1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
|
diff --git a/third_party/libyuv/source/row_neon64.cc b/third_party/libyuv/source/row_neon64.cc
index 350c964..2aab413 100644
--- a/third_party/libyuv/source/row_neon64.cc
+++ b/third_party/libyuv/source/row_neon64.cc
@@ -1835,7 +1835,7 @@ void ARGBToAB64Row_NEON(const uint8_t* src_argb,
: "+r"(src_argb), // %0
"+r"(dst_ab64), // %1
"+r"(width) // %2
- : "m"(kShuffleARGBToABGR) // %3
+ : "Q"(kShuffleARGBToABGR) // %3
: "cc", "memory", "v0", "v1", "v2", "v3", "v4");
}
@@ -1859,7 +1859,7 @@ void AR64ToARGBRow_NEON(const uint16_t* src_ar64,
: "+r"(src_ar64), // %0
"+r"(dst_argb), // %1
"+r"(width) // %2
- : "m"(kShuffleAR64ToARGB) // %3
+ : "Q"(kShuffleAR64ToARGB) // %3
: "cc", "memory", "v0", "v1", "v2", "v3", "v4");
}
@@ -1883,7 +1883,7 @@ void AB64ToARGBRow_NEON(const uint16_t* src_ab64,
: "+r"(src_ab64), // %0
"+r"(dst_argb), // %1
"+r"(width) // %2
- : "m"(kShuffleAB64ToARGB) // %3
+ : "Q"(kShuffleAB64ToARGB) // %3
: "cc", "memory", "v0", "v1", "v2", "v3", "v4");
}
diff --git a/third_party/libyuv/source/scale_neon64.cc b/third_party/libyuv/source/scale_neon64.cc
index 8656fec..9f9636e 100644
--- a/third_party/libyuv/source/scale_neon64.cc
+++ b/third_party/libyuv/source/scale_neon64.cc
@@ -601,8 +601,8 @@ void ScaleRowUp2_Bilinear_NEON(const uint8_t* src_ptr,
"umlal v4.8h, v1.8b, v31.8b \n" // 3*near+far (2, odd)
"umlal v5.8h, v0.8b, v31.8b \n" // 3*near+far (2, even)
- "mov v0.8h, v4.8h \n"
- "mov v1.8h, v5.8h \n"
+ "mov v0.16b, v4.16b \n"
+ "mov v1.16b, v5.16b \n"
"mla v4.8h, v2.8h, v30.8h \n" // 9 3 3 1 (1, odd)
"mla v5.8h, v3.8h, v30.8h \n" // 9 3 3 1 (1, even)
"mla v2.8h, v0.8h, v30.8h \n" // 9 3 3 1 (2, odd)
@@ -642,7 +642,7 @@ void ScaleRowUp2_Linear_12_NEON(const uint16_t* src_ptr,
"ld1 {v1.8h}, [%1], #16 \n" // 12345678 (16b)
"prfm pldl1keep, [%0, 448] \n" // prefetch 7 lines ahead
- "mov v2.8h, v0.8h \n"
+ "mov v2.16b, v0.16b \n"
"mla v0.8h, v1.8h, v31.8h \n" // 3*near+far (odd)
"mla v1.8h, v2.8h, v31.8h \n" // 3*near+far (even)
@@ -679,7 +679,7 @@ void ScaleRowUp2_Bilinear_12_NEON(const uint16_t* src_ptr,
"ld1 {v3.8h}, [%2], #16 \n" // 12345678 (16b)
"prfm pldl1keep, [%0, 448] \n" // prefetch 7 lines ahead
- "mov v0.8h, v2.8h \n"
+ "mov v0.16b, v2.16b \n"
"mla v2.8h, v3.8h, v31.8h \n" // 3*near+far (odd)
"mla v3.8h, v0.8h, v31.8h \n" // 3*near+far (even)
@@ -687,12 +687,12 @@ void ScaleRowUp2_Bilinear_12_NEON(const uint16_t* src_ptr,
"ld1 {v5.8h}, [%3], #16 \n" // 12345678 (16b)
"prfm pldl1keep, [%1, 448] \n" // prefetch 7 lines ahead
- "mov v0.8h, v4.8h \n"
+ "mov v0.16b, v4.16b \n"
"mla v4.8h, v5.8h, v31.8h \n" // 3*near+far (odd)
"mla v5.8h, v0.8h, v31.8h \n" // 3*near+far (even)
- "mov v0.8h, v4.8h \n"
- "mov v1.8h, v5.8h \n"
+ "mov v0.16b, v4.16b \n"
+ "mov v1.16b, v5.16b \n"
"mla v4.8h, v2.8h, v31.8h \n" // 9 3 3 1 (1, odd)
"mla v5.8h, v3.8h, v31.8h \n" // 9 3 3 1 (1, even)
"mla v2.8h, v0.8h, v31.8h \n" // 9 3 3 1 (2, odd)
@@ -887,8 +887,8 @@ void ScaleUVRowUp2_Bilinear_NEON(const uint8_t* src_ptr,
"umlal v4.8h, v1.8b, v31.8b \n" // 3*near+far (2, odd)
"umlal v5.8h, v0.8b, v31.8b \n" // 3*near+far (2, even)
- "mov v0.8h, v4.8h \n"
- "mov v1.8h, v5.8h \n"
+ "mov v0.16b, v4.16b \n"
+ "mov v1.16b, v5.16b \n"
"mla v4.8h, v2.8h, v30.8h \n" // 9 3 3 1 (1, odd)
"mla v5.8h, v3.8h, v30.8h \n" // 9 3 3 1 (1, even)
"mla v2.8h, v0.8h, v30.8h \n" // 9 3 3 1 (2, odd)
|
