MythTV: dsputil_iwmmxt_rnd.h Source File

00001 /*
00002  * iWMMXt optimized DSP utils
00003  * copyright (c) 2004 AGAWA Koji
00004  *
00005  * This file is part of FFmpeg.
00006  *
00007  * FFmpeg is free software; you can redistribute it and/or
00008  * modify it under the terms of the GNU Lesser General Public
00009  * License as published by the Free Software Foundation; either
00010  * version 2.1 of the License, or (at your option) any later version.
00011  *
00012  * FFmpeg is distributed in the hope that it will be useful,
00013  * but WITHOUT ANY WARRANTY; without even the implied warranty of
00014  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
00015  * Lesser General Public License for more details.
00016  *
00017  * You should have received a copy of the GNU Lesser General Public
00018  * License along with FFmpeg; if not, write to the Free Software
00019  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
00020  */
00021 
00022 #ifndef FFMPEG_DSPUTIL_IWMMXT_RND_H
00023 #define FFMPEG_DSPUTIL_IWMMXT_RND_H
00024 
00025 void DEF(put, pixels8)(uint8_t *block, const uint8_t *pixels, const int line_size, int h)
00026 {
00027     int stride = line_size;
00028     __asm__ __volatile__ (
00029         "and r12, %[pixels], #7 \n\t"
00030         "bic %[pixels], %[pixels], #7 \n\t"
00031         "tmcr wcgr1, r12 \n\t"
00032         "add r4, %[pixels], %[line_size] \n\t"
00033         "add r5, %[block], %[line_size] \n\t"
00034         "mov %[line_size], %[line_size], lsl #1 \n\t"
00035         "1: \n\t"
00036         "wldrd wr0, [%[pixels]] \n\t"
00037         "subs %[h], %[h], #2 \n\t"
00038         "wldrd wr1, [%[pixels], #8] \n\t"
00039         "add %[pixels], %[pixels], %[line_size] \n\t"
00040         "wldrd wr3, [r4] \n\t"
00041         "pld [%[pixels]] \n\t"
00042         "pld [%[pixels], #32] \n\t"
00043         "wldrd wr4, [r4, #8] \n\t"
00044         "add r4, r4, %[line_size] \n\t"
00045         "walignr1 wr8, wr0, wr1 \n\t"
00046         "pld [r4] \n\t"
00047         "pld [r4, #32] \n\t"
00048         "walignr1 wr10, wr3, wr4 \n\t"
00049         "wstrd wr8, [%[block]] \n\t"
00050         "add %[block], %[block], %[line_size] \n\t"
00051         "wstrd wr10, [r5] \n\t"
00052         "add r5, r5, %[line_size] \n\t"
00053         "bne 1b \n\t"
00054         : [block]"+r"(block), [pixels]"+r"(pixels), [line_size]"+r"(stride), [h]"+r"(h)
00055         :
00056         : "memory", "r4", "r5", "r12");
00057 }
00058 
00059 void DEF(avg, pixels8)(uint8_t *block, const uint8_t *pixels, const int line_size, int h)
00060 {
00061     int stride = line_size;
00062     __asm__ __volatile__ (
00063         "and r12, %[pixels], #7 \n\t"
00064         "bic %[pixels], %[pixels], #7 \n\t"
00065         "tmcr wcgr1, r12 \n\t"
00066         "add r4, %[pixels], %[line_size] \n\t"
00067         "add r5, %[block], %[line_size] \n\t"
00068         "mov %[line_size], %[line_size], lsl #1 \n\t"
00069         "1: \n\t"
00070         "wldrd wr0, [%[pixels]] \n\t"
00071         "subs %[h], %[h], #2 \n\t"
00072         "wldrd wr1, [%[pixels], #8] \n\t"
00073         "add %[pixels], %[pixels], %[line_size] \n\t"
00074         "wldrd wr3, [r4] \n\t"
00075         "pld [%[pixels]] \n\t"
00076         "pld [%[pixels], #32] \n\t"
00077         "wldrd wr4, [r4, #8] \n\t"
00078         "add r4, r4, %[line_size] \n\t"
00079         "walignr1 wr8, wr0, wr1 \n\t"
00080         "wldrd wr0, [%[block]] \n\t"
00081         "wldrd wr2, [r5] \n\t"
00082         "pld [r4] \n\t"
00083         "pld [r4, #32] \n\t"
00084         "walignr1 wr10, wr3, wr4 \n\t"
00085         WAVG2B" wr8, wr8, wr0 \n\t"
00086         WAVG2B" wr10, wr10, wr2 \n\t"
00087         "wstrd wr8, [%[block]] \n\t"
00088         "add %[block], %[block], %[line_size] \n\t"
00089         "wstrd wr10, [r5] \n\t"
00090         "pld [%[block]] \n\t"
00091         "pld [%[block], #32] \n\t"
00092         "add r5, r5, %[line_size] \n\t"
00093         "pld [r5] \n\t"
00094         "pld [r5, #32] \n\t"
00095         "bne 1b \n\t"
00096         : [block]"+r"(block), [pixels]"+r"(pixels), [line_size]"+r"(stride), [h]"+r"(h)
00097         :
00098         : "memory", "r4", "r5", "r12");
00099 }
00100 
00101 void DEF(put, pixels16)(uint8_t *block, const uint8_t *pixels, const int line_size, int h)
00102 {
00103     int stride = line_size;
00104     __asm__ __volatile__ (
00105         "and r12, %[pixels], #7 \n\t"
00106         "bic %[pixels], %[pixels], #7 \n\t"
00107         "tmcr wcgr1, r12 \n\t"
00108         "add r4, %[pixels], %[line_size] \n\t"
00109         "add r5, %[block], %[line_size] \n\t"
00110         "mov %[line_size], %[line_size], lsl #1 \n\t"
00111         "1: \n\t"
00112         "wldrd wr0, [%[pixels]] \n\t"
00113         "wldrd wr1, [%[pixels], #8] \n\t"
00114         "subs %[h], %[h], #2 \n\t"
00115         "wldrd wr2, [%[pixels], #16] \n\t"
00116         "add %[pixels], %[pixels], %[line_size] \n\t"
00117         "wldrd wr3, [r4] \n\t"
00118         "pld [%[pixels]] \n\t"
00119         "pld [%[pixels], #32] \n\t"
00120         "walignr1 wr8, wr0, wr1 \n\t"
00121         "wldrd wr4, [r4, #8] \n\t"
00122         "walignr1 wr9, wr1, wr2 \n\t"
00123         "wldrd wr5, [r4, #16] \n\t"
00124         "add r4, r4, %[line_size] \n\t"
00125         "pld [r4] \n\t"
00126         "pld [r4, #32] \n\t"
00127         "walignr1 wr10, wr3, wr4 \n\t"
00128         "wstrd wr8, [%[block]] \n\t"
00129         "walignr1 wr11, wr4, wr5 \n\t"
00130         "wstrd wr9, [%[block], #8] \n\t"
00131         "add %[block], %[block], %[line_size] \n\t"
00132         "wstrd wr10, [r5] \n\t"
00133         "wstrd wr11, [r5, #8] \n\t"
00134         "add r5, r5, %[line_size] \n\t"
00135         "bne 1b \n\t"
00136         : [block]"+r"(block), [pixels]"+r"(pixels), [line_size]"+r"(stride), [h]"+r"(h)
00137         :
00138         : "memory", "r4", "r5", "r12");
00139 }
00140 
00141 void DEF(avg, pixels16)(uint8_t *block, const uint8_t *pixels, const int line_size, int h)
00142 {
00143     int stride = line_size;
00144     __asm__ __volatile__ (
00145         "pld [%[pixels]]                \n\t"
00146         "pld [%[pixels], #32]           \n\t"
00147         "pld [%[block]]                 \n\t"
00148         "pld [%[block], #32]            \n\t"
00149         "and r12, %[pixels], #7         \n\t"
00150         "bic %[pixels], %[pixels], #7   \n\t"
00151         "tmcr wcgr1, r12                \n\t"
00152         "add r4, %[pixels], %[line_size]\n\t"
00153         "add r5, %[block], %[line_size] \n\t"
00154         "mov %[line_size], %[line_size], lsl #1 \n\t"
00155         "1:                             \n\t"
00156         "wldrd wr0, [%[pixels]]         \n\t"
00157         "wldrd wr1, [%[pixels], #8]     \n\t"
00158         "subs %[h], %[h], #2            \n\t"
00159         "wldrd wr2, [%[pixels], #16]    \n\t"
