Skip to content

Double Vector128 for SpanHelpers.IndexOf(byte,byte,int) on ARM64 #66993

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

Closed
EgorBo wants to merge 2 commits into from
Closed

Double Vector128 for SpanHelpers.IndexOf(byte,byte,int) on ARM64 #66993

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
de3ba9e
Add double-vector128 path for arm64
EgorBo Mar 22, 2022
4352c1c
Clean up
EgorBo Mar 22, 2022
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
52 changes: 44 additions & 8 deletions src/libraries/System.Private.CoreLib/src/System/SpanHelpers.Byte.cs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -611,20 +611,52 @@ public static unsafe int IndexOf(ref byte searchSpace, byte value, int length)
}
else if (AdvSimd.Arm64.IsSupported)
{
Vector128<byte> mask = Vector128.Create((ushort)0x1001).AsByte();
// Mask to help find the first lane in compareResult that is set.
// MSB 0x10 corresponds to 1st lane, 0x01 corresponds to 0th lane and so forth.
Vector128<byte> values = Vector128.Create(value);
int matchedLane = 0;

if (offset < (nuint)(uint)length)
{
lengthToExamine = GetByteVector128SpanLength(offset, length);
// Try to process data using two Vector128 since we don't have Vector256 on ARM
lengthToExamine = GetByteTwoVector128SpanLength(offset, length);
if (lengthToExamine > offset)
{
while (lengthToExamine > offset)
{
Vector128<byte> search1 = Vector128.LoadUnsafe(ref searchSpace, offset);
Vector128<byte> search2 = Vector128.LoadUnsafe(ref searchSpace, offset + (nuint)Vector128<byte>.Count);

// Mask to help find the first lane in compareResult that is set.
// MSB 0x10 corresponds to 1st lane, 0x01 corresponds to 0th lane and so forth.
Vector128<byte> mask = Vector128.Create((ushort)0x1001).AsByte();
int matchedLane = 0;
Vector128<byte> compareResult1 = Vector128.Equals(values, search1);
Vector128<byte> compareResult2 = Vector128.Equals(values, search2);

Vector128<byte> values = Vector128.Create(value);
// Fast path: no matches in both comparisons
if ((compareResult1 | compareResult2) == Vector128<byte>.Zero)
{
offset += (nuint)Vector128<byte>.Count * 2;
continue;
}

if (!TryFindFirstMatchedLane(mask, compareResult1, ref matchedLane))
{
// The match is in the second comparison
offset += (nuint)Vector128<byte>.Count;
bool found = TryFindFirstMatchedLane(mask, compareResult2, ref matchedLane);
Debug.Assert(found);
}
return (int)(offset + (uint)matchedLane);
}
}
}

if (offset < (nuint)(uint)length)
{
lengthToExamine = GetByteVector128SpanLength(offset, length);
while (lengthToExamine > offset)
{
Vector128<byte> search = LoadVector128(ref searchSpace, offset);
Vector128<byte> compareResult = AdvSimd.CompareEqual(values, search);
Vector128<byte> search = Vector128.LoadUnsafe(ref searchSpace, offset);
Vector128<byte> compareResult = Vector128.Equals(values, search);

if (!TryFindFirstMatchedLane(mask, compareResult, ref matchedLane))
{
Expand Down Expand Up @@ -2213,6 +2245,10 @@ private static nuint GetByteVectorSpanLength(nuint offset, int length)
private static nuint GetByteVector128SpanLength(nuint offset, int length)
=> (nuint)(uint)((length - (int)offset) & ~(Vector128<byte>.Count - 1));

[MethodImpl(MethodImplOptions.AggressiveInlining)]
private static nuint GetByteTwoVector128SpanLength(nuint offset, int length)
=> (nuint)(uint)((length - (int)offset) & ~(Vector128<byte>.Count * 2 - 1));

[MethodImpl(MethodImplOptions.AggressiveInlining)]
private static nuint GetByteVector256SpanLength(nuint offset, int length)
=> (nuint)(uint)((length - (int)offset) & ~(Vector256<byte>.Count - 1));
Expand Down