JUCE
|
A wrapper around the platform's native SIMD register type. More...
Public Types | |
typedef Type | ElementType |
The type that represents the individual constituents of the SIMD Register. More... | |
typedef ElementType | value_type |
STL compatible value_type definition (same as ElementType). More... | |
typedef SIMDInternal::MaskTypeFor< ElementType >::type | MaskType |
The corresponding primitive integer type, for example, this will be int32_t if type is a float. More... | |
typedef SIMDInternal::PrimitiveType< ElementType >::type | PrimitiveType |
The native primitive type (used internally). More... | |
typedef SIMDNativeOps< PrimitiveType > | NativeOps |
The native operations for this platform and type combination (used internally) More... | |
typedef NativeOps::vSIMDType | vSIMDType |
The native type (used internally). More... | |
typedef SIMDRegister< MaskType > | vMaskType |
The corresponding integer SIMDRegister type (used internally). More... | |
typedef vMaskType::vSIMDType | vMaskSIMDType |
The internal native type for the corresponding mask type (used internally). More... | |
typedef CmplxSIMDOps< ElementType > | CmplxOps |
Wrapper for operations which need to be handled differently for complex and scalar types (used internally). More... | |
Public Member Functions | |
SIMDRegister () noexcept | |
Default constructor. More... | |
SIMDRegister (vSIMDType a) noexcept | |
Constructs an object from the native SIMD type. More... | |
SIMDRegister (Type s) noexcept | |
Constructs an object from a scalar type by broadcasting it to all elements. More... | |
~SIMDRegister () noexcept | |
Destrutor. More... | |
void JUCE_VECTOR_CALLTYPE | copyToRawArray (ElementType *a) const noexcept |
Copies the elements of the SIMDRegister to a scalar array in memory. More... | |
ElementType JUCE_VECTOR_CALLTYPE | operator[] (size_t idx) const noexcept |
Returns the idx-th element of the receiver. More... | |
ElementType &JUCE_VECTOR_CALLTYPE | operator[] (size_t idx) noexcept |
Returns the idx-th element of the receiver. More... | |
SIMDRegister &JUCE_VECTOR_CALLTYPE | operator+= (SIMDRegister v) noexcept |
Adds another SIMDRegister to the receiver. More... | |
SIMDRegister &JUCE_VECTOR_CALLTYPE | operator-= (SIMDRegister v) noexcept |
Subtracts another SIMDRegister to the receiver. More... | |
SIMDRegister &JUCE_VECTOR_CALLTYPE | operator*= (SIMDRegister v) noexcept |
Multiplies another SIMDRegister to the receiver. More... | |
SIMDRegister &JUCE_VECTOR_CALLTYPE | operator= (ElementType s) noexcept |
Broadcasts the scalar to all elements of the receiver. More... | |
SIMDRegister &JUCE_VECTOR_CALLTYPE | operator+= (ElementType s) noexcept |
Adds a scalar to the receiver. More... | |
SIMDRegister &JUCE_VECTOR_CALLTYPE | operator-= (ElementType s) noexcept |
Subtracts a scalar to the receiver. More... | |
SIMDRegister &JUCE_VECTOR_CALLTYPE | operator*= (ElementType s) noexcept |
Multiplies a scalar to the receiver. More... | |
SIMDRegister &JUCE_VECTOR_CALLTYPE | operator &= (vMaskType v) noexcept |
Bit-and the reciver with SIMDRegister v and store the result in the receiver. More... | |
SIMDRegister &JUCE_VECTOR_CALLTYPE | operator|= (vMaskType v) noexcept |
Bit-or the reciver with SIMDRegister v and store the result in the receiver. More... | |
SIMDRegister &JUCE_VECTOR_CALLTYPE | operator^= (vMaskType v) noexcept |
Bit-xor the reciver with SIMDRegister v and store the result in the receiver. More... | |
SIMDRegister &JUCE_VECTOR_CALLTYPE | operator &= (MaskType s) noexcept |
Bit-and each element of the reciver with the scalar s and store the result in the receiver. More... | |
SIMDRegister &JUCE_VECTOR_CALLTYPE | operator|= (MaskType s) noexcept |
Bit-or each element of the reciver with the scalar s and store the result in the receiver. More... | |
