49 |
* * |
* * |
50 |
******************************************************************************/ |
******************************************************************************/ |
51 |
|
|
52 |
|
#include <string.h> |
53 |
|
|
54 |
#include "../portab.h" |
#include "../portab.h" |
55 |
#include "mbfunctions.h" |
#include "mbfunctions.h" |
56 |
|
|
74 |
MACROBLOCK * pMB, |
MACROBLOCK * pMB, |
75 |
const uint32_t x_pos, |
const uint32_t x_pos, |
76 |
const uint32_t y_pos, |
const uint32_t y_pos, |
77 |
int16_t data[][64], |
int16_t data[6*64], |
78 |
int16_t qcoeff[][64], |
int16_t qcoeff[6*64], |
79 |
IMAGE * const pCurrent) |
IMAGE * const pCurrent) |
80 |
|
|
81 |
{ |
{ |
82 |
|
|
83 |
const uint32_t stride = pParam->edged_width; |
const uint32_t stride = pParam->edged_width; |
84 |
uint32_t i; |
uint32_t i; |
85 |
uint32_t iQuant = pParam->quant; |
uint32_t iQuant = pParam->quant; |
90 |
pV_Cur = pCurrent->v + (y_pos << 3) * (stride >> 1) + (x_pos << 3); |
pV_Cur = pCurrent->v + (y_pos << 3) * (stride >> 1) + (x_pos << 3); |
91 |
|
|
92 |
start_timer(); |
start_timer(); |
93 |
transfer_8to16copy(data[0], pY_Cur, stride); |
transfer_8to16copy(&data[0*64], pY_Cur, stride); |
94 |
transfer_8to16copy(data[1], pY_Cur + 8, stride); |
transfer_8to16copy(&data[1*64], pY_Cur + 8, stride); |
95 |
transfer_8to16copy(data[2], pY_Cur + 8 * stride, stride); |
transfer_8to16copy(&data[2*64], pY_Cur + 8 * stride, stride); |
96 |
transfer_8to16copy(data[3], pY_Cur + 8 * stride + 8, stride); |
transfer_8to16copy(&data[3*64], pY_Cur + 8 * stride + 8, stride); |
97 |
transfer_8to16copy(data[4], pU_Cur, stride / 2); |
transfer_8to16copy(&data[4*64], pU_Cur, stride / 2); |
98 |
transfer_8to16copy(data[5], pV_Cur, stride / 2); |
transfer_8to16copy(&data[5*64], pV_Cur, stride / 2); |
99 |
stop_transfer_timer(); |
stop_transfer_timer(); |
100 |
|
|
101 |
start_timer(); |
start_timer(); |
111 |
uint32_t iDcScaler = get_dc_scaler(iQuant, i < 4); |
uint32_t iDcScaler = get_dc_scaler(iQuant, i < 4); |
112 |
|
|
113 |
start_timer(); |
start_timer(); |
114 |
fdct(data[i]); |
fdct(&data[i*64]); |
115 |
stop_dct_timer(); |
stop_dct_timer(); |
116 |
|
|
117 |
if (pParam->quant_type == H263_QUANT) |
if (pParam->quant_type == H263_QUANT) |
118 |
{ |
{ |
119 |
start_timer(); |
start_timer(); |
120 |
quant_intra(qcoeff[i], data[i], iQuant, iDcScaler); |
quant_intra(&qcoeff[i*64], &data[i*64], iQuant, iDcScaler); |
121 |
stop_quant_timer(); |
stop_quant_timer(); |
122 |
|
|
123 |
start_timer(); |
start_timer(); |
124 |
dequant_intra(data[i], qcoeff[i], iQuant, iDcScaler); |
dequant_intra(&data[i*64], &qcoeff[i*64], iQuant, iDcScaler); |
