public virtual unsafe int FindFirstCharacterToEncodeUtf8(ReadOnlySpan <byte> utf8Text)
{
if (!_isAsciiCacheInitialized)
{
InitializeAsciiCache();
}
// Loop through the input text, terminating when we see ill-formed UTF-8 or when we decode a scalar value
// that must be encoded. If we see either of these things then we'll return its index in the original
// input sequence. If we consume the entire text without seeing either of these, return -1 to indicate
// that the text can be copied as-is without escaping.
fixed(byte *ptr = utf8Text)
{
int idx = 0;
#if NETCOREAPP
if ((Sse2.IsSupported || AdvSimd.Arm64.IsSupported) && utf8Text.Length - 16 >= idx)
{
// Hoist these outside the loop, as the JIT won't do it.
Vector128 <sbyte> bitMaskLookupAsciiNeedsEscaping = _bitMaskLookupAsciiNeedsEscaping;
Vector128 <sbyte> bitPosLookup = Ssse3Helper.s_bitPosLookup;
Vector128 <sbyte> nibbleMaskSByte = Ssse3Helper.s_nibbleMaskSByte;
Vector128 <sbyte> nullMaskSByte = Ssse3Helper.s_nullMaskSByte;
sbyte *startingAddress = (sbyte *)ptr;
do
{
Debug.Assert(startingAddress >= ptr && startingAddress <= (ptr + utf8Text.Length - 16));
// Load the next 16 bytes.
Vector128 <sbyte> sourceValue;
bool containsNonAsciiBytes;
// Check for ASCII text. Any byte that's not in the ASCII range will already be negative when
// casted to signed byte.
if (Sse2.IsSupported)
{
sourceValue = Sse2.LoadVector128(startingAddress);
containsNonAsciiBytes = Sse2Helper.ContainsNonAsciiByte(sourceValue);
}
else if (AdvSimd.Arm64.IsSupported)
{
sourceValue = AdvSimd.LoadVector128(startingAddress);
containsNonAsciiBytes = AdvSimdHelper.ContainsNonAsciiByte(sourceValue);
}
else
{
throw new PlatformNotSupportedException();
}
if (!containsNonAsciiBytes)
{
// All of the following 16 bytes is ASCII.
// TODO AdvSimd: optimization maybe achievable using VectorTableLookup and/or VectorTableLookupExtension
if (Ssse3.IsSupported)
{
Vector128 <sbyte> mask = Ssse3Helper.CreateEscapingMask(sourceValue, bitMaskLookupAsciiNeedsEscaping, bitPosLookup, nibbleMaskSByte, nullMaskSByte);
int index = Sse2Helper.GetIndexOfFirstNonAsciiByte(mask.AsByte());
if (index < 16)
{
idx += index;
goto Return;
}
}
else
{
byte *p = (byte *)startingAddress;
if (DoesAsciiNeedEncoding(p[0]))
{
goto Return;
}
if (DoesAsciiNeedEncoding(p[1]))
{
goto Return1;
}
if (DoesAsciiNeedEncoding(p[2]))
{
goto Return2;
}
if (DoesAsciiNeedEncoding(p[3]))
{
goto Return3;
}
if (DoesAsciiNeedEncoding(p[4]))
{
goto Return4;
}
if (DoesAsciiNeedEncoding(p[5]))
{
goto Return5;
}
if (DoesAsciiNeedEncoding(p[6]))
{
goto Return6;
}
if (DoesAsciiNeedEncoding(p[7]))
{
goto Return7;
}
if (DoesAsciiNeedEncoding(p[8]))
{
goto Return8;
}
if (DoesAsciiNeedEncoding(p[9]))
{
goto Return9;
}
if (DoesAsciiNeedEncoding(p[10]))
{
goto Return10;
}
if (DoesAsciiNeedEncoding(p[11]))
{
goto Return11;
}
if (DoesAsciiNeedEncoding(p[12]))
{
goto Return12;
}
if (DoesAsciiNeedEncoding(p[13]))
{
goto Return13;
}
if (DoesAsciiNeedEncoding(p[14]))
{
goto Return14;
}
if (DoesAsciiNeedEncoding(p[15]))
{
goto Return15;
}
}
idx += 16;
}
else
{
// At least one of the following 16 bytes is non-ASCII.
int processNextSixteen = idx + 16;
Debug.Assert(processNextSixteen <= utf8Text.Length);
while (idx < processNextSixteen)
{
Debug.Assert((ptr + idx) <= (ptr + utf8Text.Length));
if (UnicodeUtility.IsAsciiCodePoint(ptr[idx]))
{
if (DoesAsciiNeedEncoding(ptr[idx]))
{
goto Return;
}
idx++;
}
else
{
OperationStatus opStatus = UnicodeHelpers.DecodeScalarValueFromUtf8(utf8Text.Slice(idx), out uint nextScalarValue, out int utf8BytesConsumedForScalar);
Debug.Assert(nextScalarValue <= int.MaxValue);
if (opStatus != OperationStatus.Done || WillEncode((int)nextScalarValue))
{
goto Return;
}
Debug.Assert(opStatus == OperationStatus.Done);
idx += utf8BytesConsumedForScalar;
}
}
}
startingAddress = (sbyte *)ptr + idx;
}while (utf8Text.Length - 16 >= idx);
// Process the remaining bytes.
