Skip to content

Add XtsMode128<TDataCipher, TTweakCipher> with zero-abstraction overhead #110

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
HMBSbige merged 3 commits into from
Dec 2, 2025
Merged

Add XtsMode128<TDataCipher, TTweakCipher> with zero-abstraction overhead #110

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
0a9d93b
Initial plan
Copilot Dec 1, 2025
da3c62e
Add new XtsMode128<TDataCipher, TTweakCipher> with zero-abstraction o…
Copilot Dec 1, 2025
02e9b79
refactor: xts mode
HMBSbige Dec 2, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
1 change: 1 addition & 0 deletions .editorconfig
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -27,6 +27,7 @@ resharper_for_simple_types = use_explicit_type
resharper_fsharp_insert_final_newline = false
resharper_html_insert_final_newline = false
resharper_instance_members_qualify_declared_in =
resharper_keep_existing_attribute_arrangement = true
resharper_keep_existing_initializer_arrangement = false
resharper_max_initializer_elements_on_line = 1
resharper_place_accessorholder_attribute_on_same_line = false
Expand Down
12 changes: 0 additions & 12 deletions src/CryptoBase.Abstractions/Vectors/VectorBuffer128.cs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -17,9 +17,6 @@ public ref struct VectorBuffer128
[FieldOffset(2 * 32)] public Vector256<byte> V256_2;
[FieldOffset(3 * 32)] public Vector256<byte> V256_3;

[FieldOffset(0 * 64)] public Vector512<byte> V512_0;
[FieldOffset(1 * 64)] public Vector512<byte> V512_1;

[FieldOffset(0 * 64)] public VectorBuffer64 Lower;
[FieldOffset(1 * 64)] public VectorBuffer64 Upper;

Expand All @@ -32,15 +29,6 @@ public static implicit operator Span<byte>(in VectorBuffer128 value)
[MethodImpl(MethodImplOptions.AggressiveInlining)]
public static VectorBuffer128 operator ^(scoped in VectorBuffer128 left, scoped in VectorBuffer128 right)
{
if (Vector512.IsHardwareAccelerated)
{
return new VectorBuffer128
{
V512_0 = left.V512_0 ^ right.V512_0,
V512_1 = left.V512_1 ^ right.V512_1
};
}

if (Vector256.IsHardwareAccelerated)
{
return new VectorBuffer128
Expand Down
7 changes: 0 additions & 7 deletions src/CryptoBase.Abstractions/Vectors/VectorBuffer64.cs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -11,8 +11,6 @@ public ref struct VectorBuffer64
[FieldOffset(0 * 32)] public Vector256<byte> V256_0;
[FieldOffset(1 * 32)] public Vector256<byte> V256_1;

[FieldOffset(0 * 64)] public Vector512<byte> V512;

[FieldOffset(0 * 32)] public VectorBuffer32 Lower;
[FieldOffset(1 * 32)] public VectorBuffer32 Upper;

Expand All @@ -25,11 +23,6 @@ public static implicit operator Span<byte>(in VectorBuffer64 value)
[MethodImpl(MethodImplOptions.AggressiveInlining)]
public static VectorBuffer64 operator ^(scoped in VectorBuffer64 left, scoped in VectorBuffer64 right)
{
if (Vector512.IsHardwareAccelerated)
{
return new VectorBuffer64 { V512 = left.V512 ^ right.V512 };
}

if (Vector256.IsHardwareAccelerated)
{
return new VectorBuffer64
Expand Down
103 changes: 46 additions & 57 deletions src/CryptoBase/SymmetricCryptos/BlockCryptoModes/XtsMode.Avx2.cs
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,9 +1,9 @@
namespace CryptoBase.SymmetricCryptos.BlockCryptoModes;

public sealed partial class XtsMode
public sealed partial class XtsMode<TBlockCipher>
{
[MethodImpl(MethodImplOptions.AggressiveInlining)]
private static Vector256<byte> Gf128Mul(in Vector256<byte> tweak, [ConstantExpected(Min = 1, Max = 64)] int x)
private static Vector256<byte> Gf128MulAvx2(Vector256<byte> tweak, [ConstantExpected(Min = 1, Max = 64)] int x)
{
Vector256<ulong> tmp1 = tweak.AsUInt64() >>> 64 - x;

