71 |
|
|
72 |
#define TOOSMALL_LIMIT 1 /* skip blocks having a coefficient sum below this value */ |
#define TOOSMALL_LIMIT 1 /* skip blocks having a coefficient sum below this value */ |
73 |
|
|
74 |
|
static __inline void |
75 |
|
MBfDCT(int16_t data[6 * 64]) |
76 |
|
{ |
77 |
|
start_timer(); |
78 |
|
fdct(&data[0 * 64]); |
79 |
|
fdct(&data[1 * 64]); |
80 |
|
fdct(&data[2 * 64]); |
81 |
|
fdct(&data[3 * 64]); |
82 |
|
fdct(&data[4 * 64]); |
83 |
|
fdct(&data[5 * 64]); |
84 |
|
stop_dct_timer(); |
85 |
|
} |
86 |
|
|
87 |
|
|
88 |
|
static __inline uint32_t |
89 |
|
QuantizeInterBlock( int16_t qcoeff[64], |
90 |
|
const int16_t data[64], |
91 |
|
const uint32_t iQuant, |
92 |
|
const uint32_t quant_type) |
93 |
|
{ |
94 |
|
uint32_t sum; |
95 |
|
|
96 |
|
start_timer(); |
97 |
|
if (quant_type == H263_QUANT) |
98 |
|
sum = quant_inter(qcoeff, data, iQuant); |
99 |
|
else |
100 |
|
sum = quant4_inter(qcoeff, data, iQuant); |
101 |
|
|
102 |
|
stop_quant_timer(); |
103 |
|
return sum; |
104 |
|
} |
105 |
|
|
106 |
void |
void |
107 |
MBTransQuantIntra(const MBParam * pParam, |
MBTransQuantIntra(const MBParam * const pParam, |
108 |
FRAMEINFO * frame, |
FRAMEINFO * const frame, |
109 |
MACROBLOCK * pMB, |
MACROBLOCK * const pMB, |
110 |
const uint32_t x_pos, |
const uint32_t x_pos, |
111 |
const uint32_t y_pos, |
const uint32_t y_pos, |
112 |
int16_t data[6 * 64], |
int16_t data[6 * 64], |
114 |
{ |
{ |
115 |
|
|
116 |
uint32_t stride = pParam->edged_width; |
uint32_t stride = pParam->edged_width; |
117 |
uint32_t stride2 = stride / 2; |
const uint32_t stride2 = stride / 2; |
118 |
uint32_t next_block = stride * ((frame->global_flags & XVID_REDUCED)?16:8); |
uint32_t next_block = stride * ((frame->global_flags & XVID_REDUCED)?16:8); |
119 |
uint32_t i; |
int i; |
120 |
uint32_t iQuant = frame->quant; |
const uint32_t iQuant = pMB->quant; |
121 |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
122 |
IMAGE *pCurrent = &frame->image; |
const IMAGE * const pCurrent = &frame->image; |
123 |
|
|
124 |
start_timer(); |
start_timer(); |
125 |
if ((frame->global_flags & XVID_REDUCED)) |
if ((frame->global_flags & XVID_REDUCED)) |
158 |
} |
} |
159 |
stop_interlacing_timer(); |
stop_interlacing_timer(); |
160 |
|
|
161 |
for (i = 0; i < 6; i++) { |
MBfDCT(data); |
|
uint32_t iDcScaler = get_dc_scaler(iQuant, i < 4); |
|
162 |
|
|
163 |
start_timer(); |
for (i = 0; i < 6; i++) { |
164 |
fdct(&data[i * 64]); |
const uint32_t iDcScaler = get_dc_scaler(iQuant, i < 4); |
|
stop_dct_timer(); |
|
165 |
|
|
|
if (pParam->m_quant_type == H263_QUANT) { |
|
166 |
start_timer(); |
start_timer(); |
167 |
|
if (pParam->m_quant_type == H263_QUANT) |
168 |
quant_intra(&qcoeff[i * 64], &data[i * 64], iQuant, iDcScaler); |
