Chromium Code Reviews
chromiumcodereview-hr@appspot.gserviceaccount.com (chromiumcodereview-hr) | Please choose your nickname with Settings | Help | Chromium Project | Gerrit Changes | Sign out
(12)

Issues Search
    My Issues | Starred     Open | Closed | All

Side by Side Diff: source/row_gcc.cc

Issue 1533643005: Add rounding to InterpolateRow for improved quality and consistency. (Closed) Base URL: https://chromium.googlesource.com/libyuv/libyuv@master
Patch Set: lint fix in unittest Created 5 years 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 | « source/row_common.cc ('k') | source/row_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 // VERSION 2 1 // VERSION 2
2 /* 2 /*
3 * Copyright 2011 The LibYuv Project Authors. All rights reserved. 3 * Copyright 2011 The LibYuv Project Authors. All rights reserved.
4 * 4 *
5 * Use of this source code is governed by a BSD-style license 5 * Use of this source code is governed by a BSD-style license
6 * that can be found in the LICENSE file in the root of the source 6 * that can be found in the LICENSE file in the root of the source
7 * tree. An additional intellectual property rights grant can be found 7 * tree. An additional intellectual property rights grant can be found
8 * in the file PATENTS. All contributing project authors may 8 * in the file PATENTS. All contributing project authors may
9 * be found in the AUTHORS file in the root of the source tree. 9 * be found in the AUTHORS file in the root of the source tree.
10 */ 10 */
(...skipping 4776 matching lines...) Expand 10 before | Expand all | Expand 10 after
4787 #ifdef HAS_INTERPOLATEROW_SSSE3 4787 #ifdef HAS_INTERPOLATEROW_SSSE3
4788 // Bilinear filter 16x2 -> 16x1 4788 // Bilinear filter 16x2 -> 16x1
4789 void InterpolateRow_SSSE3(uint8* dst_ptr, const uint8* src_ptr, 4789 void InterpolateRow_SSSE3(uint8* dst_ptr, const uint8* src_ptr,
4790 ptrdiff_t src_stride, int dst_width, 4790 ptrdiff_t src_stride, int dst_width,
4791 int source_y_fraction) { 4791 int source_y_fraction) {
4792 asm volatile ( 4792 asm volatile (
4793 "sub %1,%0 \n" 4793 "sub %1,%0 \n"
4794 "shr %3 \n" 4794 "shr %3 \n"
4795 "cmp $0x0,%3 \n" 4795 "cmp $0x0,%3 \n"
4796 "je 100f \n" 4796 "je 100f \n"
4797 "cmp $0x20,%3 \n"
4798 "je 75f \n"
4799 "cmp $0x40,%3 \n" 4797 "cmp $0x40,%3 \n"
4800 "je 50f \n" 4798 "je 50f \n"
4801 "cmp $0x60,%3 \n"
4802 "je 25f \n"
4803 4799
4804 "movd %3,%%xmm0 \n" 4800 "movd %3,%%xmm0 \n"
4805 "neg %3 \n" 4801 "neg %3 \n"
4806 "add $0x80,%3 \n" 4802 "add $0x80,%3 \n"
4807 "movd %3,%%xmm5 \n" 4803 "movd %3,%%xmm5 \n"
4808 "punpcklbw %%xmm0,%%xmm5 \n" 4804 "punpcklbw %%xmm0,%%xmm5 \n"
4809 "punpcklwd %%xmm5,%%xmm5 \n" 4805 "punpcklwd %%xmm5,%%xmm5 \n"
4810 "pshufd $0x0,%%xmm5,%%xmm5 \n" 4806 "pshufd $0x0,%%xmm5,%%xmm5 \n"
4807 "mov $0x400040,%%eax \n"
4808 "movd %%eax,%%xmm4 \n"
4809 "pshufd $0x0,%%xmm4,%%xmm4 \n"
4811 4810
4812 // General purpose row blend. 4811 // General purpose row blend.