Debug.Assert(utf8Text.Length - idx < 16);
}
#endif
while (idx < utf8Text.Length)
{
Debug.Assert((ptr + idx) <= (ptr + utf8Text.Length));
if (UnicodeUtility.IsAsciiCodePoint(ptr[idx]))
{
if (DoesAsciiNeedEncoding(ptr[idx]))
{
goto Return;
}
idx++;
}
else
{
OperationStatus opStatus = UnicodeHelpers.DecodeScalarValueFromUtf8(utf8Text.Slice(idx), out uint nextScalarValue, out int utf8BytesConsumedForScalar);
Debug.Assert(nextScalarValue <= int.MaxValue);
if (opStatus != OperationStatus.Done || WillEncode((int)nextScalarValue))
{
goto Return;
}
Debug.Assert(opStatus == OperationStatus.Done);
idx += utf8BytesConsumedForScalar;
}
}
Debug.Assert(idx == utf8Text.Length);
idx = -1; // All bytes are allowed.
goto Return;
#if NETCOREAPP
Return15:
return(idx + 15);
Return14:
return(idx + 14);
Return13:
return(idx + 13);
Return12:
return(idx + 12);
Return11:
return(idx + 11);
Return10:
return(idx + 10);
Return9:
return(idx + 9);
Return8:
return(idx + 8);
Return7:
return(idx + 7);
Return6:
return(idx + 6);
Return5:
return(idx + 5);
Return4:
return(idx + 4);
Return3:
return(idx + 3);
Return2:
return(idx + 2);
Return1:
return(idx + 1);
#endif
Return:
return(idx);
}
}
public override unsafe int FindFirstCharacterToEncode(char *text, int textLength)
{
if (text == null)
{
throw new ArgumentNullException(nameof(text));
}
int idx = 0;
#if NETCOREAPP
if (Sse2.IsSupported || AdvSimd.Arm64.IsSupported)
{
short *startingAddress = (short *)text;
while (textLength - 8 >= idx)
{
Debug.Assert(startingAddress >= text && startingAddress <= (text + textLength - 8));
// Load the next 8 characters.
Vector128 <short> sourceValue;
Vector128 <short> mask;
bool containsNonAsciiChars;
if (Sse2.IsSupported)
{
sourceValue = Sse2.LoadVector128(startingAddress);
mask = Sse2Helper.CreateAsciiMask(sourceValue);
containsNonAsciiChars = Sse2Helper.ContainsNonAsciiByte(mask.AsSByte());
}
else if (AdvSimd.Arm64.IsSupported)
{
sourceValue = AdvSimd.LoadVector128(startingAddress);
mask = AdvSimdHelper.CreateAsciiMask(sourceValue);
containsNonAsciiChars = AdvSimdHelper.ContainsNonAsciiByte(mask.AsSByte());
}
else
{
throw new PlatformNotSupportedException();
}
if (containsNonAsciiChars)
{
// At least one of the following 8 characters is non-ASCII.
int processNextEight = idx + 8;
Debug.Assert(processNextEight <= textLength);
for (; idx < processNextEight; idx++)
{
Debug.Assert((text + idx) <= (text + textLength));
if (!_allowedCharacters.IsCharacterAllowed(*(text + idx)))
{
goto Return;
}
}
startingAddress += 8;
}
else
{
int index;
// Check if any of the 8 characters need to be escaped.
if (Sse2.IsSupported)
{
mask = Sse2Helper.CreateEscapingMask_UnsafeRelaxedJavaScriptEncoder(sourceValue);
index = Sse2Helper.GetIndexOfFirstNonAsciiByte(mask.AsByte());
}
else if (AdvSimd.Arm64.IsSupported)
{
mask = AdvSimdHelper.CreateEscapingMask_UnsafeRelaxedJavaScriptEncoder(sourceValue);
index = AdvSimdHelper.GetIndexOfFirstNonAsciiByte(mask.AsByte());
}
else
{
throw new PlatformNotSupportedException();
}
// If index >= 16, that means none of the 8 characters needed to be escaped.
if (index < 16)
{
// Found at least one character that needs to be escaped, figure out the index of
// the first one found that needed to be escaped within the 8 characters.
