265 |
uint32_t quarterpel, |
uint32_t quarterpel, |
266 |
uint32_t rounding) |
uint32_t rounding) |
267 |
{ |
{ |
268 |
const uint32_t offset = EDGE_SIZE * (edged_width + 1); |
const uint32_t offset = EDGE_SIZE2 * (edged_width + 1); // we only interpolate half of the edge area |
269 |
const uint32_t stride_add = 7 * edged_width; |
const uint32_t stride_add = 7 * edged_width; |
270 |
|
/* |
271 |
#ifdef BFRAMES |
#ifdef BFRAMES |
272 |
const uint32_t edged_width2 = edged_width / 2; |
const uint32_t edged_width2 = edged_width / 2; |
273 |
const uint32_t edged_height2 = edged_height / 2; |
const uint32_t edged_height2 = edged_height / 2; |
274 |
const uint32_t offset2 = EDGE_SIZE2 * (edged_width2 + 1); |
const uint32_t offset2 = EDGE_SIZE2 * (edged_width2 + 1); |
275 |
const uint32_t stride_add2 = 7 * edged_width2; |
const uint32_t stride_add2 = 7 * edged_width2; |
276 |
#endif |
#endif |
277 |
|
*/ |
278 |
uint8_t *n_ptr, *h_ptr, *v_ptr, *hv_ptr; |
uint8_t *n_ptr, *h_ptr, *v_ptr, *hv_ptr; |
279 |
uint32_t x, y; |
uint32_t x, y; |
280 |
|
|
291 |
|
|
292 |
if(quarterpel) { |
if(quarterpel) { |
293 |
|
|
294 |
for (y = 0; y < edged_height; y += 8) { |
for (y = 0; y < (edged_height - EDGE_SIZE); y += 8) { |
295 |
for (x = 0; x < edged_width; x += 8) { |
for (x = 0; x < (edged_width - EDGE_SIZE); x += 8) { |
296 |
interpolate8x8_6tap_lowpass_h(h_ptr, n_ptr, edged_width, rounding); |
interpolate8x8_6tap_lowpass_h(h_ptr, n_ptr, edged_width, rounding); |
297 |
interpolate8x8_6tap_lowpass_v(v_ptr, n_ptr, edged_width, rounding); |
interpolate8x8_6tap_lowpass_v(v_ptr, n_ptr, edged_width, rounding); |
298 |
|
|
301 |
v_ptr += 8; |
v_ptr += 8; |
302 |
} |
} |
303 |
|
|
304 |
|
n_ptr += EDGE_SIZE; |
305 |
|
h_ptr += EDGE_SIZE; |
306 |
|
v_ptr += EDGE_SIZE; |
307 |
|
|
308 |
h_ptr += stride_add; |
h_ptr += stride_add; |
309 |
v_ptr += stride_add; |
v_ptr += stride_add; |
310 |
n_ptr += stride_add; |
n_ptr += stride_add; |
313 |
h_ptr = refh->y; |
h_ptr = refh->y; |
314 |
h_ptr -= offset; |
h_ptr -= offset; |
315 |
|
|
316 |
for (y = 0; y < edged_height; y = y + 8) { |
for (y = 0; y < (edged_height - EDGE_SIZE); y = y + 8) { |
317 |
for (x = 0; x < edged_width; x = x + 8) { |
for (x = 0; x < (edged_width - EDGE_SIZE); x = x + 8) { |
318 |
interpolate8x8_6tap_lowpass_v(hv_ptr, h_ptr, edged_width, rounding); |
interpolate8x8_6tap_lowpass_v(hv_ptr, h_ptr, edged_width, rounding); |
319 |
hv_ptr += 8; |
hv_ptr += 8; |
320 |
h_ptr += 8; |
h_ptr += 8; |
321 |
} |
} |
322 |
|
|
323 |
|
hv_ptr += EDGE_SIZE2; |
324 |
|
h_ptr += EDGE_SIZE2; |
325 |
|
|
326 |
hv_ptr += stride_add; |
hv_ptr += stride_add; |
327 |
h_ptr += stride_add; |
h_ptr += stride_add; |
328 |
} |
} |
329 |
} |
} |
330 |
else { |
else { |
331 |
|
|
332 |
for (y = 0; y < edged_height; y += 8) { |
for (y = 0; y < (edged_height - EDGE_SIZE); y += 8) { |
333 |
for (x = 0; x < edged_width; x += 8) { |
for (x = 0; x < (edged_width - EDGE_SIZE); x += 8) { |
334 |
interpolate8x8_halfpel_h(h_ptr, n_ptr, edged_width, rounding); |
interpolate8x8_halfpel_h(h_ptr, n_ptr, edged_width, rounding); |
335 |
interpolate8x8_halfpel_v(v_ptr, n_ptr, edged_width, rounding); |
interpolate8x8_halfpel_v(v_ptr, n_ptr, edged_width, rounding); |
336 |
interpolate8x8_halfpel_hv(hv_ptr, n_ptr, edged_width, rounding); |
interpolate8x8_halfpel_hv(hv_ptr, n_ptr, edged_width, rounding); |
341 |
hv_ptr += 8; |
hv_ptr += 8; |
342 |
} |
} |
343 |
|
|
344 |
|
h_ptr += EDGE_SIZE; |
345 |
|
v_ptr += EDGE_SIZE; |
346 |
|
hv_ptr += EDGE_SIZE; |
347 |
|
n_ptr += EDGE_SIZE; |
348 |
|
|
349 |
h_ptr += stride_add; |
h_ptr += stride_add; |
350 |
v_ptr += stride_add; |
v_ptr += stride_add; |
351 |
hv_ptr += stride_add; |
hv_ptr += stride_add; |