SIMDRegister &JUCE_VECTOR_CALLTYPE | operator^= (MaskType s) noexcept |
Bit-xor each element of the reciver with the scalar s and store the result in the receiver. More... | |
SIMDRegister JUCE_VECTOR_CALLTYPE | operator+ (SIMDRegister v) const noexcept |
Returns the sum of the receiver and v. More... | |
SIMDRegister JUCE_VECTOR_CALLTYPE | operator- (SIMDRegister v) const noexcept |
Returns the difference of the receiver and v. More... | |
SIMDRegister JUCE_VECTOR_CALLTYPE | operator* (SIMDRegister v) const noexcept |
Returns the product of the receiver and v. More... | |
SIMDRegister JUCE_VECTOR_CALLTYPE | operator+ (ElementType s) const noexcept |
Returns a vector where each element is the sum of the corresponding element in the receiver and the scalar s. More... | |
SIMDRegister JUCE_VECTOR_CALLTYPE | operator- (ElementType s) const noexcept |
Returns a vector where each element is the difference of the corresponding element in the receiver and the scalar s. More... | |
SIMDRegister JUCE_VECTOR_CALLTYPE | operator* (ElementType s) const noexcept |
Returns a vector where each element is the product of the corresponding element in the receiver and the scalar s. More... | |
SIMDRegister JUCE_VECTOR_CALLTYPE | operator & (vMaskType v) const noexcept |
Returns the bit-and of the receiver and v. More... | |
SIMDRegister JUCE_VECTOR_CALLTYPE | operator| (vMaskType v) const noexcept |
Returns the bit-or of the receiver and v. More... | |
SIMDRegister JUCE_VECTOR_CALLTYPE | operator^ (vMaskType v) const noexcept |
Returns the bit-xor of the receiver and v. More... | |
SIMDRegister JUCE_VECTOR_CALLTYPE | operator~ () const noexcept |
Returns a vector where each element is the bit-inverted value of the corresponding element in the receiver. More... | |
SIMDRegister JUCE_VECTOR_CALLTYPE | operator & (MaskType s) const noexcept |
Returns a vector where each element is the bit-and'd value of the corresponding element in the receiver and the scalar s. More... | |
SIMDRegister JUCE_VECTOR_CALLTYPE | operator| (MaskType s) const noexcept |
Returns a vector where each element is the bit-or'd value of the corresponding element in the receiver and the scalar s. More... | |
SIMDRegister JUCE_VECTOR_CALLTYPE | operator^ (MaskType s) const noexcept |
Returns a vector where each element is the bit-xor'd value of the corresponding element in the receiver and the scalar s. More... | |
bool JUCE_VECTOR_CALLTYPE | operator== (SIMDRegister other) const noexcept |
Returns true if all elements-wise comparisons return true. More... | |
bool JUCE_VECTOR_CALLTYPE | operator!= (SIMDRegister other) const noexcept |
Returns true if any elements-wise comparisons return false. More... | |
bool JUCE_VECTOR_CALLTYPE | operator== (Type s) const noexcept |
Returns true if all elements are equal to the scalar. More... | |
bool JUCE_VECTOR_CALLTYPE | operator!= (Type s) const noexcept |
Returns true if any elements are not equal to the scalar. More... | |
ElementType | sum () const noexcept |
Returns a scalar which is the sum of all elements of the receiver. More... | |
Static Public Member Functions | |
static constexpr size_t | size () noexcept |
Returns the number of elements in this vector. More... | |
static SIMDRegister JUCE_VECTOR_CALLTYPE | expand (ElementType s) noexcept |
Creates a new SIMDRegister from the corresponding scalar primitive. More... | |
static SIMDRegister JUCE_VECTOR_CALLTYPE | fromNative (vSIMDType a) noexcept |
Creates a new SIMDRegister from the internal SIMD type (for example __mm128 for single-precision floating point on SSE architectures). More... | |
static SIMDRegister JUCE_VECTOR_CALLTYPE | fromRawArray (const ElementType *a) noexcept |
Creates a new SIMDRegister from the first SIMDNumElements of a scalar array. More... | |
static vMaskType JUCE_VECTOR_CALLTYPE | equal (SIMDRegister a, SIMDRegister b) noexcept |