Expand All @@ -14,99 +14,88 @@ private static Vector256<byte> Gf128Mul(in Vector256<byte> tweak, [ConstantExpec
return (tweak.AsUInt64() << x ^ tmp1 ^ tmp2).AsByte();
}

[SkipLocalsInit]
[MethodImpl(MethodImplOptions.AggressiveInlining)]
private static void GetInitTweak8Avx2(in ReadOnlySpan<byte> tweakBuffer, in Span<byte> buffer)
private static VectorBuffer128 GetInitTweak8Avx2(Vector128<byte> tweak)
{
ref byte ptr = ref buffer.GetReference();
ref readonly Vector128<byte> x0 = ref Unsafe.As<byte, Vector128<byte>>(ref tweakBuffer.GetReference());
ref Vector256<byte> tweak0 = ref Unsafe.As<byte, Vector256<byte>>(ref Unsafe.Add(ref ptr, 0 * 32));
ref Vector256<byte> tweak1 = ref Unsafe.As<byte, Vector256<byte>>(ref Unsafe.Add(ref ptr, 1 * 32));
ref Vector256<byte> tweak2 = ref Unsafe.As<byte, Vector256<byte>>(ref Unsafe.Add(ref ptr, 2 * 32));
ref Vector256<byte> tweak3 = ref Unsafe.As<byte, Vector256<byte>>(ref Unsafe.Add(ref ptr, 3 * 32));

Vector128<byte> x1 = Gf128Mul(x0);
tweak0 = Vector256.Create(x0, x1);

tweak1 = Gf128Mul(tweak0, 2);
tweak2 = Gf128Mul(tweak0, 4);
tweak3 = Gf128Mul(tweak0, 6);
Unsafe.SkipInit(out VectorBuffer128 r);
r.V128_0 = tweak;
r.V128_1 = Gf128MulSse2(tweak, 1);

r.V256_1 = Gf128MulAvx2(r.V256_0, 2);
r.V256_2 = Gf128MulAvx2(r.V256_0, 4);
r.V256_3 = Gf128MulAvx2(r.V256_0, 6);

return r;
}

[MethodImpl(MethodImplOptions.AggressiveInlining)]
private static void Gf128Mul8Avx2(in Span<byte> buffer)
private static void Gf128Mul8Avx2(ref VectorBuffer128 tweak)
{
ref byte ptr = ref buffer.GetReference();
ref Vector256<byte> tweak0 = ref Unsafe.As<byte, Vector256<byte>>(ref Unsafe.Add(ref ptr, 0 * 32));
ref Vector256<byte> tweak1 = ref Unsafe.As<byte, Vector256<byte>>(ref Unsafe.Add(ref ptr, 1 * 32));
ref Vector256<byte> tweak2 = ref Unsafe.As<byte, Vector256<byte>>(ref Unsafe.Add(ref ptr, 2 * 32));
ref Vector256<byte> tweak3 = ref Unsafe.As<byte, Vector256<byte>>(ref Unsafe.Add(ref ptr, 3 * 32));

tweak0 = Gf128Mul(tweak0, 8);
tweak1 = Gf128Mul(tweak1, 8);
tweak2 = Gf128Mul(tweak2, 8);
tweak3 = Gf128Mul(tweak3, 8);
tweak.V256_0 = Gf128MulAvx2(tweak.V256_0, 8);
tweak.V256_1 = Gf128MulAvx2(tweak.V256_1, 8);
tweak.V256_2 = Gf128MulAvx2(tweak.V256_2, 8);
tweak.V256_3 = Gf128MulAvx2(tweak.V256_3, 8);
}

