Skip to content

L
libyuv
Commit 84456171 authored by Lei Zhang's avatar Lei Zhang Committed by Commit Bot
Browse files
Options

Mark a bunch of kArray variables as const. 

This allows the linker to move the variables from the .data section to
the .rodata section.

Bug: libyuv:254
Test: out/Release/libyuv_unittest --gtest_filter=* --libyuv_width=1280 --libyuv_height=720 --libyuv_repeat=999 --libyuv_flags=-1 --libyuv_cpu_info=-1

Change-Id: I6998570f1af4337d7b80313d9e18e36aa20d6ec0
Reviewed-on: https://chromium-review.googlesource.com/777033
Commit-Queue: Frank Barchard <fbarchard@chromium.org>
Reviewed-by: 's avatarFrank Barchard <fbarchard@chromium.org>
parent 26173eb7
Expand all Hide whitespace changes
Inline Side-by-side
Showing with 121 additions and 113 deletions
source/compare_gcc.cc
View file @ 84456171
......@@ -111,9 +111,9 @@ uint32 HammingDistance_SSE42(const uint8* src_a,
}
#endif
static vec8 kNibbleMask = {15, 15, 15, 15, 15, 15, 15, 15,
15, 15, 15, 15, 15, 15, 15, 15};
static vec8 kBitCount = {0, 1, 1, 2, 1, 2, 2, 3, 1, 2, 2, 3, 2, 3, 3, 4};
static const vec8 kNibbleMask = {15, 15, 15, 15, 15, 15, 15, 15,
15, 15, 15, 15, 15, 15, 15, 15};
static const vec8 kBitCount = {0, 1, 1, 2, 1, 2, 2, 3, 1, 2, 2, 3, 2, 3, 3, 4};
uint32 HammingDistance_SSSE3(const uint8* src_a,
const uint8* src_b,
......@@ -267,26 +267,26 @@ uint32 SumSquareError_SSE2(const uint8* src_a, const uint8* src_b, int count) {
return sse;
}
static uvec32 kHash16x33 = {0x92d9e201, 0, 0, 0}; // 33 ^ 16
static uvec32 kHashMul0 = {
static const uvec32 kHash16x33 = {0x92d9e201, 0, 0, 0}; // 33 ^ 16
static const uvec32 kHashMul0 = {
0x0c3525e1, // 33 ^ 15
0xa3476dc1, // 33 ^ 14
0x3b4039a1, // 33 ^ 13
0x4f5f0981, // 33 ^ 12
};
static uvec32 kHashMul1 = {
static const uvec32 kHashMul1 = {
0x30f35d61, // 33 ^ 11
0x855cb541, // 33 ^ 10
0x040a9121, // 33 ^ 9
0x747c7101, // 33 ^ 8
};
static uvec32 kHashMul2 = {
static const uvec32 kHashMul2 = {
0xec41d4e1, // 33 ^ 7
0x4cfa3cc1, // 33 ^ 6
0x025528a1, // 33 ^ 5
0x00121881, // 33 ^ 4
};
static uvec32 kHashMul3 = {
static const uvec32 kHashMul3 = {
0x00008c61, // 33 ^ 3
0x00000441, // 33 ^ 2
0x00000021, // 33 ^ 1
......
source/convert_argb.cc
View file @ 84456171
......@@ -995,16 +995,16 @@ int J400ToARGB(const uint8* src_y,
}
// Shuffle table for converting BGRA to ARGB.
static uvec8 kShuffleMaskBGRAToARGB = {3u, 2u, 1u, 0u, 7u, 6u, 5u, 4u,
11u, 10u, 9u, 8u, 15u, 14u, 13u, 12u};
static const uvec8 kShuffleMaskBGRAToARGB = {
3u, 2u, 1u, 0u, 7u, 6u, 5u, 4u, 11u, 10u, 9u, 8u, 15u, 14u, 13u, 12u};
// Shuffle table for converting ABGR to ARGB.
static uvec8 kShuffleMaskABGRToARGB = {2u, 1u, 0u, 3u, 6u, 5u, 4u, 7u,
10u, 9u, 8u, 11u, 14u, 13u, 12u, 15u};
static const uvec8 kShuffleMaskABGRToARGB = {
2u, 1u, 0u, 3u, 6u, 5u, 4u, 7u, 10u, 9u, 8u, 11u, 14u, 13u, 12u, 15u};
// Shuffle table for converting RGBA to ARGB.
