Cinelerra: hvirtual/quicktime/ffmpeg/libavcodec/armv4l/dsputil_iwmmxt

00001 void DEF(put, pixels8)(uint8_t *block, const uint8_t *pixels, const int line_size, int h)
00002 {
00003     int stride = line_size;
00004     __asm__ __volatile__ (
00005         "and r12, %[pixels], #7 \n\t"
00006         "bic %[pixels], %[pixels], #7 \n\t"
00007         "tmcr wcgr1, r12 \n\t"
00008         "add r4, %[pixels], %[line_size] \n\t"
00009         "add r5, %[block], %[line_size] \n\t"
00010         "mov %[line_size], %[line_size], lsl #1 \n\t"
00011         "1: \n\t"
00012         "wldrd wr0, [%[pixels]] \n\t"
00013         "subs %[h], %[h], #2 \n\t"
00014         "wldrd wr1, [%[pixels], #8] \n\t"
00015         "add %[pixels], %[pixels], %[line_size] \n\t"
00016         "wldrd wr3, [r4] \n\t"
00017         "pld [%[pixels]] \n\t"
00018         "pld [%[pixels], #32] \n\t"
00019         "wldrd wr4, [r4, #8] \n\t"
00020         "add r4, r4, %[line_size] \n\t"
00021         "walignr1 wr8, wr0, wr1 \n\t"
00022         "pld [r4] \n\t"
00023         "pld [r4, #32] \n\t"
00024         "walignr1 wr10, wr3, wr4 \n\t"
00025         "wstrd wr8, [%[block]] \n\t"
00026         "add %[block], %[block], %[line_size] \n\t"
00027         "wstrd wr10, [r5] \n\t"
00028         "add r5, r5, %[line_size] \n\t"
00029         "bne 1b \n\t"
00030         : [block]"+r"(block), [pixels]"+r"(pixels), [line_size]"+r"(stride), [h]"+r"(h)
00031         :
00032         : "memory", "r4", "r5", "r12");
00033 }
00034 
00035 void DEF(avg, pixels8)(uint8_t *block, const uint8_t *pixels, const int line_size, int h)
00036 {
00037     int stride = line_size;
00038     __asm__ __volatile__ (
00039         "and r12, %[pixels], #7 \n\t"
00040         "bic %[pixels], %[pixels], #7 \n\t"
00041         "tmcr wcgr1, r12 \n\t"
00042         "add r4, %[pixels], %[line_size] \n\t"
00043         "add r5, %[block], %[line_size] \n\t"
00044         "mov %[line_size], %[line_size], lsl #1 \n\t"
00045         "1: \n\t"
00046         "wldrd wr0, [%[pixels]] \n\t"
00047         "subs %[h], %[h], #2 \n\t"
00048         "wldrd wr1, [%[pixels], #8] \n\t"
00049         "add %[pixels], %[pixels], %[line_size] \n\t"
00050         "wldrd wr3, [r4] \n\t"
00051         "pld [%[pixels]] \n\t"
00052         "pld [%[pixels], #32] \n\t"
00053         "wldrd wr4, [r4, #8] \n\t"
00054         "add r4, r4, %[line_size] \n\t"
00055         "walignr1 wr8, wr0, wr1 \n\t"
00056         "wldrd wr0, [%[block]] \n\t"
00057         "wldrd wr2, [r5] \n\t"
00058         "pld [r4] \n\t"
00059         "pld [r4, #32] \n\t"
00060         "walignr1 wr10, wr3, wr4 \n\t"
00061         WAVG2B" wr8, wr8, wr0 \n\t"
00062         WAVG2B" wr10, wr10, wr2 \n\t"
00063         "wstrd wr8, [%[block]] \n\t"
00064         "add %[block], %[block], %[line_size] \n\t"
00065         "wstrd wr10, [r5] \n\t"
00066         "pld [%[block]] \n\t"
00067         "pld [%[block], #32] \n\t"
00068         "add r5, r5, %[line_size] \n\t"
00069         "pld [r5] \n\t"
00070         "pld [r5, #32] \n\t"
00071         "bne 1b \n\t"
00072         : [block]"+r"(block), [pixels]"+r"(pixels), [line_size]"+r"(stride), [h]"+r"(h)
00073         :
00074         : "memory", "r4", "r5", "r12");
00075 }
00076 
00077 void DEF(put, pixels16)(uint8_t *block, const uint8_t *pixels, const int line_size, int h)
00078 {
00079     int stride = line_size;
00080     __asm__ __volatile__ (
00081         "and r12, %[pixels], #7 \n\t"
00082         "bic %[pixels], %[pixels], #7 \n\t"
00083         "tmcr wcgr1, r12 \n\t"
00084         "add r4, %[pixels], %[line_size] \n\t"
00085         "add r5, %[block], %[line_size] \n\t"
00086         "mov %[line_size], %[line_size], lsl #1 \n\t"
00087         "1: \n\t"
00088         "wldrd wr0, [%[pixels]] \n\t"
00089         "wldrd wr1, [%[pixels], #8] \n\t"
00090         "subs %[h], %[h], #2 \n\t"
00091         "wldrd wr2, [%[pixels], #16] \n\t"
00092         "add %[pixels], %[pixels], %[line_size] \n\t"
00093         "wldrd wr3, [r4] \n\t"
00094         "pld [%[pixels]] \n\t"
00095         "pld [%[pixels], #32] \n\t"
00096         "walignr1 wr8, wr0, wr1 \n\t"
00097         "wldrd wr4, [r4, #8] \n\t"
00098         "walignr1 wr9, wr1, wr2 \n\t"
00099         "wldrd wr5, [r4, #16] \n\t"
00100         "add r4, r4, %[line_size] \n\t"
00101         "pld [r4] \n\t"
00102         "pld [r4, #32] \n\t"
00103         "walignr1 wr10, wr3, wr4 \n\t"
00104         "wstrd wr8, [%[block]] \n\t"
00105         "walignr1 wr11, wr4, wr5 \n\t"
00106         "wstrd wr9, [%[block], #8] \n\t"
00107         "add %[block], %[block], %[line_size] \n\t"
00108         "wstrd wr10, [r5] \n\t"
00109         "wstrd wr11, [r5, #8] \n\t"
00110         "add r5, r5, %[line_size] \n\t"
00111         "bne 1b \n\t"
00112         : [block]"+r"(block), [pixels]"+r"(pixels), [line_size]"+r"(stride), [h]"+r"(h)
00113         :
00114         : "memory", "r4", "r5", "r12");
00115 }
00116 
00117 void DEF(avg, pixels16)(uint8_t *block, const uint8_t *pixels, const int line_size, int h)
00118 {
00119     int stride = line_size;
00120     __asm__ __volatile__ (
00121         "pld [%[pixels]]                \n\t"
00122         "pld [%[pixels], #32]           \n\t"
00123         "pld [%[block]]                 \n\t"
00124         "pld [%[block], #32]            \n\t"
00125         "and r12, %[pixels], #7         \n\t"
00126         "bic %[pixels], %[pixels], #7   \n\t"
00127         "tmcr wcgr1, r12                \n\t"
00128         "add r4, %[pixels], %[line_size]\n\t"
00129         "add r5, %[block], %[line_size] \n\t"
00130         "mov %[line_size], %[line_size], lsl #1 \n\t"
00131         "1:                             \n\t"
00132         "wldrd wr0, [%[pixels]]         \n\t"
00133         "wldrd wr1, [%[pixels], #8]     \n\t"
00134         "subs %[h], %[h], #2            \n\t"
00135         "wldrd wr2, [%[pixels], #16]    \n\t"
00136         "add %[pixels], %[pixels], %[line_size] \n\t"
00137         "wldrd wr3, [r4]                \n\t"
00138         "pld [%[pixels]]                \n\t"
00139         "pld [%[pixels], #32]           \n\t"
00140         "walignr1 wr8, wr0, wr1         \n\t"
00141         "wldrd wr4, [r4, #8]            \n\t"
00142         "walignr1 wr9, wr1, wr2         \n\t"
00143         "wldrd wr5, [r4, #16]           \n\t"
00144         "add r4, r4, %[line_size]       \n\t"
00145         "wldrd wr0, [%[block]]          \n\t"
00146         "pld [r4]                       \n\t"
00147         "wldrd wr1, [%[block], #8]      \n\t"
00148         "pld [r4, #32]                  \n\t"
00149         "wldrd wr2, [r5]                \n\t"
00150         "walignr1 wr10, wr3, wr4        \n\t"
00151         "wldrd wr3, [r5, #8]            \n\t"
