[C#] Bgr24彩色位图转为Gray8灰度位图的跨平台SIMD硬件加速向量算法 - IT评测·应用市场-qidao123.com

public static unsafe void ScalarDo(BitmapData src, BitmapData dst) {
const int cbPixel = 3; // Bgr24
const int shiftPoint = 16;
const int mulPoint = 1 << shiftPoint; // 0x10000
const int mulRed = (int)(0.299 * mulPoint + 0.5); // 19595
const int mulGreen = (int)(0.587 * mulPoint + 0.5); // 38470
const int mulBlue = mulPoint - mulRed - mulGreen; // 7471
int width = src.Width;
int height = src.Height;
int strideSrc = src.Stride;
int strideDst = dst.Stride;
byte* pRow = (byte*)src.Scan0.ToPointer();
byte* qRow = (byte*)dst.Scan0.ToPointer();
for (int i = 0; i < height; i++) {
byte* p = pRow;
byte* q = qRow;
for (int j = 0; j < width; j++) {
*q = (byte)((p[2] * mulRed + p[1] * mulGreen + p[0] * mulBlue) >> shiftPoint);
p += cbPixel; // Bgr24
q += 1; // Gray8
}
pRow += strideSrc;
qRow += strideDst;
}
}

复制代码

private static readonly Random _random = new Random(1);
private BitmapData _sourceBitmapData = null;
private BitmapData _destinationBitmapData = null;
private BitmapData _expectedBitmapData = null;
[Params(1024, 2048, 4096)]
public int Width { get; set; }
public int Height { get; set; }
~Bgr24ToGrayBgr24Benchmark() {
Dispose(false);
}
public void Dispose() {
Dispose(true);
GC.SuppressFinalize(this);
}
private void Dispose(bool disposing) {
if (_disposed) return;
_disposed = true;
if (disposing) {
Cleanup();
}
}
private BitmapData AllocBitmapData(int width, int height, PixelFormat format) {
const int strideAlign = 4;
if (width <= 0) throw new ArgumentOutOfRangeException($"The width({width}) need > 0!");
if (height <= 0) throw new ArgumentOutOfRangeException($"The width({height}) need > 0!");
int stride = 0;
switch (format) {
case PixelFormat.Format8bppIndexed:
stride = width * 1;
break;
case PixelFormat.Format24bppRgb:
stride = width * 3;
break;
}
if (stride <= 0) throw new ArgumentOutOfRangeException($"Invalid pixel format({format})!");
if (0 != (stride % strideAlign)) {
stride = stride - (stride % strideAlign) + strideAlign;
}
BitmapData bitmapData = new BitmapData();
bitmapData.Width = width;
bitmapData.Height = height;
bitmapData.PixelFormat = format;
bitmapData.Stride = stride;
bitmapData.Scan0 = Marshal.AllocHGlobal(stride * height);
return bitmapData;
}
private void FreeBitmapData(BitmapData bitmapData) {
if (null == bitmapData) return;
if (IntPtr.Zero == bitmapData.Scan0) return;
Marshal.FreeHGlobal(bitmapData.Scan0);
bitmapData.Scan0 = IntPtr.Zero;
}
[GlobalCleanup]
public void Cleanup() {
FreeBitmapData(_sourceBitmapData); _sourceBitmapData = null;
FreeBitmapData(_destinationBitmapData); _destinationBitmapData = null;
FreeBitmapData(_expectedBitmapData); _expectedBitmapData = null;
}
[GlobalSetup]
public void Setup() {
Height = Width;
// Create.
Cleanup();
_sourceBitmapData = AllocBitmapData(Width, Height, PixelFormat.Format24bppRgb);
_destinationBitmapData = AllocBitmapData(Width, Height, PixelFormat.Format8bppIndexed);
_expectedBitmapData = AllocBitmapData(Width, Height, PixelFormat.Format8bppIndexed);
RandomFillBitmapData(_sourceBitmapData, _random);
}

复制代码

[Benchmark(Baseline = true)]
public void Scalar() {
ScalarDo(_sourceBitmapData, _destinationBitmapData);
}

