Chromium Code Reviews
chromiumcodereview-hr@appspot.gserviceaccount.com (chromiumcodereview-hr) | Please choose your nickname with Settings | Help | Chromium Project | Gerrit Changes | Sign out
(154)

Side by Side Diff: source/scale_common.cc

Issue 2093913004: rounding for arm filter (Closed) Base URL: https://chromium.googlesource.com/libyuv/libyuv@master
Patch Set: arm or aarch64 Created 4 years, 6 months ago
Use n/p to move between diff chunks; N/P to move between comments. Draft comments are only viewable by you.
Jump to:
View unified diff | Download patch
« no previous file with comments | « include/libyuv/version.h ('k') | source/scale_neon.cc » ('j') | no next file with comments »
Toggle Intra-line Diffs ('i') | Expand Comments ('e') | Collapse Comments ('c') | Show Comments Hide Comments ('s')
OLDNEW
1 /* 1 /*
2 * Copyright 2013 The LibYuv Project Authors. All rights reserved. 2 * Copyright 2013 The LibYuv 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
(...skipping 399 matching lines...) Expand 10 before | Expand all | Expand 10 after
410 dst_ptr[1] = dst_ptr[0] = src_ptr[0]; 410 dst_ptr[1] = dst_ptr[0] = src_ptr[0];
411 src_ptr += 1; 411 src_ptr += 1;
412 dst_ptr += 2; 412 dst_ptr += 2;
413 } 413 }
414 if (dst_width & 1) { 414 if (dst_width & 1) {
415 dst_ptr[0] = src_ptr[0]; 415 dst_ptr[0] = src_ptr[0];
416 } 416 }
417 } 417 }
418 418
419 // (1-f)a + fb can be replaced with a + f(b-a) 419 // (1-f)a + fb can be replaced with a + f(b-a)
420 #if defined(__arm__) 420 #if defined(__arm__) || defined(__aarch64__)
421 // arm uses 16 bit math with truncation.
422 // TODO(fbarchard): add rounding.
423 #define BLENDER(a, b, f) (uint8)((int)(a) + \ 421 #define BLENDER(a, b, f) (uint8)((int)(a) + \
424 (((int)((f)) * ((int)(b) - (int)(a))) >> 16)) 422 ((((int)((f)) * ((int)(b) - (int)(a))) + 0x8000) >> 16))
425 #else 423 #else
426 // inteluses 7 bit math with rounding. 424 // inteluses 7 bit math with rounding.
427 #define BLENDER(a, b, f) (uint8)((int)(a) + \ 425 #define BLENDER(a, b, f) (uint8)((int)(a) + \
428 (((int)((f) >> 9) * ((int)(b) - (int)(a)) + 0x40) >> 7)) 426 (((int)((f) >> 9) * ((int)(b) - (int)(a)) + 0x40) >> 7))
429 #endif 427 #endif
430 428
431 void ScaleFilterCols_C(uint8* dst_ptr, const uint8* src_ptr, 429 void ScaleFilterCols_C(uint8* dst_ptr, const uint8* src_ptr,
432 int dst_width, int x, int dx) { 430 int dst_width, int x, int dx) {
433 int j; 431 int j;
434 for (j = 0; j < dst_width - 1; j += 2) { 432 for (j = 0; j < dst_width - 1; j += 2) {
(...skipping 38 matching lines...) Expand 10 before | Expand all | Expand 10 after
473 int64 xi = x >> 16; 471 int64 xi = x >> 16;
474 int a = src_ptr[xi]; 472 int a = src_ptr[xi];
475 int b = src_ptr[xi + 1]; 473 int b = src_ptr[xi + 1];
476 dst_ptr[0] = BLENDER(a, b, x & 0xffff); 474 dst_ptr[0] = BLENDER(a, b, x & 0xffff);
477 } 475 }
478 } 476 }
479 #undef BLENDER 477 #undef BLENDER
480 478
481 // Same as 8 bit arm blender but return is cast to uint16 479 // Same as 8 bit arm blender but return is cast to uint16