00152         WAVG2B" wr8, wr8, wr0           \n\t"
00153         WAVG2B" wr9, wr9, wr1           \n\t"
00154         WAVG2B" wr10, wr10, wr2         \n\t"
00155         "wstrd wr8, [%[block]]          \n\t"
00156         "walignr1 wr11, wr4, wr5        \n\t"
00157         WAVG2B" wr11, wr11, wr3         \n\t"
00158         "wstrd wr9, [%[block], #8]      \n\t"
00159         "add %[block], %[block], %[line_size] \n\t"
00160         "wstrd wr10, [r5]               \n\t"
00161         "pld [%[block]]                 \n\t"
00162         "pld [%[block], #32]            \n\t"
00163         "wstrd wr11, [r5, #8]           \n\t"
00164         "add r5, r5, %[line_size]       \n\t"
00165         "pld [r5]                       \n\t"
00166         "pld [r5, #32]                  \n\t"
00167         "bne 1b \n\t"
00168         : [block]"+r"(block), [pixels]"+r"(pixels), [line_size]"+r"(stride), [h]"+r"(h)
00169         :
00170         : "memory", "r4", "r5", "r12");
00171 }
00172 
00173 void DEF(put, pixels8_x2)(uint8_t *block, const uint8_t *pixels, const int line_size, int h)
00174 {
00175     int stride = line_size;
00176     // [wr0 wr1 wr2 wr3] for previous line
00177     // [wr4 wr5 wr6 wr7] for current line
00178     SET_RND(wr15); // =2 for rnd  and  =1 for no_rnd version
00179     __asm__ __volatile__(
00180         "pld [%[pixels]]                \n\t"
00181         "pld [%[pixels], #32]           \n\t"
00182         "and r12, %[pixels], #7         \n\t"
00183         "bic %[pixels], %[pixels], #7   \n\t"
00184         "tmcr wcgr1, r12                \n\t"
00185         "add r12, r12, #1               \n\t"
00186         "add r4, %[pixels], %[line_size]\n\t"
00187         "tmcr wcgr2, r12                \n\t"
00188         "add r5, %[block], %[line_size] \n\t"
00189         "mov %[line_size], %[line_size], lsl #1 \n\t"
00190 
00191         "1:                             \n\t"
00192         "wldrd wr10, [%[pixels]]        \n\t"
00193         "cmp r12, #8                    \n\t"
00194         "wldrd wr11, [%[pixels], #8]    \n\t"
00195         "add %[pixels], %[pixels], %[line_size] \n\t"
00196         "wldrd wr13, [r4]               \n\t"
00197         "pld [%[pixels]]                \n\t"
00198         "wldrd wr14, [r4, #8]           \n\t"
00199         "pld [%[pixels], #32]           \n\t"
00200         "add r4, r4, %[line_size]       \n\t"
00201         "walignr1 wr0, wr10, wr11       \n\t"
00202         "pld [r4]                       \n\t"
00203         "pld [r4, #32]                  \n\t"
00204         "walignr1 wr2, wr13, wr14       \n\t"
00205         "wmoveq wr4, wr11               \n\t"
00206         "wmoveq wr6, wr14               \n\t"
00207         "walignr2ne wr4, wr10, wr11     \n\t"
00208         "walignr2ne wr6, wr13, wr14     \n\t"
00209         WAVG2B" wr0, wr0, wr4           \n\t"
00210         WAVG2B" wr2, wr2, wr6           \n\t"
00211         "wstrd wr0, [%[block]]          \n\t"
00212         "subs %[h], %[h], #2            \n\t"
00213         "wstrd wr2, [r5]                \n\t"
00214         "add %[block], %[block], %[line_size]   \n\t"
00215         "add r5, r5, %[line_size]       \n\t"
00216         "bne 1b                         \n\t"
00217         : [h]"+r"(h), [pixels]"+r"(pixels), [block]"+r"(block), [line_size]"+r"(stride)
00218         :
00219         : "r4", "r5", "r12", "memory");
00220 }
00221 
00222 void DEF(put, pixels16_x2)(uint8_t *block, const uint8_t *pixels, const int line_size, int h)
00223 {
00224     int stride = line_size;
00225     // [wr0 wr1 wr2 wr3] for previous line
00226     // [wr4 wr5 wr6 wr7] for current line
00227     SET_RND(wr15); // =2 for rnd  and  =1 for no_rnd version
00228     __asm__ __volatile__(
00229         "pld [%[pixels]]                \n\t"
00230         "pld [%[pixels], #32]           \n\t"
00231         "and r12, %[pixels], #7         \n\t"
00232         "bic %[pixels], %[pixels], #7   \n\t"
00233         "tmcr wcgr1, r12                \n\t"
00234         "add r12, r12, #1               \n\t"
00235         "add r4, %[pixels], %[line_size]\n\t"
00236         "tmcr wcgr2, r12                \n\t"
00237         "add r5, %[block], %[line_size] \n\t"
00238         "mov %[line_size], %[line_size], lsl #1 \n\t"
00239 
00240         "1:                             \n\t"
00241         "wldrd wr10, [%[pixels]]        \n\t"
00242         "cmp r12, #8                    \n\t"
00243         "wldrd wr11, [%[pixels], #8]    \n\t"
00244         "wldrd wr12, [%[pixels], #16]   \n\t"
00245         "add %[pixels], %[pixels], %[line_size] \n\t"
00246         "wldrd wr13, [r4]               \n\t"
00247         "pld [%[pixels]]                \n\t"
00248         "wldrd wr14, [r4, #8]           \n\t"
00249         "pld [%[pixels], #32]           \n\t"
00250         "wldrd wr15, [r4, #16]          \n\t"
00251         "add r4, r4, %[line_size]       \n\t"
00252         "walignr1 wr0, wr10, wr11       \n\t"
00253         "pld [r4]                       \n\t"
00254         "pld [r4, #32]                  \n\t"
00255         "walignr1 wr1, wr11, wr12       \n\t"
00256         "walignr1 wr2, wr13, wr14       \n\t"
00257         "walignr1 wr3, wr14, wr15       \n\t"
00258         "wmoveq wr4, wr11               \n\t"
00259         "wmoveq wr5, wr12               \n\t"
00260         "wmoveq wr6, wr14               \n\t"
00261         "wmoveq wr7, wr15               \n\t"
00262         "walignr2ne wr4, wr10, wr11     \n\t"
00263         "walignr2ne wr5, wr11, wr12     \n\t"
00264         "walignr2ne wr6, wr13, wr14     \n\t"
00265         "walignr2ne wr7, wr14, wr15     \n\t"
00266         WAVG2B" wr0, wr0, wr4           \n\t"
00267         WAVG2B" wr1, wr1, wr5           \n\t"
00268         "wstrd wr0, [%[block]]          \n\t"
00269         WAVG2B" wr2, wr2, wr6           \n\t"
00270         "wstrd wr1, [%[block], #8]      \n\t"
00271         WAVG2B" wr3, wr3, wr7           \n\t"
00272         "add %[block], %[block], %[line_size]   \n\t"
00273         "wstrd wr2, [r5]                \n\t"
00274         "subs %[h], %[h], #2            \n\t"
00275         "wstrd wr3, [r5, #8]            \n\t"
00276         "add r5, r5, %[line_size]       \n\t"
00277         "bne 1b                         \n\t"
00278         : [h]"+r"(h), [pixels]"+r"(pixels), [block]"+r"(block), [line_size]"+r"(stride)
00279         :
00280         : "r4", "r5", "r12", "memory");
00281 }
00282 
00283 void DEF(avg, pixels8_x2)(uint8_t *block, const uint8_t *pixels, const int line_size, int h)
00284 {
00285     int stride = line_size;
00286     // [wr0 wr1 wr2 wr3] for previous line
00287     // [wr4 wr5 wr6 wr7] for current line
