Mercurial > libavcodec.hg

comparison rv30dsp.c @ 8367:84a066542cd4 libavcodec

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
Unfortunately RV30 luma thirdpel MC functions cannot be substituted with applying two separate filtering passes, so implement them explicitly.
author kostya
date Wed, 17 Dec 2008 19:37:45 +0000
parents 7185e3bb0614
children e9d9d946f213
comparison
equal deleted inserted replaced
8366:3bbfd02865d7 8367:84a066542cd4
32 const int h=8;\ 32 const int h=8;\
33 uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\ 33 uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\
34 int i;\ 34 int i;\
35 for(i=0; i<h; i++)\ 35 for(i=0; i<h; i++)\
36 {\ 36 {\
37 OP(dst[0], -(src[-1]+src[2]) + src[0]*C1 + src[1]*C2);\ 37 OP(dst[0], (-(src[-1]+src[2]) + src[0]*C1 + src[1]*C2 + 8)>>4);\
38 OP(dst[1], -(src[ 0]+src[3]) + src[1]*C1 + src[2]*C2);\ 38 OP(dst[1], (-(src[ 0]+src[3]) + src[1]*C1 + src[2]*C2 + 8)>>4);\
39 OP(dst[2], -(src[ 1]+src[4]) + src[2]*C1 + src[3]*C2);\ 39 OP(dst[2], (-(src[ 1]+src[4]) + src[2]*C1 + src[3]*C2 + 8)>>4);\
40 OP(dst[3], -(src[ 2]+src[5]) + src[3]*C1 + src[4]*C2);\ 40 OP(dst[3], (-(src[ 2]+src[5]) + src[3]*C1 + src[4]*C2 + 8)>>4);\
41 OP(dst[4], -(src[ 3]+src[6]) + src[4]*C1 + src[5]*C2);\ 41 OP(dst[4], (-(src[ 3]+src[6]) + src[4]*C1 + src[5]*C2 + 8)>>4);\
42 OP(dst[5], -(src[ 4]+src[7]) + src[5]*C1 + src[6]*C2);\ 42 OP(dst[5], (-(src[ 4]+src[7]) + src[5]*C1 + src[6]*C2 + 8)>>4);\
43 OP(dst[6], -(src[ 5]+src[8]) + src[6]*C1 + src[7]*C2);\ 43 OP(dst[6], (-(src[ 5]+src[8]) + src[6]*C1 + src[7]*C2 + 8)>>4);\
44 OP(dst[7], -(src[ 6]+src[9]) + src[7]*C1 + src[8]*C2);\ 44 OP(dst[7], (-(src[ 6]+src[9]) + src[7]*C1 + src[8]*C2 + 8)>>4);\
45 dst+=dstStride;\ 45 dst+=dstStride;\
46 src+=srcStride;\ 46 src+=srcStride;\
47 }\ 47 }\
48 }\ 48 }\
49 \ 49 \
62 const int src5= src[5 *srcStride];\ 62 const int src5= src[5 *srcStride];\
63 const int src6= src[6 *srcStride];\ 63 const int src6= src[6 *srcStride];\
64 const int src7= src[7 *srcStride];\ 64 const int src7= src[7 *srcStride];\
65 const int src8= src[8 *srcStride];\ 65 const int src8= src[8 *srcStride];\
66 const int src9= src[9 *srcStride];\ 66 const int src9= src[9 *srcStride];\
67 OP(dst[0*dstStride], -(srcA+src2) + src0*C1 + src1*C2);\ 67 OP(dst[0*dstStride], (-(srcA+src2) + src0*C1 + src1*C2 + 8)>>4);\
68 OP(dst[1*dstStride], -(src0+src3) + src1*C1 + src2*C2);\ 68 OP(dst[1*dstStride], (-(src0+src3) + src1*C1 + src2*C2 + 8)>>4);\
69 OP(dst[2*dstStride], -(src1+src4) + src2*C1 + src3*C2);\ 69 OP(dst[2*dstStride], (-(src1+src4) + src2*C1 + src3*C2 + 8)>>4);\
70 OP(dst[3*dstStride], -(src2+src5) + src3*C1 + src4*C2);\ 70 OP(dst[3*dstStride], (-(src2+src5) + src3*C1 + src4*C2 + 8)>>4);\