125 |
stop_iquant_timer(); |
stop_iquant_timer(); |
126 |
} |
} |
127 |
else |
else |
128 |
{ |
{ |
129 |
start_timer(); |
start_timer(); |
130 |
quant4_intra(qcoeff[i], data[i], iQuant, iDcScaler); |
quant4_intra(&qcoeff[i*64], &data[i*64], iQuant, iDcScaler); |
131 |
stop_quant_timer(); |
stop_quant_timer(); |
132 |
|
|
133 |
start_timer(); |
start_timer(); |
134 |
dequant4_intra(data[i], qcoeff[i], iQuant, iDcScaler); |
dequant4_intra(&data[i*64], &qcoeff[i*64], iQuant, iDcScaler); |
135 |
stop_iquant_timer(); |
stop_iquant_timer(); |
136 |
} |
} |
137 |
|
|
138 |
start_timer(); |
start_timer(); |
139 |
idct(data[i]); |
idct(&data[i*64]); |
140 |
stop_idct_timer(); |
stop_idct_timer(); |
141 |
} |
} |
142 |
|
|
148 |
stop_interlacing_timer(); |
stop_interlacing_timer(); |
149 |
|
|
150 |
start_timer(); |
start_timer(); |
151 |
transfer_16to8copy(pY_Cur, data[0], stride); |
transfer_16to8copy(pY_Cur, &data[0*64], stride); |
152 |
transfer_16to8copy(pY_Cur + 8, data[1], stride); |
transfer_16to8copy(pY_Cur + 8, &data[1*64], stride); |
153 |
transfer_16to8copy(pY_Cur + 8 * stride, data[2], stride); |
transfer_16to8copy(pY_Cur + 8 * stride, &data[2*64], stride); |
154 |
transfer_16to8copy(pY_Cur + 8 + 8 * stride, data[3], stride); |
transfer_16to8copy(pY_Cur + 8 + 8 * stride, &data[3*64], stride); |
155 |
transfer_16to8copy(pU_Cur, data[4], stride / 2); |
transfer_16to8copy(pU_Cur, &data[4*64], stride / 2); |
156 |
transfer_16to8copy(pV_Cur, data[5], stride / 2); |
transfer_16to8copy(pV_Cur, &data[5*64], stride / 2); |
157 |
stop_transfer_timer(); |
stop_transfer_timer(); |
158 |
|
|
159 |
} |
} |
160 |
|
|
161 |
|
|
162 |
uint8_t MBTransQuantInter(const MBParam *pParam, |
uint8_t MBTransQuantInter(const MBParam *pParam, |
163 |
MACROBLOCK * pMB, |
MACROBLOCK * pMB, |
164 |
const uint32_t x_pos, const uint32_t y_pos, |
const uint32_t x_pos, const uint32_t y_pos, |
165 |
int16_t data[][64], |
int16_t data[6*64], |
166 |
int16_t qcoeff[][64], |
int16_t qcoeff[6*64], |
167 |
IMAGE * const pCurrent) |
IMAGE * const pCurrent) |
168 |
|
|
169 |
{ |
{ |
170 |
|
|
171 |
const uint32_t stride = pParam->edged_width; |
const uint32_t stride = pParam->edged_width; |
172 |
uint32_t i; |
uint32_t i; |
173 |
uint32_t iQuant = pParam->quant; |
uint32_t iQuant = pParam->quant; |
190 |
for(i = 0; i < 6; i++) |
for(i = 0; i < 6; i++) |
191 |
{ |
{ |
192 |
/* |
/* |
193 |
no need to transfer 8->16-bit |
* no need to transfer 8->16-bit |
194 |
(this is performed already in motion compensation) |
* (this is performed already in motion compensation) |
195 |
*/ |
*/ |
196 |
start_timer(); |