00288     SET_RND(wr15); // =2 for rnd  and  =1 for no_rnd version
00289     __asm__ __volatile__(
00290         "pld [%[pixels]]                \n\t"
00291         "pld [%[pixels], #32]           \n\t"
00292         "pld [%[block]]                 \n\t"
00293         "pld [%[block], #32]            \n\t"
00294         "and r12, %[pixels], #7         \n\t"
00295         "bic %[pixels], %[pixels], #7   \n\t"
00296         "tmcr wcgr1, r12                \n\t"
00297         "add r12, r12, #1               \n\t"
00298         "add r4, %[pixels], %[line_size]\n\t"
00299         "tmcr wcgr2, r12                \n\t"
00300         "add r5, %[block], %[line_size] \n\t"
00301         "mov %[line_size], %[line_size], lsl #1 \n\t"
00302         "pld [r5]                       \n\t"
00303         "pld [r5, #32]                  \n\t"
00304 
00305         "1:                             \n\t"
00306         "wldrd wr10, [%[pixels]]        \n\t"
00307         "cmp r12, #8                    \n\t"
00308         "wldrd wr11, [%[pixels], #8]    \n\t"
00309         "add %[pixels], %[pixels], %[line_size] \n\t"
00310         "wldrd wr13, [r4]               \n\t"
00311         "pld [%[pixels]]                \n\t"
00312         "wldrd wr14, [r4, #8]           \n\t"
00313         "pld [%[pixels], #32]           \n\t"
00314         "add r4, r4, %[line_size]       \n\t"
00315         "walignr1 wr0, wr10, wr11       \n\t"
00316         "pld [r4]                       \n\t"
00317         "pld [r4, #32]                  \n\t"
00318         "walignr1 wr2, wr13, wr14       \n\t"
00319         "wmoveq wr4, wr11               \n\t"
00320         "wmoveq wr6, wr14               \n\t"
00321         "walignr2ne wr4, wr10, wr11     \n\t"
00322         "wldrd wr10, [%[block]]         \n\t"
00323         "walignr2ne wr6, wr13, wr14     \n\t"
00324         "wldrd wr12, [r5]               \n\t"
00325         WAVG2B" wr0, wr0, wr4           \n\t"
00326         WAVG2B" wr2, wr2, wr6           \n\t"
00327         WAVG2B" wr0, wr0, wr10          \n\t"
00328         WAVG2B" wr2, wr2, wr12          \n\t"
00329         "wstrd wr0, [%[block]]          \n\t"
00330         "subs %[h], %[h], #2            \n\t"
00331         "wstrd wr2, [r5]                \n\t"
00332         "add %[block], %[block], %[line_size]   \n\t"
00333         "add r5, r5, %[line_size]       \n\t"
00334         "pld [%[block]]                 \n\t"
00335         "pld [%[block], #32]            \n\t"
00336         "pld [r5]                       \n\t"
00337         "pld [r5, #32]                  \n\t"
00338         "bne 1b                         \n\t"
00339         : [h]"+r"(h), [pixels]"+r"(pixels), [block]"+r"(block), [line_size]"+r"(stride)
00340         :
00341         : "r4", "r5", "r12", "memory");
00342 }
00343 
00344 void DEF(avg, pixels16_x2)(uint8_t *block, const uint8_t *pixels, const int line_size, int h)
00345 {
00346     int stride = line_size;
00347     // [wr0 wr1 wr2 wr3] for previous line
00348     // [wr4 wr5 wr6 wr7] for current line
00349     SET_RND(wr15); // =2 for rnd  and  =1 for no_rnd version
00350     __asm__ __volatile__(
00351         "pld [%[pixels]]                \n\t"
00352         "pld [%[pixels], #32]           \n\t"
00353         "pld [%[block]]                 \n\t"
00354         "pld [%[block], #32]            \n\t"
00355         "and r12, %[pixels], #7         \n\t"
00356         "bic %[pixels], %[pixels], #7   \n\t"
00357         "tmcr wcgr1, r12                \n\t"
00358         "add r12, r12, #1               \n\t"
00359         "add r4, %[pixels], %[line_size]\n\t"
00360         "tmcr wcgr2, r12                \n\t"
00361         "add r5, %[block], %[line_size] \n\t"
00362         "mov %[line_size], %[line_size], lsl #1 \n\t"
00363         "pld [r5]                       \n\t"
00364         "pld [r5, #32]                  \n\t"
00365 
00366         "1:                             \n\t"
00367         "wldrd wr10, [%[pixels]]        \n\t"
00368         "cmp r12, #8                    \n\t"
00369         "wldrd wr11, [%[pixels], #8]    \n\t"
00370         "wldrd wr12, [%[pixels], #16]   \n\t"
00371         "add %[pixels], %[pixels], %[line_size] \n\t"
00372         "wldrd wr13, [r4]               \n\t"
00373         "pld [%[pixels]]                \n\t"
00374         "wldrd wr14, [r4, #8]           \n\t"
00375         "pld [%[pixels], #32]           \n\t"
00376         "wldrd wr15, [r4, #16]          \n\t"
00377         "add r4, r4, %[line_size]       \n\t"
00378         "walignr1 wr0, wr10, wr11       \n\t"
00379         "pld [r4]                       \n\t"
00380         "pld [r4, #32]                  \n\t"
00381         "walignr1 wr1, wr11, wr12       \n\t"
00382         "walignr1 wr2, wr13, wr14       \n\t"
00383         "walignr1 wr3, wr14, wr15       \n\t"
00384         "wmoveq wr4, wr11               \n\t"
00385         "wmoveq wr5, wr12               \n\t"
00386         "wmoveq wr6, wr14               \n\t"
00387         "wmoveq wr7, wr15               \n\t"
00388         "walignr2ne wr4, wr10, wr11     \n\t"
00389         "walignr2ne wr5, wr11, wr12     \n\t"
00390         "walignr2ne wr6, wr13, wr14     \n\t"
00391         "walignr2ne wr7, wr14, wr15     \n\t"
00392         "wldrd wr10, [%[block]]         \n\t"
00393         WAVG2B" wr0, wr0, wr4           \n\t"
00394         "wldrd wr11, [%[block], #8]     \n\t"
00395         WAVG2B" wr1, wr1, wr5           \n\t"
00396         "wldrd wr12, [r5]               \n\t"
00397         WAVG2B" wr2, wr2, wr6           \n\t"
00398         "wldrd wr13, [r5, #8]           \n\t"
00399         WAVG2B" wr3, wr3, wr7           \n\t"
00400         WAVG2B" wr0, wr0, wr10          \n\t"
00401         WAVG2B" wr1, wr1, wr11          \n\t"
00402         WAVG2B" wr2, wr2, wr12          \n\t"
00403         WAVG2B" wr3, wr3, wr13          \n\t"
00404         "wstrd wr0, [%[block]]          \n\t"
00405         "subs %[h], %[h], #2            \n\t"
00406         "wstrd wr1, [%[block], #8]      \n\t"
00407         "add %[block], %[block], %[line_size]   \n\t"
00408         "wstrd wr2, [r5]                \n\t"
00409         "pld [%[block]]                 \n\t"
00410         "wstrd wr3, [r5, #8]            \n\t"
00411         "add r5, r5, %[line_size]       \n\t"
00412         "pld [%[block], #32]            \n\t"
00413         "pld [r5]                       \n\t"
00414         "pld [r5, #32]                  \n\t"
00415         "bne 1b                         \n\t"
00416         : [h]"+r"(h), [pixels]"+r"(pixels), [block]"+r"(block), [line_size]"+r"(stride)
00417         :
00418         :"r4", "r5", "r12", "memory");
00419 }
00420 
00421 void DEF(avg, pixels8_y2)(uint8_t *block, const uint8_t *pixels, const int line_size, int h)