[MethodImpl(MethodImplOptions.AggressiveInlining)]
private int Encrypt8Avx2(in Span<byte> tweak, in ReadOnlySpan<byte> source, in Span<byte> destination)
private int Encrypt8Avx2(ref Vector128<byte> tweak, ReadOnlySpan<byte> source, Span<byte> destination)
{
int length = source.Length;
int offset = 0;

const int blockSize = 8 * 16;
using CryptoBuffer<byte> buffer = new(blockSize);
Span<byte> tweakBuffer = buffer.Span;
ref readonly byte sourceRef = ref source.GetReference();
ref byte destinationRef = ref destination.GetReference();

GetInitTweak8Avx2(tweak, tweakBuffer);
VectorBuffer128 tweakBuffer = GetInitTweak8Avx2(tweak);

while (length >= 8 * BlockSize)
while (length >= 8 * BlockBytesSize)
{
ReadOnlySpan<byte> src = source.Slice(offset, blockSize);
Span<byte> dst = destination.Slice(offset, blockSize);
VectorBuffer128 src = Unsafe.Add(ref Unsafe.AsRef(in sourceRef), offset).AsVectorBuffer128();
ref VectorBuffer128 dst = ref Unsafe.Add(ref destinationRef, offset).AsVectorBuffer128();

FastUtils.Xor(src, tweakBuffer, dst, blockSize);
_dataCrypto.Encrypt8(dst, dst);
FastUtils.Xor(dst, tweakBuffer, blockSize);
VectorBuffer128 tmp = src ^ tweakBuffer;
tmp = _dataCipher.Encrypt(tmp);
dst = tmp ^ tweakBuffer;

Gf128Mul8Avx2(tweakBuffer);
Gf128Mul8Avx2(ref tweakBuffer);

offset += blockSize;
length -= blockSize;
offset += 8 * BlockBytesSize;
length -= 8 * BlockBytesSize;
}

tweakBuffer.Slice(0, BlockSize).CopyTo(tweak);
tweak = tweakBuffer.V128_0;

return offset;
}

[MethodImpl(MethodImplOptions.AggressiveInlining)]
private int Decrypt8Avx2(in Span<byte> tweak, in ReadOnlySpan<byte> source, in Span<byte> destination)
private int Decrypt8Avx2(ref Vector128<byte> tweak, ReadOnlySpan<byte> source, Span<byte> destination)
{
int length = source.Length;
int offset = 0;

const int blockSize = 8 * 16;
using CryptoBuffer<byte> buffer = new(blockSize);
Span<byte> tweakBuffer = buffer.Span;
ref readonly byte sourceRef = ref source.GetReference();
ref byte destinationRef = ref destination.GetReference();

GetInitTweak8Avx2(tweak, tweakBuffer);
VectorBuffer128 tweakBuffer = GetInitTweak8Avx2(tweak);

while (length >= 8 * BlockSize)
while (length >= 8 * BlockBytesSize)
{
ReadOnlySpan<byte> src = source.Slice(offset, blockSize);
Span<byte> dst = destination.Slice(offset, blockSize);
VectorBuffer128 src = Unsafe.Add(ref Unsafe.AsRef(in sourceRef), offset).AsVectorBuffer128();
ref VectorBuffer128 dst = ref Unsafe.Add(ref destinationRef, offset).AsVectorBuffer128();

FastUtils.Xor(src, tweakBuffer, dst, blockSize);
_dataCrypto.Decrypt8(dst, dst);
FastUtils.Xor(dst, tweakBuffer, blockSize);
VectorBuffer128 tmp = src ^ tweakBuffer;
tmp = _dataCipher.Decrypt(tmp);
dst = tmp ^ tweakBuffer;

Gf128Mul8Avx2(tweakBuffer);
Gf128Mul8Avx2(ref tweakBuffer);

offset += blockSize;
length -= blockSize;
offset += 8 * BlockBytesSize;
length -= 8 * BlockBytesSize;
}

tweakBuffer.Slice(0, BlockSize).CopyTo(tweak);
tweak = tweakBuffer.V128_0;

return offset;
}
Expand Down
Loading