quant_intra(&qcoeff[i * 64], &data[i * 64], iQuant, iDcScaler); |
169 |
stop_quant_timer(); |
else |
|
} else { |
|
|
start_timer(); |
|
170 |
quant4_intra(&qcoeff[i * 64], &data[i * 64], iQuant, iDcScaler); |
quant4_intra(&qcoeff[i * 64], &data[i * 64], iQuant, iDcScaler); |
171 |
stop_quant_timer(); |
stop_quant_timer(); |
|
} |
|
172 |
|
|
173 |
/* speedup: dont decode when encoding only ivops */ |
/* speedup: dont decode when encoding only ivops */ |
174 |
if (pParam->iMaxKeyInterval != 1 || pParam->max_bframes > 0) |
if (pParam->iMaxKeyInterval != 1 || pParam->max_bframes > 0) |
175 |
{ |
{ |
|
if (pParam->m_quant_type == H263_QUANT) { |
|
176 |
start_timer(); |
start_timer(); |
177 |
|
if (pParam->m_quant_type == H263_QUANT) |
178 |
dequant_intra(&data[i * 64], &qcoeff[i * 64], iQuant, iDcScaler); |
dequant_intra(&data[i * 64], &qcoeff[i * 64], iQuant, iDcScaler); |
179 |
stop_iquant_timer(); |
else |
|
} else { |
|
|
start_timer(); |
|
180 |
dequant4_intra(&data[i * 64], &qcoeff[i * 64], iQuant, iDcScaler); |
dequant4_intra(&data[i * 64], &qcoeff[i * 64], iQuant, iDcScaler); |
181 |
stop_iquant_timer(); |
stop_iquant_timer(); |
|
} |
|
182 |
|
|
183 |
start_timer(); |
start_timer(); |
184 |
idct(&data[i * 64]); |
idct(&data[i * 64]); |
196 |
} |
} |
197 |
|
|
198 |
start_timer(); |
start_timer(); |
199 |
if ((frame->global_flags & XVID_REDUCED)) |
if ((frame->global_flags & XVID_REDUCED)) { |
|
{ |
|
200 |
copy_upsampled_8x8_16to8(pY_Cur, &data[0 * 64], stride); |
copy_upsampled_8x8_16to8(pY_Cur, &data[0 * 64], stride); |
201 |
copy_upsampled_8x8_16to8(pY_Cur + 16, &data[1 * 64], stride); |
copy_upsampled_8x8_16to8(pY_Cur + 16, &data[1 * 64], stride); |
202 |
copy_upsampled_8x8_16to8(pY_Cur + next_block, &data[2 * 64], stride); |
copy_upsampled_8x8_16to8(pY_Cur + next_block, &data[2 * 64], stride); |
203 |
copy_upsampled_8x8_16to8(pY_Cur + next_block + 16, &data[3 * 64], stride); |
copy_upsampled_8x8_16to8(pY_Cur + next_block + 16, &data[3 * 64], stride); |
204 |
copy_upsampled_8x8_16to8(pU_Cur, &data[4 * 64], stride2); |
copy_upsampled_8x8_16to8(pU_Cur, &data[4 * 64], stride2); |
205 |
copy_upsampled_8x8_16to8(pV_Cur, &data[5 * 64], stride2); |
copy_upsampled_8x8_16to8(pV_Cur, &data[5 * 64], stride2); |
|
|
|
206 |
}else{ |
}else{ |
207 |
transfer_16to8copy(pY_Cur, &data[0 * 64], stride); |
transfer_16to8copy(pY_Cur, &data[0 * 64], stride); |
208 |
transfer_16to8copy(pY_Cur + 8, &data[1 * 64], stride); |
transfer_16to8copy(pY_Cur + 8, &data[1 * 64], stride); |
216 |
|
|
217 |
} |
} |
218 |
|
|
|
|
|
219 |
uint8_t |
uint8_t |
220 |
MBTransQuantInter(const MBParam * pParam, |
MBTransQuantInter(const MBParam * const pParam, |
221 |
FRAMEINFO * frame, |
FRAMEINFO * const frame, |
222 |
MACROBLOCK * pMB, |
MACROBLOCK * const pMB, |
223 |
const uint32_t x_pos, |