00422 {
00423     int stride = line_size;
00424     // [wr0 wr1 wr2 wr3] for previous line
00425     // [wr4 wr5 wr6 wr7] for current line
00426     __asm__ __volatile__(
00427         "pld            [%[pixels]]                             \n\t"
00428         "pld            [%[pixels], #32]                        \n\t"
00429         "and            r12, %[pixels], #7                      \n\t"
00430         "tmcr           wcgr1, r12                              \n\t"
00431         "bic            %[pixels], %[pixels], #7                \n\t"
00432 
00433         "wldrd          wr10, [%[pixels]]                       \n\t"
00434         "wldrd          wr11, [%[pixels], #8]                   \n\t"
00435         "pld            [%[block]]                              \n\t"
00436         "add            %[pixels], %[pixels], %[line_size]      \n\t"
00437         "walignr1       wr0, wr10, wr11                         \n\t"
00438         "pld            [%[pixels]]                             \n\t"
00439         "pld            [%[pixels], #32]                        \n\t"
00440 
00441       "1:                                                       \n\t"
00442         "wldrd          wr10, [%[pixels]]                       \n\t"
00443         "wldrd          wr11, [%[pixels], #8]                   \n\t"
00444         "add            %[pixels], %[pixels], %[line_size]      \n\t"
00445         "pld            [%[pixels]]                             \n\t"
00446         "pld            [%[pixels], #32]                        \n\t"
00447         "walignr1       wr4, wr10, wr11                         \n\t"
00448         "wldrd          wr10, [%[block]]                        \n\t"
00449          WAVG2B"        wr8, wr0, wr4                           \n\t"
00450          WAVG2B"        wr8, wr8, wr10                          \n\t"
00451         "wstrd          wr8, [%[block]]                         \n\t"
00452         "add            %[block], %[block], %[line_size]        \n\t"
00453 
00454         "wldrd          wr10, [%[pixels]]                       \n\t"
00455         "wldrd          wr11, [%[pixels], #8]                   \n\t"
00456         "pld            [%[block]]                              \n\t"
00457         "add            %[pixels], %[pixels], %[line_size]      \n\t"
00458         "pld            [%[pixels]]                             \n\t"
00459         "pld            [%[pixels], #32]                        \n\t"
00460         "walignr1       wr0, wr10, wr11                         \n\t"
00461         "wldrd          wr10, [%[block]]                        \n\t"
00462          WAVG2B"        wr8, wr0, wr4                           \n\t"
00463          WAVG2B"        wr8, wr8, wr10                          \n\t"
00464         "wstrd          wr8, [%[block]]                         \n\t"
00465         "add            %[block], %[block], %[line_size]        \n\t"
00466 
00467         "subs           %[h], %[h], #2                          \n\t"
00468         "pld            [%[block]]                              \n\t"
00469         "bne            1b                                      \n\t"
00470         : [h]"+r"(h), [pixels]"+r"(pixels), [block]"+r"(block), [line_size]"+r"(stride)
00471         :
00472         : "cc", "memory", "r12");
00473 }
00474 
00475 void DEF(put, pixels16_y2)(uint8_t *block, const uint8_t *pixels, const int line_size, int h)
00476 {
00477     int stride = line_size;
00478     // [wr0 wr1 wr2 wr3] for previous line
00479     // [wr4 wr5 wr6 wr7] for current line
00480     __asm__ __volatile__(
00481         "pld [%[pixels]]                \n\t"
00482         "pld [%[pixels], #32]           \n\t"
00483         "and r12, %[pixels], #7         \n\t"
00484         "tmcr wcgr1, r12                \n\t"
00485         "bic %[pixels], %[pixels], #7   \n\t"
00486 
00487         "wldrd wr10, [%[pixels]]        \n\t"
00488         "wldrd wr11, [%[pixels], #8]    \n\t"
00489         "wldrd wr12, [%[pixels], #16]   \n\t"
00490         "add %[pixels], %[pixels], %[line_size] \n\t"
00491         "pld [%[pixels]]                \n\t"
00492         "pld [%[pixels], #32]           \n\t"
00493         "walignr1 wr0, wr10, wr11       \n\t"
00494         "walignr1 wr1, wr11, wr12       \n\t"
00495 
00496         "1:                             \n\t"
00497         "wldrd wr10, [%[pixels]]        \n\t"
00498         "wldrd wr11, [%[pixels], #8]    \n\t"
00499         "wldrd wr12, [%[pixels], #16]   \n\t"
00500         "add %[pixels], %[pixels], %[line_size] \n\t"
00501         "pld [%[pixels]]                \n\t"
00502         "pld [%[pixels], #32]           \n\t"
00503         "walignr1 wr4, wr10, wr11       \n\t"
00504         "walignr1 wr5, wr11, wr12       \n\t"
00505         WAVG2B" wr8, wr0, wr4           \n\t"
00506         WAVG2B" wr9, wr1, wr5           \n\t"
00507         "wstrd wr8, [%[block]]          \n\t"
00508         "wstrd wr9, [%[block], #8]      \n\t"
00509         "add %[block], %[block], %[line_size]   \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         WAVG2B" wr8, wr0, wr4           \n\t"
00520         WAVG2B" wr9, wr1, wr5           \n\t"
00521         "wstrd wr8, [%[block]]          \n\t"
00522         "wstrd wr9, [%[block], #8]      \n\t"
00523         "add %[block], %[block], %[line_size]   \n\t"
00524 
00525         "subs %[h], %[h], #2            \n\t"
00526         "bne 1b                         \n\t"
00527         : [h]"+r"(h), [pixels]"+r"(pixels), [block]"+r"(block), [line_size]"+r"(stride)
00528         :
00529         : "r4", "r5", "r12", "memory");
00530 }
00531 
00532 void DEF(avg, pixels16_y2)(uint8_t *block, const uint8_t *pixels, const int line_size, int h)
00533 {
00534     int stride = line_size;
00535     // [wr0 wr1 wr2 wr3] for previous line
00536     // [wr4 wr5 wr6 wr7] for current line
00537     __asm__ __volatile__(
00538         "pld [%[pixels]]                \n\t"
00539         "pld [%[pixels], #32]           \n\t"
00540         "and r12, %[pixels], #7         \n\t"
00541         "tmcr wcgr1, r12                \n\t"
00542         "bic %[pixels], %[pixels], #7   \n\t"
00543 
00544         "wldrd wr10, [%[pixels]]        \n\t"
00545         "wldrd wr11, [%[pixels], #8]    \n\t"
00546         "pld [%[block]]                 \n\t"
00547         "wldrd wr12, [%[pixels], #16]   \n\t"
00548         "add %[pixels], %[pixels], %[line_size] \n\t"
00549         "pld [%[pixels]]                \n\t"
00550         "pld [%[pixels], #32]           \n\t"
00551         "walignr1 wr0, wr10, wr11       \n\t"
00552         "walignr1 wr1, wr11, wr12       \n\t"
00553 
00554         "1:                             \n\t"
