Skip to content
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

Jump to bottom

[API Proposal]: Arm64: FEAT_BF16 #94028

Open
a74nh opened this issue Oct 26, 2023 · 8 comments
Open

[API Proposal]: Arm64: FEAT_BF16 #94028

a74nh opened this issue Oct 26, 2023 · 8 comments
Labels
api-suggestion Early API idea and discussion, it is NOT ready for implementation arch-arm64 area-System.Runtime.Intrinsics
Milestone
Future

Comments

@a74nh
Copy link
Contributor

a74nh commented Oct 26, 2023 •
edited
Loading 

namespace System.Runtime.Intrinsics.Arm;

/// VectorT Summary
public abstract partial class SveBf16 : AdvSimd /// Feature: FEAT_BF16
{

  public static unsafe Vector<float> Bfloat16DotProduct(Vector<float> addend, Vector<bfloat16> left, Vector<bfloat16> right); // BFDOT // MOVPRFX

  public static unsafe Vector<float> Bfloat16MatrixMultiplyAccumulate(Vector<float> op1, Vector<bfloat16> op2, Vector<bfloat16> op3); // BFMMLA // MOVPRFX

  public static unsafe Vector<float> Bfloat16MultiplyAddWideningToSinglePrecisionLower(Vector<float> op1, Vector<bfloat16> op2, Vector<bfloat16> op3); // BFMLALB // MOVPRFX

  public static unsafe Vector<float> Bfloat16MultiplyAddWideningToSinglePrecisionLower(Vector<float> op1, Vector<bfloat16> op2, Vector<bfloat16> op3, ulong imm_index); // BFMLALB // MOVPRFX

  public static unsafe Vector<float> Bfloat16MultiplyAddWideningToSinglePrecisionUpper(Vector<float> op1, Vector<bfloat16> op2, Vector<bfloat16> op3); // BFMLALT // MOVPRFX

  public static unsafe Vector<float> Bfloat16MultiplyAddWideningToSinglePrecisionUpper(Vector<float> op1, Vector<bfloat16> op2, Vector<bfloat16> op3, ulong imm_index); // BFMLALT // MOVPRFX

  public static unsafe Vector<bfloat16> ConcatenateEvenInt128FromTwoInputs(Vector<bfloat16> left, Vector<bfloat16> right); // UZP1

  public static unsafe Vector<bfloat16> ConcatenateOddInt128FromTwoInputs(Vector<bfloat16> left, Vector<bfloat16> right); // UZP2

  public static unsafe Vector<bfloat16> ConditionalExtractAfterLastActiveElement(Vector<bfloat16> mask, Vector<bfloat16> defaultValue, Vector<bfloat16> data); // CLASTA // MOVPRFX

  public static unsafe bfloat16 ConditionalExtractAfterLastActiveElement(Vector<bfloat16> mask, bfloat16 defaultValues, Vector<bfloat16> data); // CLASTA // MOVPRFX

  public static unsafe Vector<bfloat16> ConditionalExtractAfterLastActiveElementAndReplicate(Vector<bfloat16> mask, Vector<bfloat16> defaultScalar, Vector<bfloat16> data); // CLASTA // MOVPRFX

  public static unsafe Vector<bfloat16> ConditionalExtractLastActiveElement(Vector<bfloat16> mask, Vector<bfloat16> defaultValue, Vector<bfloat16> data); // CLASTB // MOVPRFX

  public static unsafe bfloat16 ConditionalExtractLastActiveElement(Vector<bfloat16> mask, bfloat16 defaultValues, Vector<bfloat16> data); // CLASTB // MOVPRFX

  public static unsafe Vector<bfloat16> ConditionalExtractLastActiveElementAndReplicate(Vector<bfloat16> mask, Vector<bfloat16> fallback, Vector<bfloat16> data); // CLASTB // MOVPRFX

  public static unsafe Vector<bfloat16> ConditionalSelect(Vector<bfloat16> mask, Vector<bfloat16> left, Vector<bfloat16> right); // SEL

  public static unsafe Vector<bfloat16> ConvertToBFloat16(Vector<float> value); // BFCVT // predicated, MOVPRFX

  public static unsafe Vector<bfloat16> CreateFalseMaskBFloat16(); // PFALSE

  public static unsafe Vector<bfloat16> CreateTrueMaskBFloat16([ConstantExpected] SveMaskPattern pattern = SveMaskPattern.All); // PTRUE

  public static unsafe Vector<bfloat16> CreateWhileReadAfterWriteMask(bfloat16* left, bfloat16* right); // WHILERW

  public static unsafe Vector<bfloat16> CreateWhileWriteAfterReadMask(bfloat16* left, bfloat16* right); // WHILEWR

  public static unsafe Vector<float> DotProductBySelectedScalar(Vector<float> addend, Vector<bfloat16> left, Vector<bfloat16> right, [ConstantExpected] byte rightIndex); // BFDOT // MOVPRFX

  public static unsafe Vector<bfloat16> DownConvertNarrowingUpper(Vector<float> value); // BFCVTNT // predicated

  public static unsafe Vector<bfloat16> DuplicateSelectedScalarToVector(Vector<bfloat16> data, [ConstantExpected] byte index); // DUP or TBL

  public static unsafe bfloat16 ExtractAfterLastScalar(Vector<bfloat16> value); // LASTA // predicated

  public static unsafe Vector<bfloat16> ExtractAfterLastVector(Vector<bfloat16> value); // LASTA // predicated

  public static unsafe bfloat16 ExtractLastScalar(Vector<bfloat16> value); // LASTB // predicated

  public static unsafe Vector<bfloat16> ExtractLastVector(Vector<bfloat16> value); // LASTB // predicated

  public static unsafe Vector<bfloat16> ExtractVector(Vector<bfloat16> upper, Vector<bfloat16> lower, [ConstantExpected] byte index); // EXT // MOVPRFX

  public static unsafe ulong GetActiveElementCount(Vector<bfloat16> mask, Vector<bfloat16> from); // CNTP

  public static unsafe Vector<bfloat16> InsertIntoShiftedVector(Vector<bfloat16> left, bfloat16 right); // INSR

  public static unsafe Vector<bfloat16> InterleaveEvenInt128FromTwoInputs(Vector<bfloat16> left, Vector<bfloat16> right); // TRN1

  public static unsafe Vector<bfloat16> InterleaveInt128FromHighHalvesOfTwoInputs(Vector<bfloat16> left, Vector<bfloat16> right); // ZIP2

  public static unsafe Vector<bfloat16> InterleaveInt128FromLowHalvesOfTwoInputs(Vector<bfloat16> left, Vector<bfloat16> right); // ZIP1

  public static unsafe Vector<bfloat16> InterleaveOddInt128FromTwoInputs(Vector<bfloat16> left, Vector<bfloat16> right); // TRN2

  public static unsafe Vector<bfloat16> LoadVector(Vector<bfloat16> mask, bfloat16* address); // LD1H

  public static unsafe Vector<bfloat16> LoadVector128AndReplicateToVector(Vector<bfloat16> mask, bfloat16* address); // LD1RQH

  public static unsafe Vector<bfloat16> LoadVector256AndReplicateToVector(Vector<bfloat16> mask, bfloat16* address); // LD1ROH

  public static unsafe Vector<bfloat16> LoadVectorFirstFaulting(Vector<bfloat16> mask, bfloat16* address); // LDFF1H

  public static unsafe Vector<bfloat16> LoadVectorNonFaulting(bfloat16* address); // LDNF1H // predicated

  public static unsafe Vector<bfloat16> LoadVectorNonTemporal(Vector<bfloat16> mask, bfloat16* address); // LDNT1H

  public static unsafe (Vector<bfloat16>, Vector<bfloat16>) LoadVectorx2(Vector<bfloat16> mask, bfloat16* address); // LD2H

  public static unsafe (Vector<bfloat16>, Vector<bfloat16>, Vector<bfloat16>) LoadVectorx3(Vector<bfloat16> mask, bfloat16* address); // LD3H

  public static unsafe (Vector<bfloat16>, Vector<bfloat16>, Vector<bfloat16>, Vector<bfloat16>) LoadVectorx4(Vector<bfloat16> mask, bfloat16* address); // LD4H

  public static unsafe Vector<ushort> PopCount(Vector<bfloat16> value); // CNT // predicated, MOVPRFX

  public static unsafe Vector<bfloat16> ReverseElement(Vector<bfloat16> value); // REV

  public static unsafe Vector<bfloat16> Splice(Vector<bfloat16> mask, Vector<bfloat16> left, Vector<bfloat16> right); // SPLICE // MOVPRFX