4813 LABELALIGN 4812 LABELALIGN
4814 "1: \n" 4813 "1: \n"
4815 "movdqu " MEMACCESS(1) ",%%xmm0 \n" 4814 "movdqu " MEMACCESS(1) ",%%xmm0 \n"
4816 MEMOPREG(movdqu,0x00,1,4,1,xmm2) 4815 MEMOPREG(movdqu,0x00,1,4,1,xmm2)
4817 "movdqa %%xmm0,%%xmm1 \n" 4816 "movdqa %%xmm0,%%xmm1 \n"
4818 "punpcklbw %%xmm2,%%xmm0 \n" 4817 "punpcklbw %%xmm2,%%xmm0 \n"
4819 "punpckhbw %%xmm2,%%xmm1 \n" 4818 "punpckhbw %%xmm2,%%xmm1 \n"
4820 "pmaddubsw %%xmm5,%%xmm0 \n" 4819 "pmaddubsw %%xmm5,%%xmm0 \n"
4821 "pmaddubsw %%xmm5,%%xmm1 \n" 4820 "pmaddubsw %%xmm5,%%xmm1 \n"
4821 "paddw %%xmm4,%%xmm0 \n"
4822 "paddw %%xmm4,%%xmm1 \n"
4822 "psrlw $0x7,%%xmm0 \n" 4823 "psrlw $0x7,%%xmm0 \n"
4823 "psrlw $0x7,%%xmm1 \n" 4824 "psrlw $0x7,%%xmm1 \n"
4824 "packuswb %%xmm1,%%xmm0 \n" 4825 "packuswb %%xmm1,%%xmm0 \n"
4825 MEMOPMEM(movdqu,xmm0,0x00,1,0,1) 4826 MEMOPMEM(movdqu,xmm0,0x00,1,0,1)
4826 "lea " MEMLEA(0x10,1) ",%1 \n" 4827 "lea " MEMLEA(0x10,1) ",%1 \n"
4827 "sub $0x10,%2 \n" 4828 "sub $0x10,%2 \n"
4828 "jg 1b \n" 4829 "jg 1b \n"
4829 "jmp 99f \n" 4830 "jmp 99f \n"
4830 4831
4831 // Blend 25 / 75.
4832 LABELALIGN
4833 "25: \n"
4834 "movdqu " MEMACCESS(1) ",%%xmm0 \n"
4835 MEMOPREG(movdqu,0x00,1,4,1,xmm1)
4836 "pavgb %%xmm1,%%xmm0 \n"
4837 "pavgb %%xmm1,%%xmm0 \n"
4838 MEMOPMEM(movdqu,xmm0,0x00,1,0,1)
4839 "lea " MEMLEA(0x10,1) ",%1 \n"
4840 "sub $0x10,%2 \n"
4841 "jg 25b \n"
4842 "jmp 99f \n"
4843
4844 // Blend 50 / 50. 4832 // Blend 50 / 50.
4845 LABELALIGN 4833 LABELALIGN
4846 "50: \n" 4834 "50: \n"
4847 "movdqu " MEMACCESS(1) ",%%xmm0 \n" 4835 "movdqu " MEMACCESS(1) ",%%xmm0 \n"
4848 MEMOPREG(movdqu,0x00,1,4,1,xmm1) 4836 MEMOPREG(movdqu,0x00,1,4,1,xmm1)
4849 "pavgb %%xmm1,%%xmm0 \n" 4837 "pavgb %%xmm1,%%xmm0 \n"
4850 MEMOPMEM(movdqu,xmm0,0x00,1,0,1) 4838 MEMOPMEM(movdqu,xmm0,0x00,1,0,1)
4851 "lea " MEMLEA(0x10,1) ",%1 \n" 4839 "lea " MEMLEA(0x10,1) ",%1 \n"
4852 "sub $0x10,%2 \n" 4840 "sub $0x10,%2 \n"
4853 "jg 50b \n" 4841 "jg 50b \n"
4854 "jmp 99f \n" 4842 "jmp 99f \n"
4855 4843
4856 // Blend 75 / 25.