const uint32_t x_pos, |
224 |
const uint32_t y_pos, |
const uint32_t y_pos, |
225 |
int16_t data[6 * 64], |
int16_t data[6 * 64], |
226 |
int16_t qcoeff[6 * 64]) |
int16_t qcoeff[6 * 64]) |
227 |
{ |
{ |
|
|
|
228 |
uint32_t stride = pParam->edged_width; |
uint32_t stride = pParam->edged_width; |
229 |
uint32_t stride2 = stride / 2; |
const uint32_t stride2 = stride / 2; |
230 |
uint32_t next_block = stride * ((frame->global_flags & XVID_REDUCED)?16:8); |
uint32_t next_block = stride * ((frame->global_flags & XVID_REDUCED)?16:8); |
231 |
uint32_t i; |
int i; |
232 |
uint32_t iQuant = frame->quant; |
const uint32_t iQuant = pMB->quant; |
233 |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
234 |
uint8_t cbp = 0; |
int cbp = 0; |
235 |
uint32_t sum; |
uint32_t sum; |
236 |
IMAGE *pCurrent = &frame->image; |
const IMAGE * const pCurrent = &frame->image; |
237 |
|
|
238 |
if ((frame->global_flags & XVID_REDUCED)) |
if ((frame->global_flags & XVID_REDUCED)) { |
|
{ |
|
239 |
pY_Cur = pCurrent->y + (y_pos << 5) * stride + (x_pos << 5); |
pY_Cur = pCurrent->y + (y_pos << 5) * stride + (x_pos << 5); |
240 |
pU_Cur = pCurrent->u + (y_pos << 4) * stride2 + (x_pos << 4); |
pU_Cur = pCurrent->u + (y_pos << 4) * stride2 + (x_pos << 4); |
241 |
pV_Cur = pCurrent->v + (y_pos << 4) * stride2 + (x_pos << 4); |
pV_Cur = pCurrent->v + (y_pos << 4) * stride2 + (x_pos << 4); |
254 |
} |
} |
255 |
stop_interlacing_timer(); |
stop_interlacing_timer(); |
256 |
|
|
257 |
for (i = 0; i < 6; i++) { |
MBfDCT(data); |
|
uint32_t increase_limit = (iQuant == 1) ? 1 : 0; |
|
258 |
|
|
259 |
|
for (i = 0; i < 6; i++) { |
260 |
|
const uint32_t limit = TOOSMALL_LIMIT + ((iQuant == 1) ? 1 : 0); |
261 |
/* |
/* |
262 |
* no need to transfer 8->16-bit |
* no need to transfer 8->16-bit |
263 |
* (this is performed already in motion compensation) |
* (this is performed already in motion compensation) |
264 |
*/ |
*/ |
|
start_timer(); |
|
|
fdct(&data[i * 64]); |
|
|
stop_dct_timer(); |
|
265 |
|
|
266 |
if (pParam->m_quant_type == 0) { |
sum = QuantizeInterBlock(&qcoeff[i * 64], &data[i * 64], iQuant, pParam->m_quant_type); |
|
start_timer(); |
|
|
sum = quant_inter(&qcoeff[i * 64], &data[i * 64], iQuant); |
|
|
stop_quant_timer(); |
|
|
} else { |
|
|
start_timer(); |
|
|
sum = quant4_inter(&qcoeff[i * 64], &data[i * 64], iQuant); |
|
|
stop_quant_timer(); |
|
|
} |
|
267 |
|
|
268 |
if ((sum >= TOOSMALL_LIMIT + increase_limit) || (qcoeff[i*64] != 0) || |
if (sum >= limit) { |
|
(qcoeff[i*64+1] != 0) || (qcoeff[i*64+8] != 0)) { |
|
269 |
|
|
|
if (pParam->m_quant_type == H263_QUANT) { |
|
270 |
start_timer(); |
start_timer(); |
271 |
|
if (pParam->m_quant_type == H263_QUANT) |
272 |
dequant_inter(&data[i * 64], &qcoeff[i * 64], iQuant); |