Debug.Assert(index % 2 == 0);
idx += index >> 1;
goto Return;
}
idx += 8;
startingAddress += 8;
}
}
// Process the remaining characters.
Debug.Assert(textLength - idx < 8);
}
#endif
for (; idx < textLength; idx++)
{
Debug.Assert((text + idx) <= (text + textLength));
if (!_allowedCharacters.IsCharacterAllowed(*(text + idx)))
{
goto Return;
}
}
idx = -1; // All characters are allowed.
Return:
return(idx);
}
public override unsafe int FindFirstCharacterToEncodeUtf8(ReadOnlySpan <byte> utf8Text)
{
fixed(byte *ptr = utf8Text)
{
int idx = 0;
#if NETCOREAPP
if (Sse2.IsSupported || AdvSimd.Arm64.IsSupported)
{
sbyte *startingAddress = (sbyte *)ptr;
while (utf8Text.Length - 16 >= idx)
{
Debug.Assert(startingAddress >= ptr && startingAddress <= (ptr + utf8Text.Length - 16));
bool containsNonAsciiBytes;
// Load the next 16 bytes, and check for ASCII text.
// Any byte that's not in the ASCII range will already be negative when casted to signed byte.
if (Sse2.IsSupported)
{
Vector128 <sbyte> sourceValue = Sse2.LoadVector128(startingAddress);
containsNonAsciiBytes = Sse2Helper.ContainsNonAsciiByte(sourceValue);
}
else if (AdvSimd.Arm64.IsSupported)
{
Vector128 <sbyte> sourceValue = AdvSimd.LoadVector128(startingAddress);
containsNonAsciiBytes = AdvSimdHelper.ContainsNonAsciiByte(sourceValue);
}
else
{
throw new PlatformNotSupportedException();
}
if (containsNonAsciiBytes)
{
// At least one of the following 16 bytes is non-ASCII.
int processNextSixteen = idx + 16;
Debug.Assert(processNextSixteen <= utf8Text.Length);
while (idx < processNextSixteen)
{
Debug.Assert((ptr + idx) <= (ptr + utf8Text.Length));
if (UnicodeUtility.IsAsciiCodePoint(ptr[idx]))
{
if (DoesAsciiNeedEncoding(ptr[idx]) == 1)
{
goto Return;
}
idx++;
}
else
{
OperationStatus opStatus = UnicodeHelpers.DecodeScalarValueFromUtf8(utf8Text.Slice(idx), out uint nextScalarValue, out int utf8BytesConsumedForScalar);
Debug.Assert(nextScalarValue <= int.MaxValue);
if (opStatus != OperationStatus.Done || WillEncode((int)nextScalarValue))
{
goto Return;
}
Debug.Assert(opStatus == OperationStatus.Done);
idx += utf8BytesConsumedForScalar;
}
}
}
else
{
if (DoesAsciiNeedEncoding(ptr[idx]) == 1
|| DoesAsciiNeedEncoding(ptr[++idx]) == 1 ||
DoesAsciiNeedEncoding(ptr[++idx]) == 1 ||
DoesAsciiNeedEncoding(ptr[++idx]) == 1 ||
DoesAsciiNeedEncoding(ptr[++idx]) == 1
|| DoesAsciiNeedEncoding(ptr[++idx]) == 1 ||
DoesAsciiNeedEncoding(ptr[++idx]) == 1 ||
DoesAsciiNeedEncoding(ptr[++idx]) == 1 ||
DoesAsciiNeedEncoding(ptr[++idx]) == 1
|| DoesAsciiNeedEncoding(ptr[++idx]) == 1 ||
DoesAsciiNeedEncoding(ptr[++idx]) == 1 ||
DoesAsciiNeedEncoding(ptr[++idx]) == 1 ||
DoesAsciiNeedEncoding(ptr[++idx]) == 1
|| DoesAsciiNeedEncoding(ptr[++idx]) == 1 ||
DoesAsciiNeedEncoding(ptr[++idx]) == 1 ||
DoesAsciiNeedEncoding(ptr[++idx]) == 1)
{
goto Return;
}
idx++;
}
startingAddress = (sbyte *)ptr + idx;
}
// Process the remaining bytes.
Debug.Assert(utf8Text.Length - idx < 16);
}
#endif
while (idx < utf8Text.Length)
{
Debug.Assert((ptr + idx) <= (ptr + utf8Text.Length));
if (UnicodeUtility.IsAsciiCodePoint(ptr[idx]))
{
if (DoesAsciiNeedEncoding(ptr[idx]) == 1)
{
goto Return;
}
idx++;
}
else
{
OperationStatus opStatus = UnicodeHelpers.DecodeScalarValueFromUtf8(utf8Text.Slice(idx), out uint nextScalarValue, out int utf8BytesConsumedForScalar);
Debug.Assert(nextScalarValue <= int.MaxValue);
if (opStatus != OperationStatus.Done || WillEncode((int)nextScalarValue))
{
goto Return;
}
Debug.Assert(opStatus == OperationStatus.Done);
idx += utf8BytesConsumedForScalar;
}
}
Debug.Assert(idx == utf8Text.Length);
idx = -1; // All bytes are allowed.