start_timer(); |
197 |
fdct(data[i]); |
fdct(&data[i*64]); |
198 |
stop_dct_timer(); |
stop_dct_timer(); |
199 |
|
|
200 |
if (pParam->quant_type == 0) |
if (pParam->quant_type == 0) |
201 |
{ |
{ |
202 |
start_timer(); |
start_timer(); |
203 |
sum = quant_inter(qcoeff[i], data[i], iQuant); |
sum = quant_inter(&qcoeff[i*64], &data[i*64], iQuant); |
204 |
stop_quant_timer(); |
stop_quant_timer(); |
205 |
} |
} |
206 |
else |
else |
207 |
{ |
{ |
208 |
start_timer(); |
start_timer(); |
209 |
sum = quant4_inter(qcoeff[i], data[i], iQuant); |
sum = quant4_inter(&qcoeff[i*64], &data[i*64], iQuant); |
210 |
stop_quant_timer(); |
stop_quant_timer(); |
211 |
} |
} |
212 |
|
|
215 |
if (pParam->quant_type == H263_QUANT) |
if (pParam->quant_type == H263_QUANT) |
216 |
{ |
{ |
217 |
start_timer(); |
start_timer(); |
218 |
dequant_inter(data[i], qcoeff[i], iQuant); |
dequant_inter(&data[i*64], &qcoeff[i*64], iQuant); |
219 |
stop_iquant_timer(); |
stop_iquant_timer(); |
220 |
} |
} |
221 |
else |
else |
222 |
{ |
{ |
223 |
start_timer(); |
start_timer(); |
224 |
dequant4_inter(data[i], qcoeff[i], iQuant); |
dequant4_inter(&data[i*64], &qcoeff[i*64], iQuant); |
225 |
stop_iquant_timer(); |
stop_iquant_timer(); |
226 |
} |
} |
227 |
|
|
228 |
cbp |= 1 << (5 - i); |
cbp |= 1 << (5 - i); |
229 |
|
|
230 |
start_timer(); |
start_timer(); |
231 |
idct(data[i]); |
idct(&data[i*64]); |
232 |
stop_idct_timer(); |
stop_idct_timer(); |
233 |
} |
} |
234 |
} |
} |
242 |
|
|
243 |
start_timer(); |
start_timer(); |
244 |
if (cbp & 32) |
if (cbp & 32) |
245 |
transfer_16to8add(pY_Cur, data[0], stride); |
transfer_16to8add(pY_Cur, &data[0*64], stride); |
246 |
if (cbp & 16) |
if (cbp & 16) |
247 |
transfer_16to8add(pY_Cur + 8, data[1], stride); |
transfer_16to8add(pY_Cur + 8, &data[1*64], stride); |
248 |
if (cbp & 8) |
if (cbp & 8) |
249 |
transfer_16to8add(pY_Cur + 8 * stride, data[2], stride); |
transfer_16to8add(pY_Cur + 8 * stride, &data[2*64], stride); |
250 |
if (cbp & 4) |
if (cbp & 4) |
251 |
transfer_16to8add(pY_Cur + 8 + 8 * stride, data[3], stride); |
transfer_16to8add(pY_Cur + 8 + 8 * stride, &data[3*64], stride); |
252 |
if (cbp & 2) |
if (cbp & 2) |
253 |
transfer_16to8add(pU_Cur, data[4], stride / 2); |
transfer_16to8add(pU_Cur, &data[4*64], stride / 2); |
254 |
if (cbp & 1) |
if (cbp & 1) |
255 |
transfer_16to8add(pV_Cur, data[5], stride / 2); |
transfer_16to8add(pV_Cur, &data[5*64], stride / 2); |
256 |
stop_transfer_timer(); |
stop_transfer_timer(); |
257 |
|
|
258 |
return cbp; |
return cbp; |
259 |
|
|
260 |
} |
} |
261 |
|
|
262 |
|
|
264 |
|
|
265 |
#define ABS(X) (X)<0 ? -(X) : (X) |