static uvec8 kShuffleMaskRGBAToARGB = {1u, 2u, 3u, 0u, 5u, 6u, 7u, 4u,
9u, 10u, 11u, 8u, 13u, 14u, 15u, 12u};
static const uvec8 kShuffleMaskRGBAToARGB = {
1u, 2u, 3u, 0u, 5u, 6u, 7u, 4u, 9u, 10u, 11u, 8u, 13u, 14u, 15u, 12u};
// Convert BGRA to ARGB.
LIBYUV_API
......
source/convert_from_argb.cc
View file @ 84456171
......@@ -906,8 +906,8 @@ int ARGBToI400(const uint8* src_argb,
}
// Shuffle table for converting ARGB to RGBA.
static uvec8 kShuffleMaskARGBToRGBA = {3u, 0u, 1u, 2u, 7u, 4u, 5u, 6u,
11u, 8u, 9u, 10u, 15u, 12u, 13u, 14u};
static const uvec8 kShuffleMaskARGBToRGBA = {
3u, 0u, 1u, 2u, 7u, 4u, 5u, 6u, 11u, 8u, 9u, 10u, 15u, 12u, 13u, 14u};
// Convert ARGB to RGBA.
LIBYUV_API
......
source/rotate_neon.cc
View file @ 84456171
......@@ -21,8 +21,8 @@ extern "C" {
#if !defined(LIBYUV_DISABLE_NEON) && defined(__ARM_NEON__) && \
!defined(__aarch64__)
static uvec8 kVTbl4x4Transpose = {0, 4, 8, 12, 1, 5, 9, 13,
2, 6, 10, 14, 3, 7, 11, 15};
static const uvec8 kVTbl4x4Transpose = {0, 4, 8, 12, 1, 5, 9, 13,
2, 6, 10, 14, 3, 7, 11, 15};
void TransposeWx8_NEON(const uint8* src,
int src_stride,
......@@ -189,8 +189,8 @@ void TransposeWx8_NEON(const uint8* src,
: "memory", "cc", "q0", "q1", "q2", "q3");
}
static uvec8 kVTbl4x4TransposeDi = {0, 8, 1, 9, 2, 10, 3, 11,
4, 12, 5, 13, 6, 14, 7, 15};
static const uvec8 kVTbl4x4TransposeDi = {0, 8, 1, 9, 2, 10, 3, 11,
4, 12, 5, 13, 6, 14, 7, 15};
void TransposeUVWx8_NEON(const uint8* src,
int src_stride,
......
source/rotate_neon64.cc
View file @ 84456171
......@@ -21,8 +21,8 @@ extern "C" {
// This module is for GCC Neon armv8 64 bit.
#if !defined(LIBYUV_DISABLE_NEON) && defined(__aarch64__)
static uvec8 kVTbl4x4Transpose = {0, 4, 8, 12, 1, 5, 9, 13,
2, 6, 10, 14, 3, 7, 11, 15};
static const uvec8 kVTbl4x4Transpose = {0, 4, 8, 12, 1, 5, 9, 13,
2, 6, 10, 14, 3, 7, 11, 15};
void TransposeWx8_NEON(const uint8* src,
int src_stride,
......@@ -196,7 +196,7 @@ void TransposeWx8_NEON(const uint8* src,
"v17", "v18", "v19", "v20", "v21", "v22", "v23");
}
static uint8 kVTbl4x4TransposeDi[32] = {
static const uint8 kVTbl4x4TransposeDi[32] = {
0, 16, 32, 48, 2, 18, 34, 50, 4, 20, 36, 52, 6, 22, 38, 54,
1, 17, 33, 49, 3, 19, 35, 51, 5, 21, 37, 53, 7, 23, 39, 55};
......
source/row_gcc.cc
View file @ 84456171
This diff is collapsed.
