00001
00002
00003
00004
00005
00006
00007
00008
00009
00010
00011
00012
00013
00014
00015
00016
00017
00018
00019
00020
00021
00022
00023
00031 #include "dsputil.h"
00032 #include "faandct.h"
00033
00034 #define FLOAT float
00035 #ifdef FAAN_POSTSCALE
00036 # define SCALE(x) postscale[x]
00037 #else
00038 # define SCALE(x) 1
00039 #endif
00040
00041
00042
00043
00044
00045
00046
00047 #define B0 1.00000000000000000000
00048 #define B1 0.72095982200694791383 // (cos(pi*1/16)sqrt(2))^-1
00049 #define B2 0.76536686473017954350 // (cos(pi*2/16)sqrt(2))^-1
00050 #define B3 0.85043009476725644878 // (cos(pi*3/16)sqrt(2))^-1
00051 #define B4 1.00000000000000000000 // (cos(pi*4/16)sqrt(2))^-1
00052 #define B5 1.27275858057283393842 // (cos(pi*5/16)sqrt(2))^-1
00053 #define B6 1.84775906502257351242 // (cos(pi*6/16)sqrt(2))^-1
00054 #define B7 3.62450978541155137218 // (cos(pi*7/16)sqrt(2))^-1
00055
00056
00057 #define A1 0.70710678118654752438 // cos(pi*4/16)
00058 #define A2 0.54119610014619698435 // cos(pi*6/16)sqrt(2)
00059 #define A5 0.38268343236508977170 // cos(pi*6/16)
00060 #define A4 1.30656296487637652774 // cos(pi*2/16)sqrt(2)
00061
00062 static FLOAT postscale[64]={
00063 B0*B0, B0*B1, B0*B2, B0*B3, B0*B4, B0*B5, B0*B6, B0*B7,
00064 B1*B0, B1*B1, B1*B2, B1*B3, B1*B4, B1*B5, B1*B6, B1*B7,
00065 B2*B0, B2*B1, B2*B2, B2*B3, B2*B4, B2*B5, B2*B6, B2*B7,
00066 B3*B0, B3*B1, B3*B2, B3*B3, B3*B4, B3*B5, B3*B6, B3*B7,
00067 B4*B0, B4*B1, B4*B2, B4*B3, B4*B4, B4*B5, B4*B6, B4*B7,
00068 B5*B0, B5*B1, B5*B2, B5*B3, B5*B4, B5*B5, B5*B6, B5*B7,
00069 B6*B0, B6*B1, B6*B2, B6*B3, B6*B4, B6*B5, B6*B6, B6*B7,
00070 B7*B0, B7*B1, B7*B2, B7*B3, B7*B4, B7*B5, B7*B6, B7*B7,
00071 };
00072
00073 static av_always_inline void row_fdct(FLOAT temp[64], DCTELEM * data)
00074 {
00075 FLOAT tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
00076 FLOAT tmp10, tmp11, tmp12, tmp13;
00077 FLOAT z1, z2, z3, z4, z5, z11, z13;
00078 int i;
00079
00080 for (i=0; i<8*8; i+=8) {
00081 tmp0= data[0 + i] + data[7 + i];
00082 tmp7= data[0 + i] - data[7 + i];
00083 tmp1= data[1 + i] + data[6 + i];
00084 tmp6= data[1 + i] - data[6 + i];
00085 tmp2= data[2 + i] + data[5 + i];
00086 tmp5= data[2 + i] - data[5 + i];
00087 tmp3= data[3 + i] + data[4 + i];
00088 tmp4= data[3 + i] - data[4 + i];
00089
00090 tmp10= tmp0 + tmp3;
00091 tmp13= tmp0 - tmp3;
00092 tmp11= tmp1 + tmp2;
00093 tmp12= tmp1 - tmp2;
00094
00095 temp[0 + i]= tmp10 + tmp11;
00096 temp[4 + i]= tmp10 - tmp11;
00097
00098 z1= (tmp12 + tmp13)*A1;
00099 temp[2 + i]= tmp13 + z1;
00100 temp[6 + i]= tmp13 - z1;
00101
00102 tmp10= tmp4 + tmp5;
00103 tmp11= tmp5 + tmp6;
00104 tmp12= tmp6 + tmp7;
00105
00106 z5= (tmp10 - tmp12) * A5;
00107 z2= tmp10*A2 + z5;
00108 z4= tmp12*A4 + z5;
00109 z3= tmp11*A1;
00110
00111 z11= tmp7 + z3;
00112 z13= tmp7 - z3;
00113
00114 temp[5 + i]= z13 + z2;
00115 temp[3 + i]= z13 - z2;
00116 temp[1 + i]= z11 + z4;
00117 temp[7 + i]= z11 - z4;
00118 }
00119 }
00120
00121 void ff_faandct(DCTELEM * data)
