Detailed description

template< typename T, int LOGICAL_WARP_THREADS = CUB_PTX_WARP_THREADS, int PTX_ARCH = CUB_PTX_ARCH>
class cub::WarpReduce< T, LOGICAL_WARP_THREADS, PTX_ARCH >

The WarpReduce class provides collective methods for computing a parallel reduction of items partitioned across a CUDA thread warp.

.

Template Parameters

T	The reduction input/output element type
LOGICAL_WARP_THREADS	[optional] The number of threads per "logical" warp (may be less than the number of hardware warp threads). Default is the warp size of the targeted CUDA compute-capability (e.g., 32 threads for SM20).
PTX_ARCH	[optional] The PTX compute capability for which to to specialize this collective, formatted as per the `CUDA_ARCH` macro (e.g., 350 for sm_35). Useful for determining the collective's storage requirements for a given device from the host. (Default: the value of `CUDA_ARCH` during the current compiler pass)

Overview

A reduction (or fold) uses a binary combining operator to compute a single aggregate from a list of input elements.
Supports "logical" warps smaller than the physical warp size (e.g., logical warps of 8 threads)
The number of entrant threads must be an multiple of LOGICAL_WARP_THREADS

Performance Considerations

Uses special instructions when applicable (e.g., warp SHFL instructions)
Uses synchronization-free communication between warp lanes when applicable
Incurs zero bank conflicts for most types
Computation is slightly more efficient (i.e., having lower instruction overhead) for:
- Summation (vs. generic reduction)
- The architecture's warp size is a whole multiple of LOGICAL_WARP_THREADS

Simple Examples: Every thread in the warp uses the WarpReduce class by first specializing the WarpReduce type, then instantiating an instance with parameters for communication, and finally invoking or more collective member functions.

: The code snippet below illustrates four concurrent warp sum reductions within a block of 128 threads (one per each of the 32-thread warps).

: #include <cub/cub.cuh>

__global__ void ExampleKernel(...)

{

// Specialize WarpReduce for type int

typedef cub::WarpReduce<int> WarpReduce;

// Allocate WarpReduce shared memory for 4 warps

__shared__ typename WarpReduce::TempStorage temp_storage[4];

// Obtain one input item per thread

int thread_data = ...

// Return the warp-wide sums to each lane0 (threads 0, 32, 64, and 96)

int warp_id = threadIdx.x / 32;

int aggregate = WarpReduce(temp_storage[warp_id]).Sum(thread_data);

: Suppose the set of input thread_data across the block of threads is {0, 1, 2, 3, ..., 127}. The corresponding output aggregate in threads 0, 32, 64, and 96 will 496, 1520, 2544, and 3568, respectively (and is undefined in other threads).

: The code snippet below illustrates a single warp sum reduction within a block of 128 threads.

: #include <cub/cub.cuh>

__global__ void ExampleKernel(...)

{

// Specialize WarpReduce for type int

typedef cub::WarpReduce<int> WarpReduce;

// Allocate WarpReduce shared memory for one warp

__shared__ typename WarpReduce::TempStorage temp_storage;

...

// Only the first warp performs a reduction

if (threadIdx.x < 32)

{

// Obtain one input item per thread

int thread_data = ...

// Return the warp-wide sum to lane0

int aggregate = WarpReduce(temp_storage).Sum(thread_data);

: Suppose the set of input thread_data across the warp of threads is {0, 1, 2, 3, ..., 31}. The corresponding output aggregate in thread0 will be 496 (and is undefined in other threads).

Classes
struct	TempStorage
	The operations exposed by WarpReduce require a temporary memory allocation of this nested type for thread communication. This opaque storage can be allocated directly using the `__shared__` keyword. Alternatively, it can be aliased to externally allocated memory (shared or global) or `union`'d with other storage allocation types to facilitate memory reuse. More...

Public Methods
Collective constructors
__device__ __forceinline__	WarpReduce (TempStorage &temp_storage)
	Collective constructor using the specified memory allocation as temporary storage. Logical warp and lane identifiers are constructed from `threadIdx.x`. More...

Summation reductions
__device__ __forceinline__ T	Sum (T input)
	Computes a warp-wide sum in the calling warp. The output is valid in warp lane₀. More...