00160         "add %[pixels], %[pixels], %[line_size] \n\t"
00161         "wldrd wr3, [r4]                \n\t"
00162         "pld [%[pixels]]                \n\t"
00163         "pld [%[pixels], #32]           \n\t"
00164         "walignr1 wr8, wr0, wr1         \n\t"
00165         "wldrd wr4, [r4, #8]            \n\t"
00166         "walignr1 wr9, wr1, wr2         \n\t"
00167         "wldrd wr5, [r4, #16]           \n\t"
00168         "add r4, r4, %[line_size]       \n\t"
00169         "wldrd wr0, [%[block]]          \n\t"
00170         "pld [r4]                       \n\t"
00171         "wldrd wr1, [%[block], #8]      \n\t"
00172         "pld [r4, #32]                  \n\t"
00173         "wldrd wr2, [r5]                \n\t"
00174         "walignr1 wr10, wr3, wr4        \n\t"
00175         "wldrd wr3, [r5, #8]            \n\t"
00176         WAVG2B" wr8, wr8, wr0           \n\t"
00177         WAVG2B" wr9, wr9, wr1           \n\t"
00178         WAVG2B" wr10, wr10, wr2         \n\t"
00179         "wstrd wr8, [%[block]]          \n\t"
00180         "walignr1 wr11, wr4, wr5        \n\t"
00181         WAVG2B" wr11, wr11, wr3         \n\t"
00182         "wstrd wr9, [%[block], #8]      \n\t"
00183         "add %[block], %[block], %[line_size] \n\t"
00184         "wstrd wr10, [r5]               \n\t"
00185         "pld [%[block]]                 \n\t"
00186         "pld [%[block], #32]            \n\t"
00187         "wstrd wr11, [r5, #8]           \n\t"
00188         "add r5, r5, %[line_size]       \n\t"
00189         "pld [r5]                       \n\t"
00190         "pld [r5, #32]                  \n\t"
00191         "bne 1b \n\t"
00192         : [block]"+r"(block), [pixels]"+r"(pixels), [line_size]"+r"(stride), [h]"+r"(h)
00193         :
00194         : "memory", "r4", "r5", "r12");
00195 }
00196 
00197 void DEF(put, pixels8_x2)(uint8_t *block, const uint8_t *pixels, const int line_size, int h)
00198 {
00199     int stride = line_size;
00200     // [wr0 wr1 wr2 wr3] for previous line
00201     // [wr4 wr5 wr6 wr7] for current line
00202     SET_RND(wr15); // =2 for rnd  and  =1 for no_rnd version
00203     __asm__ __volatile__(
00204         "pld [%[pixels]]                \n\t"
00205         "pld [%[pixels], #32]           \n\t"
00206         "and r12, %[pixels], #7         \n\t"
00207         "bic %[pixels], %[pixels], #7   \n\t"
00208         "tmcr wcgr1, r12                \n\t"
00209         "add r12, r12, #1               \n\t"
00210         "add r4, %[pixels], %[line_size]\n\t"
00211         "tmcr wcgr2, r12                \n\t"
00212         "add r5, %[block], %[line_size] \n\t"
00213         "mov %[line_size], %[line_size], lsl #1 \n\t"
00214 
00215         "1:                             \n\t"
00216         "wldrd wr10, [%[pixels]]        \n\t"
00217         "cmp r12, #8                    \n\t"
00218         "wldrd wr11, [%[pixels], #8]    \n\t"
00219         "add %[pixels], %[pixels], %[line_size] \n\t"
00220         "wldrd wr13, [r4]               \n\t"
00221         "pld [%[pixels]]                \n\t"
00222         "wldrd wr14, [r4, #8]           \n\t"
00223         "pld [%[pixels], #32]           \n\t"
00224         "add r4, r4, %[line_size]       \n\t"
00225         "walignr1 wr0, wr10, wr11       \n\t"
00226         "pld [r4]                       \n\t"
00227         "pld [r4, #32]                  \n\t"
00228         "walignr1 wr2, wr13, wr14       \n\t"
00229         "wmoveq wr4, wr11               \n\t"
00230         "wmoveq wr6, wr14               \n\t"
00231         "walignr2ne wr4, wr10, wr11     \n\t"
00232         "walignr2ne wr6, wr13, wr14     \n\t"
00233         WAVG2B" wr0, wr0, wr4           \n\t"
00234         WAVG2B" wr2, wr2, wr6           \n\t"
00235         "wstrd wr0, [%[block]]          \n\t"
00236         "subs %[h], %[h], #2            \n\t"
00237         "wstrd wr2, [r5]                \n\t"
00238         "add %[block], %[block], %[line_size]   \n\t"
00239         "add r5, r5, %[line_size]       \n\t"
00240         "bne 1b                         \n\t"
00241         : [h]"+r"(h), [pixels]"+r"(pixels), [block]"+r"(block), [line_size]"+r"(stride)
00242         :
00243         : "r4", "r5", "r12", "memory");
00244 }
00245 
00246 void DEF(put, pixels16_x2)(uint8_t *block, const uint8_t *pixels, const int line_size, int h)
00247 {
00248     int stride = line_size;
00249     // [wr0 wr1 wr2 wr3] for previous line
00250     // [wr4 wr5 wr6 wr7] for current line
00251     SET_RND(wr15); // =2 for rnd  and  =1 for no_rnd version
00252     __asm__ __volatile__(
00253         "pld [%[pixels]]                \n\t"
00254         "pld [%[pixels], #32]           \n\t"
00255         "and r12, %[pixels], #7         \n\t"
00256         "bic %[pixels], %[pixels], #7   \n\t"
00257         "tmcr wcgr1, r12                \n\t"
00258         "add r12, r12, #1               \n\t"
00259         "add r4, %[pixels], %[line_size]\n\t"
00260         "tmcr wcgr2, r12                \n\t"
00261         "add r5, %[block], %[line_size] \n\t"
00262         "mov %[line_size], %[line_size], lsl #1 \n\t"
00263 
00264         "1:                             \n\t"
00265         "wldrd wr10, [%[pixels]]        \n\t"
00266         "cmp r12, #8                    \n\t"
00267         "wldrd wr11, [%[pixels], #8]    \n\t"
00268         "wldrd wr12, [%[pixels], #16]   \n\t"
00269         "add %[pixels], %[pixels], %[line_size] \n\t"
00270         "wldrd wr13, [r4]               \n\t"
00271         "pld [%[pixels]]                \n\t"
00272         "wldrd wr14, [r4, #8]           \n\t"
00273         "pld [%[pixels], #32]           \n\t"
00274         "wldrd wr15, [r4, #16]          \n\t"
00275         "add r4, r4, %[line_size]       \n\t"
00276         "walignr1 wr0, wr10, wr11       \n\t"
00277         "pld [r4]                       \n\t"
00278         "pld [r4, #32]                  \n\t"
00279         "walignr1 wr1, wr11, wr12       \n\t"
00280         "walignr1 wr2, wr13, wr14       \n\t"
00281         "walignr1 wr3, wr14, wr15       \n\t"
00282         "wmoveq wr4, wr11               \n\t"
00283         "wmoveq wr5, wr12               \n\t"
00284         "wmoveq wr6, wr14               \n\t"
00285         "wmoveq wr7, wr15               \n\t"
00286         "walignr2ne wr4, wr10, wr11     \n\t"
00287         "walignr2ne wr5, wr11, wr12     \n\t"
00288         "walignr2ne wr6, wr13, wr14     \n\t"
00289         "walignr2ne wr7, wr14, wr15     \n\t"
00290         WAVG2B" wr0, wr0, wr4           \n\t"
00291         WAVG2B" wr1, wr1, wr5           \n\t"
00292         "wstrd wr0, [%[block]]          \n\t"
00293         WAVG2B" wr2, wr2, wr6           \n\t"
00294         "wstrd wr1, [%[block], #8]      \n\t"
00295         WAVG2B" wr3, wr3, wr7           \n\t"
00296         "add %[block], %[block], %[line_size]   \n\t"
00297         "wstrd wr2, [r5]                \n\t"