71 OP(dst[4*dstStride], -(src3+src6) + src4*C1 + src5*C2);\ 71 OP(dst[4*dstStride], (-(src3+src6) + src4*C1 + src5*C2 + 8)>>4);\
72 OP(dst[5*dstStride], -(src4+src7) + src5*C1 + src6*C2);\ 72 OP(dst[5*dstStride], (-(src4+src7) + src5*C1 + src6*C2 + 8)>>4);\
73 OP(dst[6*dstStride], -(src5+src8) + src6*C1 + src7*C2);\ 73 OP(dst[6*dstStride], (-(src5+src8) + src6*C1 + src7*C2 + 8)>>4);\
74 OP(dst[7*dstStride], -(src6+src9) + src7*C1 + src8*C2);\ 74 OP(dst[7*dstStride], (-(src6+src9) + src7*C1 + src8*C2 + 8)>>4);\
75 dst++;\ 75 dst++;\
76 src++;\ 76 src++;\
77 }\ 77 }\
78 }\ 78 }\
79 \ 79 \
80 static void OPNAME ## rv30_tpel8_h3_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
81 const int h=8+2;\
82 uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\
83 int i;\
84 for(i=0; i<h; i++)\
85 {\
86 OP(dst[0], 6*src[0]+9*src[1]+src[2]);\
87 OP(dst[1], 6*src[1]+9*src[2]+src[3]);\
88 OP(dst[2], 6*src[2]+9*src[3]+src[4]);\
89 OP(dst[3], 6*src[3]+9*src[4]+src[5]);\
90 OP(dst[4], 6*src[4]+9*src[5]+src[6]);\
91 OP(dst[5], 6*src[5]+9*src[6]+src[7]);\
92 OP(dst[6], 6*src[6]+9*src[7]+src[8]);\
93 OP(dst[7], 6*src[7]+9*src[8]+src[9]);\
94 dst+=dstStride;\
95 src+=srcStride;\
96 }\
97 }\
98 \
99 static void OPNAME ## rv30_tpel8_v3_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
100 const int w=8;\
101 uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\
102 int i;\
103 for(i=0; i<w; i++)\
104 {\
105 const int src0= src[0 *srcStride];\
106 const int src1= src[1 *srcStride];\
107 const int src2= src[2 *srcStride];\
108 const int src3= src[3 *srcStride];\
109 const int src4= src[4 *srcStride];\
110 const int src5= src[5 *srcStride];\
111 const int src6= src[6 *srcStride];\
112 const int src7= src[7 *srcStride];\
113 const int src8= src[8 *srcStride];\
114 const int src9= src[9 *srcStride];\
115 OP(dst[0*dstStride], 6*src0 + 9*src1 + src2);\
116 OP(dst[1*dstStride], 6*src1 + 9*src2 + src3);\
117 OP(dst[2*dstStride], 6*src2 + 9*src3 + src4);\
118 OP(dst[3*dstStride], 6*src3 + 9*src4 + src5);\
119 OP(dst[4*dstStride], 6*src4 + 9*src5 + src6);\
120 OP(dst[5*dstStride], 6*src5 + 9*src6 + src7);\
121 OP(dst[6*dstStride], 6*src6 + 9*src7 + src8);\
122 OP(dst[7*dstStride], 6*src7 + 9*src8 + src9);\
123 dst ++;\
124 src ++;\
125 }\
126 }\
127 \
128 static void OPNAME ## rv30_tpel8_hv_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\ 80 static void OPNAME ## rv30_tpel8_hv_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
129 uint8_t half[8*10];\ 81 const int w = 8;\
130 put_rv30_tpel8_h3_lowpass(half, src, 8, srcStride);\ 82 const int h = 8;\
131 OPNAME ## rv30_tpel8_v3_lowpass(dst, half, dstStride, 8);\ 83 uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\
84 int i, j;\