Returns a SIMDRegister of the corresponding integral type where each element has each bit set if the corresponding element of a is equal to the corresponding element of b, or zero otherwise. More... | |
static vMaskType JUCE_VECTOR_CALLTYPE | notEqual (SIMDRegister a, SIMDRegister b) noexcept |
Returns a SIMDRegister of the corresponding integral type where each element has each bit set if the corresponding element of a is not equal to the corresponding element of b, or zero otherwise. More... | |
static vMaskType JUCE_VECTOR_CALLTYPE | lessThan (SIMDRegister a, SIMDRegister b) noexcept |
Returns a SIMDRegister of the corresponding integral type where each element has each bit set if the corresponding element of a is less than to the corresponding element of b, or zero otherwise. More... | |
static vMaskType JUCE_VECTOR_CALLTYPE | lessThanOrEqual (SIMDRegister a, SIMDRegister b) noexcept |
Returns a SIMDRegister of the corresponding integral type where each element has each bit set if the corresponding element of a is than or equal to the corresponding element of b, or zero otherwise. More... | |
static vMaskType JUCE_VECTOR_CALLTYPE | greaterThan (SIMDRegister a, SIMDRegister b) noexcept |
Returns a SIMDRegister of the corresponding integral type where each element has each bit set if the corresponding element of a is greater than to the corresponding element of b, or zero otherwise. More... | |
static vMaskType JUCE_VECTOR_CALLTYPE | greaterThanOrEqual (SIMDRegister a, SIMDRegister b) noexcept |
Returns a SIMDRegister of the corresponding integral type where each element has each bit set if the corresponding element of a is greater than or equal to the corresponding element of b, or zero otherwise. More... | |
static SIMDRegister JUCE_VECTOR_CALLTYPE | min (SIMDRegister a, SIMDRegister b) noexcept |
Returns a new vector where each element is the minimum of the corresponding element of a and b. More... | |
static SIMDRegister JUCE_VECTOR_CALLTYPE | max (SIMDRegister a, SIMDRegister b) noexcept |
Returns a new vector where each element is the maximum of the corresponding element of a and b. More... | |
static SIMDRegister JUCE_VECTOR_CALLTYPE | multiplyAdd (SIMDRegister a, const SIMDRegister b, SIMDRegister c) noexcept |
Multiplies b and c and adds the result to a. More... | |
static bool | isSIMDAligned (const ElementType *ptr) noexcept |
Checks if the given pointer is suffeciently aligned for using SIMD operations. More... | |
static ElementType * | getNextSIMDAlignedPtr (ElementType *ptr) noexcept |
Returns the next position in memory where isSIMDAligned returns true. More... | |
Public Attributes | |
vSIMDType | value |
Static Public Attributes | |
static constexpr size_t | SIMDRegisterSize = sizeof (vSIMDType) |
The size in bytes of this register. More... | |
static constexpr size_t | SIMDNumElements = SIMDRegisterSize / sizeof (ElementType) |
The number of elements that this vector can hold. More... | |
A wrapper around the platform's native SIMD register type.
This class is only availabe on SIMD machines. Use JUCE_USE_SIMD to query if SIMD is avaialble for your system.
SIMDRegister<Type> is a templated class representing the native vectorized version of FloatingType. SIMDRegister supports all numerical primitive types and std:complex<float> and std::complex<double> supports and most operations of the corresponding primitive type. Additionally, SIMDRegister can be accessed like an array to extract the individual elements.
If you are using SIMDRegister as a pointer, then you must ensure that the memory is suffeciently aligned for SIMD vector operations. Failing to do so will result in crashes or very slow code. Use SIMDRegister::isSIMDAligned to query if a pointer is suffeciently aligned for SIMD vector operations.