  public static unsafe void Store(Vector<bfloat16> mask, bfloat16* address, Vector<bfloat16> data); // ST1H

  public static unsafe void Store(Vector<bfloat16> mask, bfloat16* address, (Vector<bfloat16> Value1, Vector<bfloat16> Value2) data); // ST2H

  public static unsafe void Store(Vector<bfloat16> mask, bfloat16* address, (Vector<bfloat16> Value1, Vector<bfloat16> Value2, Vector<bfloat16> Value3) data); // ST3H

  public static unsafe void Store(Vector<bfloat16> mask, bfloat16* address, (Vector<bfloat16> Value1, Vector<bfloat16> Value2, Vector<bfloat16> Value3, Vector<bfloat16> Value4) data); // ST4H

  public static unsafe void StoreNonTemporal(Vector<bfloat16> mask, bfloat16* address, Vector<bfloat16> data); // STNT1H

  public static unsafe Vector<bfloat16> TransposeEven(Vector<bfloat16> left, Vector<bfloat16> right); // TRN1

  public static unsafe Vector<bfloat16> TransposeOdd(Vector<bfloat16> left, Vector<bfloat16> right); // TRN2

  public static unsafe Vector<bfloat16> UnzipEven(Vector<bfloat16> left, Vector<bfloat16> right); // UZP1

  public static unsafe Vector<bfloat16> UnzipOdd(Vector<bfloat16> left, Vector<bfloat16> right); // UZP2

  public static unsafe Vector<bfloat16> VectorTableLookup(Vector<bfloat16> data, Vector<ushort> indices); // TBL

  public static unsafe Vector<bfloat16> VectorTableLookup((Vector<bfloat16> data1, Vector<bfloat16> data2), Vector<ushort> indices); // TBL

  public static unsafe Vector<bfloat16> VectorTableLookupExtension(Vector<bfloat16> fallback, Vector<bfloat16> data, Vector<ushort> indices); // TBX

  public static unsafe Vector<bfloat16> ZipHigh(Vector<bfloat16> left, Vector<bfloat16> right); // ZIP2

  public static unsafe Vector<bfloat16> ZipLow(Vector<bfloat16> left, Vector<bfloat16> right); // ZIP1

  /// total method signatures: 60


  /// Optional Entries:

  public static unsafe Vector<float> Bfloat16DotProduct(Vector<float> addend, Vector<bfloat16> left, bfloat16 right); // BFDOT // MOVPRFX

  public static unsafe Vector<float> Bfloat16MultiplyAddWideningToSinglePrecisionLower(Vector<float> op1, Vector<bfloat16> op2, bfloat16 op3); // BFMLALB // MOVPRFX

  public static unsafe Vector<float> Bfloat16MultiplyAddWideningToSinglePrecisionUpper(Vector<float> op1, Vector<bfloat16> op2, bfloat16 op3); // BFMLALT // MOVPRFX

  public static unsafe bfloat16 ConditionalExtractAfterLastActiveElement(Vector<bfloat16> mask, bfloat16 defaultValue, Vector<bfloat16> data); // CLASTA

  public static unsafe bfloat16 ConditionalExtractAfterLastActiveElementAndReplicate(Vector<bfloat16> mask, bfloat16 defaultScalar, Vector<bfloat16> data); // CLASTA

  public static unsafe bfloat16 ConditionalExtractLastActiveElement(Vector<bfloat16> mask, bfloat16 defaultValue, Vector<bfloat16> data); // CLASTB

  public static unsafe bfloat16 ConditionalExtractLastActiveElementAndReplicate(Vector<bfloat16> mask, bfloat16 fallback, Vector<bfloat16> data); // CLASTB

  /// total optional method signatures: 7

}
@ghost ghost added the untriaged New issue has not been triaged by the area owner label Oct 26, 2023
@ghost
Copy link

ghost commented Oct 26, 2023

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

Issue Details 
namespace System.Runtime.Intrinsics.Arm

/// VectorT Summary
public abstract class Sve : AdvSimd /// Feature: FEAT_BF16
{
  public static unsafe Vector<float> Bfloat16DotProduct(Vector<float> op1, Vector<bfloat16> op2, Vector<bfloat16> op3);

  public static unsafe Vector<float> Bfloat16DotProduct(Vector<float> op1, Vector<bfloat16> op2, Vector<bfloat16> op3, ulong imm_index);

  public static unsafe Vector<float> Bfloat16MatrixMultiplyAccumulate(Vector<float> op1, Vector<bfloat16> op2, Vector<bfloat16> op3);

  public static unsafe Vector<float> Bfloat16MultiplyAddWideningToSinglePrecisionLower(Vector<float> op1, Vector<bfloat16> op2, Vector<bfloat16> op3);

  public static unsafe Vector<float> Bfloat16MultiplyAddWideningToSinglePrecisionLower(Vector<float> op1, Vector<bfloat16> op2, Vector<bfloat16> op3, ulong imm_index);

  public static unsafe Vector<float> Bfloat16MultiplyAddWideningToSinglePrecisionUpper(Vector<float> op1, Vector<bfloat16> op2, Vector<bfloat16> op3);

  public static unsafe Vector<float> Bfloat16MultiplyAddWideningToSinglePrecisionUpper(Vector<float> op1, Vector<bfloat16> op2, Vector<bfloat16> op3, ulong imm_index);

  public static unsafe Vector<bfloat16> ConditionalExtractAfterLast(Vector<bfloat16> mask, Vector<bfloat16> fallback, Vector<bfloat16> data);

  public static unsafe Vector<bfloat16> ConditionalExtractLast(Vector<bfloat16> mask, Vector<bfloat16> fallback, Vector<bfloat16> data);

  public static unsafe Vector<bfloat16> ConditionalSelect(Vector<bfloat16> mask, Vector<bfloat16> op1, Vector<bfloat16> op2);

  public static unsafe Vector<ushort> CountNonzeroBits(Vector<bfloat16> value);

  public static unsafe ulong CountTheNumberOfElementsInAFullVector(Vector<bfloat16> value);

  public static unsafe Vector<bfloat16> DownConvertNarrowingUpper(Vector<bfloat16> even, Vector<float> op);

  public static unsafe Vector<bfloat16> DuplicateToVector(bfloat16 value);

  public static unsafe Vector<bfloat16> DuplicateToVector(Vector<bfloat16> data, ushort index);

  public static unsafe Vector<bfloat16> DuplicateToVector(Vector<bfloat16> data, ulong index);

  public static unsafe bfloat16 ExtractAfterLast(Vector<bfloat16> value);

  public static unsafe bfloat16 ExtractLast(Vector<bfloat16> value);

  public static unsafe Vector<bfloat16> ExtractVector(Vector<bfloat16> upper, Vector<bfloat16> lower, ulong index);

  public static unsafe Vector<bfloat16> FloatingPointConvert(Vector<float> value);

  public static unsafe Vector<bfloat16> InsertIntoShiftedVector(Vector<bfloat16> left, bfloat16 right);

  public static unsafe Vector<bfloat16> Load128ReplicateToVector(const bfloat16 *base);

  public static unsafe Vector<bfloat16> LoadVector(const bfloat16 *base);

  public static unsafe Vector<bfloat16> LoadVector(const bfloat16 *base, long vnum);

  public static unsafe Vector<bfloat16> LoadVectorFirstFaulting(const bfloat16 *base);

  public static unsafe Vector<bfloat16> LoadVectorFirstFaulting(const bfloat16 *base, long vnum);

  public static unsafe Vector<bfloat16> LoadVectorNonFaulting(const bfloat16 *base);

  public static unsafe Vector<bfloat16> LoadVectorNonFaulting(const bfloat16 *base, long vnum);

  public static unsafe Vector<bfloat16> LoadVectorNonTemporal(const bfloat16 *base);

  public static unsafe Vector<bfloat16> LoadVectorNonTemporal(const bfloat16 *base, long vnum);

  public static unsafe (Vector<bfloat16>, Vector<bfloat16>) LoadVectorx2(const bfloat16 *base);

  public static unsafe (Vector<bfloat16>, Vector<bfloat16>) LoadVectorx2(const bfloat16 *base, long vnum);

  public static unsafe (Vector<bfloat16>, Vector<bfloat16>, Vector<bfloat16>) LoadVectorx3(const bfloat16 *base);

  public static unsafe (Vector<bfloat16>, Vector<bfloat16>, Vector<bfloat16>) LoadVectorx3(const bfloat16 *base, long vnum);

  public static unsafe (Vector<bfloat16>, Vector<bfloat16>, Vector<bfloat16>, Vector<bfloat16>) LoadVectorx4(const bfloat16 *base);