4857 LABELALIGN
4858 "75: \n"
4859 "movdqu " MEMACCESS(1) ",%%xmm1 \n"
4860 MEMOPREG(movdqu,0x00,1,4,1,xmm0)
4861 "pavgb %%xmm1,%%xmm0 \n"
4862 "pavgb %%xmm1,%%xmm0 \n"
4863 MEMOPMEM(movdqu,xmm0,0x00,1,0,1)
4864 "lea " MEMLEA(0x10,1) ",%1 \n"
4865 "sub $0x10,%2 \n"
4866 "jg 75b \n"
4867 "jmp 99f \n"
4868
4869 // Blend 100 / 0 - Copy row unchanged. 4844 // Blend 100 / 0 - Copy row unchanged.
4870 LABELALIGN 4845 LABELALIGN
4871 "100: \n" 4846 "100: \n"
4872 "movdqu " MEMACCESS(1) ",%%xmm0 \n" 4847 "movdqu " MEMACCESS(1) ",%%xmm0 \n"
4873 MEMOPMEM(movdqu,xmm0,0x00,1,0,1) 4848 MEMOPMEM(movdqu,xmm0,0x00,1,0,1)
4874 "lea " MEMLEA(0x10,1) ",%1 \n" 4849 "lea " MEMLEA(0x10,1) ",%1 \n"
4875 "sub $0x10,%2 \n" 4850 "sub $0x10,%2 \n"
4876 "jg 100b \n" 4851 "jg 100b \n"
4877 4852
4878 "99: \n" 4853 "99: \n"
4879 : "+r"(dst_ptr), // %0 4854 : "+r"(dst_ptr), // %0
4880 "+r"(src_ptr), // %1 4855 "+r"(src_ptr), // %1
4881 "+r"(dst_width), // %2 4856 "+r"(dst_width), // %2
4882 "+r"(source_y_fraction) // %3 4857 "+r"(source_y_fraction) // %3
4883 : "r"((intptr_t)(src_stride)) // %4 4858 : "r"((intptr_t)(src_stride)) // %4
4884 : "memory", "cc", NACL_R14 4859 : "memory", "cc", "eax", NACL_R14
4885 "xmm0", "xmm1", "xmm2", "xmm5" 4860 "xmm0", "xmm1", "xmm2", "xmm4", "xmm5"
4886 ); 4861 );
4887 } 4862 }
4888 #endif // HAS_INTERPOLATEROW_SSSE3 4863 #endif // HAS_INTERPOLATEROW_SSSE3
4889 4864
4890 #ifdef HAS_INTERPOLATEROW_AVX2 4865 #ifdef HAS_INTERPOLATEROW_AVX2
4891 // Bilinear filter 32x2 -> 32x1 4866 // Bilinear filter 32x2 -> 32x1
4892 void InterpolateRow_AVX2(uint8* dst_ptr, const uint8* src_ptr, 4867 void InterpolateRow_AVX2(uint8* dst_ptr, const uint8* src_ptr,
4893 ptrdiff_t src_stride, int dst_width, 4868 ptrdiff_t src_stride, int dst_width,
4894 int source_y_fraction) { 4869 int source_y_fraction) {
4895 asm volatile ( 4870 asm volatile (
4896 "shr %3 \n" 4871 "shr %3 \n"
4897 "cmp $0x0,%3 \n" 4872 "cmp $0x0,%3 \n"
4898 "je 100f \n" 4873 "je 100f \n"
4899 "sub %1,%0 \n" 4874 "sub %1,%0 \n"
4900 "cmp $0x20,%3 \n"
4901 "je 75f \n"
4902 "cmp $0x40,%3 \n" 4875 "cmp $0x40,%3 \n"
4903 "je 50f \n" 4876 "je 50f \n"
4904 "cmp $0x60,%3 \n"
4905 "je 25f \n"
4906 4877
4907 "vmovd %3,%%xmm0 \n" 4878 "vmovd %3,%%xmm0 \n"
4908 "neg %3 \n" 4879 "neg %3 \n"
4909 "add $0x80,%3 \n" 4880 "add $0x80,%3 \n"
4910 "vmovd %3,%%xmm5 \n" 4881 "vmovd %3,%%xmm5 \n"
4911 "vpunpcklbw %%xmm0,%%xmm5,%%xmm5 \n" 4882 "vpunpcklbw %%xmm0,%%xmm5,%%xmm5 \n"
4912 "vpunpcklwd %%xmm5,%%xmm5,%%xmm5 \n" 4883 "vpunpcklwd %%xmm5,%%xmm5,%%xmm5 \n"
4913 "vpxor %%ymm0,%%ymm0,%%ymm0 \n" 4884 "vpxor %%ymm0,%%ymm0,%%ymm0 \n"
4914 "vpermd %%ymm5,%%ymm0,%%ymm5 \n" 4885 "vpermd %%ymm5,%%ymm0,%%ymm5 \n"
4886 "mov $0x400040,%%eax \n"
4887 "vmovd %%eax,%%xmm4 \n"
4888 "vbroadcastss %%xmm4,%%ymm4 \n"
4915 4889
4916 // General purpose row blend. 4890 // General purpose row blend.