00555         "wldrd wr10, [%[pixels]]        \n\t"
00556         "wldrd wr11, [%[pixels], #8]    \n\t"
00557         "wldrd wr12, [%[pixels], #16]   \n\t"
00558         "add %[pixels], %[pixels], %[line_size] \n\t"
00559         "pld [%[pixels]]                \n\t"
00560         "pld [%[pixels], #32]           \n\t"
00561         "walignr1 wr4, wr10, wr11       \n\t"
00562         "walignr1 wr5, wr11, wr12       \n\t"
00563         "wldrd wr10, [%[block]]         \n\t"
00564         "wldrd wr11, [%[block], #8]     \n\t"
00565         WAVG2B" wr8, wr0, wr4           \n\t"
00566         WAVG2B" wr9, wr1, wr5           \n\t"
00567         WAVG2B" wr8, wr8, wr10          \n\t"
00568         WAVG2B" wr9, wr9, wr11          \n\t"
00569         "wstrd wr8, [%[block]]          \n\t"
00570         "wstrd wr9, [%[block], #8]      \n\t"
00571         "add %[block], %[block], %[line_size]   \n\t"
00572 
00573         "wldrd wr10, [%[pixels]]        \n\t"
00574         "wldrd wr11, [%[pixels], #8]    \n\t"
00575         "pld [%[block]]                 \n\t"
00576         "wldrd wr12, [%[pixels], #16]   \n\t"
00577         "add %[pixels], %[pixels], %[line_size] \n\t"
00578         "pld [%[pixels]]                \n\t"
00579         "pld [%[pixels], #32]           \n\t"
00580         "walignr1 wr0, wr10, wr11       \n\t"
00581         "walignr1 wr1, wr11, wr12       \n\t"
00582         "wldrd wr10, [%[block]]         \n\t"
00583         "wldrd wr11, [%[block], #8]     \n\t"
00584         WAVG2B" wr8, wr0, wr4           \n\t"
00585         WAVG2B" wr9, wr1, wr5           \n\t"
00586         WAVG2B" wr8, wr8, wr10          \n\t"
00587         WAVG2B" wr9, wr9, wr11          \n\t"
00588         "wstrd wr8, [%[block]]          \n\t"
00589         "wstrd wr9, [%[block], #8]      \n\t"
00590         "add %[block], %[block], %[line_size]   \n\t"
00591 
00592         "subs %[h], %[h], #2            \n\t"
00593         "pld [%[block]]                 \n\t"
00594         "bne 1b                         \n\t"
00595         : [h]"+r"(h), [pixels]"+r"(pixels), [block]"+r"(block), [line_size]"+r"(stride)
00596         :
00597         : "r4", "r5", "r12", "memory");
00598 }
00599 
00600 void DEF(put, pixels8_xy2)(uint8_t *block, const uint8_t *pixels, const int line_size, int h)
00601 {
00602     // [wr0 wr1 wr2 wr3] for previous line
00603     // [wr4 wr5 wr6 wr7] for current line
00604     SET_RND(wr15); // =2 for rnd  and  =1 for no_rnd version
00605     __asm__ __volatile__(
00606         "pld [%[pixels]]                \n\t"
00607         "mov r12, #2                    \n\t"
00608         "pld [%[pixels], #32]           \n\t"
00609         "tmcr wcgr0, r12                \n\t" /* for shift value */
00610         "and r12, %[pixels], #7         \n\t"
00611         "bic %[pixels], %[pixels], #7   \n\t"
00612         "tmcr wcgr1, r12                \n\t"
00613 
00614         // [wr0 wr1 wr2 wr3] <= *
00615         // [wr4 wr5 wr6 wr7]
00616         "wldrd wr12, [%[pixels]]        \n\t"
00617         "add r12, r12, #1               \n\t"
00618         "wldrd wr13, [%[pixels], #8]    \n\t"
00619         "tmcr wcgr2, r12                \n\t"
00620         "add %[pixels], %[pixels], %[line_size] \n\t"
00621         "cmp r12, #8                    \n\t"
00622         "pld [%[pixels]]                \n\t"
00623         "pld [%[pixels], #32]           \n\t"
00624         "walignr1 wr2, wr12, wr13       \n\t"
00625         "wmoveq wr10, wr13              \n\t"
00626         "walignr2ne wr10, wr12, wr13    \n\t"
00627         "wunpckelub wr0, wr2            \n\t"
00628         "wunpckehub wr1, wr2            \n\t"
00629         "wunpckelub wr8, wr10           \n\t"
00630         "wunpckehub wr9, wr10           \n\t"
00631         "waddhus wr0, wr0, wr8          \n\t"
00632         "waddhus wr1, wr1, wr9          \n\t"
00633 
00634         "1:                             \n\t"
00635         // [wr0 wr1 wr2 wr3]
00636         // [wr4 wr5 wr6 wr7] <= *
00637         "wldrd wr12, [%[pixels]]        \n\t"
00638         "cmp r12, #8                    \n\t"
00639         "wldrd wr13, [%[pixels], #8]    \n\t"
00640         "add %[pixels], %[pixels], %[line_size] \n\t"
00641         "walignr1 wr6, wr12, wr13       \n\t"
00642         "pld [%[pixels]]                \n\t"
00643         "pld [%[pixels], #32]           \n\t"
00644         "wmoveq wr10, wr13              \n\t"
00645         "walignr2ne wr10, wr12, wr13    \n\t"
00646         "wunpckelub wr4, wr6            \n\t"
00647         "wunpckehub wr5, wr6            \n\t"
00648         "wunpckelub wr8, wr10           \n\t"
00649         "wunpckehub wr9, wr10           \n\t"
00650         "waddhus wr4, wr4, wr8          \n\t"
00651         "waddhus wr5, wr5, wr9          \n\t"
00652         "waddhus wr8, wr0, wr4          \n\t"
00653         "waddhus wr9, wr1, wr5          \n\t"
00654         "waddhus wr8, wr8, wr15         \n\t"
00655         "waddhus wr9, wr9, wr15         \n\t"
00656         "wsrlhg wr8, wr8, wcgr0         \n\t"
00657         "wsrlhg wr9, wr9, wcgr0         \n\t"
00658         "wpackhus wr8, wr8, wr9         \n\t"
00659         "wstrd wr8, [%[block]]          \n\t"
00660         "add %[block], %[block], %[line_size]   \n\t"
00661 
00662         // [wr0 wr1 wr2 wr3] <= *
00663         // [wr4 wr5 wr6 wr7]
00664         "wldrd wr12, [%[pixels]]        \n\t"
00665         "wldrd wr13, [%[pixels], #8]    \n\t"
00666         "add %[pixels], %[pixels], %[line_size] \n\t"
00667         "walignr1 wr2, wr12, wr13       \n\t"
00668         "pld [%[pixels]]                \n\t"
00669         "pld [%[pixels], #32]           \n\t"
00670         "wmoveq wr10, wr13              \n\t"
00671         "walignr2ne wr10, wr12, wr13    \n\t"
00672         "wunpckelub wr0, wr2            \n\t"
00673         "wunpckehub wr1, wr2            \n\t"
00674         "wunpckelub wr8, wr10           \n\t"
00675         "wunpckehub wr9, wr10           \n\t"
00676         "waddhus wr0, wr0, wr8          \n\t"
00677         "waddhus wr1, wr1, wr9          \n\t"
00678         "waddhus wr8, wr0, wr4          \n\t"
00679         "waddhus wr9, wr1, wr5          \n\t"
00680         "waddhus wr8, wr8, wr15         \n\t"
00681         "waddhus wr9, wr9, wr15         \n\t"
00682         "wsrlhg wr8, wr8, wcgr0         \n\t"
00683         "wsrlhg wr9, wr9, wcgr0         \n\t"
00684         "wpackhus wr8, wr8, wr9         \n\t"
00685         "subs %[h], %[h], #2            \n\t"
00686         "wstrd wr8, [%[block]]          \n\t"
00687         "add %[block], %[block], %[line_size]   \n\t"
00688         "bne 1b                         \n\t"
00689         : [h]"+r"(h), [pixels]"+r"(pixels), [block]"+r"(block)