  public static unsafe (Vector<bfloat16>, Vector<bfloat16>, Vector<bfloat16>, Vector<bfloat16>) LoadVectorx4(const bfloat16 *base, long vnum);

  public static unsafe Vector<bfloat16> ReverseElement(Vector<bfloat16> value);

  public static unsafe Vector<bfloat16> Splice(Vector<bfloat16> left, Vector<bfloat16> right);

  public static unsafe void StoreVector(bfloat16 *base, Vector<bfloat16> data);

  public static unsafe void StoreVector(bfloat16 *base, long vnum, Vector<bfloat16> data);

  public static unsafe void StoreVectorNonTemporal(bfloat16 *base, Vector<bfloat16> data);

  public static unsafe void StoreVectorNonTemporal(bfloat16 *base, long vnum, Vector<bfloat16> data);

  public static unsafe void StoreVectorx2(bfloat16 *base, (Vector<bfloat16> data1, Vector<bfloat16> data2));

  public static unsafe void StoreVectorx2(bfloat16 *base, long vnum, (Vector<bfloat16> data1, Vector<bfloat16> data2));

  public static unsafe void StoreVectorx3(bfloat16 *base, (Vector<bfloat16> data1, Vector<bfloat16> data2, Vector<bfloat16> data3));

  public static unsafe void StoreVectorx3(bfloat16 *base, long vnum, (Vector<bfloat16> data1, Vector<bfloat16> data2, Vector<bfloat16> data3));

  public static unsafe void StoreVectorx4(bfloat16 *base, (Vector<bfloat16> data1, Vector<bfloat16> data2, Vector<bfloat16> data3, Vector<bfloat16> data4));

  public static unsafe void StoreVectorx4(bfloat16 *base, long vnum, (Vector<bfloat16> data1, Vector<bfloat16> data2, Vector<bfloat16> data3, Vector<bfloat16> data4));

  public static unsafe Vector<bfloat16> TransposeEven(Vector<bfloat16> left, Vector<bfloat16> right);

  public static unsafe Vector<bfloat16> TransposeOdd(Vector<bfloat16> left, Vector<bfloat16> right);

  public static unsafe Vector<bfloat16> UnzipEven(Vector<bfloat16> left, Vector<bfloat16> right);

  public static unsafe Vector<bfloat16> UnzipOdd(Vector<bfloat16> left, Vector<bfloat16> right);

  public static unsafe Vector<bfloat16> VectorTableLookup(Vector<bfloat16> data, Vector<ushort> indices);

  public static unsafe Vector<bfloat16> VectorTableLookup((Vector<bfloat16> data1, Vector<bfloat16> data2), Vector<ushort> indices);

  public static unsafe Vector<bfloat16> VectorTableLookupExtension(Vector<bfloat16> fallback, Vector<bfloat16> data, Vector<ushort> indices);

  public static unsafe Vector<bfloat16> WhileReadAfterWrite(const bfloat16 *op1, const bfloat16 *op2);

  public static unsafe Vector<bfloat16> WhileWriteAfterRead(const bfloat16 *op1, const bfloat16 *op2);

  public static unsafe Vector<bfloat16> ZipHigh(Vector<bfloat16> left, Vector<bfloat16> right);

  public static unsafe Vector<bfloat16> ZipLow(Vector<bfloat16> left, Vector<bfloat16> right);

  /// total method signatures: 59
}
Author: a74nh
Assignees: -
Labels:

area-System.Numerics
Milestone: -

@a74nh
Copy link
Contributor Author

a74nh commented Oct 26, 2023 •
edited
Loading 

/// Full API
public abstract partial class SveBf16 : AdvSimd /// Feature: FEAT_BF16
{
    /// Bfloat16DotProduct : BFloat16 dot product

    /// svfloat32_t svbfdot[_f32](svfloat32_t op1, svbfloat16_t op2, svbfloat16_t op3) : "BFDOT Ztied1.S, Zop2.H, Zop3.H" or "MOVPRFX Zresult, Zop1; BFDOT Zresult.S, Zop2.H, Zop3.H"
  public static unsafe Vector<float> Bfloat16DotProduct(Vector<float> addend, Vector<bfloat16> left, Vector<bfloat16> right);


    /// Bfloat16MatrixMultiplyAccumulate : BFloat16 matrix multiply-accumulate

    /// svfloat32_t svbfmmla[_f32](svfloat32_t op1, svbfloat16_t op2, svbfloat16_t op3) : "BFMMLA Ztied1.S, Zop2.H, Zop3.H" or "MOVPRFX Zresult, Zop1; BFMMLA Zresult.S, Zop2.H, Zop3.H"
  public static unsafe Vector<float> Bfloat16MatrixMultiplyAccumulate(Vector<float> op1, Vector<bfloat16> op2, Vector<bfloat16> op3);


    /// Bfloat16MultiplyAddWideningToSinglePrecisionLower : BFloat16 multiply-add long to single-precision (bottom)

    /// svfloat32_t svbfmlalb[_f32](svfloat32_t op1, svbfloat16_t op2, svbfloat16_t op3) : "BFMLALB Ztied1.S, Zop2.H, Zop3.H" or "MOVPRFX Zresult, Zop1; BFMLALB Zresult.S, Zop2.H, Zop3.H"
  public static unsafe Vector<float> Bfloat16MultiplyAddWideningToSinglePrecisionLower(Vector<float> op1, Vector<bfloat16> op2, Vector<bfloat16> op3);

    /// svfloat32_t svbfmlalb_lane[_f32](svfloat32_t op1, svbfloat16_t op2, svbfloat16_t op3, uint64_t imm_index) : "BFMLALB Ztied1.S, Zop2.H, Zop3.H[imm_index]" or "MOVPRFX Zresult, Zop1; BFMLALB Zresult.S, Zop2.H, Zop3.H[imm_index]"
  public static unsafe Vector<float> Bfloat16MultiplyAddWideningToSinglePrecisionLower(Vector<float> op1, Vector<bfloat16> op2, Vector<bfloat16> op3, ulong imm_index);


    /// Bfloat16MultiplyAddWideningToSinglePrecisionUpper : BFloat16 multiply-add long to single-precision (top)

    /// svfloat32_t svbfmlalt[_f32](svfloat32_t op1, svbfloat16_t op2, svbfloat16_t op3) : "BFMLALT Ztied1.S, Zop2.H, Zop3.H" or "MOVPRFX Zresult, Zop1; BFMLALT Zresult.S, Zop2.H, Zop3.H"
  public static unsafe Vector<float> Bfloat16MultiplyAddWideningToSinglePrecisionUpper(Vector<float> op1, Vector<bfloat16> op2, Vector<bfloat16> op3);

    /// svfloat32_t svbfmlalt_lane[_f32](svfloat32_t op1, svbfloat16_t op2, svbfloat16_t op3, uint64_t imm_index) : "BFMLALT Ztied1.S, Zop2.H, Zop3.H[imm_index]" or "MOVPRFX Zresult, Zop1; BFMLALT Zresult.S, Zop2.H, Zop3.H[imm_index]"
  public static unsafe Vector<float> Bfloat16MultiplyAddWideningToSinglePrecisionUpper(Vector<float> op1, Vector<bfloat16> op2, Vector<bfloat16> op3, ulong imm_index);


    /// ConcatenateEvenInt128FromTwoInputs : Concatenate even quadwords from two inputs

    /// svbfloat16_t svuzp1q[_bf16](svbfloat16_t op1, svbfloat16_t op2) : "UZP1 Zresult.Q, Zop1.Q, Zop2.Q"
  public static unsafe Vector<bfloat16> ConcatenateEvenInt128FromTwoInputs(Vector<bfloat16> left, Vector<bfloat16> right);


    /// ConcatenateOddInt128FromTwoInputs : Concatenate odd quadwords from two inputs

    /// svbfloat16_t svuzp2q[_bf16](svbfloat16_t op1, svbfloat16_t op2) : "UZP2 Zresult.Q, Zop1.Q, Zop2.Q"
  public static unsafe Vector<bfloat16> ConcatenateOddInt128FromTwoInputs(Vector<bfloat16> left, Vector<bfloat16> right);


    /// ConditionalExtractAfterLastActiveElement : Conditionally extract element after last

    /// svbfloat16_t svclasta[_bf16](svbool_t pg, svbfloat16_t fallback, svbfloat16_t data) : "CLASTA Ztied.H, Pg, Ztied.H, Zdata.H" or "MOVPRFX Zresult, Zfallback; CLASTA Zresult.H, Pg, Zresult.H, Zdata.H"
  public static unsafe Vector<bfloat16> ConditionalExtractAfterLastActiveElement(Vector<bfloat16> mask, Vector<bfloat16> defaultValue, Vector<bfloat16> data);