dequant_inter(&data[i * 64], &qcoeff[i * 64], iQuant); |
273 |
stop_iquant_timer(); |
else |
|
} else { |
|
|
start_timer(); |
|
274 |
dequant4_inter(&data[i * 64], &qcoeff[i * 64], iQuant); |
dequant4_inter(&data[i * 64], &qcoeff[i * 64], iQuant); |
275 |
stop_iquant_timer(); |
stop_iquant_timer(); |
|
} |
|
276 |
|
|
277 |
cbp |= 1 << (5 - i); |
cbp |= 1 << (5 - i); |
278 |
|
|
288 |
} |
} |
289 |
|
|
290 |
start_timer(); |
start_timer(); |
291 |
if ((frame->global_flags & XVID_REDUCED)) |
if ((frame->global_flags & XVID_REDUCED)) { |
|
{ |
|
292 |
if (cbp & 32) |
if (cbp & 32) |
293 |
add_upsampled_8x8_16to8(pY_Cur, &data[0 * 64], stride); |
add_upsampled_8x8_16to8(pY_Cur, &data[0 * 64], stride); |
294 |
if (cbp & 16) |
if (cbp & 16) |
317 |
} |
} |
318 |
stop_transfer_timer(); |
stop_transfer_timer(); |
319 |
|
|
320 |
return cbp; |
return (uint8_t) cbp; |
|
|
|
|
} |
|
|
|
|
|
void |
|
|
MBTransQuantIntra2(const MBParam * pParam, |
|
|
FRAMEINFO * frame, |
|
|
MACROBLOCK * pMB, |
|
|
const uint32_t x_pos, |
|
|
const uint32_t y_pos, |
|
|
int16_t data[6 * 64], |
|
|
int16_t qcoeff[6 * 64]) |
|
|
{ |
|
|
MBTrans(pParam,frame,pMB,x_pos,y_pos,data); |
|
|
MBfDCT(pParam,frame,pMB,data); |
|
|
MBQuantIntra(pParam,frame,pMB,data,qcoeff); |
|
|
MBDeQuantIntra(pParam,frame->quant,data,qcoeff); |
|
|
MBiDCT(data,0x3F); |
|
|
MBTransAdd(pParam,frame,pMB,x_pos,y_pos,data,0x3F); |
|
|
} |
|
|
|
|
|
|
|
|
uint8_t |
|
|
MBTransQuantInter2(const MBParam * pParam, |
|
|
FRAMEINFO * frame, |
|
|
MACROBLOCK * pMB, |
|
|
const uint32_t x_pos, |
|
|
const uint32_t y_pos, |
|
|
int16_t data[6 * 64], |
|
|
int16_t qcoeff[6 * 64]) |
|
|
{ |
|
|
uint8_t cbp; |
|
|
|
|
|
/* there is no MBTrans for Inter block, that's done in motion compensation already */ |
|
|
|
|
|
MBfDCT(pParam,frame,pMB,data); |
|
|
cbp = MBQuantInter(pParam,frame->quant,data,qcoeff); |
|
|
MBDeQuantInter(pParam,frame->quant,data,qcoeff,cbp); |
|
|
MBiDCT(data,cbp); |
|
|
MBTransAdd(pParam,frame,pMB,x_pos,y_pos,data,cbp); |
|
|
|
|
|
return cbp; |
|
321 |
} |
} |
322 |
|
|
323 |
uint8_t |
uint8_t |
327 |
int16_t data[6 * 64], |
int16_t data[6 * 64], |
328 |
int16_t qcoeff[6 * 64]) |
int16_t qcoeff[6 * 64]) |
329 |
{ |
{ |
330 |
uint8_t cbp; |
int cbp = 0; |
|
|
|
|
/* there is no MBTrans for Inter block, that's done in motion compensation already */ |
|
|
|
|
|
MBfDCT(pParam,frame,pMB,data); |
|
|
cbp = MBQuantInter(pParam,frame->quant,data,qcoeff); |
|
|
|
|
|
/* we don't have to DeQuant, iDCT and Transfer back data for B-frames */ |
|
|
|
|
|
return cbp; |
|
|
} |
|
|
|
|
|
|
|
|
void |
|
|
MBfDCT(const MBParam * pParam, |
|
|
FRAMEINFO * frame, |
|
|
MACROBLOCK * pMB, |
|
|
int16_t data[6 * 64]) |
|
|
{ |
|
331 |
int i; |
int i; |
332 |
|
|
333 |
start_timer(); |
