| 1 | /* |
| 2 | * Alpha optimized DSP utils |
| 3 | * Copyright (c) 2002 Falk Hueffner <falk@debian.org> |
| 4 | * |
| 5 | * This file is part of FFmpeg. |
| 6 | * |
| 7 | * FFmpeg is free software; you can redistribute it and/or |
| 8 | * modify it under the terms of the GNU Lesser General Public |
| 9 | * License as published by the Free Software Foundation; either |
| 10 | * version 2.1 of the License, or (at your option) any later version. |
| 11 | * |
| 12 | * FFmpeg is distributed in the hope that it will be useful, |
| 13 | * but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 14 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
| 15 | * Lesser General Public License for more details. |
| 16 | * |
| 17 | * You should have received a copy of the GNU Lesser General Public |
| 18 | * License along with FFmpeg; if not, write to the Free Software |
| 19 | * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA |
| 20 | */ |
| 21 | |
| 22 | #include "libavutil/attributes.h" |
| 23 | #include "libavcodec/hpeldsp.h" |
| 24 | #include "hpeldsp_alpha.h" |
| 25 | #include "asm.h" |
| 26 | |
| 27 | static inline uint64_t avg2_no_rnd(uint64_t a, uint64_t b) |
| 28 | { |
| 29 | return (a & b) + (((a ^ b) & BYTE_VEC(0xfe)) >> 1); |
| 30 | } |
| 31 | |
| 32 | static inline uint64_t avg2(uint64_t a, uint64_t b) |
| 33 | { |
| 34 | return (a | b) - (((a ^ b) & BYTE_VEC(0xfe)) >> 1); |
| 35 | } |
| 36 | |
| 37 | #if 0 |
| 38 | /* The XY2 routines basically utilize this scheme, but reuse parts in |
| 39 | each iteration. */ |
| 40 | static inline uint64_t avg4(uint64_t l1, uint64_t l2, uint64_t l3, uint64_t l4) |
| 41 | { |
| 42 | uint64_t r1 = ((l1 & ~BYTE_VEC(0x03)) >> 2) |
| 43 | + ((l2 & ~BYTE_VEC(0x03)) >> 2) |
| 44 | + ((l3 & ~BYTE_VEC(0x03)) >> 2) |
| 45 | + ((l4 & ~BYTE_VEC(0x03)) >> 2); |
| 46 | uint64_t r2 = (( (l1 & BYTE_VEC(0x03)) |
| 47 | + (l2 & BYTE_VEC(0x03)) |
| 48 | + (l3 & BYTE_VEC(0x03)) |
| 49 | + (l4 & BYTE_VEC(0x03)) |
| 50 | + BYTE_VEC(0x02)) >> 2) & BYTE_VEC(0x03); |
| 51 | return r1 + r2; |
| 52 | } |
| 53 | #endif |
| 54 | |
| 55 | #define OP(LOAD, STORE) \ |
| 56 | do { \ |
| 57 | STORE(LOAD(pixels), block); \ |
| 58 | pixels += line_size; \ |
| 59 | block += line_size; \ |
| 60 | } while (--h) |
| 61 | |
| 62 | #define OP_X2(LOAD, STORE) \ |
| 63 | do { \ |
| 64 | uint64_t pix1, pix2; \ |
| 65 | \ |
| 66 | pix1 = LOAD(pixels); \ |
| 67 | pix2 = pix1 >> 8 | ((uint64_t) pixels[8] << 56); \ |
| 68 | STORE(AVG2(pix1, pix2), block); \ |
| 69 | pixels += line_size; \ |
| 70 | block += line_size; \ |
| 71 | } while (--h) |
| 72 | |
| 73 | #define OP_Y2(LOAD, STORE) \ |
| 74 | do { \ |
| 75 | uint64_t pix = LOAD(pixels); \ |