    /// svbfloat16_t svclasta[_bf16](svbool_t pg, svbfloat16_t fallback, svbfloat16_t data) : "CLASTA Ztied.H, Pg, Ztied.H, Zdata.H" or "MOVPRFX Zresult, Zfallback; CLASTA Zresult.H, Pg, Zresult.H, Zdata.H"
    /// bfloat16_t svclasta[_n_bf16](svbool_t pg, bfloat16_t fallback, svbfloat16_t data) : "CLASTA Wtied, Pg, Wtied, Zdata.H" or "CLASTA Htied, Pg, Htied, Zdata.H"
  public static unsafe bfloat16 ConditionalExtractAfterLastActiveElement(Vector<bfloat16> mask, bfloat16 defaultValues, Vector<bfloat16> data);


    /// ConditionalExtractAfterLastActiveElementAndReplicate : Conditionally extract element after last

    /// svbfloat16_t svclasta[_bf16](svbool_t pg, svbfloat16_t fallback, svbfloat16_t data) : "CLASTA Ztied.H, Pg, Ztied.H, Zdata.H" or "MOVPRFX Zresult, Zfallback; CLASTA Zresult.H, Pg, Zresult.H, Zdata.H"
  public static unsafe Vector<bfloat16> ConditionalExtractAfterLastActiveElementAndReplicate(Vector<bfloat16> mask, Vector<bfloat16> defaultScalar, Vector<bfloat16> data);


    /// ConditionalExtractLastActiveElement : Conditionally extract last element

    /// svbfloat16_t svclastb[_bf16](svbool_t pg, svbfloat16_t fallback, svbfloat16_t data) : "CLASTB Ztied.H, Pg, Ztied.H, Zdata.H" or "MOVPRFX Zresult, Zfallback; CLASTB Zresult.H, Pg, Zresult.H, Zdata.H"
  public static unsafe Vector<bfloat16> ConditionalExtractLastActiveElement(Vector<bfloat16> mask, Vector<bfloat16> defaultValue, Vector<bfloat16> data);

    /// svbfloat16_t svclastb[_bf16](svbool_t pg, svbfloat16_t fallback, svbfloat16_t data) : "CLASTB Ztied.H, Pg, Ztied.H, Zdata.H" or "MOVPRFX Zresult, Zfallback; CLASTB Zresult.H, Pg, Zresult.H, Zdata.H"
    /// bfloat16_t svclastb[_n_bf16](svbool_t pg, bfloat16_t fallback, svbfloat16_t data) : "CLASTB Wtied, Pg, Wtied, Zdata.H" or "CLASTB Htied, Pg, Htied, Zdata.H"
  public static unsafe bfloat16 ConditionalExtractLastActiveElement(Vector<bfloat16> mask, bfloat16 defaultValues, Vector<bfloat16> data);


    /// ConditionalExtractLastActiveElementAndReplicate : Conditionally extract last element

    /// svbfloat16_t svclastb[_bf16](svbool_t pg, svbfloat16_t fallback, svbfloat16_t data) : "CLASTB Ztied.H, Pg, Ztied.H, Zdata.H" or "MOVPRFX Zresult, Zfallback; CLASTB Zresult.H, Pg, Zresult.H, Zdata.H"
  public static unsafe Vector<bfloat16> ConditionalExtractLastActiveElementAndReplicate(Vector<bfloat16> mask, Vector<bfloat16> fallback, Vector<bfloat16> data);


    /// ConditionalSelect : Conditionally select elements

    /// svbfloat16_t svsel[_bf16](svbool_t pg, svbfloat16_t op1, svbfloat16_t op2) : "SEL Zresult.H, Pg, Zop1.H, Zop2.H"
  public static unsafe Vector<bfloat16> ConditionalSelect(Vector<bfloat16> mask, Vector<bfloat16> left, Vector<bfloat16> right);


    /// ConvertToBFloat16 : Floating-point convert

    /// svbfloat16_t svcvt_bf16[_f32]_m(svbfloat16_t inactive, svbool_t pg, svfloat32_t op) : "BFCVT Ztied.H, Pg/M, Zop.S" or "MOVPRFX Zresult, Zinactive; BFCVT Zresult.H, Pg/M, Zop.S"
    /// svbfloat16_t svcvt_bf16[_f32]_x(svbool_t pg, svfloat32_t op) : "BFCVT Ztied.H, Pg/M, Ztied.S" or "MOVPRFX Zresult, Zop; BFCVT Zresult.H, Pg/M, Zop.S"
    /// svbfloat16_t svcvt_bf16[_f32]_z(svbool_t pg, svfloat32_t op) : "MOVPRFX Zresult.S, Pg/Z, Zop.S; BFCVT Zresult.H, Pg/M, Zop.S"
  public static unsafe Vector<bfloat16> ConvertToBFloat16(Vector<float> value);


    /// CreateFalseMaskBFloat16 : Set all predicate elements to false

    /// svbool_t svpfalse[_b]() : "PFALSE Presult.B"
  public static unsafe Vector<bfloat16> CreateFalseMaskBFloat16();


    /// CreateTrueMaskBFloat16 : Set predicate elements to true

    /// svbool_t svptrue_pat_b8(enum svpattern pattern) : "PTRUE Presult.B, pattern"
  public static unsafe Vector<bfloat16> CreateTrueMaskBFloat16([ConstantExpected] SveMaskPattern pattern = SveMaskPattern.All);


    /// CreateWhileReadAfterWriteMask : While free of read-after-write conflicts

    /// svbool_t svwhilerw[_bf16](const bfloat16_t *op1, const bfloat16_t *op2) : "WHILERW Presult.H, Xop1, Xop2"
  public static unsafe Vector<bfloat16> CreateWhileReadAfterWriteMask(bfloat16* left, bfloat16* right);


    /// CreateWhileWriteAfterReadMask : While free of write-after-read conflicts

    /// svbool_t svwhilewr[_bf16](const bfloat16_t *op1, const bfloat16_t *op2) : "WHILEWR Presult.H, Xop1, Xop2"
  public static unsafe Vector<bfloat16> CreateWhileWriteAfterReadMask(bfloat16* left, bfloat16* right);


    /// DotProductBySelectedScalar : BFloat16 dot product

    /// svfloat32_t svbfdot_lane[_f32](svfloat32_t op1, svbfloat16_t op2, svbfloat16_t op3, uint64_t imm_index) : "BFDOT Ztied1.S, Zop2.H, Zop3.H[imm_index]" or "MOVPRFX Zresult, Zop1; BFDOT Zresult.S, Zop2.H, Zop3.H[imm_index]"
  public static unsafe Vector<float> DotProductBySelectedScalar(Vector<float> addend, Vector<bfloat16> left, Vector<bfloat16> right, [ConstantExpected] byte rightIndex);


    /// DownConvertNarrowingUpper : Down convert and narrow (top)

    /// svbfloat16_t svcvtnt_bf16[_f32]_m(svbfloat16_t even, svbool_t pg, svfloat32_t op) : "BFCVTNT Ztied.H, Pg/M, Zop.S"
    /// svbfloat16_t svcvtnt_bf16[_f32]_x(svbfloat16_t even, svbool_t pg, svfloat32_t op) : "BFCVTNT Ztied.H, Pg/M, Zop.S"
  public static unsafe Vector<bfloat16> DownConvertNarrowingUpper(Vector<float> value);


    /// DuplicateSelectedScalarToVector : Broadcast a scalar value

    /// svbfloat16_t svdup_lane[_bf16](svbfloat16_t data, uint16_t index) : "DUP Zresult.H, Zdata.H[index]" or "TBL Zresult.H, Zdata.H, Zindex.H"
    /// svbfloat16_t svdupq_lane[_bf16](svbfloat16_t data, uint64_t index) : "DUP Zresult.Q, Zdata.Q[index]" or "TBL Zresult.D, Zdata.D, Zindices_d.D"
  public static unsafe Vector<bfloat16> DuplicateSelectedScalarToVector(Vector<bfloat16> data, [ConstantExpected] byte index);


    /// ExtractAfterLastScalar : Extract element after last

    /// bfloat16_t svlasta[_bf16](svbool_t pg, svbfloat16_t op) : "LASTA Wresult, Pg, Zop.H" or "LASTA Hresult, Pg, Zop.H"
  public static unsafe bfloat16 ExtractAfterLastScalar(Vector<bfloat16> value);


    /// ExtractAfterLastVector : Extract element after last

    /// bfloat16_t svlasta[_bf16](svbool_t pg, svbfloat16_t op) : "LASTA Wresult, Pg, Zop.H" or "LASTA Hresult, Pg, Zop.H"
  public static unsafe Vector<bfloat16> ExtractAfterLastVector(Vector<bfloat16> value);