Note that using SIMDRegister without enabling optimizations will result in code with very poor performance.
{DSP}
typedef Type dsp::SIMDRegister< Type >::ElementType |
The type that represents the individual constituents of the SIMD Register.
typedef ElementType dsp::SIMDRegister< Type >::value_type |
STL compatible value_type definition (same as ElementType).
typedef SIMDInternal::MaskTypeFor<ElementType>::type dsp::SIMDRegister< Type >::MaskType |
The corresponding primitive integer type, for example, this will be int32_t if type is a float.
typedef SIMDInternal::PrimitiveType<ElementType>::type dsp::SIMDRegister< Type >::PrimitiveType |
The native primitive type (used internally).
typedef SIMDNativeOps<PrimitiveType> dsp::SIMDRegister< Type >::NativeOps |
The native operations for this platform and type combination (used internally)
typedef NativeOps::vSIMDType dsp::SIMDRegister< Type >::vSIMDType |
The native type (used internally).
typedef SIMDRegister<MaskType> dsp::SIMDRegister< Type >::vMaskType |
The corresponding integer SIMDRegister type (used internally).
typedef vMaskType::vSIMDType dsp::SIMDRegister< Type >::vMaskSIMDType |
The internal native type for the corresponding mask type (used internally).
typedef CmplxSIMDOps<ElementType> dsp::SIMDRegister< Type >::CmplxOps |
Wrapper for operations which need to be handled differently for complex and scalar types (used internally).
|
noexcept |
Default constructor.
|
noexcept |
Constructs an object from the native SIMD type.
|
noexcept |
Constructs an object from a scalar type by broadcasting it to all elements.
|
noexcept |
Destrutor.
|
staticnoexcept |
Returns the number of elements in this vector.
References dsp::SIMDRegister< Type >::SIMDNumElements.
|
staticnoexcept |
Creates a new SIMDRegister from the corresponding scalar primitive.
The scalar is extended to all elements of the vector.
Referenced by dsp::SIMDRegister< Type >::getNextSIMDAlignedPtr(), and dsp::SIMDRegister< Type >::operator==().
|
staticnoexcept |
Creates a new SIMDRegister from the internal SIMD type (for example __mm128 for single-precision floating point on SSE architectures).
Referenced by dsp::SIMDRegister< Type >::getNextSIMDAlignedPtr().
|
staticnoexcept |
Creates a new SIMDRegister from the first SIMDNumElements of a scalar array.
References dsp::SIMDRegister< Type >::isSIMDAligned(), and jassert.
|
noexcept |
Copies the elements of the SIMDRegister to a scalar array in memory.
References dsp::SIMDRegister< Type >::isSIMDAligned(), and jassert.
|
noexcept |
Returns the idx-th element of the receiver.
Note that this does not check if idx is larger than the native register size.
References jassert, and dsp::SIMDRegister< Type >::value.
|
noexcept |
Returns the idx-th element of the receiver.
Note that this does not check if idx is larger than the native register size.
References jassert, and dsp::SIMDRegister< Type >::value.
|
noexcept |
Adds another SIMDRegister to the receiver.
|
noexcept |
Subtracts another SIMDRegister to the receiver.
|
noexcept |
Multiplies another SIMDRegister to the receiver.
|
noexcept |
Broadcasts the scalar to all elements of the receiver.
|
noexcept |
Adds a scalar to the receiver.
|
noexcept |
Subtracts a scalar to the receiver.
|
noexcept |
Multiplies a scalar to the receiver.
|
noexcept |
Bit-and the reciver with SIMDRegister v and store the result in the receiver.
|
noexcept |
Bit-or the reciver with SIMDRegister v and store the result in the receiver.
|
noexcept |
Bit-xor the reciver with SIMDRegister v and store the result in the receiver.
|
noexcept |
Bit-and each element of the reciver with the scalar s and store the result in the receiver.
|
noexcept |
Bit-or each element of the reciver with the scalar s and store the result in the receiver.
|
noexcept |
Bit-xor each element of the reciver with the scalar s and store the result in the receiver.
|
noexcept |
Returns the sum of the receiver and v.
