Loading...
Searching...
No Matches
Argon< ScalarType > Class Template Reference
A 128-bit SIMD vector wrapping a scalar type, providing arithmetic, logical, and data-movement operations. More...
#include <argon_full.hpp>
Inheritance diagram for Argon< ScalarType >:
Collaboration diagram for Argon< ScalarType >:
Public Types | |
| using | vector_type = simd::Vec128_t<ScalarType> |
| using | lane_type = const argon::Lane<vector_type> |
| Public Types inherited from argon::Vector< simd::Vec128_t< ScalarType > > | |
| using | const_lane_type |
| The type of a single lane of the SIMD vector. | |
| using | lane_type |
| The type of a single lane of the SIMD vector. | |
| using | scalar_type |
| The scalar type of the SIMD vector. | |
| using | vector_type |
| The SIMD vector type. | |
| using | argon_type |
| The Argon type for the SIMD vector. | |
| using | predicate_type |
| The type of a boolean SIMD vector. | |
| using | argon_bool_type |
| The Argon type for the boolean vector. | |
Public Member Functions | |
| ace | Argon (argon::Vector< vector_type > vec) |
| Construct from an underlying argon::Vector. | |
| ace | Argon (std::array< ScalarType, 4 > value_list) |
| Construct from a four-element array (loaded as a 128-bit vector). | |
| ace | Argon (ArgonHalf< ScalarType > low, ArgonHalf< ScalarType > high) |
| Construct by combining a low and high 64-bit half-vector. | |
| ace | Argon (argon::Lane< vector_type > b) |
| ace | Argon (argon::ConstLane< 0, vector_type > b) |
| template<simd::is_vector_type intrinsic_type> | |
| ace | Argon (argon::Lane< intrinsic_type > b) |
| template<typename new_scalar_type> | |
| ace Argon< new_scalar_type > | As () const |
| Reinterpret the vector bits as a vector of a different element type. | |
| ace Argon< ScalarType > | Reverse () |
| Reverse the order of all elements across the full 128-bit vector. | |
| template<typename U> | |
| ace Argon< ScalarType > | MultiplyAddLong (ArgonHalf< U > b, ArgonHalf< U > c) |
| Multiply two narrower half-vectors and add the widened result to this vector (vector × vector). | |
| template<typename U, typename C> | |
| ace Argon< ScalarType > | MultiplyAddLong (ArgonHalf< U > b, C c) |
| Multiply two narrower vectors and add the widened result (vector × raw intrinsic). | |
| template<typename U> | |
| ace Argon< ScalarType > | MultiplyAddLong (ArgonHalf< U > b, U c) |
| Multiply a half-vector by a scalar and add the widened result (vector × scalar). | |
| template<typename U> | |
| ace Argon< ScalarType > | MultiplyAddLong (ArgonHalf< U > b, typename ArgonHalf< U >::lane_type c) |
| Multiply a half-vector by a lane and add the widened result (vector × lane). | |
| template<typename U> | |
| ace Argon< ScalarType > | MultiplySubtractLong (ArgonHalf< U > b, ArgonHalf< U > c) |
| Multiply two narrower half-vectors and subtract the widened result from this vector (vector × vector). | |
| template<typename U> | |
| ace Argon< ScalarType > | MultiplySubtractLong (ArgonHalf< U > b, U c) |
| Multiply a half-vector by a scalar and subtract the widened result (vector × scalar). | |
| template<typename U> | |
| ace Argon< ScalarType > | MultiplySubtractLong (ArgonHalf< U > b, typename ArgonHalf< U >::lane_type c) |
| Multiply a half-vector by a lane and subtract the widened result (vector × lane). | |
| ace auto | AddNarrow (Argon< ScalarType > b) const |
| Add and narrow: add b to this vector and truncate each lane to the next-smaller type. | |
| ace auto | AddRoundNarrow (Argon< ScalarType > b) const |
| Add, round, and narrow: add b, round the result, and truncate to the next-smaller type. | |
| ace auto | SubtractNarrow (Argon< ScalarType > b) const |
| Subtract and narrow: subtract b and truncate each lane to the next-smaller type. | |
| ace auto | SubtractRoundNarrow (Argon< ScalarType > b) const |
| Subtract, round, and narrow: subtract b, round the result, and truncate to the next-smaller type. | |