Return:
return(idx);
}
}
public override unsafe int FindFirstCharacterToEncodeUtf8(ReadOnlySpan <byte> utf8Text)
{
fixed(byte *ptr = utf8Text)
{
int idx = 0;
#if NETCOREAPP
if (Sse2.IsSupported || AdvSimd.Arm64.IsSupported)
{
sbyte *startingAddress = (sbyte *)ptr;
while (utf8Text.Length - 16 >= idx)
{
Debug.Assert(startingAddress >= ptr && startingAddress <= (ptr + utf8Text.Length - 16));
// Load the next 16 bytes.
Vector128 <sbyte> sourceValue;
bool containsNonAsciiBytes;
// Check for ASCII text. Any byte that's not in the ASCII range will already be negative when
// casted to signed byte.
if (Sse2.IsSupported)
{
sourceValue = Sse2.LoadVector128(startingAddress);
containsNonAsciiBytes = Sse2Helper.ContainsNonAsciiByte(sourceValue);
}
else
{
sourceValue = AdvSimd.LoadVector128(startingAddress);
containsNonAsciiBytes = AdvSimdHelper.ContainsNonAsciiByte(sourceValue);
}
if (containsNonAsciiBytes)
{
// At least one of the following 16 bytes is non-ASCII.
int processNextSixteen = idx + 16;
Debug.Assert(processNextSixteen <= utf8Text.Length);
while (idx < processNextSixteen)
{
Debug.Assert((ptr + idx) <= (ptr + utf8Text.Length));
if (UnicodeUtility.IsAsciiCodePoint(ptr[idx]))
{
if (!_allowedCharacters.IsUnicodeScalarAllowed(ptr[idx]))
{
goto Return;
}
idx++;
}
else
{
OperationStatus opStatus = UnicodeHelpers.DecodeScalarValueFromUtf8(utf8Text.Slice(idx), out uint nextScalarValue, out int utf8BytesConsumedForScalar);
Debug.Assert(nextScalarValue <= int.MaxValue);
if (opStatus != OperationStatus.Done || WillEncode((int)nextScalarValue))
{
goto Return;
}
Debug.Assert(opStatus == OperationStatus.Done);
idx += utf8BytesConsumedForScalar;
}
}
startingAddress = (sbyte *)ptr + idx;
}
else
{
// Check if any of the 16 bytes need to be escaped.
int index;
if (Sse2.IsSupported)
{
Vector128 <sbyte> mask = Sse2Helper.CreateEscapingMask_UnsafeRelaxedJavaScriptEncoder(sourceValue);
index = Sse2Helper.GetIndexOfFirstNonAsciiByte(mask.AsByte());
}
else
{
Vector128 <sbyte> mask = AdvSimdHelper.CreateEscapingMask_UnsafeRelaxedJavaScriptEncoder(sourceValue);
index = AdvSimdHelper.GetIndexOfFirstNonAsciiByte(mask.AsByte());
}
// If index >= 16, that means none of the 16 bytes needed to be escaped.
if (index < 16)
{
// Found at least one byte that needs to be escaped, figure out the index of
// the first one found that needed to be escaped within the 16 bytes.
idx += index;
goto Return;
}
idx += 16;
startingAddress += 16;
}
}
// Process the remaining bytes.
Debug.Assert(utf8Text.Length - idx < 16);
}
#endif
while (idx < utf8Text.Length)
{
Debug.Assert((ptr + idx) <= (ptr + utf8Text.Length));
if (UnicodeUtility.IsAsciiCodePoint(ptr[idx]))
{
if (!_allowedCharacters.IsUnicodeScalarAllowed(ptr[idx]))
{
goto Return;
}
idx++;
}
else
{
OperationStatus opStatus = UnicodeHelpers.DecodeScalarValueFromUtf8(utf8Text.Slice(idx), out uint nextScalarValue, out int utf8BytesConsumedForScalar);
Debug.Assert(nextScalarValue <= int.MaxValue);
if (opStatus != OperationStatus.Done || WillEncode((int)nextScalarValue))
{
goto Return;
}
Debug.Assert(opStatus == OperationStatus.Done);
idx += utf8BytesConsumedForScalar;
}
}
Debug.Assert(idx == utf8Text.Length);
idx = -1; // All bytes are allowed.
Return:
return(idx);
}
}