4917 LABELALIGN 4891 LABELALIGN
4918 "1: \n" 4892 "1: \n"
4919 "vmovdqu " MEMACCESS(1) ",%%ymm0 \n" 4893 "vmovdqu " MEMACCESS(1) ",%%ymm0 \n"
4920 MEMOPREG(vmovdqu,0x00,1,4,1,ymm2) 4894 MEMOPREG(vmovdqu,0x00,1,4,1,ymm2)
4921 "vpunpckhbw %%ymm2,%%ymm0,%%ymm1 \n" 4895 "vpunpckhbw %%ymm2,%%ymm0,%%ymm1 \n"
4922 "vpunpcklbw %%ymm2,%%ymm0,%%ymm0 \n" 4896 "vpunpcklbw %%ymm2,%%ymm0,%%ymm0 \n"
4923 "vpmaddubsw %%ymm5,%%ymm0,%%ymm0 \n" 4897 "vpmaddubsw %%ymm5,%%ymm0,%%ymm0 \n"
4924 "vpmaddubsw %%ymm5,%%ymm1,%%ymm1 \n" 4898 "vpmaddubsw %%ymm5,%%ymm1,%%ymm1 \n"
4899 "vpaddw %%ymm4,%%ymm0,%%ymm0 \n"
4900 "vpaddw %%ymm4,%%ymm1,%%ymm1 \n"
4925 "vpsrlw $0x7,%%ymm0,%%ymm0 \n" 4901 "vpsrlw $0x7,%%ymm0,%%ymm0 \n"
4926 "vpsrlw $0x7,%%ymm1,%%ymm1 \n" 4902 "vpsrlw $0x7,%%ymm1,%%ymm1 \n"
4927 "vpackuswb %%ymm1,%%ymm0,%%ymm0 \n" 4903 "vpackuswb %%ymm1,%%ymm0,%%ymm0 \n"
4928 MEMOPMEM(vmovdqu,ymm0,0x00,1,0,1) 4904 MEMOPMEM(vmovdqu,ymm0,0x00,1,0,1)
4929 "lea " MEMLEA(0x20,1) ",%1 \n" 4905 "lea " MEMLEA(0x20,1) ",%1 \n"
4930 "sub $0x20,%2 \n" 4906 "sub $0x20,%2 \n"
4931 "jg 1b \n" 4907 "jg 1b \n"
4932 "jmp 99f \n" 4908 "jmp 99f \n"
4933 4909
4934 // Blend 25 / 75.
4935 LABELALIGN
4936 "25: \n"
4937 "vmovdqu " MEMACCESS(1) ",%%ymm0 \n"
4938 MEMOPREG(vmovdqu,0x00,1,4,1,ymm1)
4939 "vpavgb %%ymm1,%%ymm0,%%ymm0 \n"
4940 "vpavgb %%ymm1,%%ymm0,%%ymm0 \n"
4941 MEMOPMEM(vmovdqu,ymm0,0x00,1,0,1)
4942 "lea " MEMLEA(0x20,1) ",%1 \n"
4943 "sub $0x20,%2 \n"
4944 "jg 25b \n"
4945 "jmp 99f \n"
4946
4947 // Blend 50 / 50. 4910 // Blend 50 / 50.