| template<size_t n> | |
| ace auto | ShiftRightNarrow () const |
| Shift each lane right by n bits and narrow the result to the next-smaller type. | |
| template<size_t n> | |
| ace auto | ShiftRightSaturateNarrow () const |
| Shift right, saturate, and narrow: shift each lane right by n bits with unsigned saturation. | |
| template<size_t n> | |
| ace auto | ShiftRightRoundSaturateNarrow () const |
| Shift right, round, saturate, and narrow. | |
| template<size_t n> | |
| ace auto | ShiftRightRoundNarrow () const |
| Shift right, round, and narrow. | |
| ace auto | Narrow () const |
| Truncate each lane to the next-smaller element type (no saturation). | |
| ace auto | SaturateNarrow () const |
| Saturate and narrow each lane to the next-smaller element type. | |
| template<typename NextSmallerType> | |
| ace Argon< ScalarType > | MultiplyDoubleAddSaturateLong (ArgonHalf< NextSmallerType > b, ArgonHalf< NextSmallerType > c) |
| Multiply, double, add, and saturate long: this + saturate(2 * b * c) widening to ScalarType. | |
| ace ArgonHalf< ScalarType > | GetHigh () const |
| Return the upper 64 bits as an ArgonHalf. | |
| ace ArgonHalf< ScalarType > | GetLow () const |
| Return the lower 64 bits as an ArgonHalf. | |
| template<typename U> | |
| ace Argon< U > | ConvertTo () const |
| Convert each lane to a different element type. | |
| template<typename U, int fracbits> | |
| ace Argon< U > | ConvertTo () const |
| Convert each lane to a different type using a fixed-point fractional bit count. | |
| ace Argon< ScalarType > | Reverse () const |
| Reverse the order of all elements in the 128-bit vector. | |
| template<typename CommutableOpType> | |
| ScalarType | Reduce (CommutableOpType op) |
| Fold all lanes into a single scalar using a commutative binary operation. | |
| ScalarType | ReduceAdd () |
| Sum all lanes and return the scalar result. | |
| ScalarType | ReduceMax () |
| Return the maximum value across all lanes. | |
| ScalarType | ReduceMin () |
| Return the minimum value across all lanes. | |
| ace Argon< ScalarType > | SwapDoublewords () |
| Swap the upper and lower 64-bit halves of the vector. | |
| Public Member Functions inherited from argon::Vector< simd::Vec128_t< ScalarType > > | |
| constexpr | Vector ()=default |
| The default constructor for the Vector class. | |
| constexpr | Vector (Vector &&other)=default |
| Move constructor for the Vector class. | |
| constexpr | Vector (const Vector &other)=default |
| Copy constructor for the Vector class. | |
| constexpr Vector & | operator= (Vector &&other)=default |
| Move assignment operator for the Vector class. | |
| constexpr Vector & | operator= (const Vector &other)=default |
| Copy assignment operator for the Vector class. | |
| ace | Vector (simd::Vec128_t< ScalarType > vector) |
| Constructs a Vector from a SIMD vector type. | |
| ace | Vector (scalar_type scalar) |
| Constructs a Vector from a scalar value. | |
| ace | Vector (argon::Lane< simd::Vec128_t< ScalarType > > lane) |
| Constructs a Vector from a Lane object. | |
| ace | Vector (argon::ConstLane< LaneIndex, simd::Vec128_t< ScalarType > > lane) |
| Constructs a Vector from a ConstLane object. | |
| ace | Vector (ArgTypes... args) |
| ace argon_type | operator- () const |
| Negate the SIMD vector and return the result. | |
| ace argon_type | operator+ (argon_type b) const |
| Add a vector and return the result. | |
| ace argon_type | operator- (argon_type b) const |
| Subtract a vector and return the result. | |
| ace argon_type | operator* (argon_type b) const |
| Multiply a vector and return the result. | |
| ace argon_type | operator/ (argon_type b) const |
| Divide a vector and return the result. | |
| ace argon_bool_type | operator== (argon_type b) const |
| Compare two vectors for equality. | |
| ace argon_bool_type | operator!= (argon_type b) const |
| Compare two vectors for inequality. | |
