Mercurial > cgi-bin > hgwebdir.cgi > PR > Applications > VSs > VSs__H264__App

annotate libavcodec/arm/dsputil_iwmmxt.c @ 6:55fb61482128

find changesets by author, revision, files, or words in the commit message
VSs working
author Nina Engelhardt <nengel@mailbox.tu-berlin.de>
date Wed, 06 Mar 2013 14:35:39 +0100
parents
children
rev   line source
nengel@2 1 /*
nengel@2 2 * iWMMXt optimized DSP utils
nengel@2 3 * Copyright (c) 2004 AGAWA Koji
nengel@2 4 *
nengel@2 5 * This file is part of FFmpeg.
nengel@2 6 *
nengel@2 7 * FFmpeg is free software; you can redistribute it and/or
nengel@2 8 * modify it under the terms of the GNU Lesser General Public
nengel@2 9 * License as published by the Free Software Foundation; either
nengel@2 10 * version 2.1 of the License, or (at your option) any later version.
nengel@2 11 *
nengel@2 12 * FFmpeg is distributed in the hope that it will be useful,
nengel@2 13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
nengel@2 14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
nengel@2 15 * Lesser General Public License for more details.
nengel@2 16 *
nengel@2 17 * You should have received a copy of the GNU Lesser General Public
nengel@2 18 * License along with FFmpeg; if not, write to the Free Software
nengel@2 19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
nengel@2 20 */
nengel@2 21
nengel@2 22 #include "libavcodec/dsputil.h"
nengel@2 23
nengel@2 24 #define DEF(x, y) x ## _no_rnd_ ## y ##_iwmmxt
nengel@2 25 #define SET_RND(regd) __asm__ volatile ("mov r12, #1 \n\t tbcsth " #regd ", r12":::"r12");
nengel@2 26 #define WAVG2B "wavg2b"
nengel@2 27 #include "dsputil_iwmmxt_rnd_template.c"
nengel@2 28 #undef DEF
nengel@2 29 #undef SET_RND
nengel@2 30 #undef WAVG2B
nengel@2 31
nengel@2 32 #define DEF(x, y) x ## _ ## y ##_iwmmxt
nengel@2 33 #define SET_RND(regd) __asm__ volatile ("mov r12, #2 \n\t tbcsth " #regd ", r12":::"r12");
nengel@2 34 #define WAVG2B "wavg2br"
nengel@2 35 #include "dsputil_iwmmxt_rnd_template.c"
nengel@2 36 #undef DEF
nengel@2 37 #undef SET_RND
nengel@2 38 #undef WAVG2BR
nengel@2 39
nengel@2 40 // need scheduling
nengel@2 41 #define OP(AVG) \
nengel@2 42 __asm__ volatile ( \
nengel@2 43 /* alignment */ \
nengel@2 44 "and r12, %[pixels], #7 \n\t" \
nengel@2 45 "bic %[pixels], %[pixels], #7 \n\t" \
nengel@2 46 "tmcr wcgr1, r12 \n\t" \
nengel@2 47 \
nengel@2 48 "wldrd wr0, [%[pixels]] \n\t" \
nengel@2 49 "wldrd wr1, [%[pixels], #8] \n\t" \
nengel@2 50 "add %[pixels], %[pixels], %[line_size] \n\t" \
nengel@2 51 "walignr1 wr4, wr0, wr1 \n\t" \
nengel@2 52 \
nengel@2 53 "1: \n\t" \
nengel@2 54 \
nengel@2 55 "wldrd wr2, [%[pixels]] \n\t" \
nengel@2 56 "wldrd wr3, [%[pixels], #8] \n\t" \
nengel@2 57 "add %[pixels], %[pixels], %[line_size] \n\t" \
nengel@2 58 "pld [%[pixels]] \n\t" \
nengel@2 59 "walignr1 wr5, wr2, wr3 \n\t" \
nengel@2 60 AVG " wr6, wr4, wr5 \n\t" \
nengel@2 61 "wstrd wr6, [%[block]] \n\t" \
nengel@2 62 "add %[block], %[block], %[line_size] \n\t" \
nengel@2 63 \
nengel@2 64 "wldrd wr0, [%[pixels]] \n\t" \
nengel@2 65 "wldrd wr1, [%[pixels], #8] \n\t" \