4948 LABELALIGN 4911 LABELALIGN
4949 "50: \n" 4912 "50: \n"
4950 "vmovdqu " MEMACCESS(1) ",%%ymm0 \n" 4913 "vmovdqu " MEMACCESS(1) ",%%ymm0 \n"
4951 VMEMOPREG(vpavgb,0x00,1,4,1,ymm0,ymm0) // vpavgb (%1,%4,1),%%ymm0,%%ymm0 4914 VMEMOPREG(vpavgb,0x00,1,4,1,ymm0,ymm0) // vpavgb (%1,%4,1),%%ymm0,%%ymm0
4952 MEMOPMEM(vmovdqu,ymm0,0x00,1,0,1) 4915 MEMOPMEM(vmovdqu,ymm0,0x00,1,0,1)
4953 "lea " MEMLEA(0x20,1) ",%1 \n" 4916 "lea " MEMLEA(0x20,1) ",%1 \n"
4954 "sub $0x20,%2 \n" 4917 "sub $0x20,%2 \n"
4955 "jg 50b \n" 4918 "jg 50b \n"
4956 "jmp 99f \n" 4919 "jmp 99f \n"
4957 4920
4958 // Blend 75 / 25.
4959 LABELALIGN
4960 "75: \n"
4961 "vmovdqu " MEMACCESS(1) ",%%ymm1 \n"
4962 MEMOPREG(vmovdqu,0x00,1,4,1,ymm0)
4963 "vpavgb %%ymm1,%%ymm0,%%ymm0 \n"
4964 "vpavgb %%ymm1,%%ymm0,%%ymm0 \n"
4965 MEMOPMEM(vmovdqu,ymm0,0x00,1,0,1)
4966 "lea " MEMLEA(0x20,1) ",%1 \n"
4967 "sub $0x20,%2 \n"
4968 "jg 75b \n"
4969 "jmp 99f \n"
4970
4971 // Blend 100 / 0 - Copy row unchanged. 4921 // Blend 100 / 0 - Copy row unchanged.
4972 LABELALIGN 4922 LABELALIGN
4973 "100: \n" 4923 "100: \n"
4974 "rep movsb " MEMMOVESTRING(1,0) " \n" 4924 "rep movsb " MEMMOVESTRING(1,0) " \n"
4975 "jmp 999f \n" 4925 "jmp 999f \n"
4976 4926
4977 "99: \n" 4927 "99: \n"
4978 "vzeroupper \n" 4928 "vzeroupper \n"
4979 "999: \n" 4929 "999: \n"
4980 : "+D"(dst_ptr), // %0 4930 : "+D"(dst_ptr), // %0
4981 "+S"(src_ptr), // %1 4931 "+S"(src_ptr), // %1
4982 "+c"(dst_width), // %2 4932 "+c"(dst_width), // %2
4983 "+r"(source_y_fraction) // %3 4933 "+r"(source_y_fraction) // %3
4984 : "r"((intptr_t)(src_stride)) // %4 4934 : "r"((intptr_t)(src_stride)) // %4
4985 : "memory", "cc", NACL_R14 4935 : "memory", "cc", "eax", NACL_R14
4986 "xmm0", "xmm1", "xmm2", "xmm5" 4936 "xmm0", "xmm1", "xmm2", "xmm4", "xmm5"
4987 ); 4937 );
4988 } 4938 }
4989 #endif // HAS_INTERPOLATEROW_AVX2 4939 #endif // HAS_INTERPOLATEROW_AVX2
4990 4940
4991 #ifdef HAS_INTERPOLATEROW_SSE2
4992 // Bilinear filter 16x2 -> 16x1
4993 void InterpolateRow_SSE2(uint8* dst_ptr, const uint8* src_ptr,
4994 ptrdiff_t src_stride, int dst_width,
4995 int source_y_fraction) {
4996 asm volatile (
4997 "sub %1,%0 \n"
4998 "shr %3 \n"
4999 "cmp $0x0,%3 \n"
5000 "je 100f \n"
5001 "cmp $0x20,%3 \n"
5002 "je 75f \n"
5003 "cmp $0x40,%3 \n"
5004 "je 50f \n"
5005 "cmp $0x60,%3 \n"
5006 "je 25f \n"
5007
5008 "movd %3,%%xmm0 \n"
5009 "neg %3 \n"
5010 "add $0x80,%3 \n"
5011 "movd %3,%%xmm5 \n"
5012 "punpcklbw %%xmm0,%%xmm5 \n"
5013 "punpcklwd %%xmm5,%%xmm5 \n"
5014 "pshufd $0x0,%%xmm5,%%xmm5 \n"
5015 "pxor %%xmm4,%%xmm4 \n"
5016
5017 // General purpose row blend.