85 for(j = 0; j < h; j++){\
86 for(i = 0; i < w; i++){\
87 OP(dst[i], (\
88 src[srcStride*-1+i-1] -12*src[srcStride*-1+i] -6*src[srcStride*-1+i+1] +src[srcStride*-1+i+2]+\
89 -12*src[srcStride* 0+i-1] +144*src[srcStride* 0+i] +72*src[srcStride* 0+i+1] -12*src[srcStride* 0+i+2] +\
90 -6*src[srcStride* 1+i-1] +72*src[srcStride* 1+i] +36*src[srcStride* 1+i+1] -6*src[srcStride* 1+i+2] +\
91 src[srcStride* 2+i-1] -12*src[srcStride* 2+i] -6*src[srcStride* 2+i+1] +src[srcStride* 2+i+2] +\
92 128)>>8);\
93 }\
94 src += srcStride;\
95 dst += dstStride;\
96 }\
97 }\
98 \
99 static void OPNAME ## rv30_tpel8_hhv_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
100 const int w = 8;\
101 const int h = 8;\
102 uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\
103 int i, j;\
104 for(j = 0; j < h; j++){\
105 for(i = 0; i < w; i++){\
106 OP(dst[i], (\
107 src[srcStride*-1+i-1] -12*src[srcStride*-1+i+1] -6*src[srcStride*-1+i] +src[srcStride*-1+i+2]+\
108 -12*src[srcStride* 0+i-1] +144*src[srcStride* 0+i+1] +72*src[srcStride* 0+i] -12*src[srcStride* 0+i+2]+\
109 -6*src[srcStride* 1+i-1] +72*src[srcStride* 1+i+1] +36*src[srcStride* 1+i] -6*src[srcStride* 1+i+2]+\
110 src[srcStride* 2+i-1] -12*src[srcStride* 2+i+1] -6*src[srcStride* 2+i] +src[srcStride* 2+i+2]+\
111 128)>>8);\
112 }\
113 src += srcStride;\
114 dst += dstStride;\
115 }\
116 }\
117 \
118 static void OPNAME ## rv30_tpel8_hvv_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
119 const int w = 8;\
120 const int h = 8;\
121 uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\
122 int i, j;\
123 for(j = 0; j < h; j++){\
124 for(i = 0; i < w; i++){\
125 OP(dst[i], (\
126 src[srcStride*-1+i-1] -12*src[srcStride*-1+i] -6*src[srcStride*-1+i+1] +src[srcStride*-1+i+2]+\
127 -6*src[srcStride* 0+i-1] +72*src[srcStride* 0+i] +36*src[srcStride* 0+i+1] -6*src[srcStride* 0+i+2]+\
128 -12*src[srcStride* 1+i-1] +144*src[srcStride* 1+i] +72*src[srcStride* 1+i+1] -12*src[srcStride* 1+i+2]+\
129 src[srcStride* 2+i-1] -12*src[srcStride* 2+i] -6*src[srcStride* 2+i+1] +src[srcStride* 2+i+2]+\
130 128)>>8);\
131 }\
132 src += srcStride;\
133 dst += dstStride;\
134 }\
135 }\
136 \
137 static void OPNAME ## rv30_tpel8_hhvv_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
138 const int w = 8;\
139 const int h = 8;\
140 uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\
141 int i, j;\
142 for(j = 0; j < h; j++){\
143 for(i = 0; i < w; i++){\
144 OP(dst[i], (\
145 36*src[i+srcStride*0] +54*src[i+1+srcStride*0] +6*src[i+2+srcStride*0]+\
146 54*src[i+srcStride*1] +81*src[i+1+srcStride*1] +9*src[i+2+srcStride*1]+\
147 6*src[i+srcStride*2] + 9*src[i+1+srcStride*2] + src[i+2+srcStride*2]+\
148 128)>>8);\
149 }\
150 src += srcStride;\
