OLD | NEW |
1 /* | 1 /* |
2 * Copyright (c) 2013 The WebM project authors. All Rights Reserved. | 2 * Copyright (c) 2013 The WebM project authors. All Rights Reserved. |
3 * | 3 * |
4 * Use of this source code is governed by a BSD-style license | 4 * Use of this source code is governed by a BSD-style license |
5 * that can be found in the LICENSE file in the root of the source | 5 * that can be found in the LICENSE file in the root of the source |
6 * tree. An additional intellectual property rights grant can be found | 6 * tree. An additional intellectual property rights grant can be found |
7 * in the file PATENTS. All contributing project authors may | 7 * in the file PATENTS. All contributing project authors may |
8 * be found in the AUTHORS file in the root of the source tree. | 8 * be found in the AUTHORS file in the root of the source tree. |
9 */ | 9 */ |
10 | 10 |
11 #include <assert.h> | 11 #include <assert.h> |
12 #include <stdio.h> | 12 #include <stdio.h> |
13 | 13 |
14 #include "./vpx_dsp_rtcd.h" | 14 #include "./vpx_dsp_rtcd.h" |
15 #include "vpx_dsp/mips/vpx_common_dspr2.h" | 15 #include "vpx_dsp/mips/convolve_common_dspr2.h" |
16 #include "vpx_dsp/vpx_convolve.h" | 16 #include "vpx_dsp/vpx_convolve.h" |
17 #include "vpx_dsp/vpx_dsp_common.h" | 17 #include "vpx_dsp/vpx_dsp_common.h" |
18 #include "vpx_ports/mem.h" | 18 #include "vpx_ports/mem.h" |
19 | 19 |
20 #if HAVE_DSPR2 | 20 #if HAVE_DSPR2 |
21 static void convolve_avg_horiz_4_dspr2(const uint8_t *src, | 21 static void convolve_avg_horiz_4_dspr2(const uint8_t *src, |
22 int32_t src_stride, | 22 int32_t src_stride, |
23 uint8_t *dst, | 23 uint8_t *dst, |
24 int32_t dst_stride, | 24 int32_t dst_stride, |
25 const int16_t *filter_x0, | 25 const int16_t *filter_x0, |
(...skipping 924 matching lines...) Expand 10 before | Expand all | Expand 10 after Loading... |
950 src_ptr += src_stride; | 950 src_ptr += src_stride; |
951 dst_ptr += dst_stride; | 951 dst_ptr += dst_stride; |
952 } | 952 } |
953 } | 953 } |
954 | 954 |
955 void vpx_convolve8_avg_horiz_dspr2(const uint8_t *src, ptrdiff_t src_stride, | 955 void vpx_convolve8_avg_horiz_dspr2(const uint8_t *src, ptrdiff_t src_stride, |
956 uint8_t *dst, ptrdiff_t dst_stride, | 956 uint8_t *dst, ptrdiff_t dst_stride, |
957 const int16_t *filter_x, int x_step_q4, | 957 const int16_t *filter_x, int x_step_q4, |
958 const int16_t *filter_y, int y_step_q4, | 958 const int16_t *filter_y, int y_step_q4, |
959 int w, int h) { | 959 int w, int h) { |
960 if (((const int32_t *)filter_x)[1] == 0x800000) { | 960 assert(x_step_q4 == 16); |
961 vpx_convolve_avg(src, src_stride, | 961 assert(((const int32_t *)filter_x)[1] != 0x800000); |
962 dst, dst_stride, | 962 |
963 filter_x, x_step_q4, | 963 if (((const int32_t *)filter_x)[0] == 0) { |
964 filter_y, y_step_q4, | |
965 w, h); | |
966 } else if (((const int32_t *)filter_x)[0] == 0) { | |
967 vpx_convolve2_avg_horiz_dspr2(src, src_stride, | 964 vpx_convolve2_avg_horiz_dspr2(src, src_stride, |