00298         "subs %[h], %[h], #2            \n\t"
00299         "wstrd wr3, [r5, #8]            \n\t"
00300         "add r5, r5, %[line_size]       \n\t"
00301         "bne 1b                         \n\t"
00302         : [h]"+r"(h), [pixels]"+r"(pixels), [block]"+r"(block), [line_size]"+r"(stride)
00303         :
00304         : "r4", "r5", "r12", "memory");
00305 }
00306 
00307 void DEF(avg, pixels8_x2)(uint8_t *block, const uint8_t *pixels, const int line_size, int h)
00308 {
00309     int stride = line_size;
00310     // [wr0 wr1 wr2 wr3] for previous line
00311     // [wr4 wr5 wr6 wr7] for current line
00312     SET_RND(wr15); // =2 for rnd  and  =1 for no_rnd version
00313     __asm__ __volatile__(
00314         "pld [%[pixels]]                \n\t"
00315         "pld [%[pixels], #32]           \n\t"
00316         "pld [%[block]]                 \n\t"
00317         "pld [%[block], #32]            \n\t"
00318         "and r12, %[pixels], #7         \n\t"
00319         "bic %[pixels], %[pixels], #7   \n\t"
00320         "tmcr wcgr1, r12                \n\t"
00321         "add r12, r12, #1               \n\t"
00322         "add r4, %[pixels], %[line_size]\n\t"
00323         "tmcr wcgr2, r12                \n\t"
00324         "add r5, %[block], %[line_size] \n\t"
00325         "mov %[line_size], %[line_size], lsl #1 \n\t"
00326         "pld [r5]                       \n\t"
00327         "pld [r5, #32]                  \n\t"
00328 
00329         "1:                             \n\t"
00330         "wldrd wr10, [%[pixels]]        \n\t"
00331         "cmp r12, #8                    \n\t"
00332         "wldrd wr11, [%[pixels], #8]    \n\t"
00333         "add %[pixels], %[pixels], %[line_size] \n\t"
00334         "wldrd wr13, [r4]               \n\t"
00335         "pld [%[pixels]]                \n\t"
00336         "wldrd wr14, [r4, #8]           \n\t"
00337         "pld [%[pixels], #32]           \n\t"
00338         "add r4, r4, %[line_size]       \n\t"
00339         "walignr1 wr0, wr10, wr11       \n\t"
00340         "pld [r4]                       \n\t"
00341         "pld [r4, #32]                  \n\t"
00342         "walignr1 wr2, wr13, wr14       \n\t"
00343         "wmoveq wr4, wr11               \n\t"
00344         "wmoveq wr6, wr14               \n\t"
00345         "walignr2ne wr4, wr10, wr11     \n\t"
00346         "wldrd wr10, [%[block]]         \n\t"
00347         "walignr2ne wr6, wr13, wr14     \n\t"
00348         "wldrd wr12, [r5]               \n\t"
00349         WAVG2B" wr0, wr0, wr4           \n\t"
00350         WAVG2B" wr2, wr2, wr6           \n\t"
00351         WAVG2B" wr0, wr0, wr10          \n\t"
00352         WAVG2B" wr2, wr2, wr12          \n\t"
00353         "wstrd wr0, [%[block]]          \n\t"
00354         "subs %[h], %[h], #2            \n\t"
00355         "wstrd wr2, [r5]                \n\t"
00356         "add %[block], %[block], %[line_size]   \n\t"
00357         "add r5, r5, %[line_size]       \n\t"
00358         "pld [%[block]]                 \n\t"
00359         "pld [%[block], #32]            \n\t"
00360         "pld [r5]                       \n\t"
00361         "pld [r5, #32]                  \n\t"
00362         "bne 1b                         \n\t"
00363         : [h]"+r"(h), [pixels]"+r"(pixels), [block]"+r"(block), [line_size]"+r"(stride)
00364         :
00365         : "r4", "r5", "r12", "memory");
00366 }
00367 
00368 void DEF(avg, pixels16_x2)(uint8_t *block, const uint8_t *pixels, const int line_size, int h)
00369 {
00370     int stride = line_size;
00371     // [wr0 wr1 wr2 wr3] for previous line
00372     // [wr4 wr5 wr6 wr7] for current line
00373     SET_RND(wr15); // =2 for rnd  and  =1 for no_rnd version
00374     __asm__ __volatile__(
00375         "pld [%[pixels]]                \n\t"
00376         "pld [%[pixels], #32]           \n\t"
00377         "pld [%[block]]                 \n\t"
00378         "pld [%[block], #32]            \n\t"
00379         "and r12, %[pixels], #7         \n\t"
00380         "bic %[pixels], %[pixels], #7   \n\t"
00381         "tmcr wcgr1, r12                \n\t"
00382         "add r12, r12, #1               \n\t"
00383         "add r4, %[pixels], %[line_size]\n\t"
00384         "tmcr wcgr2, r12                \n\t"
00385         "add r5, %[block], %[line_size] \n\t"
00386         "mov %[line_size], %[line_size], lsl #1 \n\t"
00387         "pld [r5]                       \n\t"
00388         "pld [r5, #32]                  \n\t"
00389 
00390         "1:                             \n\t"
00391         "wldrd wr10, [%[pixels]]        \n\t"
00392         "cmp r12, #8                    \n\t"
00393         "wldrd wr11, [%[pixels], #8]    \n\t"
00394         "wldrd wr12, [%[pixels], #16]   \n\t"
00395         "add %[pixels], %[pixels], %[line_size] \n\t"
00396         "wldrd wr13, [r4]               \n\t"
00397         "pld [%[pixels]]                \n\t"
00398         "wldrd wr14, [r4, #8]           \n\t"
00399         "pld [%[pixels], #32]           \n\t"
00400         "wldrd wr15, [r4, #16]          \n\t"
00401         "add r4, r4, %[line_size]       \n\t"
00402         "walignr1 wr0, wr10, wr11       \n\t"
00403         "pld [r4]                       \n\t"
00404         "pld [r4, #32]                  \n\t"
00405         "walignr1 wr1, wr11, wr12       \n\t"
00406         "walignr1 wr2, wr13, wr14       \n\t"
00407         "walignr1 wr3, wr14, wr15       \n\t"
00408         "wmoveq wr4, wr11               \n\t"
00409         "wmoveq wr5, wr12               \n\t"
00410         "wmoveq wr6, wr14               \n\t"
00411         "wmoveq wr7, wr15               \n\t"
00412         "walignr2ne wr4, wr10, wr11     \n\t"
00413         "walignr2ne wr5, wr11, wr12     \n\t"
00414         "walignr2ne wr6, wr13, wr14     \n\t"
00415         "walignr2ne wr7, wr14, wr15     \n\t"
00416         "wldrd wr10, [%[block]]         \n\t"
00417         WAVG2B" wr0, wr0, wr4           \n\t"
00418         "wldrd wr11, [%[block], #8]     \n\t"
00419         WAVG2B" wr1, wr1, wr5           \n\t"
00420         "wldrd wr12, [r5]               \n\t"
00421         WAVG2B" wr2, wr2, wr6           \n\t"
00422         "wldrd wr13, [r5, #8]           \n\t"
00423         WAVG2B" wr3, wr3, wr7           \n\t"
00424         WAVG2B" wr0, wr0, wr10          \n\t"
00425         WAVG2B" wr1, wr1, wr11          \n\t"
00426         WAVG2B" wr2, wr2, wr12          \n\t"
00427         WAVG2B" wr3, wr3, wr13          \n\t"
00428         "wstrd wr0, [%[block]]          \n\t"
00429         "subs %[h], %[h], #2            \n\t"
00430         "wstrd wr1, [%[block], #8]      \n\t"
00431         "add %[block], %[block], %[line_size]   \n\t"
00432         "wstrd wr2, [r5]                \n\t"
00433         "pld [%[block]]                 \n\t"
00434         "wstrd wr3, [r5, #8]            \n\t"
00435         "add r5, r5, %[line_size]       \n\t"