| ace argon_bool_type | operator< (argon_type b) const |
| Compare two vectors, checking if this vector is less than the other. | |
| ace argon_bool_type | operator> (argon_type b) const |
| Compare two vectors, checking if this vector is greater than the other. | |
| ace argon_bool_type | operator<= (argon_type b) const |
| Compare two vectors, checking if this vector is less than or equal to the other. | |
| ace argon_bool_type | operator>= (argon_type b) const |
| Compare two vectors, checking if this vector is greater than or equal to the other. | |
| ace argon_type | operator++ () const |
| Increment the vector by 1 and return the result. | |
| ace argon_type | operator-- () const |
| Decrement the vector by 1 and return the result. | |
| ace argon_type | operator& (argon_type b) const |
| Bitwise AND two vectors and return the result. | |
| ace argon_type | operator| (argon_type b) const |
| Bitwise OR two vectors and return the result. | |
| ace argon_type | operator^ (argon_type b) const |
| Bitwise XOR two vectors and return the result. | |
| ace argon_type | operator~ () const |
| Bitwise NOT the vector and return the result. | |
| ace Lane< const simd::Vec128_t< ScalarType > > | operator[] (const size_t i) const |
| Access a lane of the vector by index. | |
| ace lane_type | operator[] (const size_t i) |
| Access a lane of the vector by index. | |
| ace argon_type | operator>> (const int i) const |
| Shift the elements of the vector to the right by a specified number of bits. | |
| ace argon_type | operator<< (const int i) const |
| Shift the elements of the vector to the left by a specified number of bits. | |
| constexpr simd::Vec128_t< ScalarType > | vec () const |
| Get the underlying SIMD vector. | |
| constexpr | operator simd::Vec128_t< ScalarType > () const |
| Convert the vector to the underlying SIMD vector type. | |
| ace std::array< scalar_type, lanes > | to_array () |
| Convert the vector to an array of scalar values. | |
| ace const lane_type | GetLane (const size_t i) const |
| Get a single lane of the vector by index. | |
| ace lane_type | GetLane (const size_t i) |
| ace const lane_type | GetLane (const int i) const |
| Get a single lane of the vector by index. | |
| ace lane_type | GetLane (const int i) |
| ace const const_lane_type< LaneIndex > | GetLane () const |
| Get a single lane of the vector by index. | |
| ace const_lane_type< LaneIndex > | GetLane () |
| ace const_lane_type< lanes - 1 > | LastLane () |
| Get the last lane of the vector. | |
| ace argon_type | ShiftRight (const int i) const |
| Shift the elements of the vector to the right by a specified number of bits. | |
| ace argon_type | ShiftLeft (const int i) const |
| Shift the elements of the vector to the left by a specified number of bits. | |
| ace argon_type | Negate () const |
| Bitwise negate the vector and return the result. | |
| ace argon_type | Add (argon_type b) const |
| Add two vectors. | |
| ace argon_type | AddHalve (argon_type b) const |
| Adds two vectors, halving the result. | |
| ace argon_type | AddHalveRound (argon_type b) const |
| Adds two vectors, halving and rounding the result. | |
| ace argon_type | AddSaturate (argon_type b) const |
| Adds two vectors, saturating the result. | |
| ace argon_type | Subtract (argon_type b) const |
| Subtract two vectors. | |
| ace argon_type | SubtractHalve (argon_type b) const |
| Subtract two vectors, halving the result. | |
| ace argon_type | SubtractSaturate (argon_type b) const |
| Subtract two vectors, saturating the result. | |
| ace argon_type | SubtractAbs (argon_type b) const |
| Subtract two vectors, taking the absolute value of the result. | |
| ace argon_type | SubtractAbsAdd (argon_type b, argon_type c) const |
| Subtract two vectors, taking the absolute value of the result and adding a third vector. | |
| ace argon_type | Multiply (argon_type b) const |
| Multiply two vectors. | |