    /// ExtractLastScalar : Extract last element

    /// bfloat16_t svlastb[_bf16](svbool_t pg, svbfloat16_t op) : "LASTB Wresult, Pg, Zop.H" or "LASTB Hresult, Pg, Zop.H"
  public static unsafe bfloat16 ExtractLastScalar(Vector<bfloat16> value);


    /// ExtractLastVector : Extract last element

    /// bfloat16_t svlastb[_bf16](svbool_t pg, svbfloat16_t op) : "LASTB Wresult, Pg, Zop.H" or "LASTB Hresult, Pg, Zop.H"
  public static unsafe Vector<bfloat16> ExtractLastVector(Vector<bfloat16> value);


    /// ExtractVector : Extract vector from pair of vectors

    /// svbfloat16_t svext[_bf16](svbfloat16_t op1, svbfloat16_t op2, uint64_t imm3) : "EXT Ztied1.B, Ztied1.B, Zop2.B, #imm3 * 2" or "MOVPRFX Zresult, Zop1; EXT Zresult.B, Zresult.B, Zop2.B, #imm3 * 2"
  public static unsafe Vector<bfloat16> ExtractVector(Vector<bfloat16> upper, Vector<bfloat16> lower, [ConstantExpected] byte index);


    /// GetActiveElementCount : Count set predicate bits

    /// uint64_t svcntp_b8(svbool_t pg, svbool_t op) : "CNTP Xresult, Pg, Pop.B"
  public static unsafe ulong GetActiveElementCount(Vector<bfloat16> mask, Vector<bfloat16> from);


    /// InsertIntoShiftedVector : Insert scalar into shifted vector

    /// svbfloat16_t svinsr[_n_bf16](svbfloat16_t op1, bfloat16_t op2) : "INSR Ztied1.H, Wop2" or "INSR Ztied1.H, Hop2"
  public static unsafe Vector<bfloat16> InsertIntoShiftedVector(Vector<bfloat16> left, bfloat16 right);


    /// InterleaveEvenInt128FromTwoInputs : Interleave even quadwords from two inputs

    /// svbfloat16_t svtrn1q[_bf16](svbfloat16_t op1, svbfloat16_t op2) : "TRN1 Zresult.Q, Zop1.Q, Zop2.Q"
  public static unsafe Vector<bfloat16> InterleaveEvenInt128FromTwoInputs(Vector<bfloat16> left, Vector<bfloat16> right);


    /// InterleaveInt128FromHighHalvesOfTwoInputs : Interleave quadwords from high halves of two inputs

    /// svbfloat16_t svzip2q[_bf16](svbfloat16_t op1, svbfloat16_t op2) : "ZIP2 Zresult.Q, Zop1.Q, Zop2.Q"
  public static unsafe Vector<bfloat16> InterleaveInt128FromHighHalvesOfTwoInputs(Vector<bfloat16> left, Vector<bfloat16> right);


    /// InterleaveInt128FromLowHalvesOfTwoInputs : Interleave quadwords from low halves of two inputs

    /// svbfloat16_t svzip1q[_bf16](svbfloat16_t op1, svbfloat16_t op2) : "ZIP1 Zresult.Q, Zop1.Q, Zop2.Q"
  public static unsafe Vector<bfloat16> InterleaveInt128FromLowHalvesOfTwoInputs(Vector<bfloat16> left, Vector<bfloat16> right);


    /// InterleaveOddInt128FromTwoInputs : Interleave odd quadwords from two inputs

    /// svbfloat16_t svtrn2q[_bf16](svbfloat16_t op1, svbfloat16_t op2) : "TRN2 Zresult.Q, Zop1.Q, Zop2.Q"
  public static unsafe Vector<bfloat16> InterleaveOddInt128FromTwoInputs(Vector<bfloat16> left, Vector<bfloat16> right);


    /// LoadVector : Unextended load

    /// svbfloat16_t svld1[_bf16](svbool_t pg, const bfloat16_t *base) : "LD1H Zresult.H, Pg/Z, [Xarray, Xindex, LSL #1]" or "LD1H Zresult.H, Pg/Z, [Xbase, #0, MUL VL]"
  public static unsafe Vector<bfloat16> LoadVector(Vector<bfloat16> mask, bfloat16* address);


    /// LoadVector128AndReplicateToVector : Load and replicate 128 bits of data

    /// svbfloat16_t svld1rq[_bf16](svbool_t pg, const bfloat16_t *base) : "LD1RQH Zresult.H, Pg/Z, [Xarray, Xindex, LSL #1]" or "LD1RQH Zresult.H, Pg/Z, [Xarray, #index * 2]" or "LD1RQH Zresult.H, Pg/Z, [Xbase, #0]"
  public static unsafe Vector<bfloat16> LoadVector128AndReplicateToVector(Vector<bfloat16> mask, bfloat16* address);


    /// LoadVector256AndReplicateToVector : Load and replicate 256 bits of data

    /// svbfloat16_t svld1ro[_bf16](svbool_t pg, const bfloat16_t *base) : "LD1ROH Zresult.H, Pg/Z, [Xarray, Xindex, LSL #1]" or "LD1ROH Zresult.H, Pg/Z, [Xarray, #index * 2]" or "LD1ROH Zresult.H, Pg/Z, [Xbase, #0]"
  public static unsafe Vector<bfloat16> LoadVector256AndReplicateToVector(Vector<bfloat16> mask, bfloat16* address);


    /// LoadVectorFirstFaulting : Unextended load, first-faulting

    /// svbfloat16_t svldff1[_bf16](svbool_t pg, const bfloat16_t *base) : "LDFF1H Zresult.H, Pg/Z, [Xarray, Xindex, LSL #1]" or "LDFF1H Zresult.H, Pg/Z, [Xbase, XZR, LSL #1]"
  public static unsafe Vector<bfloat16> LoadVectorFirstFaulting(Vector<bfloat16> mask, bfloat16* address);


    /// LoadVectorNonFaulting : Unextended load, non-faulting

    /// svbfloat16_t svldnf1[_bf16](svbool_t pg, const bfloat16_t *base) : "LDNF1H Zresult.H, Pg/Z, [Xbase, #0, MUL VL]"
  public static unsafe Vector<bfloat16> LoadVectorNonFaulting(bfloat16* address);


    /// LoadVectorNonTemporal : Unextended load, non-temporal

    /// svbfloat16_t svldnt1[_bf16](svbool_t pg, const bfloat16_t *base) : "LDNT1H Zresult.H, Pg/Z, [Xarray, Xindex, LSL #1]" or "LDNT1H Zresult.H, Pg/Z, [Xbase, #0, MUL VL]"
  public static unsafe Vector<bfloat16> LoadVectorNonTemporal(Vector<bfloat16> mask, bfloat16* address);


    /// LoadVectorx2 : Load two-element tuples into two vectors

    /// svbfloat16x2_t svld2[_bf16](svbool_t pg, const bfloat16_t *base) : "LD2H {Zresult0.H, Zresult1.H}, Pg/Z, [Xarray, Xindex, LSL #1]" or "LD2H {Zresult0.H, Zresult1.H}, Pg/Z, [Xbase, #0, MUL VL]"
  public static unsafe (Vector<bfloat16>, Vector<bfloat16>) LoadVectorx2(Vector<bfloat16> mask, bfloat16* address);


    /// LoadVectorx3 : Load three-element tuples into three vectors

    /// svbfloat16x3_t svld3[_bf16](svbool_t pg, const bfloat16_t *base) : "LD3H {Zresult0.H - Zresult2.H}, Pg/Z, [Xarray, Xindex, LSL #1]" or "LD3H {Zresult0.H - Zresult2.H}, Pg/Z, [Xbase, #0, MUL VL]"
  public static unsafe (Vector<bfloat16>, Vector<bfloat16>, Vector<bfloat16>) LoadVectorx3(Vector<bfloat16> mask, bfloat16* address);


    /// LoadVectorx4 : Load four-element tuples into four vectors

    /// svbfloat16x4_t svld4[_bf16](svbool_t pg, const bfloat16_t *base) : "LD4H {Zresult0.H - Zresult3.H}, Pg/Z, [Xarray, Xindex, LSL #1]" or "LD4H {Zresult0.H - Zresult3.H}, Pg/Z, [Xbase, #0, MUL VL]"
  public static unsafe (Vector<bfloat16>, Vector<bfloat16>, Vector<bfloat16>, Vector<bfloat16>) LoadVectorx4(Vector<bfloat16> mask, bfloat16* address);