00436         "pld [%[block], #32]            \n\t"
00437         "pld [r5]                       \n\t"
00438         "pld [r5, #32]                  \n\t"
00439         "bne 1b                         \n\t"
00440         : [h]"+r"(h), [pixels]"+r"(pixels), [block]"+r"(block), [line_size]"+r"(stride)
00441         :
00442         :"r4", "r5", "r12", "memory");
00443 }
00444 
00445 void DEF(avg, pixels8_y2)(uint8_t *block, const uint8_t *pixels, const int line_size, int h)
00446 {
00447     int stride = line_size;
00448     // [wr0 wr1 wr2 wr3] for previous line
00449     // [wr4 wr5 wr6 wr7] for current line
00450     __asm__ __volatile__(
00451         "pld            [%[pixels]]                             \n\t"
00452         "pld            [%[pixels], #32]                        \n\t"
00453         "and            r12, %[pixels], #7                      \n\t"
00454         "tmcr           wcgr1, r12                              \n\t"
00455         "bic            %[pixels], %[pixels], #7                \n\t"
00456 
00457         "wldrd          wr10, [%[pixels]]                       \n\t"
00458         "wldrd          wr11, [%[pixels], #8]                   \n\t"
00459         "pld            [%[block]]                              \n\t"
00460         "add            %[pixels], %[pixels], %[line_size]      \n\t"
00461         "walignr1       wr0, wr10, wr11                         \n\t"
00462         "pld            [%[pixels]]                             \n\t"
00463         "pld            [%[pixels], #32]                        \n\t"
00464 
00465       "1:                                                       \n\t"
00466         "wldrd          wr10, [%[pixels]]                       \n\t"
00467         "wldrd          wr11, [%[pixels], #8]                   \n\t"
00468         "add            %[pixels], %[pixels], %[line_size]      \n\t"
00469         "pld            [%[pixels]]                             \n\t"
00470         "pld            [%[pixels], #32]                        \n\t"
00471         "walignr1       wr4, wr10, wr11                         \n\t"
00472         "wldrd          wr10, [%[block]]                        \n\t"
00473          WAVG2B"        wr8, wr0, wr4                           \n\t"
00474          WAVG2B"        wr8, wr8, wr10                          \n\t"
00475         "wstrd          wr8, [%[block]]                         \n\t"
00476         "add            %[block], %[block], %[line_size]        \n\t"
00477 
00478         "wldrd          wr10, [%[pixels]]                       \n\t"
00479         "wldrd          wr11, [%[pixels], #8]                   \n\t"
00480         "pld            [%[block]]                              \n\t"
00481         "add            %[pixels], %[pixels], %[line_size]      \n\t"
00482         "pld            [%[pixels]]                             \n\t"
00483         "pld            [%[pixels], #32]                        \n\t"
00484         "walignr1       wr0, wr10, wr11                         \n\t"
00485         "wldrd          wr10, [%[block]]                        \n\t"
00486          WAVG2B"        wr8, wr0, wr4                           \n\t"
00487          WAVG2B"        wr8, wr8, wr10                          \n\t"
00488         "wstrd          wr8, [%[block]]                         \n\t"
00489         "add            %[block], %[block], %[line_size]        \n\t"
00490 
00491         "subs           %[h], %[h], #2                          \n\t"
00492         "pld            [%[block]]                              \n\t"
00493         "bne            1b                                      \n\t"
00494         : [h]"+r"(h), [pixels]"+r"(pixels), [block]"+r"(block), [line_size]"+r"(stride)
00495         :
00496         : "cc", "memory", "r12");
00497 }
00498 
00499 void DEF(put, pixels16_y2)(uint8_t *block, const uint8_t *pixels, const int line_size, int h)
00500 {
00501     int stride = line_size;
00502     // [wr0 wr1 wr2 wr3] for previous line
00503     // [wr4 wr5 wr6 wr7] for current line
00504     __asm__ __volatile__(
00505         "pld [%[pixels]]                \n\t"
00506         "pld [%[pixels], #32]           \n\t"
00507         "and r12, %[pixels], #7         \n\t"
00508         "tmcr wcgr1, r12                \n\t"
00509         "bic %[pixels], %[pixels], #7   \n\t"
00510 
00511         "wldrd wr10, [%[pixels]]        \n\t"
00512         "wldrd wr11, [%[pixels], #8]    \n\t"
00513         "wldrd wr12, [%[pixels], #16]   \n\t"
00514         "add %[pixels], %[pixels], %[line_size] \n\t"
00515         "pld [%[pixels]]                \n\t"
00516         "pld [%[pixels], #32]           \n\t"
00517         "walignr1 wr0, wr10, wr11       \n\t"
00518         "walignr1 wr1, wr11, wr12       \n\t"
00519 
00520         "1:                             \n\t"
00521         "wldrd wr10, [%[pixels]]        \n\t"
00522         "wldrd wr11, [%[pixels], #8]    \n\t"
00523         "wldrd wr12, [%[pixels], #16]   \n\t"
00524         "add %[pixels], %[pixels], %[line_size] \n\t"
00525         "pld [%[pixels]]                \n\t"
00526         "pld [%[pixels], #32]           \n\t"
00527         "walignr1 wr4, wr10, wr11       \n\t"
00528         "walignr1 wr5, wr11, wr12       \n\t"
00529         WAVG2B" wr8, wr0, wr4           \n\t"
00530         WAVG2B" wr9, wr1, wr5           \n\t"
00531         "wstrd wr8, [%[block]]          \n\t"
00532         "wstrd wr9, [%[block], #8]      \n\t"
00533         "add %[block], %[block], %[line_size]   \n\t"
00534 
00535         "wldrd wr10, [%[pixels]]        \n\t"
00536         "wldrd wr11, [%[pixels], #8]    \n\t"
00537         "wldrd wr12, [%[pixels], #16]   \n\t"
00538         "add %[pixels], %[pixels], %[line_size] \n\t"
00539         "pld [%[pixels]]                \n\t"
00540         "pld [%[pixels], #32]           \n\t"
00541         "walignr1 wr0, wr10, wr11       \n\t"
00542         "walignr1 wr1, wr11, wr12       \n\t"
00543         WAVG2B" wr8, wr0, wr4           \n\t"
00544         WAVG2B" wr9, wr1, wr5           \n\t"
00545         "wstrd wr8, [%[block]]          \n\t"
00546         "wstrd wr9, [%[block], #8]      \n\t"
00547         "add %[block], %[block], %[line_size]   \n\t"
00548 
00549         "subs %[h], %[h], #2            \n\t"
00550         "bne 1b                         \n\t"
00551         : [h]"+r"(h), [pixels]"+r"(pixels), [block]"+r"(block), [line_size]"+r"(stride)
00552         :
00553         : "r4", "r5", "r12", "memory");
00554 }
00555 
00556 void DEF(avg, pixels16_y2)(uint8_t *block, const uint8_t *pixels, const int line_size, int h)
00557 {
00558     int stride = line_size;
00559     // [wr0 wr1 wr2 wr3] for previous line
00560     // [wr4 wr5 wr6 wr7] for current line
00561     __asm__ __volatile__(
00562         "pld [%[pixels]]                \n\t"
00563         "pld [%[pixels], #32]           \n\t"