151 dst += dstStride;\
152 }\
132 }\ 153 }\
133 \ 154 \
134 static void OPNAME ## rv30_tpel16_v_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, const int C1, const int C2){\ 155 static void OPNAME ## rv30_tpel16_v_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, const int C1, const int C2){\
135 OPNAME ## rv30_tpel8_v_lowpass(dst , src , dstStride, srcStride, C1, C2);\ 156 OPNAME ## rv30_tpel8_v_lowpass(dst , src , dstStride, srcStride, C1, C2);\
136 OPNAME ## rv30_tpel8_v_lowpass(dst+8, src+8, dstStride, srcStride, C1, C2);\ 157 OPNAME ## rv30_tpel8_v_lowpass(dst+8, src+8, dstStride, srcStride, C1, C2);\
156 dst += 8*dstStride;\ 177 dst += 8*dstStride;\
157 OPNAME ## rv30_tpel8_hv_lowpass(dst , src , dstStride, srcStride);\ 178 OPNAME ## rv30_tpel8_hv_lowpass(dst , src , dstStride, srcStride);\
158 OPNAME ## rv30_tpel8_hv_lowpass(dst+8, src+8, dstStride, srcStride);\ 179 OPNAME ## rv30_tpel8_hv_lowpass(dst+8, src+8, dstStride, srcStride);\
159 }\ 180 }\
160 \ 181 \
182 static void OPNAME ## rv30_tpel16_hhv_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
183 OPNAME ## rv30_tpel8_hhv_lowpass(dst , src , dstStride, srcStride);\
184 OPNAME ## rv30_tpel8_hhv_lowpass(dst+8, src+8, dstStride, srcStride);\
185 src += 8*srcStride;\
186 dst += 8*dstStride;\
187 OPNAME ## rv30_tpel8_hhv_lowpass(dst , src , dstStride, srcStride);\
188 OPNAME ## rv30_tpel8_hhv_lowpass(dst+8, src+8, dstStride, srcStride);\
189 }\
190 \
191 static void OPNAME ## rv30_tpel16_hvv_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
192 OPNAME ## rv30_tpel8_hvv_lowpass(dst , src , dstStride, srcStride);\
193 OPNAME ## rv30_tpel8_hvv_lowpass(dst+8, src+8, dstStride, srcStride);\
194 src += 8*srcStride;\
195 dst += 8*dstStride;\
196 OPNAME ## rv30_tpel8_hvv_lowpass(dst , src , dstStride, srcStride);\
197 OPNAME ## rv30_tpel8_hvv_lowpass(dst+8, src+8, dstStride, srcStride);\
198 }\
199 \
200 static void OPNAME ## rv30_tpel16_hhvv_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
201 OPNAME ## rv30_tpel8_hhvv_lowpass(dst , src , dstStride, srcStride);\
202 OPNAME ## rv30_tpel8_hhvv_lowpass(dst+8, src+8, dstStride, srcStride);\
203 src += 8*srcStride;\
204 dst += 8*dstStride;\
205 OPNAME ## rv30_tpel8_hhvv_lowpass(dst , src , dstStride, srcStride);\
206 OPNAME ## rv30_tpel8_hhvv_lowpass(dst+8, src+8, dstStride, srcStride);\
207 }\
208 \
161 209
162 #define RV30_MC(OPNAME, SIZE) \ 210 #define RV30_MC(OPNAME, SIZE) \
163 static void OPNAME ## rv30_tpel ## SIZE ## _mc10_c(uint8_t *dst, uint8_t *src, int stride){\ 211 static void OPNAME ## rv30_tpel ## SIZE ## _mc10_c(uint8_t *dst, uint8_t *src, int stride){\