| ace argon_type | Multiply (scalar_type b) const |
| Multiply a vector by a scalar value. | |
| ace argon_type | Multiply (lane_type b) const |
| Multiply a vector by a lane value. | |
| ace argon_type | Multiply (const_lane_type< LaneIndex > b) const |
| Multiply a vector by a lane value. | |
| ace argon_type | MultiplyAdd (argon_type b, argon_type c) const |
| Multiply two vectors and add a third vector. | |
| ace argon_type | MultiplyAdd (argon_type b, scalar_type c) const |
| Multiply a vector by a scalar value and add a third vector. | |
| ace argon_type | MultiplyAdd (scalar_type b, argon_type c) const |
| Multiply a vector by a scalar value and add a third vector. | |
| ace argon_type | MultiplyAdd (argon_type b, lane_type c) const |
| Multiply a vector by a lane value and add a third vector. | |
| ace argon_type | MultiplyAdd (lane_type b, argon_type c) const |
| Multiply a vector by a lane value and add a third vector. | |
| ace argon_type | MultiplyAdd (argon_type b, const_lane_type< LaneIndex > c) const |
| Multiply a vector by a lane value and add a third vector. | |
| ace argon_type | MultiplyAdd (const_lane_type< LaneIndex > b, argon_type c) const |
| Multiply a vector by a lane value and add a third vector. | |
| ace argon_type | MultiplySubtract (argon_type b, argon_type c) const |
| Multiply two vectors and subtract from a third vector. | |
| ace argon_type | MultiplySubtract (argon_type b, scalar_type c) const |
| Multiply a vector by a scalar value and subtract from a third vector. | |
| ace argon_type | MultiplySubtract (scalar_type b, argon_type c) const |
| Multiply a vector by a scalar value and subtract from a third vector. | |
| ace argon_type | MultiplySubtract (argon_type b, lane_type c) const |
| Multiply a vector by a lane value and subtract from a third vector. | |
| ace argon_type | MultiplyFixedQMax (argon_type v) const |
| Multiply two QMax fixed-point vectors, returning a fixed-point product. | |
| ace argon_type | MultiplyFixedQMax (scalar_type s) const |
| Multiply a QMax fixed-point vector by a scalar value, returning a fixed-point product. | |
| ace argon_type | MultiplyFixedQMax (lane_type l) const |
| Multiply a QMax fixed-point vector by a lane value, returning a fixed-point product. | |
| ace argon_type | MultiplyRoundFixedQMax (argon_type v) const |
| Multiply two fixed-point vectors, returning a fixed-point product. | |
| ace argon_type | MultiplyRoundFixedQMax (scalar_type s) const |
| Multiply a fixed-point vector by a scalar value, returning a fixed-point product. | |
| ace argon_type | MultiplyRoundFixedQMax (lane_type l) const |
| Multiply a fixed-point vector by a lane value, returning a fixed-point product. | |
| ace argon_type | Absolute () const |
| Get the absolute value of the vector. | |
| ace argon_type | ReciprocalEstimate () const |
| 1 / value, using an estimate for speed | |
| ace argon_type | ReciprocalSqrtEstimate () const |
| 1 / sqrt(value), using an estimate for speed | |
| ace argon_type | ReciprocalStep (argon_type b) const |
| Newton-Raphson step for reciprocal refinement: (2 - a * b) / 2. | |
| ace argon_type | ReciprocalSqrtStep (argon_type b) const |
| Newton-Raphson step for reciprocal-sqrt refinement: (3 - a * b) / 2. | |
| ace argon_type | ReciprocalEstimateRefine (int n_iters=1) const |
| Compute a refined reciprocal estimate using Newton-Raphson iterations. | |
| ace argon_type | ReciprocalSqrtEstimateRefine (int n_iters=1) const |
| Compute a refined reciprocal-sqrt estimate using Newton-Raphson iterations. | |
| ace argon_type | MultiplyAddFixedQMax (argon_type b, arg_type c) const |
| Multiply-add three fixed-point vectors, returning a fixed-point sum. | |
| ace argon_type | MultiplyRoundAddFixedQMax (argon_type b, arg_type c) const |
| Multiply-round-add three fixed-point vectors, returning a fixed-point sum. | |
| ace argon_type | Divide (argon_type b) const |
| Divide two vectors. | |