复制代码

static readonly Vector128<byte> YGroup3Unzip_Shuffle_Byte_X_Part0 = Vector128.Create((sbyte)0, 3, 6, 9, 12, 15, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1).AsByte();
static readonly Vector128<byte> YGroup3Unzip_Shuffle_Byte_X_Part1 = Vector128.Create((sbyte)-1, -1, -1, -1, -1, -1, 2, 5, 8, 11, 14, -1, -1, -1, -1, -1).AsByte();
static readonly Vector128<byte> YGroup3Unzip_Shuffle_Byte_X_Part2 = Vector128.Create((sbyte)-1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 1, 4, 7, 10, 13).AsByte();
static readonly Vector128<byte> YGroup3Unzip_Shuffle_Byte_Y_Part0 = Vector128.Create((sbyte)1, 4, 7, 10, 13, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1).AsByte();
static readonly Vector128<byte> YGroup3Unzip_Shuffle_Byte_Y_Part1 = Vector128.Create((sbyte)-1, -1, -1, -1, -1, 0, 3, 6, 9, 12, 15, -1, -1, -1, -1, -1).AsByte();
static readonly Vector128<byte> YGroup3Unzip_Shuffle_Byte_Y_Part2 = Vector128.Create((sbyte)-1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 2, 5, 8, 11, 14).AsByte();
static readonly Vector128<byte> YGroup3Unzip_Shuffle_Byte_Z_Part0 = Vector128.Create((sbyte)2, 5, 8, 11, 14, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1).AsByte();
static readonly Vector128<byte> YGroup3Unzip_Shuffle_Byte_Z_Part1 = Vector128.Create((sbyte)-1, -1, -1, -1, -1, 1, 4, 7, 10, 13, -1, -1, -1, -1, -1, -1).AsByte();
static readonly Vector128<byte> YGroup3Unzip_Shuffle_Byte_Z_Part2 = Vector128.Create((sbyte)-1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 0, 3, 6, 9, 12, 15).AsByte();
public static Vector128<byte> YGroup3Unzip(Vector128<byte> data0, Vector128<byte> data1, Vector128<byte> data2, out Vector128<byte> y, out Vector128<byte> z) {
var f0A = YGroup3Unzip_Shuffle_Byte_X_Part0;
var f0B = YGroup3Unzip_Shuffle_Byte_X_Part1;
var f0C = YGroup3Unzip_Shuffle_Byte_X_Part2;
var f1A = YGroup3Unzip_Shuffle_Byte_Y_Part0;
var f1B = YGroup3Unzip_Shuffle_Byte_Y_Part1;
var f1C = YGroup3Unzip_Shuffle_Byte_Y_Part2;
var f2A = YGroup3Unzip_Shuffle_Byte_Z_Part0;
var f2B = YGroup3Unzip_Shuffle_Byte_Z_Part1;
var f2C = YGroup3Unzip_Shuffle_Byte_Z_Part2;
var rt0 = Sse2.Or(Sse2.Or(Ssse3.Shuffle(data0, f0A), Ssse3.Shuffle(data1, f0B)), Ssse3.Shuffle(data2, f0C));
var rt1 = Sse2.Or(Sse2.Or(Ssse3.Shuffle(data0, f1A), Ssse3.Shuffle(data1, f1B)), Ssse3.Shuffle(data2, f1C));
var rt2 = Sse2.Or(Sse2.Or(Ssse3.Shuffle(data0, f2A), Ssse3.Shuffle(data1, f2B)), Ssse3.Shuffle(data2, f2C));
y = rt1;
z = rt2;
return rt0;
}

复制代码

public static unsafe void UseVectorsDoBatch(byte* pSrc, int strideSrc, int width, int height, byte* pDst, int strideDst) {
const int cbPixel = 3; // Bgr24
const int shiftPoint = 8;
const int mulPoint = 1 << shiftPoint; // 0x100
const ushort mulRed = (ushort)(0.299 * mulPoint + 0.5); // 77
const ushort mulGreen = (ushort)(0.587 * mulPoint + 0.5); // 150
const ushort mulBlue = mulPoint - mulRed - mulGreen; // 29
Vector<ushort> vmulRed = new Vector<ushort>(mulRed);
Vector<ushort> vmulGreen = new Vector<ushort>(mulGreen);
Vector<ushort> vmulBlue = new Vector<ushort>(mulBlue);
int vectorWidth = Vector<byte>.Count;
int maxX = width - vectorWidth;
byte* pRow = pSrc;
byte* qRow = pDst;
for (int i = 0; i < height; i++) {
Vector<byte>* pLast = (Vector<byte>*)(pRow + maxX * cbPixel);
Vector<byte>* qLast = (Vector<byte>*)(qRow + maxX * 1);
Vector<byte>* p = (Vector<byte>*)pRow;
Vector<byte>* q = (Vector<byte>*)qRow;
for (; ; ) {
Vector<byte> r, g, b, gray;
Vector<ushort> wr0, wr1, wg0, wg1, wb0, wb1;
// Load.
b = Vectors.YGroup3Unzip(p[0], p[1], p[2], out g, out r);
// widen(r) * mulRed + widen(g) * mulGreen + widen(b) * mulBlue
Vector.Widen(r, out wr0, out wr1);
Vector.Widen(g, out wg0, out wg1);
Vector.Widen(b, out wb0, out wb1);
wr0 = Vectors.Multiply(wr0, vmulRed);
wr1 = Vectors.Multiply(wr1, vmulRed);
wg0 = Vectors.Multiply(wg0, vmulGreen);
wg1 = Vectors.Multiply(wg1, vmulGreen);
wb0 = Vectors.Multiply(wb0, vmulBlue);
wb1 = Vectors.Multiply(wb1, vmulBlue);
wr0 = Vector.Add(wr0, wg0);
wr1 = Vector.Add(wr1, wg1);
wr0 = Vector.Add(wr0, wb0);
wr1 = Vector.Add(wr1, wb1);
// Shift right and narrow.
wr0 = Vectors.ShiftRightLogical_Const(wr0, shiftPoint);
wr1 = Vectors.ShiftRightLogical_Const(wr1, shiftPoint);
gray = Vector.Narrow(wr0, wr1);
// Store.
*q = gray;
// Next.
if (p >= pLast) break;
p += cbPixel;
++q;
if (p > pLast) p = pLast; // The last block is also use vector.
if (q > qLast) q = qLast;
}
pRow += strideSrc;
qRow += strideDst;
}
}

复制代码