    /// PopCount : Count nonzero bits

    /// svuint16_t svcnt[_bf16]_m(svuint16_t inactive, svbool_t pg, svbfloat16_t op) : "CNT Ztied.H, Pg/M, Zop.H" or "MOVPRFX Zresult, Zinactive; CNT Zresult.H, Pg/M, Zop.H"
    /// svuint16_t svcnt[_bf16]_x(svbool_t pg, svbfloat16_t op) : "CNT Ztied.H, Pg/M, Ztied.H" or "MOVPRFX Zresult, Zop; CNT Zresult.H, Pg/M, Zop.H"
    /// svuint16_t svcnt[_bf16]_z(svbool_t pg, svbfloat16_t op) : "MOVPRFX Zresult.H, Pg/Z, Zop.H; CNT Zresult.H, Pg/M, Zop.H"
  public static unsafe Vector<ushort> PopCount(Vector<bfloat16> value);


    /// ReverseElement : Reverse all elements

    /// svbfloat16_t svrev[_bf16](svbfloat16_t op) : "REV Zresult.H, Zop.H"
  public static unsafe Vector<bfloat16> ReverseElement(Vector<bfloat16> value);


    /// Splice : Splice two vectors under predicate control

    /// svbfloat16_t svsplice[_bf16](svbool_t pg, svbfloat16_t op1, svbfloat16_t op2) : "SPLICE Ztied1.H, Pg, Ztied1.H, Zop2.H" or "MOVPRFX Zresult, Zop1; SPLICE Zresult.H, Pg, Zresult.H, Zop2.H"
  public static unsafe Vector<bfloat16> Splice(Vector<bfloat16> mask, Vector<bfloat16> left, Vector<bfloat16> right);


    /// Store : Non-truncating store

    /// void svst1[_bf16](svbool_t pg, bfloat16_t *base, svbfloat16_t data) : "ST1H Zdata.H, Pg, [Xarray, Xindex, LSL #1]" or "ST1H Zdata.H, Pg, [Xbase, #0, MUL VL]"
  public static unsafe void Store(Vector<bfloat16> mask, bfloat16* address, Vector<bfloat16> data);

    /// void svst2[_bf16](svbool_t pg, bfloat16_t *base, svbfloat16x2_t data) : "ST2H {Zdata0.H, Zdata1.H}, Pg, [Xarray, Xindex, LSL #1]" or "ST2H {Zdata0.H, Zdata1.H}, Pg, [Xbase, #0, MUL VL]"
  public static unsafe void Store(Vector<bfloat16> mask, bfloat16* address, (Vector<bfloat16> Value1, Vector<bfloat16> Value2) data);

    /// void svst3[_bf16](svbool_t pg, bfloat16_t *base, svbfloat16x3_t data) : "ST3H {Zdata0.H - Zdata2.H}, Pg, [Xarray, Xindex, LSL #1]" or "ST3H {Zdata0.H - Zdata2.H}, Pg, [Xbase, #0, MUL VL]"
  public static unsafe void Store(Vector<bfloat16> mask, bfloat16* address, (Vector<bfloat16> Value1, Vector<bfloat16> Value2, Vector<bfloat16> Value3) data);

    /// void svst4[_bf16](svbool_t pg, bfloat16_t *base, svbfloat16x4_t data) : "ST4H {Zdata0.H - Zdata3.H}, Pg, [Xarray, Xindex, LSL #1]" or "ST4H {Zdata0.H - Zdata3.H}, Pg, [Xbase, #0, MUL VL]"
  public static unsafe void Store(Vector<bfloat16> mask, bfloat16* address, (Vector<bfloat16> Value1, Vector<bfloat16> Value2, Vector<bfloat16> Value3, Vector<bfloat16> Value4) data);


    /// StoreNonTemporal : Non-truncating store, non-temporal

    /// void svstnt1[_bf16](svbool_t pg, bfloat16_t *base, svbfloat16_t data) : "STNT1H Zdata.H, Pg, [Xarray, Xindex, LSL #1]" or "STNT1H Zdata.H, Pg, [Xbase, #0, MUL VL]"
  public static unsafe void StoreNonTemporal(Vector<bfloat16> mask, bfloat16* address, Vector<bfloat16> data);


    /// TransposeEven : Interleave even elements from two inputs

    /// svbfloat16_t svtrn1[_bf16](svbfloat16_t op1, svbfloat16_t op2) : "TRN1 Zresult.H, Zop1.H, Zop2.H"
  public static unsafe Vector<bfloat16> TransposeEven(Vector<bfloat16> left, Vector<bfloat16> right);


    /// TransposeOdd : Interleave odd elements from two inputs

    /// svbfloat16_t svtrn2[_bf16](svbfloat16_t op1, svbfloat16_t op2) : "TRN2 Zresult.H, Zop1.H, Zop2.H"
  public static unsafe Vector<bfloat16> TransposeOdd(Vector<bfloat16> left, Vector<bfloat16> right);


    /// UnzipEven : Concatenate even elements from two inputs

    /// svbfloat16_t svuzp1[_bf16](svbfloat16_t op1, svbfloat16_t op2) : "UZP1 Zresult.H, Zop1.H, Zop2.H"
  public static unsafe Vector<bfloat16> UnzipEven(Vector<bfloat16> left, Vector<bfloat16> right);


    /// UnzipOdd : Concatenate odd elements from two inputs

    /// svbfloat16_t svuzp2[_bf16](svbfloat16_t op1, svbfloat16_t op2) : "UZP2 Zresult.H, Zop1.H, Zop2.H"
  public static unsafe Vector<bfloat16> UnzipOdd(Vector<bfloat16> left, Vector<bfloat16> right);


    /// VectorTableLookup : Table lookup in single-vector table

    /// svbfloat16_t svtbl[_bf16](svbfloat16_t data, svuint16_t indices) : "TBL Zresult.H, Zdata.H, Zindices.H"
  public static unsafe Vector<bfloat16> VectorTableLookup(Vector<bfloat16> data, Vector<ushort> indices);

    /// svbfloat16_t svtbl2[_bf16](svbfloat16x2_t data, svuint16_t indices) : "TBL Zresult.H, {Zdata0.H, Zdata1.H}, Zindices.H"
  public static unsafe Vector<bfloat16> VectorTableLookup((Vector<bfloat16> data1, Vector<bfloat16> data2), Vector<ushort> indices);


    /// VectorTableLookupExtension : Table lookup in single-vector table (merging)

    /// svbfloat16_t svtbx[_bf16](svbfloat16_t fallback, svbfloat16_t data, svuint16_t indices) : "TBX Ztied.H, Zdata.H, Zindices.H"
  public static unsafe Vector<bfloat16> VectorTableLookupExtension(Vector<bfloat16> fallback, Vector<bfloat16> data, Vector<ushort> indices);


    /// ZipHigh : Interleave elements from high halves of two inputs

    /// svbfloat16_t svzip2[_bf16](svbfloat16_t op1, svbfloat16_t op2) : "ZIP2 Zresult.H, Zop1.H, Zop2.H"
  public static unsafe Vector<bfloat16> ZipHigh(Vector<bfloat16> left, Vector<bfloat16> right);


    /// ZipLow : Interleave elements from low halves of two inputs

    /// svbfloat16_t svzip1[_bf16](svbfloat16_t op1, svbfloat16_t op2) : "ZIP1 Zresult.H, Zop1.H, Zop2.H"
  public static unsafe Vector<bfloat16> ZipLow(Vector<bfloat16> left, Vector<bfloat16> right);


  /// total method signatures: 60
  /// total method names:      53
}

@a74nh
Copy link
Contributor Author

a74nh commented Oct 26, 2023 •
edited
Loading 

  /// Optional Entries:
  ///   public static unsafe Vector<float> Bfloat16DotProduct(Vector<float> addend, Vector<bfloat16> left, bfloat16 right); // svbfdot[_n_f32]
  ///   public static unsafe Vector<float> Bfloat16MultiplyAddWideningToSinglePrecisionLower(Vector<float> op1, Vector<bfloat16> op2, bfloat16 op3); // svbfmlalb[_n_f32]
  ///   public static unsafe Vector<float> Bfloat16MultiplyAddWideningToSinglePrecisionUpper(Vector<float> op1, Vector<bfloat16> op2, bfloat16 op3); // svbfmlalt[_n_f32]
  ///   public static unsafe bfloat16 ConditionalExtractAfterLastActiveElement(Vector<bfloat16> mask, bfloat16 defaultValue, Vector<bfloat16> data); // svclasta[_n_bf16]
  ///   public static unsafe bfloat16 ConditionalExtractAfterLastActiveElementAndReplicate(Vector<bfloat16> mask, bfloat16 defaultScalar, Vector<bfloat16> data); // svclasta[_n_bf16]
  ///   public static unsafe bfloat16 ConditionalExtractLastActiveElement(Vector<bfloat16> mask, bfloat16 defaultValue, Vector<bfloat16> data); // svclastb[_n_bf16]
  ///   public static unsafe bfloat16 ConditionalExtractLastActiveElementAndReplicate(Vector<bfloat16> mask, bfloat16 fallback, Vector<bfloat16> data); // svclastb[_n_bf16]
  ///   Total Maybe: 7