| ace argon_type | Modulo (argon_type b) const |
| Get the modulo of two vectors. | |
| ace argon_type | Modulo (scalar_type b) const |
| Get the modulo of a vector and a scalar value. | |
| ace argon_type | Max (argon_type b) const |
| Compare the lanes of two vectors, copying the larger of each lane to the result. | |
| ace argon_type | Min (argon_type b) const |
| Compare the lanes of two vectors, copying the smaller of each lane to the result. | |
| ace argon_bool_type | Equal (argon_type b) const |
| Compare the lanes of two vectors, setting the result lane's bits to ON if are equal. | |
| ace argon_bool_type | GreaterThanOrEqual (argon_type b) const |
| Compare the lanes of two vectors, setting the result lane's bits to ON if a is greater than or equal to b. | |
| ace argon_bool_type | LessThanOrEqual (argon_type b) const |
| Compare the lanes of two vectors, setting the result lane's bits to ON if a is less than or equal to b. | |
| ace argon_bool_type | GreaterThan (argon_type b) const |
| Compare the lanes of two vectors, setting the result lane's bits to ON if a is greater than b. | |
| ace argon_bool_type | LessThan (argon_type b) const |
| Compare the lanes of two vectors, setting the result lane's bits to ON if a is less than b. | |
| ace argon_type | ShiftLeft (helpers::ArgonFor_t< simd::make_signed_t< Bool_t< simd::Vec128_t< ScalarType > > > > b) const |
| Shift the elemnets of the vector to the left by a specified number of bits. | |
| ace argon_type | ShiftLeft (std::make_signed_t< simd::Scalar_t< Bool_t< simd::Vec128_t< ScalarType > > > > n) const |
| Shift the elements of the vector to the left by a specified number of bits. | |
| ace argon_type | ShiftLeft () const |
| Shift the elements of the vector to the left by a specified number of bits. | |
| ace argon_type | ShiftLeftSaturate (helpers::ArgonFor_t< simd::make_signed_t< Bool_t< simd::Vec128_t< ScalarType > > > > b) const |
| Shift the elements of the vector to the left by a specified number of bits, saturating the result. | |
| ace argon_type | ShiftLeftRound (argon_type b) const |
| Shift the elements of the vector to the left by a specified number of bits, rounding the result. | |
| ace argon_type | ShiftLeftRoundSaturate (argon_type b) const |
| Shift the elements of the vector to the left by a specified number of bits, rounding and saturating the result. | |
| ace argon_type | ShiftLeftSaturate () const |
| Shift the elements of the vector to the left by a specified number of bits, saturating the result. | |
| ace argon_type | ShiftLeftInsert (argon_type b) const |
| Shift the elements of the vector to the left by a specified number of bits, and then OR the result with another vector masked to the number of shift bits. | |
| ace argon_type | ShiftRight () const |
| Shift the elements of the vector to the right by a specified number of bits. | |
| ace argon_type | ShiftRightRound () const |
| Shift the elements of the vector to the right by a specified number of bits, rounding the result. | |
| ace argon_type | ShiftRightAccumulate (argon_type b) const |
| Shift the elements of the b vector to the right by a specified number of bits, and then add the result to this vector. | |
| ace argon_type | ShiftRightAccumulateRound (argon_type b) const |
| Shift the elements of the b vector to the right by a specified number of bits, and then add the result to this vector. | |
| ace argon_type | ShiftRightInsert (argon_type b) const |
| Shift the elements of the vector to the right by a specified number of bits, ORing the result with the vector masked to the number of shift bits. | |
| ace argon_type | LoadToLane (const scalar_type *ptr) |
| Load a lane from a pointer. | |
| ace void | StoreTo (scalar_type *ptr) const |
| Store the vector to a pointer. | |
| ace void | StoreLaneTo (scalar_type *ptr) |
| Store a lane of the vector to a pointer. | |
| ace argon_type | PairwiseAdd (argon_type b) const |