source/scale_gcc.cc
View file @ 84456171
......@@ -21,72 +21,73 @@ extern "C" {
(defined(__x86_64__) || (defined(__i386__) && !defined(_MSC_VER)))
// Offsets for source bytes 0 to 9
static uvec8 kShuf0 = {0, 1, 3, 4, 5, 7, 8, 9,
128, 128, 128, 128, 128, 128, 128, 128};
static const uvec8 kShuf0 = {0, 1, 3, 4, 5, 7, 8, 9,
128, 128, 128, 128, 128, 128, 128, 128};
// Offsets for source bytes 11 to 20 with 8 subtracted = 3 to 12.
static uvec8 kShuf1 = {3, 4, 5, 7, 8, 9, 11, 12,
128, 128, 128, 128, 128, 128, 128, 128};
static const uvec8 kShuf1 = {3, 4, 5, 7, 8, 9, 11, 12,
128, 128, 128, 128, 128, 128, 128, 128};
// Offsets for source bytes 21 to 31 with 16 subtracted = 5 to 31.
static uvec8 kShuf2 = {5, 7, 8, 9, 11, 12, 13, 15,
128, 128, 128, 128, 128, 128, 128, 128};
static const uvec8 kShuf2 = {5, 7, 8, 9, 11, 12, 13, 15,
128, 128, 128, 128, 128, 128, 128, 128};
// Offsets for source bytes 0 to 10
static uvec8 kShuf01 = {0, 1, 1, 2, 2, 3, 4, 5, 5, 6, 6, 7, 8, 9, 9, 10};
static const uvec8 kShuf01 = {0, 1, 1, 2, 2, 3, 4, 5, 5, 6, 6, 7, 8, 9, 9, 10};
// Offsets for source bytes 10 to 21 with 8 subtracted = 3 to 13.
static uvec8 kShuf11 = {2, 3, 4, 5, 5, 6, 6, 7, 8, 9, 9, 10, 10, 11, 12, 13};
static const uvec8 kShuf11 = {2, 3, 4, 5, 5, 6, 6, 7,
8, 9, 9, 10, 10, 11, 12, 13};
// Offsets for source bytes 21 to 31 with 16 subtracted = 5 to 31.
static uvec8 kShuf21 = {5, 6, 6, 7, 8, 9, 9, 10,
10, 11, 12, 13, 13, 14, 14, 15};
static const uvec8 kShuf21 = {5, 6, 6, 7, 8, 9, 9, 10,
10, 11, 12, 13, 13, 14, 14, 15};
// Coefficients for source bytes 0 to 10
static uvec8 kMadd01 = {3, 1, 2, 2, 1, 3, 3, 1, 2, 2, 1, 3, 3, 1, 2, 2};
static const uvec8 kMadd01 = {3, 1, 2, 2, 1, 3, 3, 1, 2, 2, 1, 3, 3, 1, 2, 2};
// Coefficients for source bytes 10 to 21
static uvec8 kMadd11 = {1, 3, 3, 1, 2, 2, 1, 3, 3, 1, 2, 2, 1, 3, 3, 1};
static const uvec8 kMadd11 = {1, 3, 3, 1, 2, 2, 1, 3, 3, 1, 2, 2, 1, 3, 3, 1};
// Coefficients for source bytes 21 to 31
static uvec8 kMadd21 = {2, 2, 1, 3, 3, 1, 2, 2, 1, 3, 3, 1, 2, 2, 1, 3};
static const uvec8 kMadd21 = {2, 2, 1, 3, 3, 1, 2, 2, 1, 3, 3, 1, 2, 2, 1, 3};
// Coefficients for source bytes 21 to 31
static vec16 kRound34 = {2, 2, 2, 2, 2, 2, 2, 2};
static const vec16 kRound34 = {2, 2, 2, 2, 2, 2, 2, 2};
static uvec8 kShuf38a = {0, 3, 6, 8, 11, 14, 128, 128,
128, 128, 128, 128, 128, 128, 128, 128};
static const uvec8 kShuf38a = {0, 3, 6, 8, 11, 14, 128, 128,
128, 128, 128, 128, 128, 128, 128, 128};
static uvec8 kShuf38b = {128, 128, 128, 128, 128, 128, 0, 3,
6, 8, 11, 14, 128, 128, 128, 128};
static const uvec8 kShuf38b = {128, 128, 128, 128, 128, 128, 0, 3,
6, 8, 11, 14, 128, 128, 128, 128};
// Arrange words 0,3,6 into 0,1,2
static uvec8 kShufAc = {0, 1, 6, 7, 12, 13, 128, 128,
128, 128, 128, 128, 128, 128, 128, 128};
static const uvec8 kShufAc = {0, 1, 6, 7, 12, 13, 128, 128,
