* HAVE_ALTIVEC */ |
| 152 | |
| 153 | av_cold void ff_mpegvideoencdsp_init_ppc(MpegvideoEncDSPContext *c, |
| 154 | AVCodecContext *avctx) |
| 155 | { |
| 156 | #if HAVE_ALTIVEC |
| 157 | if (!PPC_ALTIVEC(av_get_cpu_flags())) |
| 158 | return; |
| 159 | |
| 160 | c->pix_norm1 = pix_norm1_altivec; |
| 161 | c->pix_sum = pix_sum_altivec; |
| 162 | #endif /* HAVE_ALTIVEC */ |
| 163 | } |