00690         : [line_size]"r"(line_size)
00691         : "r12", "memory");
00692 }
00693 
00694 void DEF(put, pixels16_xy2)(uint8_t *block, const uint8_t *pixels, const int line_size, int h)
00695 {
00696     // [wr0 wr1 wr2 wr3] for previous line
00697     // [wr4 wr5 wr6 wr7] for current line
00698     SET_RND(wr15); // =2 for rnd  and  =1 for no_rnd version
00699     __asm__ __volatile__(
00700         "pld [%[pixels]]                \n\t"
00701         "mov r12, #2                    \n\t"
00702         "pld [%[pixels], #32]           \n\t"
00703         "tmcr wcgr0, r12                \n\t" /* for shift value */
00704         /* alignment */
00705         "and r12, %[pixels], #7         \n\t"
00706         "bic %[pixels], %[pixels], #7   \n\t"
00707         "tmcr wcgr1, r12                \n\t"
00708         "add r12, r12, #1               \n\t"
00709         "tmcr wcgr2, r12                \n\t"
00710 
00711         // [wr0 wr1 wr2 wr3] <= *
00712         // [wr4 wr5 wr6 wr7]
00713         "wldrd wr12, [%[pixels]]        \n\t"
00714         "cmp r12, #8                    \n\t"
00715         "wldrd wr13, [%[pixels], #8]    \n\t"
00716         "wldrd wr14, [%[pixels], #16]   \n\t"
00717         "add %[pixels], %[pixels], %[line_size] \n\t"
00718         "pld [%[pixels]]                \n\t"
00719         "walignr1 wr2, wr12, wr13       \n\t"
00720         "pld [%[pixels], #32]           \n\t"
00721         "walignr1 wr3, wr13, wr14       \n\t"
00722         "wmoveq wr10, wr13              \n\t"
00723         "wmoveq wr11, wr14              \n\t"
00724         "walignr2ne wr10, wr12, wr13    \n\t"
00725         "walignr2ne wr11, wr13, wr14    \n\t"
00726         "wunpckelub wr0, wr2            \n\t"
00727         "wunpckehub wr1, wr2            \n\t"
00728         "wunpckelub wr2, wr3            \n\t"
00729         "wunpckehub wr3, wr3            \n\t"
00730         "wunpckelub wr8, wr10           \n\t"
00731         "wunpckehub wr9, wr10           \n\t"
00732         "wunpckelub wr10, wr11          \n\t"
00733         "wunpckehub wr11, wr11          \n\t"
00734         "waddhus wr0, wr0, wr8          \n\t"
00735         "waddhus wr1, wr1, wr9          \n\t"
00736         "waddhus wr2, wr2, wr10         \n\t"
00737         "waddhus wr3, wr3, wr11         \n\t"
00738 
00739         "1:                             \n\t"
00740         // [wr0 wr1 wr2 wr3]
00741         // [wr4 wr5 wr6 wr7] <= *
00742         "wldrd wr12, [%[pixels]]        \n\t"
00743         "cmp r12, #8                    \n\t"
00744         "wldrd wr13, [%[pixels], #8]    \n\t"
00745         "wldrd wr14, [%[pixels], #16]   \n\t"
00746         "add %[pixels], %[pixels], %[line_size] \n\t"
00747         "walignr1 wr6, wr12, wr13       \n\t"
00748         "pld [%[pixels]]                \n\t"
00749         "pld [%[pixels], #32]           \n\t"
00750         "walignr1 wr7, wr13, wr14       \n\t"
00751         "wmoveq wr10, wr13              \n\t"
00752         "wmoveq wr11, wr14              \n\t"
00753         "walignr2ne wr10, wr12, wr13    \n\t"
00754         "walignr2ne wr11, wr13, wr14    \n\t"
00755         "wunpckelub wr4, wr6            \n\t"
00756         "wunpckehub wr5, wr6            \n\t"
00757         "wunpckelub wr6, wr7            \n\t"
00758         "wunpckehub wr7, wr7            \n\t"
00759         "wunpckelub wr8, wr10           \n\t"
00760         "wunpckehub wr9, wr10           \n\t"
00761         "wunpckelub wr10, wr11          \n\t"
00762         "wunpckehub wr11, wr11          \n\t"
00763         "waddhus wr4, wr4, wr8          \n\t"
00764         "waddhus wr5, wr5, wr9          \n\t"
00765         "waddhus wr6, wr6, wr10         \n\t"
00766         "waddhus wr7, wr7, wr11         \n\t"
00767         "waddhus wr8, wr0, wr4          \n\t"
00768         "waddhus wr9, wr1, wr5          \n\t"
00769         "waddhus wr10, wr2, wr6         \n\t"
00770         "waddhus wr11, wr3, wr7         \n\t"
00771         "waddhus wr8, wr8, wr15         \n\t"
00772         "waddhus wr9, wr9, wr15         \n\t"
00773         "waddhus wr10, wr10, wr15       \n\t"
00774         "waddhus wr11, wr11, wr15       \n\t"
00775         "wsrlhg wr8, wr8, wcgr0         \n\t"
00776         "wsrlhg wr9, wr9, wcgr0         \n\t"
00777         "wsrlhg wr10, wr10, wcgr0       \n\t"
00778         "wsrlhg wr11, wr11, wcgr0       \n\t"
00779         "wpackhus wr8, wr8, wr9         \n\t"
00780         "wpackhus wr9, wr10, wr11       \n\t"
00781         "wstrd wr8, [%[block]]          \n\t"
00782         "wstrd wr9, [%[block], #8]      \n\t"
00783         "add %[block], %[block], %[line_size]   \n\t"
00784 
00785         // [wr0 wr1 wr2 wr3] <= *
00786         // [wr4 wr5 wr6 wr7]
00787         "wldrd wr12, [%[pixels]]        \n\t"
00788         "wldrd wr13, [%[pixels], #8]    \n\t"
00789         "wldrd wr14, [%[pixels], #16]   \n\t"
00790         "add %[pixels], %[pixels], %[line_size] \n\t"
00791         "walignr1 wr2, wr12, wr13       \n\t"
00792         "pld [%[pixels]]                \n\t"
00793         "pld [%[pixels], #32]           \n\t"
00794         "walignr1 wr3, wr13, wr14       \n\t"
00795         "wmoveq wr10, wr13              \n\t"
00796         "wmoveq wr11, wr14              \n\t"
00797         "walignr2ne wr10, wr12, wr13    \n\t"
00798         "walignr2ne wr11, wr13, wr14    \n\t"
00799         "wunpckelub wr0, wr2            \n\t"
00800         "wunpckehub wr1, wr2            \n\t"
00801         "wunpckelub wr2, wr3            \n\t"
00802         "wunpckehub wr3, wr3            \n\t"
00803         "wunpckelub wr8, wr10           \n\t"
00804         "wunpckehub wr9, wr10           \n\t"
00805         "wunpckelub wr10, wr11          \n\t"
00806         "wunpckehub wr11, wr11          \n\t"
00807         "waddhus wr0, wr0, wr8          \n\t"
00808         "waddhus wr1, wr1, wr9          \n\t"
00809         "waddhus wr2, wr2, wr10         \n\t"
00810         "waddhus wr3, wr3, wr11         \n\t"
00811         "waddhus wr8, wr0, wr4          \n\t"
00812         "waddhus wr9, wr1, wr5          \n\t"
00813         "waddhus wr10, wr2, wr6         \n\t"
00814         "waddhus wr11, wr3, wr7         \n\t"
00815         "waddhus wr8, wr8, wr15         \n\t"
00816         "waddhus wr9, wr9, wr15         \n\t"
00817         "waddhus wr10, wr10, wr15       \n\t"
00818         "waddhus wr11, wr11, wr15       \n\t"
00819         "wsrlhg wr8, wr8, wcgr0         \n\t"
00820         "wsrlhg wr9, wr9, wcgr0         \n\t"
00821         "wsrlhg wr10, wr10, wcgr0       \n\t"
00822         "wsrlhg wr11, wr11, wcgr0       \n\t"
00823         "wpackhus wr8, wr8, wr9         \n\t"
00824         "wpackhus wr9, wr10, wr11       \n\t"