#define ABS(X) (X)<0 ? -(X) : (X) |
266 |
|
|
267 |
uint32_t MBDecideFieldDCT(int16_t data[][64]) |
uint32_t MBDecideFieldDCT(int16_t data[6*64]) |
268 |
{ |
{ |
269 |
const uint8_t blocks[] = {0, 0, 0, 0, 2, 2, 2, 2}; |
|
270 |
|
const uint8_t blocks[] = {0*64, 0*64, 0*64, 0*64, 2*64, 2*64, 2*64, 2*64}; |
271 |
const uint8_t lines[] = {0, 16, 32, 48, 0, 16, 32, 48}; |
const uint8_t lines[] = {0, 16, 32, 48, 0, 16, 32, 48}; |
272 |
|
|
273 |
int frame = 0, field = 0; |
int frame = 0, field = 0; |
277 |
{ |
{ |
278 |
for (j=0 ; j<8 ; ++j) |
for (j=0 ; j<8 ; ++j) |
279 |
{ |
{ |
280 |
frame += ABS(data[0][(i+1)*8 + j] - data[0][i*8 + j]); |
frame += ABS(data[0*64 + (i+1)*8 + j] - data[0*64 + i*8 + j]); |
281 |
frame += ABS(data[1][(i+1)*8 + j] - data[1][i*8 + j]); |
frame += ABS(data[1*64 + (i+1)*8 + j] - data[1*64 + i*8 + j]); |
282 |
frame += ABS(data[2][(i+1)*8 + j] - data[2][i*8 + j]); |
frame += ABS(data[2*64 + (i+1)*8 + j] - data[2*64 + i*8 + j]); |
283 |
frame += ABS(data[3][(i+1)*8 + j] - data[3][i*8 + j]); |
frame += ABS(data[3*64 + (i+1)*8 + j] - data[3*64 + i*8 + j]); |
284 |
|
|
285 |
field += ABS(data[blocks[i+1]][lines[i+1] + j] - data[blocks[i]][lines[i] + j]); |
field += ABS(data[blocks[i+1] + lines[i+1] + j] -\ |
286 |
field += ABS(data[blocks[i+1]][lines[i+1] + 8 + j] - data[blocks[i]][lines[i] + 8 + j]); |
data[blocks[i ] + lines[i ] + j]); |
287 |
field += ABS(data[blocks[i+1]+1][lines[i+1] + j] - data[blocks[i]+1][lines[i] + j]); |
field += ABS(data[blocks[i+1] + lines[i+1] + 8 + j] -\ |
288 |
field += ABS(data[blocks[i+1]+1][lines[i+1] + 8 + j] - data[blocks[i]+1][lines[i] + 8 + j]); |
data[blocks[i ] + lines[i ] + 8 + j]); |
289 |
|
field += ABS(data[blocks[i+1] + 64 + lines[i+1] + j] -\ |
290 |
|
data[blocks[i ] + 64 + lines[i ] + j]); |
291 |
|
field += ABS(data[blocks[i+1] + 64 + lines[i+1] + 8 + j] -\ |
292 |
|
data[blocks[i ] + 64 + lines[i ] + 8 + j]); |
293 |
} |
} |
294 |
} |
} |
295 |
|
|
305 |
/* deinterlace Y blocks vertically */ |
/* deinterlace Y blocks vertically */ |
306 |
|
|
307 |
#define MOVLINE(X,Y) memcpy(X, Y, sizeof(tmp)) |
#define MOVLINE(X,Y) memcpy(X, Y, sizeof(tmp)) |
308 |
#define LINE(X,Y) &data[X][Y*8] |
#define LINE(X,Y) &data[X*64 + Y*8] |
309 |
|
|
310 |
void MBFrameToField(int16_t data[][64]) |
void MBFrameToField(int16_t data[6*64]) |
311 |
{ |
{ |
312 |
int16_t tmp[8]; |
int16_t tmp[8]; |
313 |
|
|
371 |
|
|
372 |
/* interlace Y blocks vertically */ |
/* interlace Y blocks vertically */ |
373 |
|
|
374 |
void MBFieldToFrame(int16_t data[][64]) |
void MBFieldToFrame(int16_t data[6*64]) |
375 |
{ |
{ |
376 |
uint16_t tmp[8]; |
uint16_t tmp[8]; |
377 |
|
|