5018 LABELALIGN
5019 "1: \n"
5020 "movdqu " MEMACCESS(1) ",%%xmm0 \n"
5021 MEMOPREG(movdqu,0x00,1,4,1,xmm2) // movdqu (%1,%4,1),%%xmm2
5022 "movdqa %%xmm0,%%xmm1 \n"
5023 "movdqa %%xmm2,%%xmm3 \n"
5024 "punpcklbw %%xmm4,%%xmm2 \n"
5025 "punpckhbw %%xmm4,%%xmm3 \n"
5026 "punpcklbw %%xmm4,%%xmm0 \n"
5027 "punpckhbw %%xmm4,%%xmm1 \n"
5028 "psubw %%xmm0,%%xmm2 \n"
5029 "psubw %%xmm1,%%xmm3 \n"
5030 "paddw %%xmm2,%%xmm2 \n"
5031 "paddw %%xmm3,%%xmm3 \n"
5032 "pmulhw %%xmm5,%%xmm2 \n"
5033 "pmulhw %%xmm5,%%xmm3 \n"
5034 "paddw %%xmm2,%%xmm0 \n"
5035 "paddw %%xmm3,%%xmm1 \n"
5036 "packuswb %%xmm1,%%xmm0 \n"
5037 MEMOPMEM(movdqu,xmm0,0x00,1,0,1) // movdqu %%xmm0,(%1,%0,1)
5038 "lea " MEMLEA(0x10,1) ",%1 \n"
5039 "sub $0x10,%2 \n"
5040 "jg 1b \n"
5041 "jmp 99f \n"
5042
5043 // Blend 25 / 75.
5044 LABELALIGN
5045 "25: \n"
5046 "movdqu " MEMACCESS(1) ",%%xmm0 \n"
5047 MEMOPREG(movdqu,0x00,1,4,1,xmm1) // movdqu (%1,%4,1),%%xmm1
5048 "pavgb %%xmm1,%%xmm0 \n"
5049 "pavgb %%xmm1,%%xmm0 \n"
5050 MEMOPMEM(movdqu,xmm0,0x00,1,0,1) // movdqu %%xmm0,(%1,%0,1)
5051 "lea " MEMLEA(0x10,1) ",%1 \n"
5052 "sub $0x10,%2 \n"
5053 "jg 25b \n"
5054 "jmp 99f \n"
5055
5056 // Blend 50 / 50.
5057 LABELALIGN
5058 "50: \n"
5059 "movdqu " MEMACCESS(1) ",%%xmm0 \n"
5060 MEMOPREG(movdqu,0x00,1,4,1,xmm1) // movdqu (%1,%4,1),%%xmm1
5061 "pavgb %%xmm1,%%xmm0 \n"
5062 MEMOPMEM(movdqu,xmm0,0x00,1,0,1) // movdqu %%xmm0,(%1,%0,1)
5063 "lea " MEMLEA(0x10,1) ",%1 \n"
5064 "sub $0x10,%2 \n"
5065 "jg 50b \n"
5066 "jmp 99f \n"
5067
5068 // Blend 75 / 25.