00564         "and r12, %[pixels], #7         \n\t"
00565         "tmcr wcgr1, r12                \n\t"
00566         "bic %[pixels], %[pixels], #7   \n\t"
00567 
00568         "wldrd wr10, [%[pixels]]        \n\t"
00569         "wldrd wr11, [%[pixels], #8]    \n\t"
00570         "pld [%[block]]                 \n\t"
00571         "wldrd wr12, [%[pixels], #16]   \n\t"
00572         "add %[pixels], %[pixels], %[line_size] \n\t"
00573         "pld [%[pixels]]                \n\t"
00574         "pld [%[pixels], #32]           \n\t"
00575         "walignr1 wr0, wr10, wr11       \n\t"
00576         "walignr1 wr1, wr11, wr12       \n\t"
00577 
00578         "1:                             \n\t"
00579         "wldrd wr10, [%[pixels]]        \n\t"
00580         "wldrd wr11, [%[pixels], #8]    \n\t"
00581         "wldrd wr12, [%[pixels], #16]   \n\t"
00582         "add %[pixels], %[pixels], %[line_size] \n\t"
00583         "pld [%[pixels]]                \n\t"
00584         "pld [%[pixels], #32]           \n\t"
00585         "walignr1 wr4, wr10, wr11       \n\t"
00586         "walignr1 wr5, wr11, wr12       \n\t"
00587         "wldrd wr10, [%[block]]         \n\t"
00588         "wldrd wr11, [%[block], #8]     \n\t"
00589         WAVG2B" wr8, wr0, wr4           \n\t"
00590         WAVG2B" wr9, wr1, wr5           \n\t"
00591         WAVG2B" wr8, wr8, wr10          \n\t"
00592         WAVG2B" wr9, wr9, wr11          \n\t"
00593         "wstrd wr8, [%[block]]          \n\t"
00594         "wstrd wr9, [%[block], #8]      \n\t"
00595         "add %[block], %[block], %[line_size]   \n\t"
00596 
00597         "wldrd wr10, [%[pixels]]        \n\t"
00598         "wldrd wr11, [%[pixels], #8]    \n\t"
00599         "pld [%[block]]                 \n\t"
00600         "wldrd wr12, [%[pixels], #16]   \n\t"
00601         "add %[pixels], %[pixels], %[line_size] \n\t"
00602         "pld [%[pixels]]                \n\t"
00603         "pld [%[pixels], #32]           \n\t"
00604         "walignr1 wr0, wr10, wr11       \n\t"
00605         "walignr1 wr1, wr11, wr12       \n\t"
00606         "wldrd wr10, [%[block]]         \n\t"
00607         "wldrd wr11, [%[block], #8]     \n\t"
00608         WAVG2B" wr8, wr0, wr4           \n\t"
00609         WAVG2B" wr9, wr1, wr5           \n\t"
00610         WAVG2B" wr8, wr8, wr10          \n\t"
00611         WAVG2B" wr9, wr9, wr11          \n\t"
00612         "wstrd wr8, [%[block]]          \n\t"
00613         "wstrd wr9, [%[block], #8]      \n\t"
00614         "add %[block], %[block], %[line_size]   \n\t"
00615 
00616         "subs %[h], %[h], #2            \n\t"
00617         "pld [%[block]]                 \n\t"
00618         "bne 1b                         \n\t"
00619         : [h]"+r"(h), [pixels]"+r"(pixels), [block]"+r"(block), [line_size]"+r"(stride)
00620         :
00621         : "r4", "r5", "r12", "memory");
00622 }
00623 
00624 void DEF(put, pixels8_xy2)(uint8_t *block, const uint8_t *pixels, const int line_size, int h)
00625 {
00626     // [wr0 wr1 wr2 wr3] for previous line
00627     // [wr4 wr5 wr6 wr7] for current line
00628     SET_RND(wr15); // =2 for rnd  and  =1 for no_rnd version
00629     __asm__ __volatile__(
00630         "pld [%[pixels]]                \n\t"
00631         "mov r12, #2                    \n\t"
00632         "pld [%[pixels], #32]           \n\t"
00633         "tmcr wcgr0, r12                \n\t" /* for shift value */
00634         "and r12, %[pixels], #7         \n\t"
00635         "bic %[pixels], %[pixels], #7   \n\t"
00636         "tmcr wcgr1, r12                \n\t"
00637 
00638         // [wr0 wr1 wr2 wr3] <= *
00639         // [wr4 wr5 wr6 wr7]
00640         "wldrd wr12, [%[pixels]]        \n\t"
00641         "add r12, r12, #1               \n\t"
00642         "wldrd wr13, [%[pixels], #8]    \n\t"
00643         "tmcr wcgr2, r12                \n\t"
00644         "add %[pixels], %[pixels], %[line_size] \n\t"
00645         "cmp r12, #8                    \n\t"
00646         "pld [%[pixels]]                \n\t"
00647         "pld [%[pixels], #32]           \n\t"
00648         "walignr1 wr2, wr12, wr13       \n\t"
00649         "wmoveq wr10, wr13              \n\t"
00650         "walignr2ne wr10, wr12, wr13    \n\t"
00651         "wunpckelub wr0, wr2            \n\t"
00652         "wunpckehub wr1, wr2            \n\t"
00653         "wunpckelub wr8, wr10           \n\t"
00654         "wunpckehub wr9, wr10           \n\t"
00655         "waddhus wr0, wr0, wr8          \n\t"
00656         "waddhus wr1, wr1, wr9          \n\t"
00657 
00658         "1:                             \n\t"
00659         // [wr0 wr1 wr2 wr3]
00660         // [wr4 wr5 wr6 wr7] <= *
00661         "wldrd wr12, [%[pixels]]        \n\t"
00662         "cmp r12, #8                    \n\t"
00663         "wldrd wr13, [%[pixels], #8]    \n\t"
00664         "add %[pixels], %[pixels], %[line_size] \n\t"
00665         "walignr1 wr6, wr12, wr13       \n\t"
00666         "pld [%[pixels]]                \n\t"
00667         "pld [%[pixels], #32]           \n\t"
00668         "wmoveq wr10, wr13              \n\t"
00669         "walignr2ne wr10, wr12, wr13    \n\t"
00670         "wunpckelub wr4, wr6            \n\t"
00671         "wunpckehub wr5, wr6            \n\t"
00672         "wunpckelub wr8, wr10           \n\t"
00673         "wunpckehub wr9, wr10           \n\t"
00674         "waddhus wr4, wr4, wr8          \n\t"
00675         "waddhus wr5, wr5, wr9          \n\t"
00676         "waddhus wr8, wr0, wr4          \n\t"
00677         "waddhus wr9, wr1, wr5          \n\t"
00678         "waddhus wr8, wr8, wr15         \n\t"
00679         "waddhus wr9, wr9, wr15         \n\t"
00680         "wsrlhg wr8, wr8, wcgr0         \n\t"
00681         "wsrlhg wr9, wr9, wcgr0         \n\t"
00682         "wpackhus wr8, wr8, wr9         \n\t"
00683         "wstrd wr8, [%[block]]          \n\t"
00684         "add %[block], %[block], %[line_size]   \n\t"
00685 
00686         // [wr0 wr1 wr2 wr3] <= *
00687         // [wr4 wr5 wr6 wr7]
00688         "wldrd wr12, [%[pixels]]        \n\t"
00689         "wldrd wr13, [%[pixels], #8]    \n\t"
00690         "add %[pixels], %[pixels], %[line_size] \n\t"
00691         "walignr1 wr2, wr12, wr13       \n\t"
00692         "pld [%[pixels]]                \n\t"
00693         "pld [%[pixels], #32]           \n\t"
00694         "wmoveq wr10, wr13              \n\t"
00695         "walignr2ne wr10, wr12, wr13    \n\t"
00696         "wunpckelub wr0, wr2            \n\t"
00697         "wunpckehub wr1, wr2            \n\t"
00698         "wunpckelub wr8, wr10           \n\t"
00699         "wunpckehub wr9, wr10           \n\t"
00700         "waddhus wr0, wr0, wr8          \n\t"
00701         "waddhus wr1, wr1, wr9          \n\t"
00702         "waddhus wr8, wr0, wr4          \n\t"
00703         "waddhus wr9, wr1, wr5          \n\t"