00825         "wstrd wr8, [%[block]]          \n\t"
00826         "wstrd wr9, [%[block], #8]      \n\t"
00827         "add %[block], %[block], %[line_size]   \n\t"
00828 
00829         "subs %[h], %[h], #2            \n\t"
00830         "bne 1b                         \n\t"
00831         : [h]"+r"(h), [pixels]"+r"(pixels), [block]"+r"(block)
00832         : [line_size]"r"(line_size)
00833         : "r12", "memory");
00834 }
00835 
00836 void DEF(avg, pixels8_xy2)(uint8_t *block, const uint8_t *pixels, const int line_size, int h)
00837 {
00838     // [wr0 wr1 wr2 wr3] for previous line
00839     // [wr4 wr5 wr6 wr7] for current line
00840     SET_RND(wr15); // =2 for rnd  and  =1 for no_rnd version
00841     __asm__ __volatile__(
00842         "pld [%[block]]                 \n\t"
00843         "pld [%[block], #32]            \n\t"
00844         "pld [%[pixels]]                \n\t"
00845         "mov r12, #2                    \n\t"
00846         "pld [%[pixels], #32]           \n\t"
00847         "tmcr wcgr0, r12                \n\t" /* for shift value */
00848         "and r12, %[pixels], #7         \n\t"
00849         "bic %[pixels], %[pixels], #7   \n\t"
00850         "tmcr wcgr1, r12                \n\t"
00851 
00852         // [wr0 wr1 wr2 wr3] <= *
00853         // [wr4 wr5 wr6 wr7]
00854         "wldrd wr12, [%[pixels]]        \n\t"
00855         "add r12, r12, #1               \n\t"
00856         "wldrd wr13, [%[pixels], #8]    \n\t"
00857         "tmcr wcgr2, r12                \n\t"
00858         "add %[pixels], %[pixels], %[line_size] \n\t"
00859         "cmp r12, #8                    \n\t"
00860         "pld [%[pixels]]                \n\t"
00861         "pld [%[pixels], #32]           \n\t"
00862         "walignr1 wr2, wr12, wr13       \n\t"
00863         "wmoveq wr10, wr13              \n\t"
00864         "walignr2ne wr10, wr12, wr13    \n\t"
00865         "wunpckelub wr0, wr2            \n\t"
00866         "wunpckehub wr1, wr2            \n\t"
00867         "wunpckelub wr8, wr10           \n\t"
00868         "wunpckehub wr9, wr10           \n\t"
00869         "waddhus wr0, wr0, wr8          \n\t"
00870         "waddhus wr1, wr1, wr9          \n\t"
00871 
00872         "1:                             \n\t"
00873         // [wr0 wr1 wr2 wr3]
00874         // [wr4 wr5 wr6 wr7] <= *
00875         "wldrd wr12, [%[pixels]]        \n\t"
00876         "cmp r12, #8                    \n\t"
00877         "wldrd wr13, [%[pixels], #8]    \n\t"
00878         "add %[pixels], %[pixels], %[line_size] \n\t"
00879         "walignr1 wr6, wr12, wr13       \n\t"
00880         "pld [%[pixels]]                \n\t"
00881         "pld [%[pixels], #32]           \n\t"
00882         "wmoveq wr10, wr13              \n\t"
00883         "walignr2ne wr10, wr12, wr13    \n\t"
00884         "wunpckelub wr4, wr6            \n\t"
00885         "wunpckehub wr5, wr6            \n\t"
00886         "wunpckelub wr8, wr10           \n\t"
00887         "wunpckehub wr9, wr10           \n\t"
00888         "waddhus wr4, wr4, wr8          \n\t"
00889         "waddhus wr5, wr5, wr9          \n\t"
00890         "waddhus wr8, wr0, wr4          \n\t"
00891         "waddhus wr9, wr1, wr5          \n\t"
00892         "waddhus wr8, wr8, wr15         \n\t"
00893         "waddhus wr9, wr9, wr15         \n\t"
00894         "wldrd wr12, [%[block]]         \n\t"
00895         "wsrlhg wr8, wr8, wcgr0         \n\t"
00896         "wsrlhg wr9, wr9, wcgr0         \n\t"
00897         "wpackhus wr8, wr8, wr9         \n\t"
00898         WAVG2B" wr8, wr8, wr12          \n\t"
00899         "wstrd wr8, [%[block]]          \n\t"
00900         "add %[block], %[block], %[line_size]   \n\t"
00901         "wldrd wr12, [%[pixels]]        \n\t"
00902         "pld [%[block]]                 \n\t"
00903         "pld [%[block], #32]            \n\t"
00904 
00905         // [wr0 wr1 wr2 wr3] <= *
00906         // [wr4 wr5 wr6 wr7]
00907         "wldrd wr13, [%[pixels], #8]    \n\t"
00908         "add %[pixels], %[pixels], %[line_size] \n\t"
00909         "walignr1 wr2, wr12, wr13       \n\t"
00910         "pld [%[pixels]]                \n\t"
00911         "pld [%[pixels], #32]           \n\t"
00912         "wmoveq wr10, wr13              \n\t"
00913         "walignr2ne wr10, wr12, wr13    \n\t"
00914         "wunpckelub wr0, wr2            \n\t"
00915         "wunpckehub wr1, wr2            \n\t"
00916         "wunpckelub wr8, wr10           \n\t"
00917         "wunpckehub wr9, wr10           \n\t"
00918         "waddhus wr0, wr0, wr8          \n\t"
00919         "waddhus wr1, wr1, wr9          \n\t"
00920         "waddhus wr8, wr0, wr4          \n\t"
00921         "waddhus wr9, wr1, wr5          \n\t"
00922         "waddhus wr8, wr8, wr15         \n\t"
00923         "waddhus wr9, wr9, wr15         \n\t"
00924         "wldrd wr12, [%[block]]         \n\t"
00925         "wsrlhg wr8, wr8, wcgr0         \n\t"
00926         "wsrlhg wr9, wr9, wcgr0         \n\t"
00927         "wpackhus wr8, wr8, wr9         \n\t"
00928         "subs %[h], %[h], #2            \n\t"
00929         WAVG2B" wr8, wr8, wr12          \n\t"
00930         "wstrd wr8, [%[block]]          \n\t"
00931         "add %[block], %[block], %[line_size]   \n\t"
00932         "pld [%[block]]                 \n\t"
00933         "pld [%[block], #32]            \n\t"
00934         "bne 1b                         \n\t"
00935         : [h]"+r"(h), [pixels]"+r"(pixels), [block]"+r"(block)
00936         : [line_size]"r"(line_size)
00937         : "r12", "memory");
00938 }
00939 
00940 void DEF(avg, pixels16_xy2)(uint8_t *block, const uint8_t *pixels, const int line_size, int h)
00941 {
00942     // [wr0 wr1 wr2 wr3] for previous line
00943     // [wr4 wr5 wr6 wr7] for current line
00944     SET_RND(wr15); // =2 for rnd  and  =1 for no_rnd version
00945     __asm__ __volatile__(
00946         "pld [%[block]]                 \n\t"
00947         "pld [%[block], #32]            \n\t"
00948         "pld [%[pixels]]                \n\t"
00949         "mov r12, #2                    \n\t"
00950         "pld [%[pixels], #32]           \n\t"
00951         "tmcr wcgr0, r12                \n\t" /* for shift value */
00952         /* alignment */
00953         "and r12, %[pixels], #7         \n\t"
00954         "bic %[pixels], %[pixels], #7           \n\t"
00955         "tmcr wcgr1, r12                \n\t"
00956         "add r12, r12, #1               \n\t"
00957         "tmcr wcgr2, r12                \n\t"
00958 
00959         // [wr0 wr1 wr2 wr3] <= *
00960         // [wr4 wr5 wr6 wr7]
00961         "wldrd wr12, [%[pixels]]        \n\t"
00962         "cmp r12, #8                    \n\t"