| Pairwise ops. | |
| ace argon_type | PairwiseMax (argon_type b) const |
| Select the maximum of each pair of lanes in the two vectors. | |
| ace argon_type | PairwiseMin (argon_type b) const |
| Select the maximum of each pair of lanes in the two vectors. | |
| ace argon_type | BitwiseNot () const |
| Bitwise ops. | |
| ace argon_type | BitwiseAnd (argon_type b) const |
| Bitwise AND of the vector with another vector. | |
| ace argon_type | BitwiseOr (argon_type b) const |
| Bitwise OR of the vector with another vector. | |
| ace argon_type | BitwiseXor (argon_type b) const |
| Bitwise XOR of the vector with another vector. | |
| ace argon_type | BitwiseOrNot (argon_type b) const |
| Bitwise OR of the vector with the NOT of another vector. | |
| ace argon_type | BitwiseAndNot (argon_type b) const |
| Bitwise AND of the vector with the NOT of another vector. | |
| ace argon_type | BitwiseClear (argon_type b) const |
| Bitwise AND of the vector with the NOT of another vector. | |
| ace argon_type | BitwiseSelect (ArgType true_value, ArgType false_value) const |
| Bitwise select between two vectors, using the current vector as a mask. | |
| ace argon_type | Select (ArgType true_value, ArgType false_value) const |
| Bitwise select between two vectors, using the current vector as a mask. | |
| ace predicate_type | CompareTestNonzero (argon_type b) const |
| Ands the current vector with the given vector, then checks if nonzero. | |
| ace predicate_type | TestNonzero () const |
| Ands the current vector with the given vector, then checks if nonzero. | |
| ace helpers::ArgonFor_t< simd::make_signed_t< Bool_t< simd::Vec128_t< ScalarType > > > > | CountLeadingSignBits () const |
| Count the number of consecutive bits following the sign bit that are set to the same value as the sign bit. | |
| ace argon_type | CountLeadingZeroBits () const |
| Count the number of consecutive top bits that are set to zero. | |
| ace argon_type | CountActiveBits () const |
| Count the number of bits that are set to one in the vector. | |
| ace argon_type | Popcount () const |
| Count the number of bits that are set to one in the vector. | |
| ace argon_type | Extract (argon_type b) const |
| Extract n elements from the lower end of the operand, and the remaining elements from the top end of this vector, combining them into the result vector. | |
| ace argon_type | Reverse64bit () const |
| ace argon_type | Reverse32bit () const |
| ace argon_type | Reverse16bit () const |
| ace std::array< argon_type, 2 > | ZipWith (argon_type b) const |
| Zip two vectors together, returning two vectors of pairs. | |
| std::array< argon_type, 2 > | UnzipWith (argon_type b) |
| Unzip two vectors, returning two vectors of pairs. | |
| std::array< argon_type, 2 > | TransposeWith (argon_type b) const |
| Perform a 2x2 matrix transpose on two vectors, returning two vectors of pairs. | |
| ace argon_type | map (FuncType body) const |
| ace argon_type | map_with_index (FuncType body) const |
| ace argon_type | map2 (argon_type other, FuncType body) const |
| ace argon_type | each_lane (FuncType body) |
| ace argon_type | each_lane_with_index (FuncType body) |
| ace void | if_lane (FuncType true_branch) |
| ace void | if_else_lane (FuncType true_branch, FuncType false_branch) |
| ace void | if_lane_with_index (FuncType true_branch) |
| ace void | if_else_lane_with_index (FuncType1 true_branch, FuncType2 false_branch) |
| ace bool | any () |
| ace bool | all () |
| std::tuple_element_t< Index, argon_type > | get () |
| constexpr | operator simd::Vec128_t< ScalarType > () const |
| Convert the vector to the underlying SIMD vector type. | |
Static Public Member Functions | |
| static ace Argon< ScalarType > | Combine (ArgonHalf< ScalarType > low, ArgonHalf< ScalarType > high) |
| Combine two 64-bit half-vectors into a single 128-bit vector. | |
| Static Public Member Functions inherited from argon::Vector< simd::Vec128_t< ScalarType > > | |