/* there is no MBTrans for Inter block, that's done in motion compensation already */ |
|
pMB->field_dct = 0; |
|
|
if ((frame->global_flags & XVID_INTERLACING)) { |
|
|
pMB->field_dct = MBDecideFieldDCT(data); |
|
|
} |
|
|
stop_interlacing_timer(); |
|
|
|
|
|
for (i = 0; i < 6; i++) { |
|
|
start_timer(); |
|
|
fdct(&data[i * 64]); |
|
|
stop_dct_timer(); |
|
|
} |
|
|
} |
|
|
|
|
|
void |
|
|
MBQuantDeQuantIntra(const MBParam * pParam, |
|
|
FRAMEINFO * frame, |
|
|
MACROBLOCK * pMB, |
|
|
int16_t qcoeff[6 * 64], |
|
|
int16_t data[6*64]) |
|
|
{ |
|
|
int i; |
|
|
int iQuant = frame->quant; |
|
|
|
|
|
start_timer(); |
|
|
pMB->field_dct = 0; |
|
|
if ((frame->global_flags & XVID_INTERLACING)) { |
|
|
pMB->field_dct = MBDecideFieldDCT(data); |
|
|
} |
|
|
stop_interlacing_timer(); |
|
|
|
|
|
for (i = 0; i < 6; i++) { |
|
|
uint32_t iDcScaler = get_dc_scaler(iQuant, i < 4); |
|
|
|
|
|
if (pParam->m_quant_type == H263_QUANT) { |
|
|
start_timer(); |
|
|
quant_intra(&qcoeff[i * 64], &data[i * 64], iQuant, iDcScaler); |
|
|
stop_quant_timer(); |
|
|
|
|
|
start_timer(); |
|
|
dequant_intra(&data[i * 64], &qcoeff[i * 64], iQuant, iDcScaler); |
|
|
stop_iquant_timer(); |
|
|
} else { |
|
|
start_timer(); |
|
|
quant4_intra(&qcoeff[i * 64], &data[i * 64], iQuant, iDcScaler); |
|
|
stop_quant_timer(); |
|
|
|
|
|
start_timer(); |
|
|
dequant4_intra(&data[i * 64], &qcoeff[i * 64], iQuant, iDcScaler); |
|
|
stop_iquant_timer(); |
|
|
} |
|
|
} |
|
|
} |
|
|
|
|
|
void |
|
|
MBQuantIntra(const MBParam * pParam, |
|
|
FRAMEINFO * frame, |
|
|
MACROBLOCK *pMB, |
|
|
int16_t qcoeff[6 * 64], |
|
|
int16_t data[6*64]) |
|
|
{ |
|
|
int i; |
|
|
int iQuant = frame->quant; |
|
334 |
|
|
335 |
start_timer(); |
start_timer(); |
336 |
pMB->field_dct = 0; |
pMB->field_dct = 0; |
339 |
} |
} |
340 |
stop_interlacing_timer(); |
stop_interlacing_timer(); |
341 |
|
|
342 |
for (i = 0; i < 6; i++) { |
MBfDCT(data); |
|
uint32_t iDcScaler = get_dc_scaler(iQuant, i < 4); |
|
|
|
|
|
if (pParam->m_quant_type == H263_QUANT) { |
|
|
start_timer(); |
|
|
quant_intra(&qcoeff[i * 64], &data[i * 64], iQuant, iDcScaler); |
|
|
stop_quant_timer(); |
|
|
} else { |
|
|
start_timer(); |
|
|
quant4_intra(&qcoeff[i * 64], &data[i * 64], iQuant, iDcScaler); |
|
|
stop_quant_timer(); |
|
|
} |
|
|
} |
|
|
} |
|
|
|
|
|
void |
|
|
MBDeQuantIntra(const MBParam * pParam, |
|
|
const int iQuant, |
|
|
int16_t qcoeff[6 * 64], |
|
|
int16_t data[6*64]) |
|
|
{ |
|
|
int i; |
|
343 |
|
|
344 |
for (i = 0; i < 6; i++) { |
for (i = 0; i < 6; i++) { |
345 |
uint32_t iDcScaler = get_dc_scaler(iQuant, i < 4); |
int codedecision = 0; |
|
|
|
|
if (pParam->m_quant_type == H263_QUANT) { |
|
|
start_timer(); |
|
|
dequant_intra(&data[i * 64], &qcoeff[i * 64], iQuant, iDcScaler); |
|
|
stop_iquant_timer(); |