00704         "waddhus wr8, wr8, wr15         \n\t"
00705         "waddhus wr9, wr9, wr15         \n\t"
00706         "wsrlhg wr8, wr8, wcgr0         \n\t"
00707         "wsrlhg wr9, wr9, wcgr0         \n\t"
00708         "wpackhus wr8, wr8, wr9         \n\t"
00709         "subs %[h], %[h], #2            \n\t"
00710         "wstrd wr8, [%[block]]          \n\t"
00711         "add %[block], %[block], %[line_size]   \n\t"
00712         "bne 1b                         \n\t"
00713         : [h]"+r"(h), [pixels]"+r"(pixels), [block]"+r"(block)
00714         : [line_size]"r"(line_size)
00715         : "r12", "memory");
00716 }
00717 
00718 void DEF(put, pixels16_xy2)(uint8_t *block, const uint8_t *pixels, const int line_size, int h)
00719 {
00720     // [wr0 wr1 wr2 wr3] for previous line
00721     // [wr4 wr5 wr6 wr7] for current line
00722     SET_RND(wr15); // =2 for rnd  and  =1 for no_rnd version
00723     __asm__ __volatile__(
00724         "pld [%[pixels]]                \n\t"
00725         "mov r12, #2                    \n\t"
00726         "pld [%[pixels], #32]           \n\t"
00727         "tmcr wcgr0, r12                \n\t" /* for shift value */
00728         /* alignment */
00729         "and r12, %[pixels], #7         \n\t"
00730         "bic %[pixels], %[pixels], #7   \n\t"
00731         "tmcr wcgr1, r12                \n\t"
00732         "add r12, r12, #1               \n\t"
00733         "tmcr wcgr2, r12                \n\t"
00734 
00735         // [wr0 wr1 wr2 wr3] <= *
00736         // [wr4 wr5 wr6 wr7]
00737         "wldrd wr12, [%[pixels]]        \n\t"
00738         "cmp r12, #8                    \n\t"
00739         "wldrd wr13, [%[pixels], #8]    \n\t"
00740         "wldrd wr14, [%[pixels], #16]   \n\t"
00741         "add %[pixels], %[pixels], %[line_size] \n\t"
00742         "pld [%[pixels]]                \n\t"
00743         "walignr1 wr2, wr12, wr13       \n\t"
00744         "pld [%[pixels], #32]           \n\t"
00745         "walignr1 wr3, wr13, wr14       \n\t"
00746         "wmoveq wr10, wr13              \n\t"
00747         "wmoveq wr11, wr14              \n\t"
00748         "walignr2ne wr10, wr12, wr13    \n\t"
00749         "walignr2ne wr11, wr13, wr14    \n\t"
00750         "wunpckelub wr0, wr2            \n\t"
00751         "wunpckehub wr1, wr2            \n\t"
00752         "wunpckelub wr2, wr3            \n\t"
00753         "wunpckehub wr3, wr3            \n\t"
00754         "wunpckelub wr8, wr10           \n\t"
00755         "wunpckehub wr9, wr10           \n\t"
00756         "wunpckelub wr10, wr11          \n\t"
00757         "wunpckehub wr11, wr11          \n\t"
00758         "waddhus wr0, wr0, wr8          \n\t"
00759         "waddhus wr1, wr1, wr9          \n\t"
00760         "waddhus wr2, wr2, wr10         \n\t"
00761         "waddhus wr3, wr3, wr11         \n\t"
00762 
00763         "1:                             \n\t"
00764         // [wr0 wr1 wr2 wr3]
00765         // [wr4 wr5 wr6 wr7] <= *
00766         "wldrd wr12, [%[pixels]]        \n\t"
00767         "cmp r12, #8                    \n\t"
00768         "wldrd wr13, [%[pixels], #8]    \n\t"
00769         "wldrd wr14, [%[pixels], #16]   \n\t"
00770         "add %[pixels], %[pixels], %[line_size] \n\t"
00771         "walignr1 wr6, wr12, wr13       \n\t"
00772         "pld [%[pixels]]                \n\t"
00773         "pld [%[pixels], #32]           \n\t"
00774         "walignr1 wr7, wr13, wr14       \n\t"
00775         "wmoveq wr10, wr13              \n\t"
00776         "wmoveq wr11, wr14              \n\t"
00777         "walignr2ne wr10, wr12, wr13    \n\t"
00778         "walignr2ne wr11, wr13, wr14    \n\t"
00779         "wunpckelub wr4, wr6            \n\t"
00780         "wunpckehub wr5, wr6            \n\t"
00781         "wunpckelub wr6, wr7            \n\t"
00782         "wunpckehub wr7, wr7            \n\t"
00783         "wunpckelub wr8, wr10           \n\t"
00784         "wunpckehub wr9, wr10           \n\t"
00785         "wunpckelub wr10, wr11          \n\t"
00786         "wunpckehub wr11, wr11          \n\t"
00787         "waddhus wr4, wr4, wr8          \n\t"
00788         "waddhus wr5, wr5, wr9          \n\t"
00789         "waddhus wr6, wr6, wr10         \n\t"
00790         "waddhus wr7, wr7, wr11         \n\t"
00791         "waddhus wr8, wr0, wr4          \n\t"
00792         "waddhus wr9, wr1, wr5          \n\t"
00793         "waddhus wr10, wr2, wr6         \n\t"
00794         "waddhus wr11, wr3, wr7         \n\t"
00795         "waddhus wr8, wr8, wr15         \n\t"
00796         "waddhus wr9, wr9, wr15         \n\t"
00797         "waddhus wr10, wr10, wr15       \n\t"
00798         "waddhus wr11, wr11, wr15       \n\t"
00799         "wsrlhg wr8, wr8, wcgr0         \n\t"
00800         "wsrlhg wr9, wr9, wcgr0         \n\t"
00801         "wsrlhg wr10, wr10, wcgr0       \n\t"
00802         "wsrlhg wr11, wr11, wcgr0       \n\t"
00803         "wpackhus wr8, wr8, wr9         \n\t"
00804         "wpackhus wr9, wr10, wr11       \n\t"
00805         "wstrd wr8, [%[block]]          \n\t"
00806         "wstrd wr9, [%[block], #8]      \n\t"
00807         "add %[block], %[block], %[line_size]   \n\t"
00808 
00809         // [wr0 wr1 wr2 wr3] <= *
00810         // [wr4 wr5 wr6 wr7]
00811         "wldrd wr12, [%[pixels]]        \n\t"
00812         "wldrd wr13, [%[pixels], #8]    \n\t"
00813         "wldrd wr14, [%[pixels], #16]   \n\t"
00814         "add %[pixels], %[pixels], %[line_size] \n\t"
00815         "walignr1 wr2, wr12, wr13       \n\t"
00816         "pld [%[pixels]]                \n\t"
00817         "pld [%[pixels], #32]           \n\t"
00818         "walignr1 wr3, wr13, wr14       \n\t"
00819         "wmoveq wr10, wr13              \n\t"
00820         "wmoveq wr11, wr14              \n\t"
00821         "walignr2ne wr10, wr12, wr13    \n\t"
00822         "walignr2ne wr11, wr13, wr14    \n\t"
00823         "wunpckelub wr0, wr2            \n\t"
00824         "wunpckehub wr1, wr2            \n\t"
00825         "wunpckelub wr2, wr3            \n\t"
00826         "wunpckehub wr3, wr3            \n\t"
00827         "wunpckelub wr8, wr10           \n\t"
00828         "wunpckehub wr9, wr10           \n\t"
00829         "wunpckelub wr10, wr11          \n\t"
00830         "wunpckehub wr11, wr11          \n\t"
00831         "waddhus wr0, wr0, wr8          \n\t"
00832         "waddhus wr1, wr1, wr9          \n\t"
00833         "waddhus wr2, wr2, wr10         \n\t"
00834         "waddhus wr3, wr3, wr11         \n\t"
00835         "waddhus wr8, wr0, wr4          \n\t"
00836         "waddhus wr9, wr1, wr5          \n\t"
00837         "waddhus wr10, wr2, wr6         \n\t"
00838         "waddhus wr11, wr3, wr7         \n\t"
00839         "waddhus wr8, wr8, wr15         \n\t"
00840         "waddhus wr9, wr9, wr15         \n\t"
00841         "waddhus wr10, wr10, wr15       \n\t"