nengel@2 66 "add %[pixels], %[pixels], %[line_size] \n\t" \
nengel@2 67 "walignr1 wr4, wr0, wr1 \n\t" \
nengel@2 68 "pld [%[pixels]] \n\t" \
nengel@2 69 AVG " wr6, wr4, wr5 \n\t" \
nengel@2 70 "wstrd wr6, [%[block]] \n\t" \
nengel@2 71 "add %[block], %[block], %[line_size] \n\t" \
nengel@2 72 \
nengel@2 73 "subs %[h], %[h], #2 \n\t" \
nengel@2 74 "bne 1b \n\t" \
nengel@2 75 : [block]"+r"(block), [pixels]"+r"(pixels), [h]"+r"(h) \
nengel@2 76 : [line_size]"r"(line_size) \
nengel@2 77 : "memory", "r12");
nengel@2 78 void put_pixels8_y2_iwmmxt(uint8_t *block, const uint8_t *pixels, const int line_size, int h)
nengel@2 79 {
nengel@2 80 OP("wavg2br");
nengel@2 81 }
nengel@2 82 void put_no_rnd_pixels8_y2_iwmmxt(uint8_t *block, const uint8_t *pixels, const int line_size, int h)
nengel@2 83 {
nengel@2 84 OP("wavg2b");
nengel@2 85 }
nengel@2 86 #undef OP
nengel@2 87
nengel@2 88 void add_pixels_clamped_iwmmxt(const DCTELEM *block, uint8_t *pixels, int line_size)
nengel@2 89 {
nengel@2 90 uint8_t *pixels2 = pixels + line_size;
nengel@2 91
nengel@2 92 __asm__ volatile (
nengel@2 93 "mov r12, #4 \n\t"
nengel@2 94 "1: \n\t"
nengel@2 95 "pld [%[pixels], %[line_size2]] \n\t"
nengel@2 96 "pld [%[pixels2], %[line_size2]] \n\t"
nengel@2 97 "wldrd wr4, [%[pixels]] \n\t"
nengel@2 98 "wldrd wr5, [%[pixels2]] \n\t"
nengel@2 99 "pld [%[block], #32] \n\t"
nengel@2 100 "wunpckelub wr6, wr4 \n\t"
nengel@2 101 "wldrd wr0, [%[block]] \n\t"
nengel@2 102 "wunpckehub wr7, wr4 \n\t"
nengel@2 103 "wldrd wr1, [%[block], #8] \n\t"
nengel@2 104 "wunpckelub wr8, wr5 \n\t"
nengel@2 105 "wldrd wr2, [%[block], #16] \n\t"
nengel@2 106 "wunpckehub wr9, wr5 \n\t"
nengel@2 107 "wldrd wr3, [%[block], #24] \n\t"
nengel@2 108 "add %[block], %[block], #32 \n\t"
nengel@2 109 "waddhss wr10, wr0, wr6 \n\t"
nengel@2 110 "waddhss wr11, wr1, wr7 \n\t"
nengel@2 111 "waddhss wr12, wr2, wr8 \n\t"
nengel@2 112 "waddhss wr13, wr3, wr9 \n\t"
nengel@2 113 "wpackhus wr14, wr10, wr11 \n\t"
nengel@2 114 "wpackhus wr15, wr12, wr13 \n\t"
nengel@2 115 "wstrd wr14, [%[pixels]] \n\t"
nengel@2 116 "add %[pixels], %[pixels], %[line_size2] \n\t"
nengel@2 117 "subs r12, r12, #1 \n\t"
nengel@2 118 "wstrd wr15, [%[pixels2]] \n\t"
nengel@2 119 "add %[pixels2], %[pixels2], %[line_size2] \n\t"
nengel@2 120 "bne 1b \n\t"
nengel@2 121 : [block]"+r"(block), [pixels]"+r"(pixels), [pixels2]"+r"(pixels2)
nengel@2 122 : [line_size2]"r"(line_size << 1)
nengel@2 123 : "cc", "memory", "r12");
nengel@2 124 }
nengel@2 125
nengel@2 126 static void clear_blocks_iwmmxt(DCTELEM *blocks)
nengel@2 127 {
nengel@2 128 __asm__ volatile(
nengel@2 129 "wzero wr0 \n\t"
nengel@2 130 "mov r1, #(128 * 6 / 32) \n\t"
nengel@2 131 "1: \n\t"
nengel@2 132 "wstrd wr0, [%0] \n\t"
nengel@2 133 "wstrd wr0, [%0, #8] \n\t"
nengel@2 134 "wstrd wr0, [%0, #16] \n\t"
nengel@2 135 "wstrd wr0, [%0, #24] \n\t"
nengel@2 136 "subs r1, r1, #1 \n\t"
nengel@2 137 "add %0, %0, #32 \n\t"
nengel@2 138 "bne 1b \n\t"
nengel@2 139 : "+r"(blocks)
nengel@2 140 :
nengel@2 141 : "r1"
nengel@2 142 );
nengel@2 143 }
nengel@2 144
nengel@2 145 static void nop(uint8_t *block, const uint8_t *pixels, int line_size, int h)