128, 128, 128, 128, 128, 128, 128, 128};
// Arrange words 0,3,6 into 3,4,5
static uvec8 kShufAc3 = {128, 128, 128, 128, 128, 128, 0, 1,
6, 7, 12, 13, 128, 128, 128, 128};
static const uvec8 kShufAc3 = {128, 128, 128, 128, 128, 128, 0, 1,
6, 7, 12, 13, 128, 128, 128, 128};
// Scaling values for boxes of 3x3 and 2x3
static uvec16 kScaleAc33 = {65536 / 9, 65536 / 9, 65536 / 6, 65536 / 9,
65536 / 9, 65536 / 6, 0, 0};
static const uvec16 kScaleAc33 = {65536 / 9, 65536 / 9, 65536 / 6, 65536 / 9,
65536 / 9, 65536 / 6, 0, 0};
// Arrange first value for pixels 0,1,2,3,4,5
static uvec8 kShufAb0 = {0, 128, 3, 128, 6, 128, 8, 128,
11, 128, 14, 128, 128, 128, 128, 128};
static const uvec8 kShufAb0 = {0, 128, 3, 128, 6, 128, 8, 128,
11, 128, 14, 128, 128, 128, 128, 128};
// Arrange second value for pixels 0,1,2,3,4,5
static uvec8 kShufAb1 = {1, 128, 4, 128, 7, 128, 9, 128,
12, 128, 15, 128, 128, 128, 128, 128};
static const uvec8 kShufAb1 = {1, 128, 4, 128, 7, 128, 9, 128,
12, 128, 15, 128, 128, 128, 128, 128};
// Arrange third value for pixels 0,1,2,3,4,5
static uvec8 kShufAb2 = {2, 128, 5, 128, 128, 128, 10, 128,
13, 128, 128, 128, 128, 128, 128, 128};
static const uvec8 kShufAb2 = {2, 128, 5, 128, 128, 128, 10, 128,
13, 128, 128, 128, 128, 128, 128, 128};
// Scaling values for boxes of 3x2 and 2x2
static uvec16 kScaleAb2 = {65536 / 3, 65536 / 3, 65536 / 2, 65536 / 3,
65536 / 3, 65536 / 2, 0, 0};
static const uvec16 kScaleAb2 = {65536 / 3, 65536 / 3, 65536 / 2, 65536 / 3,
65536 / 3, 65536 / 2, 0, 0};
// GCC versions of row functions are verbatim conversions from Visual C.
// Generated using gcc disassembly on Visual C object file:
......@@ -851,12 +852,12 @@ void ScaleAddRow_AVX2(const uint8* src_ptr, uint16* dst_ptr, int src_width) {
// Constant for making pixels signed to avoid pmaddubsw
// saturation.
static uvec8 kFsub80 = {0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80};
static const uvec8 kFsub80 = {0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80};
// Constant for making pixels unsigned and adding .5 for rounding.
static uvec16 kFadd40 = {0x4040, 0x4040, 0x4040, 0x4040,
0x4040, 0x4040, 0x4040, 0x4040};
static const uvec16 kFadd40 = {0x4040, 0x4040, 0x4040, 0x4040,
0x4040, 0x4040, 0x4040, 0x4040};
// Bilinear column filtering. SSSE3 version.
void ScaleFilterCols_SSSE3(uint8* dst_ptr,
......@@ -1250,13 +1251,13 @@ void ScaleARGBColsUp2_SSE2(uint8* dst_argb,
}
// Shuffle table for arranging 2 pixels into pairs for pmaddubsw
static uvec8 kShuffleColARGB = {
static const uvec8 kShuffleColARGB = {
0u, 4u, 1u, 5u, 2u, 6u, 3u, 7u, // bbggrraa 1st pixel
8u, 12u, 9u, 13u, 10u, 14u, 11u, 15u // bbggrraa 2nd pixel
};
// Shuffle table for duplicating 2 fractions into 8 bytes each
static uvec8 kShuffleFractions = {
static const uvec8 kShuffleFractions = {
0u, 0u, 0u, 0u, 0u, 0u, 0u, 0u, 4u, 4u, 4u, 4u, 4u, 4u, 4u, 4u,
};
......