00842         "waddhus wr11, wr11, wr15       \n\t"
00843         "wsrlhg wr8, wr8, wcgr0         \n\t"
00844         "wsrlhg wr9, wr9, wcgr0         \n\t"
00845         "wsrlhg wr10, wr10, wcgr0       \n\t"
00846         "wsrlhg wr11, wr11, wcgr0       \n\t"
00847         "wpackhus wr8, wr8, wr9         \n\t"
00848         "wpackhus wr9, wr10, wr11       \n\t"
00849         "wstrd wr8, [%[block]]          \n\t"
00850         "wstrd wr9, [%[block], #8]      \n\t"
00851         "add %[block], %[block], %[line_size]   \n\t"
00852 
00853         "subs %[h], %[h], #2            \n\t"
00854         "bne 1b                         \n\t"
00855         : [h]"+r"(h), [pixels]"+r"(pixels), [block]"+r"(block)
00856         : [line_size]"r"(line_size)
00857         : "r12", "memory");
00858 }
00859 
00860 void DEF(avg, pixels8_xy2)(uint8_t *block, const uint8_t *pixels, const int line_size, int h)
00861 {
00862     // [wr0 wr1 wr2 wr3] for previous line
00863     // [wr4 wr5 wr6 wr7] for current line
00864     SET_RND(wr15); // =2 for rnd  and  =1 for no_rnd version
00865     __asm__ __volatile__(
00866         "pld [%[block]]                 \n\t"
00867         "pld [%[block], #32]            \n\t"
00868         "pld [%[pixels]]                \n\t"
00869         "mov r12, #2                    \n\t"
00870         "pld [%[pixels], #32]           \n\t"
00871         "tmcr wcgr0, r12                \n\t" /* for shift value */
00872         "and r12, %[pixels], #7         \n\t"
00873         "bic %[pixels], %[pixels], #7   \n\t"
00874         "tmcr wcgr1, r12                \n\t"
00875 
00876         // [wr0 wr1 wr2 wr3] <= *
00877         // [wr4 wr5 wr6 wr7]
00878         "wldrd wr12, [%[pixels]]        \n\t"
00879         "add r12, r12, #1               \n\t"
00880         "wldrd wr13, [%[pixels], #8]    \n\t"
00881         "tmcr wcgr2, r12                \n\t"
00882         "add %[pixels], %[pixels], %[line_size] \n\t"
00883         "cmp r12, #8                    \n\t"
00884         "pld [%[pixels]]                \n\t"
00885         "pld [%[pixels], #32]           \n\t"
00886         "walignr1 wr2, wr12, wr13       \n\t"
00887         "wmoveq wr10, wr13              \n\t"
00888         "walignr2ne wr10, wr12, wr13    \n\t"
00889         "wunpckelub wr0, wr2            \n\t"
00890         "wunpckehub wr1, wr2            \n\t"
00891         "wunpckelub wr8, wr10           \n\t"
00892         "wunpckehub wr9, wr10           \n\t"
00893         "waddhus wr0, wr0, wr8          \n\t"
00894         "waddhus wr1, wr1, wr9          \n\t"
00895 
00896         "1:                             \n\t"
00897         // [wr0 wr1 wr2 wr3]
00898         // [wr4 wr5 wr6 wr7] <= *
00899         "wldrd wr12, [%[pixels]]        \n\t"
00900         "cmp r12, #8                    \n\t"
00901         "wldrd wr13, [%[pixels], #8]    \n\t"
00902         "add %[pixels], %[pixels], %[line_size] \n\t"
00903         "walignr1 wr6, wr12, wr13       \n\t"
00904         "pld [%[pixels]]                \n\t"
00905         "pld [%[pixels], #32]           \n\t"
00906         "wmoveq wr10, wr13              \n\t"
00907         "walignr2ne wr10, wr12, wr13    \n\t"
00908         "wunpckelub wr4, wr6            \n\t"
00909         "wunpckehub wr5, wr6            \n\t"
00910         "wunpckelub wr8, wr10           \n\t"
00911         "wunpckehub wr9, wr10           \n\t"
00912         "waddhus wr4, wr4, wr8          \n\t"
00913         "waddhus wr5, wr5, wr9          \n\t"
00914         "waddhus wr8, wr0, wr4          \n\t"
00915         "waddhus wr9, wr1, wr5          \n\t"
00916         "waddhus wr8, wr8, wr15         \n\t"
00917         "waddhus wr9, wr9, wr15         \n\t"
00918         "wldrd wr12, [%[block]]         \n\t"
00919         "wsrlhg wr8, wr8, wcgr0         \n\t"
00920         "wsrlhg wr9, wr9, wcgr0         \n\t"
00921         "wpackhus wr8, wr8, wr9         \n\t"
00922         WAVG2B" wr8, wr8, wr12          \n\t"
00923         "wstrd wr8, [%[block]]          \n\t"
00924         "add %[block], %[block], %[line_size]   \n\t"
00925         "wldrd wr12, [%[pixels]]        \n\t"
00926         "pld [%[block]]                 \n\t"
00927         "pld [%[block], #32]            \n\t"
00928 
00929         // [wr0 wr1 wr2 wr3] <= *
00930         // [wr4 wr5 wr6 wr7]
00931         "wldrd wr13, [%[pixels], #8]    \n\t"
00932         "add %[pixels], %[pixels], %[line_size] \n\t"
00933         "walignr1 wr2, wr12, wr13       \n\t"
00934         "pld [%[pixels]]                \n\t"
00935         "pld [%[pixels], #32]           \n\t"
00936         "wmoveq wr10, wr13              \n\t"
00937         "walignr2ne wr10, wr12, wr13    \n\t"
00938         "wunpckelub wr0, wr2            \n\t"
00939         "wunpckehub wr1, wr2            \n\t"
00940         "wunpckelub wr8, wr10           \n\t"
00941         "wunpckehub wr9, wr10           \n\t"
00942         "waddhus wr0, wr0, wr8          \n\t"
00943         "waddhus wr1, wr1, wr9          \n\t"
00944         "waddhus wr8, wr0, wr4          \n\t"
00945         "waddhus wr9, wr1, wr5          \n\t"
00946         "waddhus wr8, wr8, wr15         \n\t"
00947         "waddhus wr9, wr9, wr15         \n\t"
00948         "wldrd wr12, [%[block]]         \n\t"
00949         "wsrlhg wr8, wr8, wcgr0         \n\t"
00950         "wsrlhg wr9, wr9, wcgr0         \n\t"
00951         "wpackhus wr8, wr8, wr9         \n\t"
00952         "subs %[h], %[h], #2            \n\t"
00953         WAVG2B" wr8, wr8, wr12          \n\t"
00954         "wstrd wr8, [%[block]]          \n\t"
00955         "add %[block], %[block], %[line_size]   \n\t"
00956         "pld [%[block]]                 \n\t"
00957         "pld [%[block], #32]            \n\t"
00958         "bne 1b                         \n\t"
00959         : [h]"+r"(h), [pixels]"+r"(pixels), [block]"+r"(block)
00960         : [line_size]"r"(line_size)
00961         : "r12", "memory");
00962 }
00963 
00964 void DEF(avg, pixels16_xy2)(uint8_t *block, const uint8_t *pixels, const int line_size, int h)
00965 {
00966     // [wr0 wr1 wr2 wr3] for previous line
00967     // [wr4 wr5 wr6 wr7] for current line
00968     SET_RND(wr15); // =2 for rnd  and  =1 for no_rnd version
00969     __asm__ __volatile__(
00970         "pld [%[block]]                 \n\t"
00971         "pld [%[block], #32]            \n\t"
00972         "pld [%[pixels]]                \n\t"
00973         "mov r12, #2                    \n\t"
00974         "pld [%[pixels], #32]           \n\t"
00975         "tmcr wcgr0, r12                \n\t" /* for shift value */
00976         /* alignment */
00977         "and r12, %[pixels], #7         \n\t"
00978         "bic %[pixels], %[pixels], #7           \n\t"
00979         "tmcr wcgr1, r12                \n\t"
00980         "add r12, r12, #1               \n\t"
00981         "tmcr wcgr2, r12                \n\t"
00982 