  /// Rejected:
  ///   public static unsafe ulong CountElementsInAFullVector(Vector<bfloat16> value); // svlen[_bf16]
  ///   public static unsafe Vector<bfloat16> CreateTrueMaskBFloat16(); // svptrue_b8
  ///   public static unsafe Vector<bfloat16> DuplicateSelectedScalarToVector(bfloat16 value); // svdup[_n]_bf16 or svdup[_n]_bf16_m or svdup[_n]_bf16_x or svdup[_n]_bf16_z
  ///   public static unsafe Vector<bfloat16> LoadVector(Vector<bfloat16> mask, bfloat16* address, long vnum); // svld1_vnum[_bf16]
  ///   public static unsafe Vector<bfloat16> LoadVectorFirstFaulting(Vector<bfloat16> mask, bfloat16* address, long vnum); // svldff1_vnum[_bf16]
  ///   public static unsafe Vector<bfloat16> LoadVectorNonFaulting(bfloat16* address, long vnum); // svldnf1_vnum[_bf16]
  ///   public static unsafe Vector<bfloat16> LoadVectorNonTemporal(Vector<bfloat16> mask, bfloat16* address, long vnum); // svldnt1_vnum[_bf16]
  ///   public static unsafe (Vector<bfloat16>, Vector<bfloat16>) LoadVectorx2(Vector<bfloat16> mask, bfloat16* address, long vnum); // svld2_vnum[_bf16]
  ///   public static unsafe (Vector<bfloat16>, Vector<bfloat16>, Vector<bfloat16>) LoadVectorx3(Vector<bfloat16> mask, bfloat16* address, long vnum); // svld3_vnum[_bf16]
  ///   public static unsafe (Vector<bfloat16>, Vector<bfloat16>, Vector<bfloat16>, Vector<bfloat16>) LoadVectorx4(Vector<bfloat16> mask, bfloat16* address, long vnum); // svld4_vnum[_bf16]
  ///   public static unsafe void Store(Vector<bfloat16> mask, bfloat16* base, long vnum, Vector<bfloat16> data); // svst1_vnum[_bf16]
  ///   public static unsafe void Store(Vector<bfloat16> mask, bfloat16* base, long vnum, (Vector<bfloat16> data1, Vector<bfloat16> data2)); // svst2_vnum[_bf16]
  ///   public static unsafe void Store(Vector<bfloat16> mask, bfloat16* base, long vnum, (Vector<bfloat16> data1, Vector<bfloat16> data2, Vector<bfloat16> data3)); // svst3_vnum[_bf16]
  ///   public static unsafe void Store(Vector<bfloat16> mask, bfloat16* base, long vnum, (Vector<bfloat16> data1, Vector<bfloat16> data2, Vector<bfloat16> data3, Vector<bfloat16> data4)); // svst4_vnum[_bf16]
  ///   public static unsafe void StoreNonTemporal(Vector<bfloat16> mask, bfloat16* base, long vnum, Vector<bfloat16> data); // svstnt1_vnum[_bf16]
  ///   Total Rejected: 15

  /// Total ACLE covered across API:      93

@a74nh
Copy link
Contributor Author

a74nh commented Oct 26, 2023

This contributes to #93095

It covers all of the instructions in FEAT_BF16. This an optional 8.2 feature and from 8.6 is mandatory in any hardware with SVE. All the methods are binaryfloat16 versions of existing double methods in SVE/SVE2. This requires 16bit binary float support adding to C# before it can be implemented.

This list was auto generated from the C ACLE for SVE, and is in three parts:

The methods list reduced down to Vector versions. All possible varaints of T are given above the method.
The complete list of all methods. The corresponding ACLE methods and SVE instructions are given above the method.
All rejected ACLE methods. These are methods we have agreed that do not need including in C#.
Where possible, existing C# naming conventions have been matched.

Many of the C functions include predicate argument(s), of type svbool_t as the first argument. These are missing from the C# method. It is expected that the Jit will create predicates where required, or combine with uses of conditionalSelect(). For more discussion see #88140 comment.

@a74nh a74nh mentioned this issue Oct 26, 2023
Closed
31 tasks
@MichalPetryka
Copy link
Contributor

I think we should open a proposal for System.BFloat16 before this is considered.

@a74nh
Copy link
Contributor Author

a74nh commented Nov 6, 2023

I think we should open a proposal for System.BFloat16 before this is considered.

Agreed. I suspect this API is of low priority.

@a74nh a74nh mentioned this issue Nov 13, 2023
Closed
@kunalspathak kunalspathak added the api-ready-for-review API is ready for review, it is NOT ready for implementation label Feb 7, 2024
@ghost
Copy link

ghost commented Feb 8, 2024

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

Issue Details 
namespace System.Runtime.Intrinsics.Arm

/// VectorT Summary
public abstract class SveBf16 : AdvSimd /// Feature: FEAT_BF16
{

  public static unsafe Vector<float> Bfloat16DotProduct(Vector<float> op1, Vector<bfloat16> op2, Vector<bfloat16> op3); // BFDOT // MOVPRFX

  public static unsafe Vector<float> Bfloat16DotProduct(Vector<float> op1, Vector<bfloat16> op2, Vector<bfloat16> op3, ulong imm_index); // BFDOT // MOVPRFX

  public static unsafe Vector<float> Bfloat16MatrixMultiplyAccumulate(Vector<float> op1, Vector<bfloat16> op2, Vector<bfloat16> op3); // BFMMLA // MOVPRFX

  public static unsafe Vector<float> Bfloat16MultiplyAddWideningToSinglePrecisionLower(Vector<float> op1, Vector<bfloat16> op2, Vector<bfloat16> op3); // BFMLALB // MOVPRFX

  public static unsafe Vector<float> Bfloat16MultiplyAddWideningToSinglePrecisionLower(Vector<float> op1, Vector<bfloat16> op2, Vector<bfloat16> op3, ulong imm_index); // BFMLALB // MOVPRFX

  public static unsafe Vector<float> Bfloat16MultiplyAddWideningToSinglePrecisionUpper(Vector<float> op1, Vector<bfloat16> op2, Vector<bfloat16> op3); // BFMLALT // MOVPRFX

  public static unsafe Vector<float> Bfloat16MultiplyAddWideningToSinglePrecisionUpper(Vector<float> op1, Vector<bfloat16> op2, Vector<bfloat16> op3, ulong imm_index); // BFMLALT // MOVPRFX

  public static unsafe Vector<bfloat16> ConditionalExtractAfterLastActiveElement(Vector<bfloat16> mask, Vector<bfloat16> fallback, Vector<bfloat16> data); // CLASTA // MOVPRFX

  public static unsafe Vector<bfloat16> ConditionalExtractLastActiveElement(Vector<bfloat16> mask, Vector<bfloat16> fallback, Vector<bfloat16> data); // CLASTB // MOVPRFX

  public static unsafe Vector<bfloat16> ConditionalSelect(Vector<bfloat16> mask, Vector<bfloat16> left, Vector<bfloat16> right); // SEL

  public static unsafe Vector<bfloat16> ConvertToBFloat16(Vector<float> value); // BFCVT // predicated, MOVPRFX

  public static unsafe Vector<bfloat16> CreateWhileReadAfterWriteMask(const bfloat16 left, const bfloat16 right); // WHILERW

  public static unsafe Vector<bfloat16> CreateWhileWriteAfterReadMask(const bfloat16 left, const bfloat16 right); // WHILEWR

  public static unsafe Vector<bfloat16> DownConvertNarrowingUpper(Vector<float> value); // BFCVTNT // predicated

  public static unsafe Vector<bfloat16> DuplicateSelectedScalarToVector(bfloat16 value); // DUP or FDUP or CPY or FCPY // predicated, MOVPRFX

  public static unsafe Vector<bfloat16> DuplicateSelectedScalarToVector(Vector<bfloat16> data, ushort index); // DUP or TBL

  public static unsafe Vector<bfloat16> DuplicateSelectedScalarToVector(Vector<bfloat16> data, ulong index); // DUP or TBL

  public static unsafe bfloat16 ExtractAfterLast(Vector<bfloat16> value); // LASTA // predicated