|
noexcept |
Returns the difference of the receiver and v.
|
noexcept |
Returns the product of the receiver and v.
|
noexcept |
Returns a vector where each element is the sum of the corresponding element in the receiver and the scalar s.
|
noexcept |
Returns a vector where each element is the difference of the corresponding element in the receiver and the scalar s.
|
noexcept |
Returns a vector where each element is the product of the corresponding element in the receiver and the scalar s.
|
noexcept |
Returns the bit-and of the receiver and v.
|
noexcept |
Returns the bit-or of the receiver and v.
|
noexcept |
Returns the bit-xor of the receiver and v.
|
noexcept |
Returns a vector where each element is the bit-inverted value of the corresponding element in the receiver.
|
noexcept |
Returns a vector where each element is the bit-and'd value of the corresponding element in the receiver and the scalar s.
|
noexcept |
Returns a vector where each element is the bit-or'd value of the corresponding element in the receiver and the scalar s.
|
noexcept |
Returns a vector where each element is the bit-xor'd value of the corresponding element in the receiver and the scalar s.
|
noexcept |
Returns true if all elements-wise comparisons return true.
|
noexcept |
Returns true if any elements-wise comparisons return false.
|
noexcept |
Returns true if all elements are equal to the scalar.
References dsp::SIMDRegister< Type >::expand().
|
noexcept |
Returns true if any elements are not equal to the scalar.
|
staticnoexcept |
Returns a SIMDRegister of the corresponding integral type where each element has each bit set if the corresponding element of a is equal to the corresponding element of b, or zero otherwise.
The result can then be used in bit operations defined above to avoid branches in vector SIMD code.
|
staticnoexcept |
Returns a SIMDRegister of the corresponding integral type where each element has each bit set if the corresponding element of a is not equal to the corresponding element of b, or zero otherwise.
The result can then be used in bit operations defined above to avoid branches in vector SIMD code.
|
staticnoexcept |
Returns a SIMDRegister of the corresponding integral type where each element has each bit set if the corresponding element of a is less than to the corresponding element of b, or zero otherwise.
The result can then be used in bit operations defined above to avoid branches in vector SIMD code.
|
staticnoexcept |
Returns a SIMDRegister of the corresponding integral type where each element has each bit set if the corresponding element of a is than or equal to the corresponding element of b, or zero otherwise.
The result can then be used in bit operations defined above to avoid branches in vector SIMD code.
|
staticnoexcept |
Returns a SIMDRegister of the corresponding integral type where each element has each bit set if the corresponding element of a is greater than to the corresponding element of b, or zero otherwise.
The result can then be used in bit operations defined above to avoid branches in vector SIMD code.
|
staticnoexcept |
Returns a SIMDRegister of the corresponding integral type where each element has each bit set if the corresponding element of a is greater than or equal to the corresponding element of b, or zero otherwise.
The result can then be used in bit operations defined above to avoid branches in vector SIMD code.
|
staticnoexcept |
|
staticnoexcept |
|
staticnoexcept |
Multiplies b and c and adds the result to a.
|
noexcept |
Returns a scalar which is the sum of all elements of the receiver.
Referenced by dsp::SIMDRegister< Type >::getNextSIMDAlignedPtr().
|
staticnoexcept |
Checks if the given pointer is suffeciently aligned for using SIMD operations.
Referenced by dsp::SIMDRegister< Type >::copyToRawArray(), and dsp::SIMDRegister< Type >::fromRawArray().
|
staticnoexcept |
Returns the next position in memory where isSIMDAligned returns true.
If the current position in memory is already aligned then this method will simply return the pointer.
References dsp::SIMDRegister< Type >::expand(), dsp::SIMDRegister< Type >::fromNative(), snapPointerToAlignment(), dsp::SIMDRegister< Type >::sum(), and dsp::SIMDRegister< Type >::value.
|
static |
The size in bytes of this register.
|
static |
The number of elements that this vector can hold.
Referenced by dsp::SIMDRegister< Type >::size().
vSIMDType dsp::SIMDRegister< Type >::value |