| 76 | do { \ |
| 77 | uint64_t next_pix; \ |
| 78 | \ |
| 79 | pixels += line_size; \ |
| 80 | next_pix = LOAD(pixels); \ |
| 81 | STORE(AVG2(pix, next_pix), block); \ |
| 82 | block += line_size; \ |
| 83 | pix = next_pix; \ |
| 84 | } while (--h); \ |
| 85 | } while (0) |
| 86 | |
| 87 | #define OP_XY2(LOAD, STORE) \ |
| 88 | do { \ |
| 89 | uint64_t pix1 = LOAD(pixels); \ |
| 90 | uint64_t pix2 = pix1 >> 8 | ((uint64_t) pixels[8] << 56); \ |
| 91 | uint64_t pix_l = (pix1 & BYTE_VEC(0x03)) \ |
| 92 | + (pix2 & BYTE_VEC(0x03)); \ |
| 93 | uint64_t pix_h = ((pix1 & ~BYTE_VEC(0x03)) >> 2) \ |
| 94 | + ((pix2 & ~BYTE_VEC(0x03)) >> 2); \ |
| 95 | \ |
| 96 | do { \ |
| 97 | uint64_t npix1, npix2; \ |
| 98 | uint64_t npix_l, npix_h; \ |
| 99 | uint64_t avg; \ |
| 100 | \ |
| 101 | pixels += line_size; \ |
| 102 | npix1 = LOAD(pixels); \ |
| 103 | npix2 = npix1 >> 8 | ((uint64_t) pixels[8] << 56); \ |
| 104 | npix_l = (npix1 & BYTE_VEC(0x03)) \ |
| 105 | + (npix2 & BYTE_VEC(0x03)); \ |
| 106 | npix_h = ((npix1 & ~BYTE_VEC(0x03)) >> 2) \ |
| 107 | + ((npix2 & ~BYTE_VEC(0x03)) >> 2); \ |
| 108 | avg = (((pix_l + npix_l + AVG4_ROUNDER) >> 2) & BYTE_VEC(0x03)) \ |
| 109 | + pix_h + npix_h; \ |
| 110 | STORE(avg, block); \ |
| 111 | \ |
| 112 | block += line_size; \ |
| 113 | pix_l = npix_l; \ |
| 114 | pix_h = npix_h; \ |
| 115 | } while (--h); \ |
| 116 | } while (0) |
| 117 | |
| 118 | #define MAKE_OP(OPNAME, SUFF, OPKIND, STORE) \ |
| 119 | static void OPNAME ## _pixels ## SUFF ## _axp \ |
| 120 | (uint8_t *restrict block, const uint8_t *restrict pixels, \ |
| 121 | ptrdiff_t line_size, int h) \ |
| 122 | { \ |
| 123 | if ((size_t) pixels & 0x7) { \ |
| 124 | OPKIND(uldq, STORE); \ |
| 125 | } else { \ |
| 126 | OPKIND(ldq, STORE); \ |
| 127 | } \ |
| 128 | } \ |
| 129 | \ |
| 130 | static void OPNAME ## _pixels16 ## SUFF ## _axp \ |
| 131 | (uint8_t *restrict block, const uint8_t *restrict pixels, \ |
| 132 | ptrdiff_t line_size, int h) \ |
| 133 | { \ |
| 134 | OPNAME ## _pixels ## SUFF ## _axp(block, pixels, line_size, h); \ |
| 135 | OPNAME ## _pixels ## SUFF ## _axp(block + 8, pixels + 8, line_size, h); \ |
| 136 | } |
| 137 | |
| 138 | #define PIXOP(OPNAME, STORE) \ |
| 139 | MAKE_OP(OPNAME, , OP, STORE) \ |
| 140 | MAKE_OP(OPNAME, _x2, OP_X2, STORE) \ |
| 141 | MAKE_OP(OPNAME, _y2, OP_Y2, STORE) \ |
| 142 | MAKE_OP(OPNAME, _xy2, OP_XY2, STORE) |
| 143 | |
| 144 | /* Rounding primitives. */ |
| 145 | #define AVG2 avg2 |
| 146 | #define AVG4 avg4 |
| 147 | #define AVG4_ROUNDER BYTE_VEC(0x02) |
| 148 | #define STORE(l, b) stq(l, b) |
| 149 | PIXOP(put, STORE); |
| 150 | |
| 151 | #undef STORE |
| 152 | #define STORE(l, b) stq(AVG2(l, ldq(b)), b); |