  public static unsafe bfloat16 ExtractLast(Vector<bfloat16> value); // LASTB // predicated

  public static unsafe Vector<bfloat16> ExtractVector(Vector<bfloat16> upper, Vector<bfloat16> lower, ulong index); // EXT // MOVPRFX

  public static unsafe Vector<bfloat16> InsertIntoShiftedVector(Vector<bfloat16> left, bfloat16 right); // INSR

  public static unsafe Vector<bfloat16> LoadVector(Vector<bfloat16> mask, const bfloat16 *base); // LD1H

  public static unsafe Vector<bfloat16> LoadVector128AndReplicateToVector(Vector<bfloat16> mask, const bfloat16 *base); // LD1RQH

  public static unsafe Vector<bfloat16> LoadVectorFirstFaulting(Vector<bfloat16> mask, const bfloat16 *base); // LDFF1H

  public static unsafe Vector<bfloat16> LoadVectorNonFaulting(Vector<bfloat16> mask, const bfloat16 *base); // LDNF1H

  public static unsafe Vector<bfloat16> LoadVectorNonTemporal(Vector<bfloat16> mask, const bfloat16 *base); // LDNT1H

  public static unsafe (Vector<bfloat16>, Vector<bfloat16>) LoadVectorx2(Vector<bfloat16> mask, const bfloat16 *base); // LD2H

  public static unsafe (Vector<bfloat16>, Vector<bfloat16>, Vector<bfloat16>) LoadVectorx3(Vector<bfloat16> mask, const bfloat16 *base); // LD3H

  public static unsafe (Vector<bfloat16>, Vector<bfloat16>, Vector<bfloat16>, Vector<bfloat16>) LoadVectorx4(Vector<bfloat16> mask, const bfloat16 *base); // LD4H

  public static unsafe Vector<ushort> PopCount(Vector<bfloat16> value); // CNT // predicated, MOVPRFX

  public static unsafe Vector<bfloat16> ReverseElement(Vector<bfloat16> value); // REV

  public static unsafe Vector<bfloat16> Splice(Vector<bfloat16> mask, Vector<bfloat16> left, Vector<bfloat16> right); // SPLICE // MOVPRFX

  public static unsafe void Store(Vector<bfloat16> mask, bfloat16 *base, Vector<bfloat16> data); // ST1H

  public static unsafe void StoreNonTemporal(Vector<bfloat16> mask, bfloat16 *base, Vector<bfloat16> data); // STNT1H

  public static unsafe void Storex2(Vector<bfloat16> mask, bfloat16 *base, (Vector<bfloat16> data1, Vector<bfloat16> data2)); // ST2H

  public static unsafe void Storex3(Vector<bfloat16> mask, bfloat16 *base, (Vector<bfloat16> data1, Vector<bfloat16> data2, Vector<bfloat16> data3)); // ST3H

  public static unsafe void Storex4(Vector<bfloat16> mask, bfloat16 *base, (Vector<bfloat16> data1, Vector<bfloat16> data2, Vector<bfloat16> data3, Vector<bfloat16> data4)); // ST4H

  public static unsafe Vector<bfloat16> TransposeEven(Vector<bfloat16> left, Vector<bfloat16> right); // TRN1

  public static unsafe Vector<bfloat16> TransposeOdd(Vector<bfloat16> left, Vector<bfloat16> right); // TRN2

  public static unsafe Vector<bfloat16> UnzipEven(Vector<bfloat16> left, Vector<bfloat16> right); // UZP1

  public static unsafe Vector<bfloat16> UnzipOdd(Vector<bfloat16> left, Vector<bfloat16> right); // UZP2

  public static unsafe Vector<bfloat16> VectorTableLookup(Vector<bfloat16> data, Vector<ushort> indices); // TBL

  public static unsafe Vector<bfloat16> VectorTableLookup((Vector<bfloat16> data1, Vector<bfloat16> data2), Vector<ushort> indices); // TBL

  public static unsafe Vector<bfloat16> VectorTableLookupExtension(Vector<bfloat16> fallback, Vector<bfloat16> data, Vector<ushort> indices); // TBX

  public static unsafe Vector<bfloat16> ZipHigh(Vector<bfloat16> left, Vector<bfloat16> right); // ZIP2

  public static unsafe Vector<bfloat16> ZipLow(Vector<bfloat16> left, Vector<bfloat16> right); // ZIP1

  /// total method signatures: 46


  /// Optional Entries:

  public static unsafe Vector<float> Bfloat16DotProduct(Vector<float> op1, Vector<bfloat16> op2, bfloat16 op3); // BFDOT // MOVPRFX

  public static unsafe Vector<float> Bfloat16MultiplyAddWideningToSinglePrecisionLower(Vector<float> op1, Vector<bfloat16> op2, bfloat16 op3); // BFMLALB // MOVPRFX

  public static unsafe Vector<float> Bfloat16MultiplyAddWideningToSinglePrecisionUpper(Vector<float> op1, Vector<bfloat16> op2, bfloat16 op3); // BFMLALT // MOVPRFX

  public static unsafe bfloat16 ConditionalExtractAfterLastActiveElement(Vector<bfloat16> mask, bfloat16 fallback, Vector<bfloat16> data); // CLASTA

  public static unsafe bfloat16 ConditionalExtractLastActiveElement(Vector<bfloat16> mask, bfloat16 fallback, Vector<bfloat16> data); // CLASTB

  public static unsafe Vector<bfloat16> LoadVector(Vector<bfloat16> mask, const bfloat16 *base, long vnum); // LD1H

  public static unsafe Vector<bfloat16> LoadVectorFirstFaulting(Vector<bfloat16> mask, const bfloat16 *base, long vnum); // LDFF1H

  public static unsafe Vector<bfloat16> LoadVectorNonFaulting(Vector<bfloat16> mask, const bfloat16 *base, long vnum); // LDNF1H

  public static unsafe Vector<bfloat16> LoadVectorNonTemporal(Vector<bfloat16> mask, const bfloat16 *base, long vnum); // LDNT1H

  public static unsafe (Vector<bfloat16>, Vector<bfloat16>) LoadVectorx2(Vector<bfloat16> mask, const bfloat16 *base, long vnum); // LD2H

  public static unsafe (Vector<bfloat16>, Vector<bfloat16>, Vector<bfloat16>) LoadVectorx3(Vector<bfloat16> mask, const bfloat16 *base, long vnum); // LD3H

  public static unsafe (Vector<bfloat16>, Vector<bfloat16>, Vector<bfloat16>, Vector<bfloat16>) LoadVectorx4(Vector<bfloat16> mask, const bfloat16 *base, long vnum); // LD4H

  public static unsafe void Store(Vector<bfloat16> mask, bfloat16 *base, long vnum, Vector<bfloat16> data); // ST1H

  public static unsafe void StoreNonTemporal(Vector<bfloat16> mask, bfloat16 *base, long vnum, Vector<bfloat16> data); // STNT1H

  public static unsafe void Storex2(Vector<bfloat16> mask, bfloat16 *base, long vnum, (Vector<bfloat16> data1, Vector<bfloat16> data2)); // ST2H

  public static unsafe void Storex3(Vector<bfloat16> mask, bfloat16 *base, long vnum, (Vector<bfloat16> data1, Vector<bfloat16> data2, Vector<bfloat16> data3)); // ST3H

  public static unsafe void Storex4(Vector<bfloat16> mask, bfloat16 *base, long vnum, (Vector<bfloat16> data1, Vector<bfloat16> data2, Vector<bfloat16> data3, Vector<bfloat16> data4)); // ST4H

  /// total optional method signatures: 17

}
Author: a74nh
Assignees: -
Labels:

area-System.Runtime.Intrinsics, untriaged, api-ready-for-review
Milestone: -

@kunalspathak kunalspathak mentioned this issue Mar 19, 2024
Closed
@kunalspathak kunalspathak added this to the Future milestone May 1, 2024
@jeffschwMSFT jeffschwMSFT removed the untriaged New issue has not been triaged by the area owner label May 3, 2024
@tannergooding tannergooding added api-suggestion Early API idea and discussion, it is NOT ready for implementation arch-arm64 and removed api-ready-for-review API is ready for review, it is NOT ready for implementation labels Aug 20, 2024
@a74nh
Copy link
Contributor Author

a74nh commented Aug 21, 2024

Updated to match implemented SVE1 methods.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees
No one assigned
Labels
api-suggestion Early API idea and discussion, it is NOT ready for implementation arch-arm64 area-System.Runtime.Intrinsics
Projects
None yet
Milestone
Future
Development

No branches or pull requests
5 participants
@a74nh @jeffschwMSFT @tannergooding @kunalspathak @MichalPetryka