|
|
} else { |
|
|
start_timer(); |
|
|
dequant4_intra(&data[i * 64], &qcoeff[i * 64], iQuant, iDcScaler); |
|
|
stop_iquant_timer(); |
|
|
} |
|
|
} |
|
|
} |
|
|
|
|
|
uint8_t |
|
|
MBQuantInter(const MBParam * pParam, |
|
|
const int iQuant, |
|
|
int16_t data[6 * 64], |
|
|
int16_t qcoeff[6 * 64]) |
|
|
{ |
|
|
|
|
|
int i; |
|
|
uint8_t cbp = 0; |
|
|
int sum; |
|
346 |
|
|
347 |
for (i = 0; i < 6; i++) { |
int sum = QuantizeInterBlock(&qcoeff[i * 64], &data[i * 64], pMB->quant, pParam->m_quant_type); |
348 |
|
|
349 |
if (pParam->m_quant_type == 0) { |
if ((sum > 2) || (qcoeff[i*64+1] != 0) || (qcoeff[i*64+8] != 0) ) codedecision = 1; |
350 |
start_timer(); |
else { |
351 |
sum = quant_inter(&qcoeff[i * 64], &data[i * 64], iQuant); |
if (pMB->mode == MODE_DIRECT || pMB->mode == MODE_DIRECT_NO4V) { |
352 |
stop_quant_timer(); |
// dark blocks prevention for direct mode |
353 |
} else { |
if ( (qcoeff[i*64] < -1) || (qcoeff[i*64] > 0) ) codedecision = 1; |
354 |
start_timer(); |
} else |
355 |
sum = quant4_inter(&qcoeff[i * 64], &data[i * 64], iQuant); |
if (qcoeff[i*64] != 0) codedecision = 1; // not direct mode |
|
stop_quant_timer(); |
|
356 |
} |
} |
357 |
|
|
358 |
if (sum >= TOOSMALL_LIMIT) { // skip block ? |
if (codedecision) cbp |= 1 << (5 - i); |
|
cbp |= 1 << (5 - i); |
|
|
} |
|
|
} |
|
|
return cbp; |
|
359 |
} |
} |
360 |
|
|
361 |
void |
/* we don't have to DeQuant, iDCT and Transfer back data for B-frames if we don't reconstruct this frame */ |
362 |
MBDeQuantInter( const MBParam * pParam, |
/* warning: reconstruction not supported yet */ |
363 |
const int iQuant, |
return (uint8_t) cbp; |
|
int16_t data[6 * 64], |
|
|
int16_t qcoeff[6 * 64], |
|
|
const uint8_t cbp) |
|
|
{ |
|
|
int i; |
|
|
|
|
|
for (i = 0; i < 6; i++) { |
|
|
if (cbp & (1 << (5 - i))) |
|
|
{ |
|
|
if (pParam->m_quant_type == H263_QUANT) { |
|
|
start_timer(); |
|
|
dequant_inter(&data[i * 64], &qcoeff[i * 64], iQuant); |
|
|
stop_iquant_timer(); |
|
|
} else { |
|
|
start_timer(); |
|
|
dequant4_inter(&data[i * 64], &qcoeff[i * 64], iQuant); |
|
|
stop_iquant_timer(); |
|
|
} |
|
|
} |
|
|
} |
|
364 |
} |
} |
365 |
|
|
|
void |
|
|
MBiDCT( int16_t data[6 * 64], |
|
|
const uint8_t cbp) |
|
|
{ |
|
|
int i; |
|
|
|
|
|
for (i = 0; i < 6; i++) { |
|
|
if (cbp & (1 << (5 - i))) |
|
|
{ |
|
|
start_timer(); |
|
|
idct(&data[i * 64]); |
|
|
stop_idct_timer(); |
|
|
|
|
|
} |
|
|
} |
|
|
} |
|
|
|
|
|
|
|
|
void |
|
|
MBTrans(const MBParam * pParam, |
|
|
FRAMEINFO * frame, |
|
|
MACROBLOCK * pMB, |
|
|
const uint32_t x_pos, |
|
|
const uint32_t y_pos, |
|
|
int16_t data[6 * 64]) |
|
|
{ |
|
|
uint32_t stride = pParam->edged_width; |
|
|
uint32_t stride2 = stride / 2; |
|
|
uint32_t next_block = stride * 8; |
|
|