| 153 | PIXOP(avg, STORE); |
| 154 | |
| 155 | /* Not rounding primitives. */ |
| 156 | #undef AVG2 |
| 157 | #undef AVG4 |
| 158 | #undef AVG4_ROUNDER |
| 159 | #undef STORE |
| 160 | #define AVG2 avg2_no_rnd |
| 161 | #define AVG4 avg4_no_rnd |
| 162 | #define AVG4_ROUNDER BYTE_VEC(0x01) |
| 163 | #define STORE(l, b) stq(l, b) |
| 164 | PIXOP(put_no_rnd, STORE); |
| 165 | |
| 166 | #undef STORE |
| 167 | #define STORE(l, b) stq(AVG2(l, ldq(b)), b); |
| 168 | PIXOP(avg_no_rnd, STORE); |
| 169 | |
| 170 | static void put_pixels16_axp_asm(uint8_t *block, const uint8_t *pixels, |
| 171 | ptrdiff_t line_size, int h) |
| 172 | { |
| 173 | put_pixels_axp_asm(block, pixels, line_size, h); |
| 174 | put_pixels_axp_asm(block + 8, pixels + 8, line_size, h); |
| 175 | } |
| 176 | |
| 177 | av_cold void ff_hpeldsp_init_alpha(HpelDSPContext *c, int flags) |
| 178 | { |
| 179 | c->put_pixels_tab[0][0] = put_pixels16_axp_asm; |
| 180 | c->put_pixels_tab[0][1] = put_pixels16_x2_axp; |
| 181 | c->put_pixels_tab[0][2] = put_pixels16_y2_axp; |
| 182 | c->put_pixels_tab[0][3] = put_pixels16_xy2_axp; |
| 183 | |
| 184 | c->put_no_rnd_pixels_tab[0][0] = put_pixels16_axp_asm; |
| 185 | c->put_no_rnd_pixels_tab[0][1] = put_no_rnd_pixels16_x2_axp; |
| 186 | c->put_no_rnd_pixels_tab[0][2] = put_no_rnd_pixels16_y2_axp; |
| 187 | c->put_no_rnd_pixels_tab[0][3] = put_no_rnd_pixels16_xy2_axp; |
| 188 | |
| 189 | c->avg_pixels_tab[0][0] = avg_pixels16_axp; |
| 190 | c->avg_pixels_tab[0][1] = avg_pixels16_x2_axp; |
| 191 | c->avg_pixels_tab[0][2] = avg_pixels16_y2_axp; |
| 192 | c->avg_pixels_tab[0][3] = avg_pixels16_xy2_axp; |
| 193 | |
| 194 | c->avg_no_rnd_pixels_tab[0] = avg_no_rnd_pixels16_axp; |
| 195 | c->avg_no_rnd_pixels_tab[1] = avg_no_rnd_pixels16_x2_axp; |
| 196 | c->avg_no_rnd_pixels_tab[2] = avg_no_rnd_pixels16_y2_axp; |
| 197 | c->avg_no_rnd_pixels_tab[3] = avg_no_rnd_pixels16_xy2_axp; |
| 198 | |
| 199 | c->put_pixels_tab[1][0] = put_pixels_axp_asm; |
| 200 | c->put_pixels_tab[1][1] = put_pixels_x2_axp; |
| 201 | c->put_pixels_tab[1][2] = put_pixels_y2_axp; |
| 202 | c->put_pixels_tab[1][3] = put_pixels_xy2_axp; |
| 203 | |
| 204 | c->put_no_rnd_pixels_tab[1][0] = put_pixels_axp_asm; |
| 205 | c->put_no_rnd_pixels_tab[1][1] = put_no_rnd_pixels_x2_axp; |
| 206 | c->put_no_rnd_pixels_tab[1][2] = put_no_rnd_pixels_y2_axp; |
| 207 | c->put_no_rnd_pixels_tab[1][3] = put_no_rnd_pixels_xy2_axp; |
| 208 | |
| 209 | c->avg_pixels_tab[1][0] = avg_pixels_axp; |
| 210 | c->avg_pixels_tab[1][1] = avg_pixels_x2_axp; |
| 211 | c->avg_pixels_tab[1][2] = avg_pixels_y2_axp; |
| 212 | c->avg_pixels_tab[1][3] = avg_pixels_xy2_axp; |
| 213 | } |