00963         "wldrd wr13, [%[pixels], #8]    \n\t"
00964         "wldrd wr14, [%[pixels], #16]   \n\t"
00965         "add %[pixels], %[pixels], %[line_size] \n\t"
00966         "pld [%[pixels]]                \n\t"
00967         "walignr1 wr2, wr12, wr13       \n\t"
00968         "pld [%[pixels], #32]           \n\t"
00969         "walignr1 wr3, wr13, wr14       \n\t"
00970         "wmoveq wr10, wr13              \n\t"
00971         "wmoveq wr11, wr14              \n\t"
00972         "walignr2ne wr10, wr12, wr13    \n\t"
00973         "walignr2ne wr11, wr13, wr14    \n\t"
00974         "wunpckelub wr0, wr2            \n\t"
00975         "wunpckehub wr1, wr2            \n\t"
00976         "wunpckelub wr2, wr3            \n\t"
00977         "wunpckehub wr3, wr3            \n\t"
00978         "wunpckelub wr8, wr10           \n\t"
00979         "wunpckehub wr9, wr10           \n\t"
00980         "wunpckelub wr10, wr11          \n\t"
00981         "wunpckehub wr11, wr11          \n\t"
00982         "waddhus wr0, wr0, wr8          \n\t"
00983         "waddhus wr1, wr1, wr9          \n\t"
00984         "waddhus wr2, wr2, wr10         \n\t"
00985         "waddhus wr3, wr3, wr11         \n\t"
00986 
00987         "1:                             \n\t"
00988         // [wr0 wr1 wr2 wr3]
00989         // [wr4 wr5 wr6 wr7] <= *
00990         "wldrd wr12, [%[pixels]]        \n\t"
00991         "cmp r12, #8                    \n\t"
00992         "wldrd wr13, [%[pixels], #8]    \n\t"
00993         "wldrd wr14, [%[pixels], #16]   \n\t"
00994         "add %[pixels], %[pixels], %[line_size] \n\t"
00995         "walignr1 wr6, wr12, wr13       \n\t"
00996         "pld [%[pixels]]                \n\t"
00997         "pld [%[pixels], #32]           \n\t"
00998         "walignr1 wr7, wr13, wr14       \n\t"
00999         "wmoveq wr10, wr13              \n\t"
01000         "wmoveq wr11, wr14              \n\t"
01001         "walignr2ne wr10, wr12, wr13    \n\t"
01002         "walignr2ne wr11, wr13, wr14    \n\t"
01003         "wunpckelub wr4, wr6            \n\t"
01004         "wunpckehub wr5, wr6            \n\t"
01005         "wunpckelub wr6, wr7            \n\t"
01006         "wunpckehub wr7, wr7            \n\t"
01007         "wunpckelub wr8, wr10           \n\t"
01008         "wunpckehub wr9, wr10           \n\t"
01009         "wunpckelub wr10, wr11          \n\t"
01010         "wunpckehub wr11, wr11          \n\t"
01011         "waddhus wr4, wr4, wr8          \n\t"
01012         "waddhus wr5, wr5, wr9          \n\t"
01013         "waddhus wr6, wr6, wr10         \n\t"
01014         "waddhus wr7, wr7, wr11         \n\t"
01015         "waddhus wr8, wr0, wr4          \n\t"
01016         "waddhus wr9, wr1, wr5          \n\t"
01017         "waddhus wr10, wr2, wr6         \n\t"
01018         "waddhus wr11, wr3, wr7         \n\t"
01019         "waddhus wr8, wr8, wr15         \n\t"
01020         "waddhus wr9, wr9, wr15         \n\t"
01021         "waddhus wr10, wr10, wr15       \n\t"
01022         "waddhus wr11, wr11, wr15       \n\t"
01023         "wsrlhg wr8, wr8, wcgr0         \n\t"
01024         "wsrlhg wr9, wr9, wcgr0         \n\t"
01025         "wldrd wr12, [%[block]]         \n\t"
01026         "wldrd wr13, [%[block], #8]     \n\t"
01027         "wsrlhg wr10, wr10, wcgr0       \n\t"
01028         "wsrlhg wr11, wr11, wcgr0       \n\t"
01029         "wpackhus wr8, wr8, wr9         \n\t"
01030         "wpackhus wr9, wr10, wr11       \n\t"
01031         WAVG2B" wr8, wr8, wr12          \n\t"
01032         WAVG2B" wr9, wr9, wr13          \n\t"
01033         "wstrd wr8, [%[block]]          \n\t"
01034         "wstrd wr9, [%[block], #8]      \n\t"
01035         "add %[block], %[block], %[line_size]   \n\t"
01036 
01037         // [wr0 wr1 wr2 wr3] <= *
01038         // [wr4 wr5 wr6 wr7]
01039         "wldrd wr12, [%[pixels]]        \n\t"
01040         "pld [%[block]]                 \n\t"
01041         "wldrd wr13, [%[pixels], #8]    \n\t"
01042         "pld [%[block], #32]            \n\t"
01043         "wldrd wr14, [%[pixels], #16]   \n\t"
01044         "add %[pixels], %[pixels], %[line_size] \n\t"
01045         "walignr1 wr2, wr12, wr13       \n\t"
01046         "pld [%[pixels]]                \n\t"
01047         "pld [%[pixels], #32]           \n\t"
01048         "walignr1 wr3, wr13, wr14       \n\t"
01049         "wmoveq wr10, wr13              \n\t"
01050         "wmoveq wr11, wr14              \n\t"
01051         "walignr2ne wr10, wr12, wr13    \n\t"
01052         "walignr2ne wr11, wr13, wr14    \n\t"
01053         "wunpckelub wr0, wr2            \n\t"
01054         "wunpckehub wr1, wr2            \n\t"
01055         "wunpckelub wr2, wr3            \n\t"
01056         "wunpckehub wr3, wr3            \n\t"
01057         "wunpckelub wr8, wr10           \n\t"
01058         "wunpckehub wr9, wr10           \n\t"
01059         "wunpckelub wr10, wr11          \n\t"
01060         "wunpckehub wr11, wr11          \n\t"
01061         "waddhus wr0, wr0, wr8          \n\t"
01062         "waddhus wr1, wr1, wr9          \n\t"
01063         "waddhus wr2, wr2, wr10         \n\t"
01064         "waddhus wr3, wr3, wr11         \n\t"
01065         "waddhus wr8, wr0, wr4          \n\t"
01066         "waddhus wr9, wr1, wr5          \n\t"
01067         "waddhus wr10, wr2, wr6         \n\t"
01068         "waddhus wr11, wr3, wr7         \n\t"
01069         "waddhus wr8, wr8, wr15         \n\t"
01070         "waddhus wr9, wr9, wr15         \n\t"
01071         "waddhus wr10, wr10, wr15       \n\t"
01072         "waddhus wr11, wr11, wr15       \n\t"
01073         "wsrlhg wr8, wr8, wcgr0         \n\t"
01074         "wsrlhg wr9, wr9, wcgr0         \n\t"
01075         "wldrd wr12, [%[block]]         \n\t"
01076         "wldrd wr13, [%[block], #8]     \n\t"
01077         "wsrlhg wr10, wr10, wcgr0       \n\t"
01078         "wsrlhg wr11, wr11, wcgr0       \n\t"
01079         "wpackhus wr8, wr8, wr9         \n\t"
01080         "wpackhus wr9, wr10, wr11       \n\t"
01081         WAVG2B" wr8, wr8, wr12          \n\t"
01082         WAVG2B" wr9, wr9, wr13          \n\t"
01083         "wstrd wr8, [%[block]]          \n\t"
01084         "wstrd wr9, [%[block], #8]      \n\t"
01085         "add %[block], %[block], %[line_size]   \n\t"
01086         "subs %[h], %[h], #2            \n\t"
01087         "pld [%[block]]                 \n\t"
01088         "pld [%[block], #32]            \n\t"
01089         "bne 1b                         \n\t"
01090         : [h]"+r"(h), [pixels]"+r"(pixels), [block]"+r"(block)
01091         : [line_size]"r"(line_size)
01092         : "r12", "memory");
01093 }
hvirtual/quicktime/ffmpeg/libavcodec/armv4l/dsputil_iwmmxt_rnd.h