| static ace argon_type | LoadScalar (const scalar_type *ptr) |
| Constructs a Vector from a scalar pointer. | |
| static ace argon_type | FromScalar (scalar_type scalar) |
| Constructs a Vector from a scalar value. | |
| static ace argon_type | FromLane (argon::Lane< IntrinsicType > lane) |
| Constructs a Vector from a Lane object. | |
| static ace argon_type | FromLane (argon::ConstLane< LaneIndex, simd::Vec128_t< ScalarType > > lane) |
| Constructs a Vector from a ConstLane object. | |
| static ace argon_type | Iota (scalar_type start) |
| Constructs a Vector from an incrementing sequence. | |
| static ace argon_type | Generate (FuncType body) |
| Constructs a Vector from a function that generates values. | |
| static ace argon_type | GenerateWithIndex (FuncType body) |
| Constructs a Vector from a function that generates values with an index. | |
| static ace argon_type | Load (const scalar_type *ptr) |
| Load a vector from a pointer. | |
| static ace argon_type | LoadCopy (const scalar_type *ptr) |
| Load a vector from a pointer, duplicating the value across all lanes. | |
| static ace argon_type | LoadGatherOffsetBytes (const scalar_type *base, helpers::ArgonFor_t< simd::make_unsigned_t< Bool_t< simd::Vec128_t< ScalarType > > > > offset_vector) |
| Using a base address and a vector of offset bytes and a base pointer, create a new vector. | |
| static ace argon_type | LoadGatherOffsetIndex (const scalar_type *base, helpers::ArgonFor_t< simd::make_unsigned_t< Bool_t< simd::Vec128_t< ScalarType > > > > offset_vector) |
| Using a base address and a vector of offset indices and a base pointer, create a new vector. | |
| static ace std::array< argon_type, stride > | LoadInterleaved (const scalar_type *ptr) |
| Load multiple vectors from a pointer, de-interleaving. | |
| static ace std::array< argon_type, stride > | LoadCopyInterleaved (const scalar_type *ptr) |
| Load multiple vectors from a pointer, duplicating the value across all lanes. | |
| static ace std::array< argon_type, Stride > | LoadToLaneInterleaved (simd::MultiVector_t< simd::Vec128_t< ScalarType >, Stride > multi, const scalar_type *ptr) |
| Load a value from a pointer into a vector at the lane index lane, de-interleaving. | |
| static ace std::array< argon_type, stride > | LoadToLaneInterleaved (std::array< argon_type, stride > multi, const scalar_type *ptr) |
| Load a value from a pointer into a vector at the lane index lane, de-interleaving. | |
| static ace std::array< argon_type, stride > | LoadGatherOffsetIndexInterleaved (const scalar_type *base_ptr, helpers::ArgonFor_t< simd::make_unsigned_t< Bool_t< simd::Vec128_t< ScalarType > > > > offset_vector) |
| Perform a Load-Gather of interleaved elements. | |
| static ace std::array< argon_type, n > | LoadMulti (const scalar_type *ptr) |
| Load n vectors from a single contiguous set of memory. | |
| static ace int | size () |
| Get the number of elements. | |
Static Public Attributes | |
| static constexpr size_t | bytes = 16 |
| static constexpr size_t | lanes = bytes / sizeof(ScalarType) |
| Static Public Attributes inherited from argon::Vector< simd::Vec128_t< ScalarType > > | |
| static constexpr size_t | lanes |
| The number of lanes in the SIMD vector. | |
Additional Inherited Members | |
| Static Protected Member Functions inherited from argon::Vector< simd::Vec128_t< ScalarType > > | |
| static ace argon_type | IotaHelper (scalar_type start, std::index_sequence< Ints... >) |
| Protected Attributes inherited from argon::Vector< simd::Vec128_t< ScalarType > > | |
| simd::Vec128_t< ScalarType > | vec_ |
Detailed Description
template<typename ScalarType>
requires std::is_arithmetic_v<ScalarType>
class Argon< ScalarType >
requires std::is_arithmetic_v<ScalarType>
class Argon< ScalarType >
A 128-bit SIMD vector wrapping a scalar type, providing arithmetic, logical, and data-movement operations.
