OLD | NEW |
1 /* | 1 /* |
2 * Copyright (c) 2013 The WebM project authors. All Rights Reserved. | 2 * Copyright (c) 2013 The WebM project authors. All Rights Reserved. |
3 * | 3 * |
4 * Use of this source code is governed by a BSD-style license | 4 * Use of this source code is governed by a BSD-style license |
5 * that can be found in the LICENSE file in the root of the source | 5 * that can be found in the LICENSE file in the root of the source |
6 * tree. An additional intellectual property rights grant can be found | 6 * tree. An additional intellectual property rights grant can be found |
7 * in the file PATENTS. All contributing project authors may | 7 * in the file PATENTS. All contributing project authors may |
8 * be found in the AUTHORS file in the root of the source tree. | 8 * be found in the AUTHORS file in the root of the source tree. |
9 */ | 9 */ |
10 | 10 |
11 #include <assert.h> | 11 #include <assert.h> |
12 #include <stdio.h> | 12 #include <stdio.h> |
13 | 13 |
14 #include "./vpx_dsp_rtcd.h" | 14 #include "./vpx_dsp_rtcd.h" |
15 #include "vpx_dsp/mips/vpx_common_dspr2.h" | 15 #include "vpx_dsp/mips/convolve_common_dspr2.h" |
16 #include "vpx_dsp/vpx_convolve.h" | 16 #include "vpx_dsp/vpx_convolve.h" |
17 #include "vpx_dsp/vpx_dsp_common.h" | 17 #include "vpx_dsp/vpx_dsp_common.h" |
18 #include "vpx_ports/mem.h" | 18 #include "vpx_ports/mem.h" |
19 | 19 |
20 #if HAVE_DSPR2 | 20 #if HAVE_DSPR2 |
21 static void convolve_avg_vert_4_dspr2(const uint8_t *src, | 21 static void convolve_avg_vert_4_dspr2(const uint8_t *src, |
22 int32_t src_stride, | 22 int32_t src_stride, |
23 uint8_t *dst, | 23 uint8_t *dst, |
24 int32_t dst_stride, | 24 int32_t dst_stride, |
25 const int16_t *filter_y, | 25 const int16_t *filter_y, |
(...skipping 314 matching lines...) Expand 10 before | Expand all | Expand 10 after Loading... |
340 src += src_stride; | 340 src += src_stride; |
341 dst += dst_stride; | 341 dst += dst_stride; |
342 } | 342 } |
343 } | 343 } |
344 | 344 |
345 void vpx_convolve8_avg_vert_dspr2(const uint8_t *src, ptrdiff_t src_stride, | 345 void vpx_convolve8_avg_vert_dspr2(const uint8_t *src, ptrdiff_t src_stride, |
346 uint8_t *dst, ptrdiff_t dst_stride, | 346 uint8_t *dst, ptrdiff_t dst_stride, |
347 const int16_t *filter_x, int x_step_q4, | 347 const int16_t *filter_x, int x_step_q4, |
348 const int16_t *filter_y, int y_step_q4, | 348 const int16_t *filter_y, int y_step_q4, |
349 int w, int h) { | 349 int w, int h) { |
350 if (((const int32_t *)filter_y)[1] == 0x800000) { | 350 assert(y_step_q4 == 16); |
351 vpx_convolve_avg(src, src_stride, | 351 assert(((const int32_t *)filter_y)[1] != 0x800000); |
352 dst, dst_stride, | 352 |
353 filter_x, x_step_q4, | 353 if (((const int32_t *)filter_y)[0] == 0) { |
354 filter_y, y_step_q4, | |
355 w, h); | |
356 } else if (((const int32_t *)filter_y)[0] == 0) { | |
357 vpx_convolve2_avg_vert_dspr2(src, src_stride, | 354 vpx_convolve2_avg_vert_dspr2(src, src_stride, |
358 dst, dst_stride, | 355 dst, dst_stride, |
359 filter_x, x_step_q4, | 356 filter_x, x_step_q4, |
360 filter_y, y_step_q4, | 357 filter_y, y_step_q4, |
361 w, h); | 358 w, h); |
362 } else { | 359 } else { |
363 if (16 == y_step_q4) { | 360 uint32_t pos = 38; |
364 uint32_t pos = 38; | |
365 | 361 |
366 /* bit positon for extract from acc */ | 362 /* bit positon for extract from acc */ |
367 __asm__ __volatile__ ( | 363 __asm__ __volatile__ ( |
368 "wrdsp %[pos], 1 \n\t" | 364 "wrdsp %[pos], 1 \n\t" |
369 : | 365 : |
370 : [pos] "r" (pos) | 366 : [pos] "r" (pos) |
371 ); | 367 ); |
372 | 368 |
373 prefetch_store(dst); | 369 prefetch_store(dst); |