00122 {
00123 FLOAT tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
00124 FLOAT tmp10, tmp11, tmp12, tmp13;
00125 FLOAT z1, z2, z3, z4, z5, z11, z13;
00126 FLOAT temp[64];
00127 int i;
00128
00129 emms_c();
00130
00131 row_fdct(temp, data);
00132
00133 for (i=0; i<8; i++) {
00134 tmp0= temp[8*0 + i] + temp[8*7 + i];
00135 tmp7= temp[8*0 + i] - temp[8*7 + i];
00136 tmp1= temp[8*1 + i] + temp[8*6 + i];
00137 tmp6= temp[8*1 + i] - temp[8*6 + i];
00138 tmp2= temp[8*2 + i] + temp[8*5 + i];
00139 tmp5= temp[8*2 + i] - temp[8*5 + i];
00140 tmp3= temp[8*3 + i] + temp[8*4 + i];
00141 tmp4= temp[8*3 + i] - temp[8*4 + i];
00142
00143 tmp10= tmp0 + tmp3;
00144 tmp13= tmp0 - tmp3;
00145 tmp11= tmp1 + tmp2;
00146 tmp12= tmp1 - tmp2;
00147
00148 data[8*0 + i]= lrintf(SCALE(8*0 + i) * (tmp10 + tmp11));
00149 data[8*4 + i]= lrintf(SCALE(8*4 + i) * (tmp10 - tmp11));
00150
00151 z1= (tmp12 + tmp13)* A1;
00152 data[8*2 + i]= lrintf(SCALE(8*2 + i) * (tmp13 + z1));
00153 data[8*6 + i]= lrintf(SCALE(8*6 + i) * (tmp13 - z1));
00154
00155 tmp10= tmp4 + tmp5;
00156 tmp11= tmp5 + tmp6;
00157 tmp12= tmp6 + tmp7;
00158
00159 z5= (tmp10 - tmp12) * A5;
00160 z2= tmp10*A2 + z5;
00161 z4= tmp12*A4 + z5;
00162 z3= tmp11*A1;
00163
00164 z11= tmp7 + z3;
00165 z13= tmp7 - z3;
00166
00167 data[8*5 + i]= lrintf(SCALE(8*5 + i) * (z13 + z2));
00168 data[8*3 + i]= lrintf(SCALE(8*3 + i) * (z13 - z2));
00169 data[8*1 + i]= lrintf(SCALE(8*1 + i) * (z11 + z4));
00170 data[8*7 + i]= lrintf(SCALE(8*7 + i) * (z11 - z4));
00171 }
00172 }
00173
00174 void ff_faandct248(DCTELEM * data)
00175 {
00176 FLOAT tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
00177 FLOAT tmp10, tmp11, tmp12, tmp13;
00178 FLOAT z1;
00179 FLOAT temp[64];
00180 int i;
00181
00182 emms_c();
00183
00184 row_fdct(temp, data);
00185
00186 for (i=0; i<8; i++) {
00187 tmp0 = temp[8*0 + i] + temp[8*1 + i];
00188 tmp1 = temp[8*2 + i] + temp[8*3 + i];
00189 tmp2 = temp[8*4 + i] + temp[8*5 + i];
00190 tmp3 = temp[8*6 + i] + temp[8*7 + i];
00191 tmp4 = temp[8*0 + i] - temp[8*1 + i];
00192 tmp5 = temp[8*2 + i] - temp[8*3 + i];
00193 tmp6 = temp[8*4 + i] - temp[8*5 + i];
00194 tmp7 = temp[8*6 + i] - temp[8*7 + i];
00195
00196 tmp10 = tmp0 + tmp3;
00197 tmp11 = tmp1 + tmp2;
00198 tmp12 = tmp1 - tmp2;
00199 tmp13 = tmp0 - tmp3;
00200
00201 data[8*0 + i] = lrintf(SCALE(8*0 + i) * (tmp10 + tmp11));
00202 data[8*4 + i] = lrintf(SCALE(8*4 + i) * (tmp10 - tmp11));
00203
00204 z1 = (tmp12 + tmp13)* A1;
00205 data[8*2 + i] = lrintf(SCALE(8*2 + i) * (tmp13 + z1));
00206 data[8*6 + i] = lrintf(SCALE(8*6 + i) * (tmp13 - z1));
00207
00208 tmp10 = tmp4 + tmp7;
00209 tmp11 = tmp5 + tmp6;
00210 tmp12 = tmp5 - tmp6;
00211 tmp13 = tmp4 - tmp7;
00212
00213 data[8*1 + i] = lrintf(SCALE(8*0 + i) * (tmp10 + tmp11));
00214 data[8*5 + i] = lrintf(SCALE(8*4 + i) * (tmp10 - tmp11));
00215
00216 z1 = (tmp12 + tmp13)* A1;
00217 data[8*3 + i] = lrintf(SCALE(8*2 + i) * (tmp13 + z1));
00218 data[8*7 + i] = lrintf(SCALE(8*6 + i) * (tmp13 - z1));
00219 }
00220 }