nengel@2 146 {
nengel@2 147 return;
nengel@2 148 }
nengel@2 149
nengel@2 150 /* A run time test is not simple. If this file is compiled in
nengel@2 151 * then we should install the functions
nengel@2 152 */
nengel@2 153 int mm_flags = FF_MM_IWMMXT; /* multimedia extension flags */
nengel@2 154
nengel@2 155 void ff_dsputil_init_iwmmxt(DSPContext* c, AVCodecContext *avctx)
nengel@2 156 {
nengel@2 157 if (avctx->dsp_mask) {
nengel@2 158 if (avctx->dsp_mask & FF_MM_FORCE)
nengel@2 159 mm_flags |= (avctx->dsp_mask & 0xffff);
nengel@2 160 else
nengel@2 161 mm_flags &= ~(avctx->dsp_mask & 0xffff);
nengel@2 162 }
nengel@2 163
nengel@2 164 if (!(mm_flags & FF_MM_IWMMXT)) return;
nengel@2 165
nengel@2 166 c->add_pixels_clamped = add_pixels_clamped_iwmmxt;
nengel@2 167
nengel@2 168 c->clear_blocks = clear_blocks_iwmmxt;
nengel@2 169
nengel@2 170 c->put_pixels_tab[0][0] = put_pixels16_iwmmxt;
nengel@2 171 c->put_pixels_tab[0][1] = put_pixels16_x2_iwmmxt;
nengel@2 172 c->put_pixels_tab[0][2] = put_pixels16_y2_iwmmxt;
nengel@2 173 c->put_pixels_tab[0][3] = put_pixels16_xy2_iwmmxt;
nengel@2 174 c->put_no_rnd_pixels_tab[0][0] = put_pixels16_iwmmxt;
nengel@2 175 c->put_no_rnd_pixels_tab[0][1] = put_no_rnd_pixels16_x2_iwmmxt;
nengel@2 176 c->put_no_rnd_pixels_tab[0][2] = put_no_rnd_pixels16_y2_iwmmxt;
nengel@2 177 c->put_no_rnd_pixels_tab[0][3] = put_no_rnd_pixels16_xy2_iwmmxt;
nengel@2 178
nengel@2 179 c->put_pixels_tab[1][0] = put_pixels8_iwmmxt;
nengel@2 180 c->put_pixels_tab[1][1] = put_pixels8_x2_iwmmxt;
nengel@2 181 c->put_pixels_tab[1][2] = put_pixels8_y2_iwmmxt;
nengel@2 182 c->put_pixels_tab[1][3] = put_pixels8_xy2_iwmmxt;
nengel@2 183 c->put_no_rnd_pixels_tab[1][0] = put_pixels8_iwmmxt;
nengel@2 184 c->put_no_rnd_pixels_tab[1][1] = put_no_rnd_pixels8_x2_iwmmxt;
nengel@2 185 c->put_no_rnd_pixels_tab[1][2] = put_no_rnd_pixels8_y2_iwmmxt;
nengel@2 186 c->put_no_rnd_pixels_tab[1][3] = put_no_rnd_pixels8_xy2_iwmmxt;
nengel@2 187
nengel@2 188 c->avg_pixels_tab[0][0] = avg_pixels16_iwmmxt;
nengel@2 189 c->avg_pixels_tab[0][1] = avg_pixels16_x2_iwmmxt;
nengel@2 190 c->avg_pixels_tab[0][2] = avg_pixels16_y2_iwmmxt;
nengel@2 191 c->avg_pixels_tab[0][3] = avg_pixels16_xy2_iwmmxt;
nengel@2 192 c->avg_no_rnd_pixels_tab[0][0] = avg_pixels16_iwmmxt;
nengel@2 193 c->avg_no_rnd_pixels_tab[0][1] = avg_no_rnd_pixels16_x2_iwmmxt;
nengel@2 194 c->avg_no_rnd_pixels_tab[0][2] = avg_no_rnd_pixels16_y2_iwmmxt;
nengel@2 195 c->avg_no_rnd_pixels_tab[0][3] = avg_no_rnd_pixels16_xy2_iwmmxt;
nengel@2 196
nengel@2 197 c->avg_pixels_tab[1][0] = avg_pixels8_iwmmxt;
nengel@2 198 c->avg_pixels_tab[1][1] = avg_pixels8_x2_iwmmxt;
nengel@2 199 c->avg_pixels_tab[1][2] = avg_pixels8_y2_iwmmxt;
nengel@2 200 c->avg_pixels_tab[1][3] = avg_pixels8_xy2_iwmmxt;
nengel@2 201 c->avg_no_rnd_pixels_tab[1][0] = avg_no_rnd_pixels8_iwmmxt;
nengel@2 202 c->avg_no_rnd_pixels_tab[1][1] = avg_no_rnd_pixels8_x2_iwmmxt;
nengel@2 203 c->avg_no_rnd_pixels_tab[1][2] = avg_no_rnd_pixels8_y2_iwmmxt;
nengel@2 204 c->avg_no_rnd_pixels_tab[1][3] = avg_no_rnd_pixels8_xy2_iwmmxt;
nengel@2 205 }