- Template Parameters
-
ScalarType The element type (e.g., int32_t, float, uint8_t).
Inherits all lane-wise operations from argon::Vector and adds 128-bit-specific operations such as narrowing, widening multiplies, cross-half operations, reductions, and (optionally) AES acceleration. Use ArgonHalf<ScalarType> for the 64-bit sibling.
Member Function Documentation
◆ AddNarrow()
template<typename ScalarType>
|
inline |
Add and narrow: add b to this vector and truncate each lane to the next-smaller type.
- Returns
- An ArgonHalf-width vector of the next-smaller element type.
◆ As()
template<typename ScalarType>
template<typename new_scalar_type>
|
inline |
Reinterpret the vector bits as a vector of a different element type.
- Template Parameters
-
new_scalar_type The target element type.
- Returns
- The same 128 bits viewed as Argon<new_scalar_type>.
◆ Combine()
template<typename ScalarType>
|
inlinestatic |
Combine two 64-bit half-vectors into a single 128-bit vector.
- Parameters
-
low The lower 64 bits. high The upper 64 bits.
◆ ConvertTo() [1/2]
template<typename ScalarType>
template<typename U, int fracbits>
|
inline |
Convert each lane to a different type using a fixed-point fractional bit count.
- Template Parameters
-
U The destination element type (float, int32_t, or uint32_t). fracbits The number of fractional bits in the fixed-point representation.
◆ ConvertTo() [2/2]
template<typename ScalarType>
template<typename U>
|
inline |
Convert each lane to a different element type.
- Template Parameters
-
U The destination element type.
◆ MultiplyAddLong()
template<typename ScalarType>
template<typename U>
|
inline |
Multiply two narrower half-vectors and add the widened result to this vector (vector × vector).
- Template Parameters
-
U The narrower element type (must be the next-smaller type of ScalarType).
◆ Reduce()
template<typename ScalarType>
template<typename CommutableOpType>
|
inline |
◆ Reverse() [1/2]
Reverse the order of all elements across the full 128-bit vector.
Reverses within each 64-bit half, then swaps the two halves.
◆ Reverse() [2/2]
Reverse the order of all elements in the 128-bit vector.
Reverses elements within each 64-bit doubleword, then swaps the two doublewords.
◆ ShiftRightNarrow()
template<typename ScalarType>
template<size_t n>
|
inline |
Shift each lane right by n bits and narrow the result to the next-smaller type.
- Template Parameters
-
n Shift amount (must be in range [1, sizeof(ScalarType)*8]).
◆ ShiftRightRoundNarrow()
template<typename ScalarType>
template<size_t n>
|
inline |
Shift right, round, and narrow.
- Template Parameters
-
n Shift amount.
◆ ShiftRightRoundSaturateNarrow()
template<typename ScalarType>
template<size_t n>
|
inline |
Shift right, round, saturate, and narrow.
- Template Parameters
-
n Shift amount.
◆ ShiftRightSaturateNarrow()
template<typename ScalarType>
template<size_t n>
|
inline |
Shift right, saturate, and narrow: shift each lane right by n bits with unsigned saturation.
- Template Parameters
-
n Shift amount.
The documentation for this class was generated from the following file:
- include/argon/argon_full.hpp