source/scale_neon.cc
View file @ 84456171
......@@ -264,13 +264,16 @@ void ScaleRowDown34_1_Box_NEON(const uint8* src_ptr,
}
#define HAS_SCALEROWDOWN38_NEON
static uvec8 kShuf38 = {0, 3, 6, 8, 11, 14, 16, 19, 22, 24, 27, 30, 0, 0, 0, 0};
static uvec8 kShuf38_2 = {0, 8, 16, 2, 10, 17, 4, 12,
18, 6, 14, 19, 0, 0, 0, 0};
static vec16 kMult38_Div6 = {65536 / 12, 65536 / 12, 65536 / 12, 65536 / 12,
65536 / 12, 65536 / 12, 65536 / 12, 65536 / 12};
static vec16 kMult38_Div9 = {65536 / 18, 65536 / 18, 65536 / 18, 65536 / 18,
65536 / 18, 65536 / 18, 65536 / 18, 65536 / 18};
static const uvec8 kShuf38 = {0, 3, 6, 8, 11, 14, 16, 19,
22, 24, 27, 30, 0, 0, 0, 0};
static const uvec8 kShuf38_2 = {0, 8, 16, 2, 10, 17, 4, 12,
18, 6, 14, 19, 0, 0, 0, 0};
static const vec16 kMult38_Div6 = {65536 / 12, 65536 / 12, 65536 / 12,
65536 / 12, 65536 / 12, 65536 / 12,
65536 / 12, 65536 / 12};
static const vec16 kMult38_Div9 = {65536 / 18, 65536 / 18, 65536 / 18,
65536 / 18, 65536 / 18, 65536 / 18,
65536 / 18, 65536 / 18};
// 32 -> 12
void ScaleRowDown38_NEON(const uint8* src_ptr,
......
source/scale_neon64.cc
View file @ 84456171
......@@ -261,13 +261,16 @@ void ScaleRowDown34_1_Box_NEON(const uint8* src_ptr,
: "v0", "v1", "v2", "v3", "v4", "v5", "v6", "v7", "v20", "memory", "cc");
}
static uvec8 kShuf38 = {0, 3, 6, 8, 11, 14, 16, 19, 22, 24, 27, 30, 0, 0, 0, 0};
static uvec8 kShuf38_2 = {0, 16, 32, 2, 18, 33, 4, 20,
34, 6, 22, 35, 0, 0, 0, 0};
static vec16 kMult38_Div6 = {65536 / 12, 65536 / 12, 65536 / 12, 65536 / 12,
65536 / 12, 65536 / 12, 65536 / 12, 65536 / 12};
static vec16 kMult38_Div9 = {65536 / 18, 65536 / 18, 65536 / 18, 65536 / 18,
65536 / 18, 65536 / 18, 65536 / 18, 65536 / 18};
static const uvec8 kShuf38 = {0, 3, 6, 8, 11, 14, 16, 19,
22, 24, 27, 30, 0, 0, 0, 0};
static const uvec8 kShuf38_2 = {0, 16, 32, 2, 18, 33, 4, 20,
34, 6, 22, 35, 0, 0, 0, 0};
static const vec16 kMult38_Div6 = {65536 / 12, 65536 / 12, 65536 / 12,
65536 / 12, 65536 / 12, 65536 / 12,
65536 / 12, 65536 / 12};
static const vec16 kMult38_Div9 = {65536 / 18, 65536 / 18, 65536 / 18,
65536 / 18, 65536 / 18, 65536 / 18,
65536 / 18, 65536 / 18};
// 32 -> 12
void ScaleRowDown38_NEON(const uint8* src_ptr,
......
source/scale_win.cc
View file @ 84456171
......@@ -20,72 +20,73 @@ extern "C" {
#if !defined(LIBYUV_DISABLE_X86) && defined(_M_IX86) && defined(_MSC_VER)
// Offsets for source bytes 0 to 9
static uvec8 kShuf0 = {0, 1, 3, 4, 5, 7, 8, 9,
128, 128, 128, 128, 128, 128, 128, 128};
static const uvec8 kShuf0 = {0, 1, 3, 4, 5, 7, 8, 9,
128, 128, 128, 128, 128, 128, 128, 128};
// Offsets for source bytes 11 to 20 with 8 subtracted = 3 to 12.