00983         // [wr0 wr1 wr2 wr3] <= *
00984         // [wr4 wr5 wr6 wr7]
00985         "wldrd wr12, [%[pixels]]        \n\t"
00986         "cmp r12, #8                    \n\t"
00987         "wldrd wr13, [%[pixels], #8]    \n\t"
00988         "wldrd wr14, [%[pixels], #16]   \n\t"
00989         "add %[pixels], %[pixels], %[line_size] \n\t"
00990         "pld [%[pixels]]                \n\t"
00991         "walignr1 wr2, wr12, wr13       \n\t"
00992         "pld [%[pixels], #32]           \n\t"
00993         "walignr1 wr3, wr13, wr14       \n\t"
00994         "wmoveq wr10, wr13              \n\t"
00995         "wmoveq wr11, wr14              \n\t"
00996         "walignr2ne wr10, wr12, wr13    \n\t"
00997         "walignr2ne wr11, wr13, wr14    \n\t"
00998         "wunpckelub wr0, wr2            \n\t"
00999         "wunpckehub wr1, wr2            \n\t"
01000         "wunpckelub wr2, wr3            \n\t"
01001         "wunpckehub wr3, wr3            \n\t"
01002         "wunpckelub wr8, wr10           \n\t"
01003         "wunpckehub wr9, wr10           \n\t"
01004         "wunpckelub wr10, wr11          \n\t"
01005         "wunpckehub wr11, wr11          \n\t"
01006         "waddhus wr0, wr0, wr8          \n\t"
01007         "waddhus wr1, wr1, wr9          \n\t"
01008         "waddhus wr2, wr2, wr10         \n\t"
01009         "waddhus wr3, wr3, wr11         \n\t"
01010 
01011         "1:                             \n\t"
01012         // [wr0 wr1 wr2 wr3]
01013         // [wr4 wr5 wr6 wr7] <= *
01014         "wldrd wr12, [%[pixels]]        \n\t"
01015         "cmp r12, #8                    \n\t"
01016         "wldrd wr13, [%[pixels], #8]    \n\t"
01017         "wldrd wr14, [%[pixels], #16]   \n\t"
01018         "add %[pixels], %[pixels], %[line_size] \n\t"
01019         "walignr1 wr6, wr12, wr13       \n\t"
01020         "pld [%[pixels]]                \n\t"
01021         "pld [%[pixels], #32]           \n\t"
01022         "walignr1 wr7, wr13, wr14       \n\t"
01023         "wmoveq wr10, wr13              \n\t"
01024         "wmoveq wr11, wr14              \n\t"
01025         "walignr2ne wr10, wr12, wr13    \n\t"
01026         "walignr2ne wr11, wr13, wr14    \n\t"
01027         "wunpckelub wr4, wr6            \n\t"
01028         "wunpckehub wr5, wr6            \n\t"
01029         "wunpckelub wr6, wr7            \n\t"
01030         "wunpckehub wr7, wr7            \n\t"
01031         "wunpckelub wr8, wr10           \n\t"
01032         "wunpckehub wr9, wr10           \n\t"
01033         "wunpckelub wr10, wr11          \n\t"
01034         "wunpckehub wr11, wr11          \n\t"
01035         "waddhus wr4, wr4, wr8          \n\t"
01036         "waddhus wr5, wr5, wr9          \n\t"
01037         "waddhus wr6, wr6, wr10         \n\t"
01038         "waddhus wr7, wr7, wr11         \n\t"
01039         "waddhus wr8, wr0, wr4          \n\t"
01040         "waddhus wr9, wr1, wr5          \n\t"
01041         "waddhus wr10, wr2, wr6         \n\t"
01042         "waddhus wr11, wr3, wr7         \n\t"
01043         "waddhus wr8, wr8, wr15         \n\t"
01044         "waddhus wr9, wr9, wr15         \n\t"
01045         "waddhus wr10, wr10, wr15       \n\t"
01046         "waddhus wr11, wr11, wr15       \n\t"
01047         "wsrlhg wr8, wr8, wcgr0         \n\t"
01048         "wsrlhg wr9, wr9, wcgr0         \n\t"
01049         "wldrd wr12, [%[block]]         \n\t"
01050         "wldrd wr13, [%[block], #8]     \n\t"
01051         "wsrlhg wr10, wr10, wcgr0       \n\t"
01052         "wsrlhg wr11, wr11, wcgr0       \n\t"
01053         "wpackhus wr8, wr8, wr9         \n\t"
01054         "wpackhus wr9, wr10, wr11       \n\t"
01055         WAVG2B" wr8, wr8, wr12          \n\t"
01056         WAVG2B" wr9, wr9, wr13          \n\t"
01057         "wstrd wr8, [%[block]]          \n\t"
01058         "wstrd wr9, [%[block], #8]      \n\t"
01059         "add %[block], %[block], %[line_size]   \n\t"
01060 
01061         // [wr0 wr1 wr2 wr3] <= *
01062         // [wr4 wr5 wr6 wr7]
01063         "wldrd wr12, [%[pixels]]        \n\t"
01064         "pld [%[block]]                 \n\t"
01065         "wldrd wr13, [%[pixels], #8]    \n\t"
01066         "pld [%[block], #32]            \n\t"
01067         "wldrd wr14, [%[pixels], #16]   \n\t"
01068         "add %[pixels], %[pixels], %[line_size] \n\t"
01069         "walignr1 wr2, wr12, wr13       \n\t"
01070         "pld [%[pixels]]                \n\t"
01071         "pld [%[pixels], #32]           \n\t"
01072         "walignr1 wr3, wr13, wr14       \n\t"
01073         "wmoveq wr10, wr13              \n\t"
01074         "wmoveq wr11, wr14              \n\t"
01075         "walignr2ne wr10, wr12, wr13    \n\t"
01076         "walignr2ne wr11, wr13, wr14    \n\t"
01077         "wunpckelub wr0, wr2            \n\t"
01078         "wunpckehub wr1, wr2            \n\t"
01079         "wunpckelub wr2, wr3            \n\t"
01080         "wunpckehub wr3, wr3            \n\t"
01081         "wunpckelub wr8, wr10           \n\t"
01082         "wunpckehub wr9, wr10           \n\t"
01083         "wunpckelub wr10, wr11          \n\t"
01084         "wunpckehub wr11, wr11          \n\t"
01085         "waddhus wr0, wr0, wr8          \n\t"
01086         "waddhus wr1, wr1, wr9          \n\t"
01087         "waddhus wr2, wr2, wr10         \n\t"
01088         "waddhus wr3, wr3, wr11         \n\t"
01089         "waddhus wr8, wr0, wr4          \n\t"
01090         "waddhus wr9, wr1, wr5          \n\t"
01091         "waddhus wr10, wr2, wr6         \n\t"
01092         "waddhus wr11, wr3, wr7         \n\t"
01093         "waddhus wr8, wr8, wr15         \n\t"
01094         "waddhus wr9, wr9, wr15         \n\t"
01095         "waddhus wr10, wr10, wr15       \n\t"
01096         "waddhus wr11, wr11, wr15       \n\t"
01097         "wsrlhg wr8, wr8, wcgr0         \n\t"
01098         "wsrlhg wr9, wr9, wcgr0         \n\t"
01099         "wldrd wr12, [%[block]]         \n\t"
01100         "wldrd wr13, [%[block], #8]     \n\t"
01101         "wsrlhg wr10, wr10, wcgr0       \n\t"
01102         "wsrlhg wr11, wr11, wcgr0       \n\t"
01103         "wpackhus wr8, wr8, wr9         \n\t"
01104         "wpackhus wr9, wr10, wr11       \n\t"
01105         WAVG2B" wr8, wr8, wr12          \n\t"
01106         WAVG2B" wr9, wr9, wr13          \n\t"
01107         "wstrd wr8, [%[block]]          \n\t"
01108         "wstrd wr9, [%[block], #8]      \n\t"
01109         "add %[block], %[block], %[line_size]   \n\t"
01110         "subs %[h], %[h], #2            \n\t"
01111         "pld [%[block]]                 \n\t"
01112         "pld [%[block], #32]            \n\t"
01113         "bne 1b                         \n\t"
01114         : [h]"+r"(h), [pixels]"+r"(pixels), [block]"+r"(block)
01115         : [line_size]"r"(line_size)
01116         : "r12", "memory");
01117 }
01118 
01119 #endif /* FFMPEG_DSPUTIL_IWMMXT_RND_H */