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
|
|
IMAGE *pCurrent = &frame->image; |
|
|
|
|
|
pY_Cur = pCurrent->y + (y_pos << 4) * stride + (x_pos << 4); |
|
|
pU_Cur = pCurrent->u + (y_pos << 3) * stride2 + (x_pos << 3); |
|
|
pV_Cur = pCurrent->v + (y_pos << 3) * stride2 + (x_pos << 3); |
|
|
|
|
|
start_timer(); |
|
|
transfer_8to16copy(&data[0 * 64], pY_Cur, stride); |
|
|
transfer_8to16copy(&data[1 * 64], pY_Cur + 8, stride); |
|
|
transfer_8to16copy(&data[2 * 64], pY_Cur + next_block, stride); |
|
|
transfer_8to16copy(&data[3 * 64], pY_Cur + next_block + 8, stride); |
|
|
transfer_8to16copy(&data[4 * 64], pU_Cur, stride2); |
|
|
transfer_8to16copy(&data[5 * 64], pV_Cur, stride2); |
|
|
stop_transfer_timer(); |
|
|
} |
|
|
|
|
|
void |
|
|
MBTransAdd(const MBParam * pParam, |
|
|
FRAMEINFO * frame, |
|
|
MACROBLOCK * pMB, |
|
|
const uint32_t x_pos, |
|
|
const uint32_t y_pos, |
|
|
int16_t data[6 * 64], |
|
|
const uint8_t cbp) |
|
|
{ |
|
|
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
|
|
uint32_t stride = pParam->edged_width; |
|
|
uint32_t stride2 = stride / 2; |
|
|
uint32_t next_block = stride * 8; |
|
|
IMAGE *pCurrent = &frame->image; |
|
|
|
|
|
pY_Cur = pCurrent->y + (y_pos << 4) * stride + (x_pos << 4); |
|
|
pU_Cur = pCurrent->u + (y_pos << 3) * stride2 + (x_pos << 3); |
|
|
pV_Cur = pCurrent->v + (y_pos << 3) * stride2 + (x_pos << 3); |
|
|
|
|
|
if (pMB->field_dct) { |
|
|
next_block = stride; |
|
|
stride *= 2; |
|
|
} |
|
|
|
|
|
start_timer(); |
|
|
if (cbp & 32) |
|
|
transfer_16to8add(pY_Cur, &data[0 * 64], stride); |
|
|
if (cbp & 16) |
|
|
transfer_16to8add(pY_Cur + 8, &data[1 * 64], stride); |
|
|
if (cbp & 8) |
|
|
transfer_16to8add(pY_Cur + next_block, &data[2 * 64], stride); |
|
|
if (cbp & 4) |
|
|
transfer_16to8add(pY_Cur + next_block + 8, &data[3 * 64], stride); |
|
|
if (cbp & 2) |
|
|
transfer_16to8add(pU_Cur, &data[4 * 64], stride2); |
|
|
if (cbp & 1) |
|
|
transfer_16to8add(pV_Cur, &data[5 * 64], stride2); |
|
|
stop_transfer_timer(); |
|
|
} |
|
|
|
|
|
|
|
|
|
|
366 |
/* permute block and return field dct choice */ |
/* permute block and return field dct choice */ |
367 |
|
|
368 |
|
static uint32_t |
|
uint32_t |
|
369 |
MBDecideFieldDCT(int16_t data[6 * 64]) |
MBDecideFieldDCT(int16_t data[6 * 64]) |
370 |
{ |
{ |
371 |
uint32_t field = MBFieldTest(data); |
const uint32_t field = MBFieldTest(data); |
372 |
|
if (field) MBFrameToField(data); |
|
if (field) { |
|
|
MBFrameToField(data); |
|
|
} |
|
373 |
|
|
374 |
return field; |
return field; |
375 |
} |
} |
376 |
|
|
|
|
|
377 |
/* if sum(diff between field lines) < sum(diff between frame lines), use field dct */ |
/* if sum(diff between field lines) < sum(diff between frame lines), use field dct */ |
378 |
|
|
379 |
uint32_t |
uint32_t |