static uvec8 kShuf1 = {3, 4, 5, 7, 8, 9, 11, 12,
128, 128, 128, 128, 128, 128, 128, 128};
static const uvec8 kShuf1 = {3, 4, 5, 7, 8, 9, 11, 12,
128, 128, 128, 128, 128, 128, 128, 128};
// Offsets for source bytes 21 to 31 with 16 subtracted = 5 to 31.
static uvec8 kShuf2 = {5, 7, 8, 9, 11, 12, 13, 15,
128, 128, 128, 128, 128, 128, 128, 128};
static const uvec8 kShuf2 = {5, 7, 8, 9, 11, 12, 13, 15,
128, 128, 128, 128, 128, 128, 128, 128};
// Offsets for source bytes 0 to 10
static uvec8 kShuf01 = {0, 1, 1, 2, 2, 3, 4, 5, 5, 6, 6, 7, 8, 9, 9, 10};
static const uvec8 kShuf01 = {0, 1, 1, 2, 2, 3, 4, 5, 5, 6, 6, 7, 8, 9, 9, 10};
// Offsets for source bytes 10 to 21 with 8 subtracted = 3 to 13.
static uvec8 kShuf11 = {2, 3, 4, 5, 5, 6, 6, 7, 8, 9, 9, 10, 10, 11, 12, 13};
static const uvec8 kShuf11 = {2, 3, 4, 5, 5, 6, 6, 7,
8, 9, 9, 10, 10, 11, 12, 13};
// Offsets for source bytes 21 to 31 with 16 subtracted = 5 to 31.
static uvec8 kShuf21 = {5, 6, 6, 7, 8, 9, 9, 10,
10, 11, 12, 13, 13, 14, 14, 15};
static const uvec8 kShuf21 = {5, 6, 6, 7, 8, 9, 9, 10,
10, 11, 12, 13, 13, 14, 14, 15};
// Coefficients for source bytes 0 to 10
static uvec8 kMadd01 = {3, 1, 2, 2, 1, 3, 3, 1, 2, 2, 1, 3, 3, 1, 2, 2};
static const uvec8 kMadd01 = {3, 1, 2, 2, 1, 3, 3, 1, 2, 2, 1, 3, 3, 1, 2, 2};
// Coefficients for source bytes 10 to 21
static uvec8 kMadd11 = {1, 3, 3, 1, 2, 2, 1, 3, 3, 1, 2, 2, 1, 3, 3, 1};
static const uvec8 kMadd11 = {1, 3, 3, 1, 2, 2, 1, 3, 3, 1, 2, 2, 1, 3, 3, 1};
// Coefficients for source bytes 21 to 31
static uvec8 kMadd21 = {2, 2, 1, 3, 3, 1, 2, 2, 1, 3, 3, 1, 2, 2, 1, 3};
static const uvec8 kMadd21 = {2, 2, 1, 3, 3, 1, 2, 2, 1, 3, 3, 1, 2, 2, 1, 3};
// Coefficients for source bytes 21 to 31
static vec16 kRound34 = {2, 2, 2, 2, 2, 2, 2, 2};
static const vec16 kRound34 = {2, 2, 2, 2, 2, 2, 2, 2};
static uvec8 kShuf38a = {0, 3, 6, 8, 11, 14, 128, 128,
128, 128, 128, 128, 128, 128, 128, 128};
static const uvec8 kShuf38a = {0, 3, 6, 8, 11, 14, 128, 128,
128, 128, 128, 128, 128, 128, 128, 128};
static uvec8 kShuf38b = {128, 128, 128, 128, 128, 128, 0, 3,
6, 8, 11, 14, 128, 128, 128, 128};
static const uvec8 kShuf38b = {128, 128, 128, 128, 128, 128, 0, 3,
6, 8, 11, 14, 128, 128, 128, 128};
// Arrange words 0,3,6 into 0,1,2
static uvec8 kShufAc = {0, 1, 6, 7, 12, 13, 128, 128,
128, 128, 128, 128, 128, 128, 128, 128};
static const uvec8 kShufAc = {0, 1, 6, 7, 12, 13, 128, 128,
128, 128, 128, 128, 128, 128, 128, 128};
// Arrange words 0,3,6 into 3,4,5
static uvec8 kShufAc3 = {128, 128, 128, 128, 128, 128, 0, 1,
6, 7, 12, 13, 128, 128, 128, 128};
static const uvec8 kShufAc3 = {128, 128, 128, 128, 128, 128, 0, 1,
6, 7, 12, 13, 128, 128, 128, 128};