482 #define BLENDER(a, b, f) (uint16)((int)(a) + \ 480 #define BLENDER(a, b, f) (uint16)((int)(a) + \
483 (((int)((f)) * ((int)(b) - (int)(a))) >> 16)) 481 ((((int)((f)) * ((int)(b) - (int)(a))) + 0x8000) >> 16))
484 482
485 void ScaleFilterCols_16_C(uint16* dst_ptr, const uint16* src_ptr, 483 void ScaleFilterCols_16_C(uint16* dst_ptr, const uint16* src_ptr,
486 int dst_width, int x, int dx) { 484 int dst_width, int x, int dx) {
487 int j; 485 int j;
488 for (j = 0; j < dst_width - 1; j += 2) { 486 for (j = 0; j < dst_width - 1; j += 2) {
489 int xi = x >> 16; 487 int xi = x >> 16;
490 int a = src_ptr[xi]; 488 int a = src_ptr[xi];
491 int b = src_ptr[xi + 1]; 489 int b = src_ptr[xi + 1];
492 dst_ptr[0] = BLENDER(a, b, x & 0xffff); 490 dst_ptr[0] = BLENDER(a, b, x & 0xffff);
493 x += dx; 491 x += dx;
(...skipping 317 matching lines...) Expand 10 before | Expand all | Expand 10 after
811 for (j = 0; j < dst_width - 1; j += 2) { 809 for (j = 0; j < dst_width - 1; j += 2) {
812 dst[1] = dst[0] = src[0]; 810 dst[1] = dst[0] = src[0];
813 src += 1; 811 src += 1;
814 dst += 2; 812 dst += 2;
815 } 813 }
816 if (dst_width & 1) { 814 if (dst_width & 1) {
817 dst[0] = src[0]; 815 dst[0] = src[0];
818 } 816 }
819 } 817 }
820 818
821 // TODO(fbarchard): Replace 0x7f ^ f with 128-f. bug=605. 819 // TODO(fbarchard): Replace 0x7f ^ f with 128-f. bug=607.
822 // Mimics SSSE3 blender 820 // Mimics SSSE3 blender
823 #define BLENDER1(a, b, f) ((a) * (0x7f ^ f) + (b) * f) >> 7 821 #define BLENDER1(a, b, f) ((a) * (0x7f ^ f) + (b) * f) >> 7
824 #define BLENDERC(a, b, f, s) (uint32)( \ 822 #define BLENDERC(a, b, f, s) (uint32)( \
825 BLENDER1(((a) >> s) & 255, ((b) >> s) & 255, f) << s) 823 BLENDER1(((a) >> s) & 255, ((b) >> s) & 255, f) << s)
826 #define BLENDER(a, b, f) \ 824 #define BLENDER(a, b, f) \
827 BLENDERC(a, b, f, 24) | BLENDERC(a, b, f, 16) | \ 825 BLENDERC(a, b, f, 24) | BLENDERC(a, b, f, 16) | \
828 BLENDERC(a, b, f, 8) | BLENDERC(a, b, f, 0) 826 BLENDERC(a, b, f, 8) | BLENDERC(a, b, f, 0)
829 827
830 void ScaleARGBFilterCols_C(uint8* dst_argb, const uint8* src_argb, 828 void ScaleARGBFilterCols_C(uint8* dst_argb, const uint8* src_argb,
831 int dst_width, int x, int dx) { 829 int dst_width, int x, int dx) {
(...skipping 320 matching lines...) Expand 10 before | Expand all | Expand 10 after
1152 *dx = -*dx; 1150 *dx = -*dx;
1153 // src_width = -src_width; // Caller must do this. 1151 // src_width = -src_width; // Caller must do this.
1154 } 1152 }
1155 } 1153 }
1156 #undef CENTERSTART 1154 #undef CENTERSTART
1157 1155
1158 #ifdef __cplusplus 1156 #ifdef __cplusplus
1159 } // extern "C" 1157 } // extern "C"
1160 } // namespace libyuv 1158 } // namespace libyuv
1161 #endif 1159 #endif
OLDNEW
« no previous file with comments | « include/libyuv/version.h ('k') | source/scale_neon.cc » ('j') | no next file with comments »

Powered by Google App Engine
This is Rietveld 408576698