__device__ __forceinline__ T	Sum (T input, int valid_items)
	Computes a partially-full warp-wide sum in the calling warp. The output is valid in warp lane₀. More...

template<typename FlagT >
__device__ __forceinline__ T	HeadSegmentedSum (T input, FlagT head_flag)
	Computes a segmented sum in the calling warp where segments are defined by head-flags. The sum of each segment is returned to the first lane in that segment (which always includes lane₀). More...

template<typename FlagT >
__device__ __forceinline__ T	TailSegmentedSum (T input, FlagT tail_flag)
	Computes a segmented sum in the calling warp where segments are defined by tail-flags. The sum of each segment is returned to the first lane in that segment (which always includes lane₀). More...

Generic reductions
template<typename ReductionOp >
__device__ __forceinline__ T	Reduce (T input, ReductionOp reduction_op)
	Computes a warp-wide reduction in the calling warp using the specified binary reduction functor. The output is valid in warp lane₀. More...

template<typename ReductionOp >
__device__ __forceinline__ T	Reduce (T input, ReductionOp reduction_op, int valid_items)
	Computes a partially-full warp-wide reduction in the calling warp using the specified binary reduction functor. The output is valid in warp lane₀. More...

template<typename ReductionOp , typename FlagT >
__device__ __forceinline__ T	HeadSegmentedReduce (T input, FlagT head_flag, ReductionOp reduction_op)
	Computes a segmented reduction in the calling warp where segments are defined by head-flags. The reduction of each segment is returned to the first lane in that segment (which always includes lane₀). More...

template<typename ReductionOp , typename FlagT >
__device__ __forceinline__ T	TailSegmentedReduce (T input, FlagT tail_flag, ReductionOp reduction_op)
	Computes a segmented reduction in the calling warp where segments are defined by tail-flags. The reduction of each segment is returned to the first lane in that segment (which always includes lane₀). More...

Constructor & Destructor Documentation

template<typename T , int LOGICAL_WARP_THREADS = CUB_PTX_WARP_THREADS, int PTX_ARCH = CUB_PTX_ARCH>

__device__ __forceinline__ cub::WarpReduce< T, LOGICAL_WARP_THREADS, PTX_ARCH >::WarpReduce ( TempStorage & temp_storage)

inline

Collective constructor using the specified memory allocation as temporary storage. Logical warp and lane identifiers are constructed from threadIdx.x.

Parameters

[in] temp_storage Reference to memory allocation having layout type TempStorage

Member Function Documentation

template<typename T , int LOGICAL_WARP_THREADS = CUB_PTX_WARP_THREADS, int PTX_ARCH = CUB_PTX_ARCH>

__device__ __forceinline__ T cub::WarpReduce< T, LOGICAL_WARP_THREADS, PTX_ARCH >::Sum ( T input)

inline

Computes a warp-wide sum in the calling warp. The output is valid in warp lane₀.

A subsequent __syncthreads() threadblock barrier should be invoked after calling this method if the collective's temporary storage (e.g., temp_storage) is to be reused or repurposed.

Snippet: The code snippet below illustrates four concurrent warp sum reductions within a block of 128 threads (one per each of the 32-thread warps).

: #include <cub/cub.cuh>

__global__ void ExampleKernel(...)

{

// Specialize WarpReduce for type int

typedef cub::WarpReduce<int> WarpReduce;

// Allocate WarpReduce shared memory for 4 warps

__shared__ typename WarpReduce::TempStorage temp_storage[4];

// Obtain one input item per thread

int thread_data = ...

// Return the warp-wide sums to each lane0

int warp_id = threadIdx.x / 32;

int aggregate = WarpReduce(temp_storage[warp_id]).Sum(thread_data);

: Suppose the set of input thread_data across the block of threads is {0, 1, 2, 3, ..., 127}. The corresponding output aggregate in threads 0, 32, 64, and 96 will 496, 1520, 2544, and 3568, respectively (and is undefined in other threads).

Parameters

[in] input Calling thread's input

template<typename T , int LOGICAL_WARP_THREADS = CUB_PTX_WARP_THREADS, int PTX_ARCH = CUB_PTX_ARCH>

__device__ __forceinline__ T cub::WarpReduce< T, LOGICAL_WARP_THREADS, PTX_ARCH >::Sum	(	T	input,
		int	valid_items
	)

inline

Computes a partially-full warp-wide sum in the calling warp. The output is valid in warp lane₀.