5069 LABELALIGN
5070 "75: \n"
5071 "movdqu " MEMACCESS(1) ",%%xmm1 \n"
5072 MEMOPREG(movdqu,0x00,1,4,1,xmm0) // movdqu (%1,%4,1),%%xmm0
5073 "pavgb %%xmm1,%%xmm0 \n"
5074 "pavgb %%xmm1,%%xmm0 \n"
5075 MEMOPMEM(movdqu,xmm0,0x00,1,0,1) // movdqu %%xmm0,(%1,%0,1)
5076 "lea " MEMLEA(0x10,1) ",%1 \n"
5077 "sub $0x10,%2 \n"
5078 "jg 75b \n"
5079 "jmp 99f \n"
5080
5081 // Blend 100 / 0 - Copy row unchanged.
5082 LABELALIGN
5083 "100: \n"
5084 "movdqu " MEMACCESS(1) ",%%xmm0 \n"
5085 MEMOPMEM(movdqu,xmm0,0x00,1,0,1) // movdqu %%xmm0,(%1,%0,1)
5086 "lea " MEMLEA(0x10,1) ",%1 \n"
5087 "sub $0x10,%2 \n"
5088 "jg 100b \n"
5089
5090 "99: \n"
5091 : "+r"(dst_ptr), // %0
5092 "+r"(src_ptr), // %1
5093 "+r"(dst_width), // %2
5094 "+r"(source_y_fraction) // %3
5095 : "r"((intptr_t)(src_stride)) // %4
5096 : "memory", "cc", NACL_R14
5097 "xmm0", "xmm1", "xmm2", "xmm3", "xmm4", "xmm5"
5098 );
5099 }
5100 #endif // HAS_INTERPOLATEROW_SSE2
5101
5102 #ifdef HAS_ARGBSHUFFLEROW_SSSE3 4941 #ifdef HAS_ARGBSHUFFLEROW_SSSE3
5103 // For BGRAToARGB, ABGRToARGB, RGBAToARGB, and ARGBToRGBA. 4942 // For BGRAToARGB, ABGRToARGB, RGBAToARGB, and ARGBToRGBA.
5104 void ARGBShuffleRow_SSSE3(const uint8* src_argb, uint8* dst_argb, 4943 void ARGBShuffleRow_SSSE3(const uint8* src_argb, uint8* dst_argb,
5105 const uint8* shuffler, int width) { 4944 const uint8* shuffler, int width) {
5106 asm volatile ( 4945 asm volatile (
5107 "movdqu " MEMACCESS(3) ",%%xmm5 \n" 4946 "movdqu " MEMACCESS(3) ",%%xmm5 \n"
5108 LABELALIGN 4947 LABELALIGN
5109 "1: \n" 4948 "1: \n"
5110 "movdqu " MEMACCESS(0) ",%%xmm0 \n" 4949 "movdqu " MEMACCESS(0) ",%%xmm0 \n"
5111 "movdqu " MEMACCESS2(0x10,0) ",%%xmm1 \n" 4950 "movdqu " MEMACCESS2(0x10,0) ",%%xmm1 \n"
(...skipping 489 matching lines...) Expand 10 before | Expand all | Expand 10 after
5601 ); 5440 );
5602 } 5441 }
5603 #endif // HAS_ARGBLUMACOLORTABLEROW_SSSE3 5442 #endif // HAS_ARGBLUMACOLORTABLEROW_SSSE3
5604 5443
5605 #endif // defined(__x86_64__) || defined(__i386__) 5444 #endif // defined(__x86_64__) || defined(__i386__)
5606 5445
5607 #ifdef __cplusplus 5446 #ifdef __cplusplus
5608 } // extern "C" 5447 } // extern "C"
5609 } // namespace libyuv 5448 } // namespace libyuv
5610 #endif 5449 #endif
OLDNEW
« no previous file with comments | « source/row_common.cc ('k') | source/row_neon.cc » ('j') | no next file with comments »

Powered by Google App Engine
This is Rietveld 408576698