// Scaling values for boxes of 3x3 and 2x3
static uvec16 kScaleAc33 = {65536 / 9, 65536 / 9, 65536 / 6, 65536 / 9,
65536 / 9, 65536 / 6, 0, 0};
static const uvec16 kScaleAc33 = {65536 / 9, 65536 / 9, 65536 / 6, 65536 / 9,
65536 / 9, 65536 / 6, 0, 0};
// Arrange first value for pixels 0,1,2,3,4,5
static uvec8 kShufAb0 = {0, 128, 3, 128, 6, 128, 8, 128,
11, 128, 14, 128, 128, 128, 128, 128};
static const uvec8 kShufAb0 = {0, 128, 3, 128, 6, 128, 8, 128,
11, 128, 14, 128, 128, 128, 128, 128};
// Arrange second value for pixels 0,1,2,3,4,5
static uvec8 kShufAb1 = {1, 128, 4, 128, 7, 128, 9, 128,
12, 128, 15, 128, 128, 128, 128, 128};
static const uvec8 kShufAb1 = {1, 128, 4, 128, 7, 128, 9, 128,
12, 128, 15, 128, 128, 128, 128, 128};
// Arrange third value for pixels 0,1,2,3,4,5
static uvec8 kShufAb2 = {2, 128, 5, 128, 128, 128, 10, 128,
13, 128, 128, 128, 128, 128, 128, 128};
static const uvec8 kShufAb2 = {2, 128, 5, 128, 128, 128, 10, 128,
13, 128, 128, 128, 128, 128, 128, 128};
// Scaling values for boxes of 3x2 and 2x2
static uvec16 kScaleAb2 = {65536 / 3, 65536 / 3, 65536 / 2, 65536 / 3,
65536 / 3, 65536 / 2, 0, 0};
static const uvec16 kScaleAb2 = {65536 / 3, 65536 / 3, 65536 / 2, 65536 / 3,
65536 / 3, 65536 / 2, 0, 0};
// Reads 32 pixels, throws half away and writes 16 pixels.
__declspec(naked) void ScaleRowDown2_SSSE3(const uint8* src_ptr,
......@@ -870,12 +871,12 @@ __declspec(naked) void ScaleAddRow_AVX2(const uint8* src_ptr,
// Constant for making pixels signed to avoid pmaddubsw
// saturation.
static uvec8 kFsub80 = {0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80};
static const uvec8 kFsub80 = {0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80};
// Constant for making pixels unsigned and adding .5 for rounding.
static uvec16 kFadd40 = {0x4040, 0x4040, 0x4040, 0x4040,
0x4040, 0x4040, 0x4040, 0x4040};
static const uvec16 kFadd40 = {0x4040, 0x4040, 0x4040, 0x4040,
0x4040, 0x4040, 0x4040, 0x4040};
// Bilinear column filtering. SSSE3 version.
__declspec(naked) void ScaleFilterCols_SSSE3(uint8* dst_ptr,
......@@ -1246,13 +1247,13 @@ __declspec(naked) void ScaleARGBCols_SSE2(uint8* dst_argb,
// TODO(fbarchard): Port to Neon
// Shuffle table for arranging 2 pixels into pairs for pmaddubsw
static uvec8 kShuffleColARGB = {
static const uvec8 kShuffleColARGB = {
0u, 4u, 1u, 5u, 2u, 6u, 3u, 7u, // bbggrraa 1st pixel
8u, 12u, 9u, 13u, 10u, 14u, 11u, 15u // bbggrraa 2nd pixel
};
// Shuffle table for duplicating 2 fractions into 8 bytes each
static uvec8 kShuffleFractions = {
static const uvec8 kShuffleFractions = {
0u, 0u, 0u, 0u, 0u, 0u, 0u, 0u, 4u, 4u, 4u, 4u, 4u, 4u, 4u, 4u,
};
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment