| !sum |
| # TODO: remove the "Const Immediate" from the documentation field, it's there only for debug purposes. |
| - go: Add |
| commutative: "true" |
| extension: "AVX.*" |
| documentation: !string |- |
| // Add adds corresponding elements of two vectors. |
| - go: SaturatedAdd |
| commutative: "true" |
| extension: "AVX.*" |
| documentation: !string |- |
| // SaturatedAdd adds corresponding elements of two vectors with saturation. |
| - go: MaskedAdd |
| masked: "true" |
| commutative: "true" |
| extension: "AVX.*" |
| documentation: !string |- |
| // MaskedAdd adds corresponding elements of two vectors. |
| - go: MaskedSaturatedAdd |
| masked: "true" |
| commutative: "true" |
| extension: "AVX.*" |
| documentation: !string |- |
| // MaskedSaturatedAdd adds corresponding elements of two vectors with saturation. |
| - go: Sub |
| commutative: "false" |
| extension: "AVX.*" |
| documentation: !string |- |
| // Sub subtracts corresponding elements of two vectors. |
| - go: SaturatedSub |
| commutative: "false" |
| extension: "AVX.*" |
| documentation: !string |- |
| // SaturatedSub subtracts corresponding elements of two vectors with saturation. |
| - go: MaskedSub |
| masked: "true" |
| commutative: "false" |
| extension: "AVX.*" |
| documentation: !string |- |
| // MaskedSub subtracts corresponding elements of two vectors. |
| - go: MaskedSaturatedSub |
| masked: "true" |
| commutative: "false" |
| extension: "AVX.*" |
| documentation: !string |- |
| // MaskedSaturatedSub subtracts corresponding elements of two vectors with saturation. |
| - go: PairwiseAdd |
| commutative: "false" |
| extension: "AVX.*" |
| documentation: !string |- |
| // PairwiseAdd horizontally adds adjacent pairs of elements. |
| // For x = [x0, x1, x2, x3, ...] and y = [y0, y1, y2, y3, ...], the result is [y0+y1, y2+y3, ..., x0+x1, x2+x3, ...]. |
| - go: PairwiseSub |
| commutative: "false" |
| extension: "AVX.*" |
| documentation: !string |- |
| // PairwiseSub horizontally subtracts adjacent pairs of elements. |
| // For x = [x0, x1, x2, x3, ...] and y = [y0, y1, y2, y3, ...], the result is [y0-y1, y2-y3, ..., x0-x1, x2-x3, ...]. |
| - go: SaturatedPairwiseAdd |
| commutative: "false" |
| extension: "AVX.*" |
| documentation: !string |- |
| // SaturatedPairwiseAdd horizontally adds adjacent pairs of elements with saturation. |
| // For x = [x0, x1, x2, x3, ...] and y = [y0, y1, y2, y3, ...], the result is [y0+y1, y2+y3, ..., x0+x1, x2+x3, ...]. |
| - go: SaturatedPairwiseSub |
| commutative: "false" |
| extension: "AVX.*" |
| documentation: !string |- |
| // SaturatedPairwiseSub horizontally subtracts adjacent pairs of elements with saturation. |
| // For x = [x0, x1, x2, x3, ...] and y = [y0, y1, y2, y3, ...], the result is [y0-y1, y2-y3, ..., x0-x1, x2-x3, ...]. |
| - go: And |
| commutative: "true" |
| extension: "AVX.*" |
| documentation: !string |- |
| // And performs a bitwise AND operation between two vectors. |
| - go: MaskedAnd |
| masked: "true" |
| commutative: "true" |
| extension: "AVX.*" |
| documentation: !string |- |
| // MaskedAnd performs a masked bitwise AND operation between two vectors. |
| - go: Or |
| commutative: "true" |
| extension: "AVX.*" |
| documentation: !string |- |
| // Or performs a bitwise OR operation between two vectors. |
| - go: MaskedOr |
| masked: "true" |
| commutative: "true" |
| extension: "AVX.*" |
| documentation: !string |- |
| // MaskedOr performs a masked bitwise OR operation between two vectors. |
| - go: AndNot |
| commutative: "false" |
| extension: "AVX.*" |
| documentation: !string |- |
| // AndNot performs a bitwise AND NOT operation between two vectors. |
| - go: MaskedAndNot |
| masked: "true" |
| commutative: "false" |
| extension: "AVX.*" |
| documentation: !string |- |
| // MaskedAndNot performs a masked bitwise AND NOT operation between two vectors. |
| - go: Xor |
| commutative: "true" |
| extension: "AVX.*" |
| documentation: !string |- |
| // Xor performs a bitwise XOR operation between two vectors. |
| - go: MaskedXor |
| masked: "true" |
| commutative: "true" |
| extension: "AVX.*" |
| documentation: !string |- |
| // MaskedXor performs a masked bitwise XOR operation between two vectors. |
| # We also have PTEST and VPTERNLOG, those should be hidden from the users |
| # and only appear in rewrite rules. |
| # const imm predicate(holds for both float and int|uint): |
| # 0: Equal |
| # 1: Less |
| # 2: LessEqual |
| # 4: NotEqual |
| # 5: GreaterEqual |
| # 6: Greater |
| - go: Equal |
| constImm: 0 |
| commutative: "true" |
| extension: "AVX.*" |
| documentation: !string |- |
| // Equal compares for equality. |
| // Const Immediate = 0. |
| - go: Less |
| constImm: 1 |
| commutative: "false" |
| extension: "AVX.*" |
| documentation: !string |- |
| // Less compares for less than. |
| // Const Immediate = 1. |
| - go: LessEqual |
| constImm: 2 |
| commutative: "false" |
| extension: "AVX.*" |
| documentation: !string |- |
| // LessEqual compares for less than or equal. |
| // Const Immediate = 2. |
| - go: IsNan # For float only. |
| constImm: 3 |
| commutative: "true" |
| extension: "AVX.*" |
| documentation: !string |- |
| // IsNan checks if elements are NaN. Use as x.IsNan(x). |
| // Const Immediate = 3. |
| - go: NotEqual |
| constImm: 4 |
| commutative: "true" |
| extension: "AVX.*" |
| documentation: !string |- |
| // NotEqual compares for inequality. |
| // Const Immediate = 4. |
| - go: GreaterEqual |
| constImm: 5 |
| commutative: "false" |
| extension: "AVX.*" |
| documentation: !string |- |
| // GreaterEqual compares for greater than or equal. |
| // Const Immediate = 5. |
| - go: Greater |
| constImm: 6 |
| commutative: "false" |
| extension: "AVX.*" |
| documentation: !string |- |
| // Greater compares for greater than. |
| // Const Immediate = 6. |
| |
| - go: MaskedEqual |
| constImm: 0 |
| masked: "true" |
| commutative: "true" |
| extension: "AVX.*" |
| documentation: !string |- |
| // MaskedEqual compares for equality, masked. |
| // Const Immediate = 0. |
| docUnmasked: !string |- |
| // Equal compares for equality. |
| // Const Immediate = 0. |
| - go: MaskedLess |
| constImm: 1 |
| masked: "true" |
| commutative: "false" |
| extension: "AVX.*" |
| documentation: !string |- |
| // MaskedLess compares for less than. |
| // Const Immediate = 1. |
| - go: MaskedLessEqual |
| constImm: 2 |
| masked: "true" |
| commutative: "false" |
| extension: "AVX.*" |
| documentation: !string |- |
| // MaskedLessEqual compares for less than or equal. |
| // Const Immediate = 2. |
| - go: MaskedIsNan # For float only. |
| constImm: 3 |
| masked: "true" |
| commutative: "true" |
| extension: "AVX.*" |
| documentation: !string |- |
| // MaskedIsNan checks if elements are NaN. Use as x.IsNan(x). |
| // Const Immediate = 3. |
| - go: MaskedNotEqual |
| constImm: 4 |
| masked: "true" |
| commutative: "true" |
| extension: "AVX.*" |
| documentation: !string |- |
| // MaskedNotEqual compares for inequality. |
| // Const Immediate = 4. |
| - go: MaskedGreaterEqual |
| constImm: 5 |
| masked: "true" |
| commutative: "false" |
| extension: "AVX.*" |
| documentation: !string |- |
| // MaskedGreaterEqual compares for greater than or equal. |
| // Const Immediate = 5. |
| - go: MaskedGreater |
| constImm: 6 |
| masked: "true" |
| commutative: "false" |
| extension: "AVX.*" |
| documentation: !string |- |
| // MaskedGreater compares for greater than. |
| // Const Immediate = 6. |
| - go: Div |
| commutative: "false" |
| extension: "AVX.*" |
| documentation: !string |- |
| // Div divides elements of two vectors. |
| - go: MaskedDiv |
| commutative: "false" |
| masked: "true" |
| extension: "AVX.*" |
| documentation: !string |- |
| // MaskedDiv divides elements of two vectors. |
| - go: Sqrt |
| commutative: "false" |
| extension: "AVX.*" |
| documentation: !string |- |
| // Sqrt computes the square root of each element. |
| - go: MaskedSqrt |
| commutative: "false" |
| masked: "true" |
| extension: "AVX.*" |
| documentation: !string |- |
| // MaskedSqrt computes the square root of each element. |
| - go: ApproximateReciprocal |
| commutative: "false" |
| extension: "AVX.*" |
| documentation: !string |- |
| // ApproximateReciprocal computes an approximate reciprocal of each element. |
| - go: MaskedApproximateReciprocal |
| commutative: "false" |
| masked: "true" |
| extension: "AVX.*" |
| documentation: !string |- |
| // MaskedApproximateReciprocal computes an approximate reciprocal of each element. |
| - go: ApproximateReciprocalOfSqrt |
| commutative: "false" |
| extension: "AVX.*" |
| documentation: !string |- |
| // ApproximateReciprocalOfSqrt computes an approximate reciprocal of the square root of each element. |
| - go: MaskedApproximateReciprocalOfSqrt |
| commutative: "false" |
| masked: "true" |
| extension: "AVX.*" |
| documentation: !string |- |
| // MaskedApproximateReciprocalOfSqrt computes an approximate reciprocal of the square root of each element. |
| - go: MaskedMulByPowOf2 # This operation is all after AVX512, the unmasked version will be generated. |
| commutative: "false" |
| masked: "true" |
| extension: "AVX.*" |
| documentation: !string |- |
| // MaskedMulByPowOf2 multiplies elements by a power of 2. |
| |
| - go: Round |
| commutative: "false" |
| extension: "AVX.*" |
| constImm: 0 |
| documentation: !string |- |
| // Round rounds elements to the nearest integer. |
| // Const Immediate = 0. |
| - go: MaskedRoundWithPrecision |
| commutative: "false" |
| extension: "AVX.*" |
| constImm: 0 |
| masked: "true" |
| documentation: !string |- |
| // MaskedRoundWithPrecision rounds elements with specified precision. |
| // Const Immediate = 0. |
| - go: MaskedRoundSuppressExceptionWithPrecision |
| commutative: "false" |
| extension: "AVX.*" |
| constImm: 8 |
| masked: "true" |
| documentation: !string |- |
| // MaskedRoundSuppressExceptionWithPrecision rounds elements with specified precision, suppressing exceptions. |
| // Const Immediate = 8. |
| - go: MaskedDiffWithRoundWithPrecision |
| commutative: "false" |
| extension: "AVX.*" |
| constImm: 0 |
| masked: "true" |
| documentation: !string |- |
| // MaskedDiffWithRoundWithPrecision computes the difference after rounding with specified precision. |
| // Const Immediate = 0. |
| - go: MaskedDiffWithRoundSuppressExceptionWithPrecision |
| commutative: "false" |
| extension: "AVX.*" |
| constImm: 8 |
| masked: "true" |
| documentation: !string |- |
| // MaskedDiffWithRoundSuppressExceptionWithPrecision computes the difference after rounding with specified precision, suppressing exceptions. |
| // Const Immediate = 8. |
| |
| - go: Floor |
| commutative: "false" |
| extension: "AVX.*" |
| constImm: 1 |
| documentation: !string |- |
| // Floor rounds elements down to the nearest integer. |
| // Const Immediate = 1. |
| - go: MaskedFloorWithPrecision |
| commutative: "false" |
| extension: "AVX.*" |
| constImm: 1 |
| masked: "true" |
| documentation: !string |- |
| // MaskedFloorWithPrecision rounds elements down with specified precision, masked. |
| // Const Immediate = 1. |
| docUnmasked: !string |- |
| // FloorWithPrecision rounds elements down with specified precision. |
| // Const Immediate = 1. |
| - go: MaskedFloorSuppressExceptionWithPrecision |
| commutative: "false" |
| extension: "AVX.*" |
| constImm: 9 |
| masked: "true" |
| documentation: !string |- |
| // MaskedFloorSuppressExceptionWithPrecision rounds elements down with specified precision, suppressing exceptions, masked. |
| // Const Immediate = 9. |
| docUnmasked: !string |- |
| // FloorSuppressExceptionWithPrecision rounds elements down with specified precision, suppressing exceptions. |
| // Const Immediate = 9. |
| - go: MaskedDiffWithFloorWithPrecision |
| commutative: "false" |
| extension: "AVX.*" |
| constImm: 1 |
| masked: "true" |
| documentation: !string |- |
| // MaskedDiffWithFloorWithPrecision computes the difference after flooring with specified precision. |
| // Const Immediate = 1. |
| - go: MaskedDiffWithFloorSuppressExceptionWithPrecision |
| commutative: "false" |
| extension: "AVX.*" |
| constImm: 9 |
| masked: "true" |
| documentation: !string |- |
| // MaskedDiffWithFloorSuppressExceptionWithPrecision computes the difference after flooring with specified precision, suppressing exceptions. |
| // Const Immediate = 9. |
| |
| - go: Ceil |
| commutative: "false" |
| extension: "AVX.*" |
| constImm: 2 |
| documentation: !string |- |
| // Ceil rounds elements up to the nearest integer. |
| // Const Immediate = 2. |
| - go: MaskedCeilWithPrecision |
| commutative: "false" |
| extension: "AVX.*" |
| constImm: 2 |
| masked: "true" |
| documentation: !string |- |
| // MaskedCeilWithPrecision rounds elements up with specified precision, masked. |
| // Const Immediate = 2. |
| docUnmasked: !string |- |
| // CeilWithPrecision rounds elements up with specified precision. |
| // Const Immediate = 2. |
| - go: MaskedCeilSuppressExceptionWithPrecision |
| commutative: "false" |
| extension: "AVX.*" |
| constImm: 10 |
| masked: "true" |
| documentation: !string |- |
| // MaskedCeilSuppressExceptionWithPrecision rounds elements up with specified precision, suppressing exceptions. |
| // Const Immediate = 10. |
| - go: MaskedDiffWithCeilWithPrecision |
| commutative: "false" |
| extension: "AVX.*" |
| constImm: 2 |
| masked: "true" |
| documentation: !string |- |
| // MaskedDiffWithCeilWithPrecision computes the difference after ceiling with specified precision. |
| // Const Immediate = 2. |
| - go: MaskedDiffWithCeilSuppressExceptionWithPrecision |
| commutative: "false" |
| extension: "AVX.*" |
| constImm: 10 |
| masked: "true" |
| documentation: !string |- |
| // MaskedDiffWithCeilSuppressExceptionWithPrecision computes the difference after ceiling with specified precision, suppressing exceptions. |
| // Const Immediate = 10. |
| |
| - go: Trunc |
| commutative: "false" |
| extension: "AVX.*" |
| constImm: 3 |
| documentation: !string |- |
| // Trunc truncates elements towards zero. |
| // Const Immediate = 3. |
| - go: MaskedTruncWithPrecision |
| commutative: "false" |
| extension: "AVX.*" |
| constImm: 3 |
| masked: "true" |
| documentation: !string |- |
| // MaskedTruncWithPrecision truncates elements with specified precision. |
| // Const Immediate = 3. |
| - go: MaskedTruncSuppressExceptionWithPrecision |
| commutative: "false" |
| extension: "AVX.*" |
| constImm: 11 |
| masked: "true" |
| documentation: !string |- |
| // MaskedTruncSuppressExceptionWithPrecision truncates elements with specified precision, suppressing exceptions. |
| // Const Immediate = 11. |
| - go: MaskedDiffWithTruncWithPrecision |
| commutative: "false" |
| extension: "AVX.*" |
| constImm: 3 |
| masked: "true" |
| documentation: !string |- |
| // MaskedDiffWithTruncWithPrecision computes the difference after truncating with specified precision. |
| // Const Immediate = 3. |
| - go: MaskedDiffWithTruncSuppressExceptionWithPrecision |
| commutative: "false" |
| extension: "AVX.*" |
| constImm: 11 |
| masked: "true" |
| documentation: !string |- |
| // MaskedDiffWithTruncSuppressExceptionWithPrecision computes the difference after truncating with specified precision, suppressing exceptions. |
| // Const Immediate = 11. |
| |
| - go: AddSub |
| commutative: "false" |
| extension: "AVX.*" |
| documentation: !string |- |
| // AddSub subtracts even elements and adds odd elements of two vectors. |
| - go: MaskedGaloisFieldAffineTransform |
| masked: "true" |
| commutative: "false" |
| extension: "AVX.*" |
| documentation: !string |- |
| // GaloisFieldAffineTransform computes an affine transformation in GF(2^8): |
| // x is a vector of 8-bit vectors, with each adjacent 8 as a group; y is a vector of 8x8 1-bit matrixes; |
| // imm is an 8-bit vector. The affine transformation is y * x + imm, with each element of y |
| // corresponding to a group of 8 elements in x. |
| - go: MaskedGaloisFieldAffineTransformInversed |
| masked: "true" |
| commutative: "false" |
| extension: "AVX.*" |
| documentation: !string |- |
| // GaloisFieldAffineTransform computes an affine transformation in GF(2^8), |
| // with x inversed with reduction polynomial x^8 + x^4 + x^3 + x + 1: |
| // x is a vector of 8-bit vectors, with each adjacent 8 as a group; y is a vector of 8x8 1-bit matrixes; |
| // imm is an 8-bit vector. The affine transformation is y * x + imm, with each element of y |
| // corresponding to a group of 8 elements in x. |
| - go: MaskedGaloisFieldMul |
| masked: "true" |
| commutative: "false" |
| extension: "AVX.*" |
| documentation: !string |- |
| // GaloisFieldMul computes element-wise GF(2^8) multiplication with |
| // reduction polynomial x^8 + x^4 + x^3 + x + 1. |
| - go: Average |
| commutative: "true" |
| extension: "AVX.*" # VPAVGB/W are available across various AVX versions |
| documentation: !string |- |
| // Average computes the rounded average of corresponding elements. |
| - go: MaskedAverage |
| commutative: "true" |
| masked: "true" |
| extension: "AVX512.*" # Masked operations are typically AVX512 |
| documentation: !string |- |
| // MaskedAverage computes the rounded average of corresponding elements. |
| |
| - go: Absolute |
| commutative: "false" |
| # Unary operation, not commutative |
| extension: "AVX.*" # VPABSB/W/D are AVX, VPABSQ is AVX512 |
| documentation: !string |- |
| // Absolute computes the absolute value of each element. |
| - go: MaskedAbsolute |
| commutative: "false" |
| masked: "true" |
| extension: "AVX512.*" |
| documentation: !string |- |
| // MaskedAbsolute computes the absolute value of each element. |
| |
| - go: Sign |
| # Applies sign of second operand to first: sign(val, sign_src) |
| commutative: "false" |
| extension: "AVX.*" |
| documentation: !string |- |
| // Sign returns the product of the first operand with -1, 0, or 1, |
| // whichever constant is nearest to the value of the second operand. |
| # Sign does not have masked version |
| |
| - go: MaskedPopCount |
| commutative: "false" |
| masked: "true" |
| extension: "AVX512.*" # VPOPCNT instructions are AVX512 (BITALG or VPOPCNTDQ) |
| documentation: !string |- |
| // MaskedPopCount counts the number of set bits in each element. |
| - go: PairDotProd |
| commutative: "false" |
| extension: "AVX.*" |
| documentation: !string |- |
| // PairDotProd multiplies the elements and add the pairs together, |
| // yielding a vector of half as many elements with twice the input element size. |
| - go: MaskedPairDotProd |
| masked: "true" |
| commutative: "false" |
| extension: "AVX.*" |
| documentation: !string |- |
| // MaskedPairDotProd multiplies the elements and add the pairs together, |
| // yielding a vector of half as many elements with twice the input element size. |
| - go: SaturatedUnsignedSignedPairDotProd |
| commutative: "false" |
| extension: "AVX.*" |
| documentation: !string |- |
| // SaturatedPairDotProd multiplies the elements and add the pairs together with saturation, |
| // yielding a vector of half as many elements with twice the input element size. |
| - go: MaskedSaturatedUnsignedSignedPairDotProd |
| masked: "true" |
| commutative: "false" |
| extension: "AVX.*" |
| documentation: !string |- |
| // MaskedSaturatedPairDotProd multiplies the elements and add the pairs together with saturation, |
| // yielding a vector of half as many elements with twice the input element size. |
| |
| # QuadDotProd, i.e. VPDPBUSD(S) are operations with src/dst on the same register, we are not supporting this as of now. |
| - go: DotProdBroadcast |
| commutative: "true" |
| extension: "AVX.*" |
| documentation: !string |- |
| // DotProdBroadcast multiplies all elements and broadcasts the sum. |
| // Const Immediate = 127. |
| - go: UnsignedSignedQuadDotProdAccumulate |
| commutative: "false" |
| extension: "AVX.*" |
| documentation: !string |- |
| // UnsignedSignedQuadDotProdAccumulate performs dot products on groups of 4 elements of y and z and accumulates the results to x. |
| - go: MaskedUnsignedSignedQuadDotProdAccumulate |
| masked: "true" |
| commutative: "false" |
| extension: "AVX.*" |
| documentation: !string |- |
| // MaskedUnsignedSignedQuadDotProdAccumulate performs dot products on groups of 4 elements of y and z and accumulates the results to x. |
| - go: SaturatedUnsignedSignedQuadDotProdAccumulate |
| commutative: "false" |
| extension: "AVX.*" |
| documentation: !string |- |
| // SaturatedUnsignedSignedQuadDotProdAccumulate multiplies performs dot products on groups of 4 elements of y and z and accumulates the results to x. |
| - go: MaskedSaturatedUnsignedSignedQuadDotProdAccumulate |
| masked: "true" |
| commutative: "false" |
| extension: "AVX.*" |
| documentation: !string |- |
| // SaturatedUnsignedSignedQuadDotProdAccumulate multiplies performs dot products on groups of 4 elements of y and z and accumulates the results to x. |
| - go: PairDotProdAccumulate |
| commutative: "false" |
| extension: "AVX.*" |
| documentation: !string |- |
| // PairDotProdAccumulate performs dot products on pairs of elements of y and z and accumulates the results to x. |
| - go: MaskedPairDotProdAccumulate |
| masked: "true" |
| commutative: "false" |
| extension: "AVX.*" |
| documentation: !string |- |
| // MaskedPairDotProdAccumulate performs dot products on pairs of elements of y and z and accumulates the results to x. |
| - go: SaturatedPairDotProdAccumulate |
| commutative: "false" |
| extension: "AVX.*" |
| documentation: !string |- |
| // SaturatedPairDotProdAccumulate performs dot products on pairs of elements of y and z and accumulates the results to x. |
| - go: MaskedSaturatedPairDotProdAccumulate |
| masked: "true" |
| commutative: "false" |
| extension: "AVX.*" |
| documentation: !string |- |
| // MaskedSaturatedPairDotProdAccumulate performs dot products on pairs of elements of y and z and accumulates the results to x. |
| - go: MaskedFusedMultiplyAdd |
| masked: "true" |
| commutative: "false" |
| extension: "AVX.*" |
| documentation: !string |- |
| // MaskedFusedMultiplyAdd performs `(v1 * v2) + v3`. |
| - go: MaskedFusedMultiplyAddSub |
| masked: "true" |
| commutative: "false" |
| extension: "AVX.*" |
| documentation: !string |- |
| // MaskedFusedMultiplyAddSub performs `(v1 * v2) - v3` for odd-indexed elements, and `(v1 * v2) + v3` for even-indexed elements. |
| - go: MaskedFusedMultiplySubAdd |
| masked: "true" |
| commutative: "false" |
| extension: "AVX.*" |
| documentation: !string |- |
| // MaskedFusedMultiplySubAdd performs `(v1 * v2) + v3` for odd-indexed elements, and `(v1 * v2) - v3` for even-indexed elements. |
| - go: Max |
| commutative: "true" |
| extension: "AVX.*" |
| documentation: !string |- |
| // Max computes the maximum of corresponding elements. |
| - go: MaskedMax |
| commutative: "true" |
| masked: "true" |
| extension: "AVX.*" |
| documentation: !string |- |
| // MaskedMax computes the maximum of corresponding elements. |
| - go: Min |
| commutative: "true" |
| extension: "AVX.*" |
| documentation: !string |- |
| // Min computes the minimum of corresponding elements. |
| - go: MaskedMin |
| commutative: "true" |
| masked: "true" |
| extension: "AVX.*" |
| documentation: !string |- |
| // MaskedMin computes the minimum of corresponding elements. |
| - go: SetElem |
| commutative: "false" |
| extension: "AVX.*" |
| documentation: !string |- |
| // SetElem sets a single constant-indexed element's value. |
| - go: GetElem |
| commutative: "false" |
| extension: "AVX.*" |
| documentation: !string |- |
| // GetElem retrieves a single constant-indexed element's value. |
| |
| - go: Mul |
| commutative: "true" |
| extension: "AVX.*" |
| documentation: !string |- |
| // Mul multiplies corresponding elements of two vectors. |
| - go: MulEvenWiden |
| commutative: "true" |
| extension: "AVX.*" |
| documentation: !string |- |
| // MulEvenWiden multiplies even-indexed elements, widening the result. |
| // Result[i] = v1.Even[i] * v2.Even[i]. |
| - go: MulHigh |
| commutative: "true" |
| extension: "AVX.*" |
| documentation: !string |- |
| // MulHigh multiplies elements and stores the high part of the result. |
| - go: MulLow |
| commutative: "true" |
| extension: "AVX.*" |
| documentation: !string |- |
| // MulLow multiplies elements and stores the low part of the result. |
| - go: MaskedMul |
| masked: "true" |
| commutative: "true" |
| extension: "AVX.*" |
| documentation: !string |- |
| // MaskedMul multiplies corresponding elements of two vectors, masked. |
| docUnmasked: !string |- |
| // Mul multiplies corresponding elements of two vectors. |
| - go: MaskedMulEvenWiden |
| masked: "true" |
| commutative: "true" |
| extension: "AVX.*" |
| documentation: !string |- |
| // MaskedMulEvenWiden multiplies even-indexed elements, widening the result, masked. |
| // Result[i] = v1.Even[i] * v2.Even[i]. |
| docUnmasked: !string |- |
| // MulEvenWiden multiplies even-indexed elements, widening the result. |
| // Result[i] = v1.Even[i] * v2.Even[i]. |
| - go: MaskedMulHigh |
| masked: "true" |
| commutative: "true" |
| extension: "AVX.*" |
| documentation: !string |- |
| // MaskedMulHigh multiplies elements and stores the high part of the result, masked. |
| docUnmasked: !string |- |
| // MulHigh multiplies elements and stores the high part of the result. |
| - go: MaskedMulLow |
| masked: "true" |
| commutative: "true" |
| extension: "AVX.*" |
| documentation: !string |- |
| // MaskedMulLow multiplies elements and stores the low part of the result, masked. |
| docUnmasked: !string |- |
| // MulLow multiplies elements and stores the low part of the result. |
| - go: ShiftAllLeft |
| nameAndSizeCheck: "true" |
| commutative: "false" |
| extension: "AVX.*" |
| documentation: !string |- |
| // ShiftAllLeft shifts each element to the left by the specified number of bits. Emptied lower bits are zeroed. |
| - go: MaskedShiftAllLeft |
| nameAndSizeCheck: "true" |
| masked: "true" |
| commutative: "false" |
| extension: "AVX.*" |
| documentation: !string |- |
| // MaskedShiftAllLeft shifts each element to the left by the specified number of bits. Emptied lower bits are zeroed. |
| - go: ShiftAllRight |
| nameAndSizeCheck: "true" |
| commutative: "false" |
| extension: "AVX.*" |
| documentation: !string |- |
| // ShiftAllRight shifts each element to the right by the specified number of bits. Emptied upper bits are zeroed. |
| - go: MaskedShiftAllRight |
| nameAndSizeCheck: "true" |
| masked: "true" |
| commutative: "false" |
| extension: "AVX.*" |
| documentation: !string |- |
| // MaskedShiftAllRight shifts each element to the right by the specified number of bits. Emptied upper bits are zeroed. |
| - go: ShiftAllRightSignExtended |
| nameAndSizeCheck: "true" |
| commutative: "false" |
| extension: "AVX.*" |
| documentation: !string |- |
| // ShiftAllRightSignExtended shifts each element to the right by the specified number of bits. Emptied upper bits are filled with the sign bit. |
| - go: MaskedShiftAllRightSignExtended |
| nameAndSizeCheck: "true" |
| masked: "true" |
| commutative: "false" |
| extension: "AVX.*" |
| documentation: !string |- |
| // MaskedShiftAllRightSignExtended shifts each element to the right by the specified number of bits. Emptied upper bits are filled with the sign bit. |
| |
| - go: ShiftLeft |
| nameAndSizeCheck: "true" |
| commutative: "false" |
| extension: "AVX.*" |
| documentation: !string |- |
| // ShiftLeft shifts each element in x to the left by the number of bits specified in y's corresponding elements. Emptied lower bits are zeroed. |
| - go: MaskedShiftLeft |
| nameAndSizeCheck: "true" |
| masked: "true" |
| commutative: "false" |
| extension: "AVX.*" |
| documentation: !string |- |
| // MaskedShiftLeft shifts each element in x to the left by the number of bits specified in y's corresponding elements. Emptied lower bits are zeroed. |
| - go: ShiftRight |
| nameAndSizeCheck: "true" |
| commutative: "false" |
| extension: "AVX.*" |
| documentation: !string |- |
| // ShiftRight shifts each element in x to the right by the number of bits specified in y's corresponding elements. Emptied upper bits are zeroed. |
| - go: MaskedShiftRight |
| nameAndSizeCheck: "true" |
| masked: "true" |
| commutative: "false" |
| extension: "AVX.*" |
| documentation: !string |- |
| // MaskedShiftRight shifts each element in x to the right by the number of bits specified in y's corresponding elements. Emptied upper bits are zeroed. |
| - go: ShiftRightSignExtended |
| nameAndSizeCheck: "true" |
| commutative: "false" |
| extension: "AVX.*" |
| documentation: !string |- |
| // ShiftRightSignExtended shifts each element in x to the right by the number of bits specified in y's corresponding elements. Emptied upper bits are filled with the sign bit. |
| - go: MaskedShiftRightSignExtended |
| nameAndSizeCheck: "true" |
| masked: "true" |
| commutative: "false" |
| extension: "AVX.*" |
| documentation: !string |- |
| // MaskedShiftRightSignExtended shifts each element in x to the right by the number of bits specified in y's corresponding elements. Emptied upper bits are filled with the sign bit. |
| |
| - go: MaskedRotateAllLeft |
| nameAndSizeCheck: "true" |
| masked: "true" |
| commutative: "false" |
| extension: "AVX.*" |
| documentation: !string |- |
| // MaskedRotateAllLeft rotates each element to the left by the number of bits specified by the immediate. |
| - go: MaskedRotateLeft |
| nameAndSizeCheck: "true" |
| masked: "true" |
| commutative: "false" |
| extension: "AVX.*" |
| documentation: !string |- |
| // MaskedRotateLeft rotates each element in x to the left by the number of bits specified by y's corresponding elements. |
| - go: MaskedRotateAllRight |
| nameAndSizeCheck: "true" |
| masked: "true" |
| commutative: "false" |
| extension: "AVX.*" |
| documentation: !string |- |
| // MaskedRotateAllRight rotates each element to the right by the number of bits specified by the immediate. |
| - go: MaskedRotateRight |
| nameAndSizeCheck: "true" |
| masked: "true" |
| commutative: "false" |
| extension: "AVX.*" |
| documentation: !string |- |
| // MaskedRotateRight rotates each element in x to the right by the number of bits specified by y's corresponding elements. |
| |
| - go: MaskedShiftAllLeftAndFillUpperFrom |
| nameAndSizeCheck: "true" |
| masked: "true" |
| commutative: "false" |
| extension: "AVX.*" |
| documentation: !string |- |
| // MaskedShiftAllLeftAndFillUpperFrom shifts each element of x to the left by the number of bits specified by the |
| // immediate(only the lower 5 bits are used), and then copies the upper bits of y to the emptied lower bits of the shifted x. |
| - go: MaskedShiftAllRightAndFillUpperFrom |
| nameAndSizeCheck: "true" |
| masked: "true" |
| commutative: "false" |
| extension: "AVX.*" |
| documentation: !string |- |
| // MaskedShiftAllRightAndFillUpperFrom shifts each element of x to the right by the number of bits specified by the |
| // immediate(only the lower 5 bits are used), and then copies the lower bits of y to the emptied upper bits of the shifted x. |
| - go: MaskedShiftLeftAndFillUpperFrom |
| nameAndSizeCheck: "true" |
| masked: "true" |
| commutative: "false" |
| extension: "AVX.*" |
| documentation: !string |- |
| // MaskedShiftLeftAndFillUpperFrom shifts each element of x to the left by the number of bits specified by the |
| // corresponding elements in y(only the lower 5 bits are used), and then copies the upper bits of z to the emptied lower bits of the shifted x. |
| - go: MaskedShiftRightAndFillUpperFrom |
| nameAndSizeCheck: "true" |
| masked: "true" |
| commutative: "false" |
| extension: "AVX.*" |
| documentation: !string |- |
| // MaskedShiftRightAndFillUpperFrom shifts each element of x to the right by the number of bits specified by the |
| // corresponding elements in y(only the lower 5 bits are used), and then copies the lower bits of z to the emptied upper bits of the shifted x. |
