Further improve ProbabilisticMap on Avx512

[MihaZupan]

This PR contains 3 separate changes that take advantage of Avx512's permute behavior.

The first change is skipping an AND instruction by taking advantage of the fact that PermuteVar64x8 will only look at the bottom 6 bits (64 values) of the control.

-Vector512<byte> index = values & Vector512.Create((byte)31);
-Vector512<byte> bitMask = Avx512Vbmi.PermuteVar64x8(charMap, index);
+Vector512<byte> bitMask = Avx512Vbmi.PermuteVar64x8(charMap, values);

Note that we were doing AND 31 instead of 63 though.
In this case this is okay because we know that the charMap is just a duplicated 256-bit lookup.
The 6th bit that we aren't masking off anymore will impact whether we pick from the first 0-31, or 32-63 values, but that doesn't matter since the two are the same.

---

The second change is also taking advantage of the above observation.
It recognizes that the values >>> 5 operation is emulated as values.AsInt32() >>> 5).AsByte() & Vector128.Create((byte)7) since there's no instruction for >>> on bytes on X86.
We can skip that & 7 operation if we swap out the shuffle for a permute. This way we're again only looking at the lower 6 bits. As before, we now have bits 4/5/6 that aren't getting masked off, but that's okay since the values are duplicated 8 times.

-Vector512<byte> shifted = values >>> 5;
-Vector512<byte> bitPositions = Avx512BW.Shuffle(Vector512.Create(0x8040201008040201).AsByte(), shifted);
+Vector512<byte> shifted = (values.AsInt32() >>> 5).AsByte();
+Vector512<byte> bitPositions = Avx512Vbmi.PermuteVar64x8(Vector512.Create(0x8040201008040201).AsByte(), shifted);

---

The third change is taking advantage of the PermuteVar32x8x2 instruction to pick alternating bytes from the two source vectors, instead of shifting the bytes around and doing a saturating pack.

-Vector512<byte> sourceLower = Avx512BW.PackUnsignedSaturate(
-    (source0 & Vector256.Create((ushort)255)).AsInt16(),
-    (source1 & Vector256.Create((ushort)255)).AsInt16());
+Vector512<byte> sourceLower = Avx512Vbmi.PermuteVar64x8x2(source0.AsByte(), Vector512.CreateSequence<byte>(0, 2), source1.AsByte());

-Vector512<byte> sourceUpper = Avx512BW.PackUnsignedSaturate(
-    (source0 >>> 8).AsInt16(),
-    (source1 >>> 8).AsInt16());
+Vector512<byte> sourceUpper = Avx512Vbmi.PermuteVar64x8x2(source0.AsByte(), Vector512.CreateSequence<byte>(1, 2), source1.AsByte());

Since PackUnsignedSaturate also shuffles inputs around a bit, we needed to reverse that by calling FixUpPackedVector512Result (another permute) if we did find any potential matches.
PermuteVar64x8x2 keeps the input order as-is, meaning we can skip that "fix up".

-if (TryFindMatchAvx512<TUseFastContains>(ref cur, PackedSpanHelpers.FixUpPackedVector512Result(result).ExtractMostSignificantBits(), ref state, out int index))
+if (TryFindMatchAvx512<TUseFastContains>(ref cur, result.ExtractMostSignificantBits(), ref state, out int index))

---

Overall, it adds up to a ~20% improvement.

public class ProbMap
{
    [Params(32, 64, 128, 256, 512, 1024, 10_000)]
    public int Length;

    private static readonly SearchValues<char> s_searchValues = SearchValues.Create("ßäöüÄÖÜ");
    private char[] _text;

    [GlobalSetup]
    public void Setup() => _text = new string('\n', Length).ToCharArray();

    [Benchmark]
    public int IndexOfAny() => _text.AsSpan().IndexOfAny(s_searchValues);
}

Method	Toolchain	Length	Mean	Error	Ratio
IndexOfAny	main	32	2.606 ns	0.0793 ns	1.00
IndexOfAny	pr	32	2.414 ns	0.0560 ns	0.93
IndexOfAny	main	64	2.750 ns	0.0505 ns	1.00
IndexOfAny	pr	64	2.540 ns	0.0271 ns	0.92
IndexOfAny	main	128	4.307 ns	0.0289 ns	1.00
IndexOfAny	pr	128	3.175 ns	0.0483 ns	0.74
IndexOfAny	main	256	5.637 ns	0.0459 ns	1.00
IndexOfAny	pr	256	4.932 ns	0.0445 ns	0.87
IndexOfAny	main	512	10.615 ns	0.0535 ns	1.00
IndexOfAny	pr	512	8.577 ns	0.1427 ns	0.81
IndexOfAny	main	1024	22.187 ns	0.1047 ns	1.00
IndexOfAny	pr	1024	18.331 ns	0.3958 ns	0.83
IndexOfAny	main	10000	231.458 ns	1.1701 ns	1.00
IndexOfAny	pr	10000	185.283 ns	0.4801 ns	0.80

[dotnet-policy-service]

Tagging subscribers to this area: @dotnet/area-system-memory
See info in area-owners.md if you want to be subscribed.

Copilot reviewed 1 out of 1 changed files in this pull request and generated no comments.