968 dst, dst_stride, | 965 dst, dst_stride, |
969 filter_x, x_step_q4, | 966 filter_x, x_step_q4, |
970 filter_y, y_step_q4, | 967 filter_y, y_step_q4, |
971 w, h); | 968 w, h); |
972 } else { | 969 } else { |
973 if (16 == x_step_q4) { | 970 uint32_t pos = 38; |
974 uint32_t pos = 38; | |
975 | 971 |
976 src -= 3; | 972 src -= 3; |
977 | 973 |
978 /* bit positon for extract from acc */ | 974 /* bit positon for extract from acc */ |
979 __asm__ __volatile__ ( | 975 __asm__ __volatile__ ( |
980 "wrdsp %[pos], 1 \n\t" | 976 "wrdsp %[pos], 1 \n\t" |
981 : | 977 : |
982 : [pos] "r" (pos) | 978 : [pos] "r" (pos) |
983 ); | 979 ); |
984 | 980 |
985 /* prefetch data to cache memory */ | 981 /* prefetch data to cache memory */ |
986 prefetch_load(src); | 982 prefetch_load(src); |
987 prefetch_load(src + 32); | 983 prefetch_load(src + 32); |
988 prefetch_store(dst); | 984 prefetch_store(dst); |
989 | 985 |
990 switch (w) { | 986 switch (w) { |
991 case 4: | 987 case 4: |
992 convolve_avg_horiz_4_dspr2(src, src_stride, | 988 convolve_avg_horiz_4_dspr2(src, src_stride, |
993 dst, dst_stride, | 989 dst, dst_stride, |
994 filter_x, h); | 990 filter_x, h); |
995 break; | 991 break; |
996 case 8: | 992 case 8: |
997 convolve_avg_horiz_8_dspr2(src, src_stride, | 993 convolve_avg_horiz_8_dspr2(src, src_stride, |
998 dst, dst_stride, | 994 dst, dst_stride, |
999 filter_x, h); | 995 filter_x, h); |
1000 break; | 996 break; |
1001 case 16: | 997 case 16: |
1002 convolve_avg_horiz_16_dspr2(src, src_stride, | 998 convolve_avg_horiz_16_dspr2(src, src_stride, |
1003 dst, dst_stride, | 999 dst, dst_stride, |
1004 filter_x, h, 1); | 1000 filter_x, h, 1); |
1005 break; | 1001 break; |
1006 case 32: | 1002 case 32: |
1007 convolve_avg_horiz_16_dspr2(src, src_stride, | 1003 convolve_avg_horiz_16_dspr2(src, src_stride, |
1008 dst, dst_stride, | 1004 dst, dst_stride, |
1009 filter_x, h, 2); | 1005 filter_x, h, 2); |
1010 break; | 1006 break; |
1011 case 64: | 1007 case 64: |
1012 prefetch_load(src + 64); | 1008 prefetch_load(src + 64); |
1013 prefetch_store(dst + 32); | 1009 prefetch_store(dst + 32); |
1014 | 1010 |
1015 convolve_avg_horiz_64_dspr2(src, src_stride, | 1011 convolve_avg_horiz_64_dspr2(src, src_stride, |
1016 dst, dst_stride, | |
1017 filter_x, h); | |
1018 break; | |
1019 default: | |
1020 vpx_convolve8_avg_horiz_c(src + 3, src_stride, | |
1021 dst, dst_stride, | 1012 dst, dst_stride, |
1022 filter_x, x_step_q4, | 1013 filter_x, h); |
1023 filter_y, y_step_q4, | 1014 break; |
1024 w, h); | 1015 default: |
1025 break; | 1016 vpx_convolve8_avg_horiz_c(src + 3, src_stride, |
1026 } | 1017 dst, dst_stride, |
1027 } else { | 1018 filter_x, x_step_q4, |
1028 vpx_convolve8_avg_horiz_c(src, src_stride, | 1019 filter_y, y_step_q4, |
1029 dst, dst_stride, | 1020 w, h); |
1030 filter_x, x_step_q4, | 1021 break; |
1031 filter_y, y_step_q4, | |
1032 w, h); | |
1033 } | 1022 } |
1034 } | 1023 } |
1035 } | 1024 } |
1036 #endif | 1025 #endif |
OLD | NEW |