374 | 370 |
375 switch (w) { | 371 switch (w) { |
376 case 4: | 372 case 4: |
377 case 8: | 373 case 8: |
378 case 16: | 374 case 16: |
379 case 32: | 375 case 32: |
380 convolve_avg_vert_4_dspr2(src, src_stride, | 376 convolve_avg_vert_4_dspr2(src, src_stride, |
381 dst, dst_stride, | 377 dst, dst_stride, |
382 filter_y, w, h); | 378 filter_y, w, h); |
383 break; | 379 break; |
384 case 64: | 380 case 64: |
385 prefetch_store(dst + 32); | 381 prefetch_store(dst + 32); |
386 convolve_avg_vert_64_dspr2(src, src_stride, | 382 convolve_avg_vert_64_dspr2(src, src_stride, |
387 dst, dst_stride, | |
388 filter_y, h); | |
389 break; | |
390 default: | |
391 vpx_convolve8_avg_vert_c(src, src_stride, | |
392 dst, dst_stride, | 383 dst, dst_stride, |
393 filter_x, x_step_q4, | 384 filter_y, h); |
394 filter_y, y_step_q4, | 385 break; |
395 w, h); | 386 default: |
396 break; | 387 vpx_convolve8_avg_vert_c(src, src_stride, |
397 } | 388 dst, dst_stride, |
398 } else { | 389 filter_x, x_step_q4, |
399 vpx_convolve8_avg_vert_c(src, src_stride, | 390 filter_y, y_step_q4, |
400 dst, dst_stride, | 391 w, h); |
401 filter_x, x_step_q4, | 392 break; |
402 filter_y, y_step_q4, | |
403 w, h); | |
404 } | 393 } |
405 } | 394 } |
406 } | 395 } |
407 | 396 |
408 void vpx_convolve8_avg_dspr2(const uint8_t *src, ptrdiff_t src_stride, | 397 void vpx_convolve8_avg_dspr2(const uint8_t *src, ptrdiff_t src_stride, |
409 uint8_t *dst, ptrdiff_t dst_stride, | 398 uint8_t *dst, ptrdiff_t dst_stride, |
410 const int16_t *filter_x, int x_step_q4, | 399 const int16_t *filter_x, int x_step_q4, |
411 const int16_t *filter_y, int y_step_q4, | 400 const int16_t *filter_y, int y_step_q4, |
412 int w, int h) { | 401 int w, int h) { |
413 /* Fixed size intermediate buffer places limits on parameters. */ | 402 /* Fixed size intermediate buffer places limits on parameters. */ |
414 DECLARE_ALIGNED(32, uint8_t, temp[64 * 135]); | 403 DECLARE_ALIGNED(32, uint8_t, temp[64 * 135]); |
415 int32_t intermediate_height = ((h * y_step_q4) >> 4) + 7; | 404 int32_t intermediate_height = ((h * y_step_q4) >> 4) + 7; |
416 | 405 |
417 assert(w <= 64); | 406 assert(w <= 64); |
418 assert(h <= 64); | 407 assert(h <= 64); |
| 408 assert(x_step_q4 == 16); |
| 409 assert(y_step_q4 == 16); |
419 | 410 |
420 if (intermediate_height < h) | 411 if (intermediate_height < h) |
421 intermediate_height = h; | 412 intermediate_height = h; |
422 | 413 |
423 if (x_step_q4 != 16 || y_step_q4 != 16) | |
424 return vpx_convolve8_avg_c(src, src_stride, | |
425 dst, dst_stride, | |
426 filter_x, x_step_q4, | |
427 filter_y, y_step_q4, | |
428 w, h); | |
429 | |
430 vpx_convolve8_horiz(src - (src_stride * 3), src_stride, | 414 vpx_convolve8_horiz(src - (src_stride * 3), src_stride, |
431 temp, 64, | 415 temp, 64, |
432 filter_x, x_step_q4, | 416 filter_x, x_step_q4, |
433 filter_y, y_step_q4, | 417 filter_y, y_step_q4, |
434 w, intermediate_height); | 418 w, intermediate_height); |
435 | 419 |
436 vpx_convolve8_avg_vert(temp + 64 * 3, 64, | 420 vpx_convolve8_avg_vert(temp + 64 * 3, 64, |
437 dst, dst_stride, | 421 dst, dst_stride, |
438 filter_x, x_step_q4, | 422 filter_x, x_step_q4, |
439 filter_y, y_step_q4, | 423 filter_y, y_step_q4, |
(...skipping 244 matching lines...) Expand 10 before | Expand all | Expand 10 after Loading... |
684 dst[x] = (dst[x] + src[x] + 1) >> 1; | 668 dst[x] = (dst[x] + src[x] + 1) >> 1; |
685 } | 669 } |
686 | 670 |
687 src += src_stride; | 671 src += src_stride; |
688 dst += dst_stride; | 672 dst += dst_stride; |
689 } | 673 } |
690 break; | 674 break; |
691 } | 675 } |
692 } | 676 } |
693 #endif | 677 #endif |
OLD | NEW |