164 OPNAME ## rv30_tpel ## SIZE ## _h_lowpass(dst, src, stride, stride, 12, 6);\ 212 OPNAME ## rv30_tpel ## SIZE ## _h_lowpass(dst, src, stride, stride, 12, 6);\
165 }\ 213 }\
175 static void OPNAME ## rv30_tpel ## SIZE ## _mc02_c(uint8_t *dst, uint8_t *src, int stride){\ 223 static void OPNAME ## rv30_tpel ## SIZE ## _mc02_c(uint8_t *dst, uint8_t *src, int stride){\
176 OPNAME ## rv30_tpel ## SIZE ## _v_lowpass(dst, src, stride, stride, 6, 12);\ 224 OPNAME ## rv30_tpel ## SIZE ## _v_lowpass(dst, src, stride, stride, 6, 12);\
177 }\ 225 }\
178 \ 226 \
179 static void OPNAME ## rv30_tpel ## SIZE ## _mc11_c(uint8_t *dst, uint8_t *src, int stride){\ 227 static void OPNAME ## rv30_tpel ## SIZE ## _mc11_c(uint8_t *dst, uint8_t *src, int stride){\
180 uint8_t half[SIZE*SIZE];\ 228 OPNAME ## rv30_tpel ## SIZE ## _hv_lowpass(dst, src, stride, stride);\
181 put_rv30_tpel ## SIZE ## _h_lowpass(half, src, SIZE, stride, 12, 6);\
182 OPNAME ## rv30_tpel ## SIZE ## _v_lowpass(dst, src, stride, stride, 12, 6);\
183 }\ 229 }\
184 \ 230 \
185 static void OPNAME ## rv30_tpel ## SIZE ## _mc12_c(uint8_t *dst, uint8_t *src, int stride){\ 231 static void OPNAME ## rv30_tpel ## SIZE ## _mc12_c(uint8_t *dst, uint8_t *src, int stride){\
186 uint8_t half[SIZE*SIZE];\ 232 OPNAME ## rv30_tpel ## SIZE ## _hvv_lowpass(dst, src, stride, stride);\
187 put_rv30_tpel ## SIZE ## _h_lowpass(half, src, SIZE, stride, 12, 6);\
188 OPNAME ## rv30_tpel ## SIZE ## _v_lowpass(dst, src, stride, stride, 6, 12);\
189 }\ 233 }\
190 \ 234 \
191 static void OPNAME ## rv30_tpel ## SIZE ## _mc21_c(uint8_t *dst, uint8_t *src, int stride){\ 235 static void OPNAME ## rv30_tpel ## SIZE ## _mc21_c(uint8_t *dst, uint8_t *src, int stride){\
192 uint8_t half[SIZE*SIZE];\ 236 OPNAME ## rv30_tpel ## SIZE ## _hhv_lowpass(dst, src, stride, stride);\
193 put_rv30_tpel ## SIZE ## _h_lowpass(half, src, SIZE, stride, 6, 12);\
194 OPNAME ## rv30_tpel ## SIZE ## _v_lowpass(dst, src, stride, stride, 12, 6);\
195 }\ 237 }\
196 \ 238 \
197 static void OPNAME ## rv30_tpel ## SIZE ## _mc22_c(uint8_t *dst, uint8_t *src, int stride){\ 239 static void OPNAME ## rv30_tpel ## SIZE ## _mc22_c(uint8_t *dst, uint8_t *src, int stride){\
198 OPNAME ## rv30_tpel ## SIZE ## _hv_lowpass(dst, src, stride, stride);\ 240 OPNAME ## rv30_tpel ## SIZE ## _hhvv_lowpass(dst, src, stride, stride);\
199 }\ 241 }\
200 \ 242 \
201 243
202 #define op_avg(a, b) a = (((a)+cm[((b) + 8)>>4]+1)>>1) 244 #define op_avg(a, b) a = (((a)+cm[b]+1)>>1)
203 #define op_put(a, b) a = cm[((b) + 8)>>4] 245 #define op_put(a, b) a = cm[b]
204 246
205 RV30_LOWPASS(put_ , op_put) 247 RV30_LOWPASS(put_ , op_put)
206 RV30_LOWPASS(avg_ , op_avg) 248 RV30_LOWPASS(avg_ , op_avg)
207 RV30_MC(put_, 8) 249 RV30_MC(put_, 8)
208 RV30_MC(put_, 16) 250 RV30_MC(put_, 16)