All threads across the calling warp must agree on the same value for valid_items. Otherwise the result is undefined.

A subsequent __syncthreads() threadblock barrier should be invoked after calling this method if the collective's temporary storage (e.g., temp_storage) is to be reused or repurposed.

Snippet: The code snippet below illustrates a sum reduction within a single, partially-full block of 32 threads (one warp).

: #include <cub/cub.cuh>

__global__ void ExampleKernel(int *d_data, int valid_items)

{

// Specialize WarpReduce for type int

typedef cub::WarpReduce<int> WarpReduce;

// Allocate WarpReduce shared memory for one warp

__shared__ typename WarpReduce::TempStorage temp_storage;

// Obtain one input item per thread if in range

int thread_data;

if (threadIdx.x < valid_items)

thread_data = d_data[threadIdx.x];

// Return the warp-wide sums to each lane0

int aggregate = WarpReduce(temp_storage).Sum(

thread_data, valid_items);

: Suppose the input d_data is {0, 1, 2, 3, 4, ... and valid_items is 4. The corresponding output aggregate in thread0 is 6 (and is undefined in other threads).

Parameters

[in]	input	Calling thread's input
[in]	valid_items	Total number of valid items in the calling thread's logical warp (may be less than `LOGICAL_WARP_THREADS`)

template<typename T , int LOGICAL_WARP_THREADS = CUB_PTX_WARP_THREADS, int PTX_ARCH = CUB_PTX_ARCH>

template<typename FlagT >

__device__ __forceinline__ T cub::WarpReduce< T, LOGICAL_WARP_THREADS, PTX_ARCH >::HeadSegmentedSum	(	T	input,
		FlagT	head_flag
	)

inline

Computes a segmented sum in the calling warp where segments are defined by head-flags. The sum of each segment is returned to the first lane in that segment (which always includes lane₀).

A subsequent __syncthreads() threadblock barrier should be invoked after calling this method if the collective's temporary storage (e.g., temp_storage) is to be reused or repurposed.

Snippet: The code snippet below illustrates a head-segmented warp sum reduction within a block of 32 threads (one warp).

: #include <cub/cub.cuh>

__global__ void ExampleKernel(...)

{

// Specialize WarpReduce for type int

typedef cub::WarpReduce<int> WarpReduce;

// Allocate WarpReduce shared memory for one warp

__shared__ typename WarpReduce::TempStorage temp_storage;

// Obtain one input item and flag per thread

int thread_data = ...

int head_flag = ...

// Return the warp-wide sums to each lane0

int aggregate = WarpReduce(temp_storage).HeadSegmentedSum(

thread_data, head_flag);

: Suppose the set of input thread_data and head_flag across the block of threads is {0, 1, 2, 3, ..., 31 and is {1, 0, 0, 0, 1, 0, 0, 0, ..., 1, 0, 0, 0, respectively. The corresponding output aggregate in threads 0, 4, 8, etc. will be 6, 22, 38, etc. (and is undefined in other threads).

Template Parameters

ReductionOp [inferred] Binary reduction operator type having member T operator()(const T &a, const T &b)

Parameters

[in]	input	Calling thread's input
[in]	head_flag	Head flag denoting whether or not `input` is the start of a new segment

template<typename T , int LOGICAL_WARP_THREADS = CUB_PTX_WARP_THREADS, int PTX_ARCH = CUB_PTX_ARCH>

template<typename FlagT >

__device__ __forceinline__ T cub::WarpReduce< T, LOGICAL_WARP_THREADS, PTX_ARCH >::TailSegmentedSum	(	T	input,
		FlagT	tail_flag
	)

inline

Computes a segmented sum in the calling warp where segments are defined by tail-flags. The sum of each segment is returned to the first lane in that segment (which always includes lane₀).

A subsequent __syncthreads() threadblock barrier should be invoked after calling this method if the collective's temporary storage (e.g., temp_storage) is to be reused or repurposed.

Snippet: The code snippet below illustrates a tail-segmented warp sum reduction within a block of 32 threads (one warp).

: #include <cub/cub.cuh>

__global__ void ExampleKernel(...)

{

// Specialize WarpReduce for type int

typedef cub::WarpReduce<int> WarpReduce;

// Allocate WarpReduce shared memory for one warp

__shared__ typename WarpReduce::TempStorage temp_storage;

// Obtain one input item and flag per thread

int thread_data = ...

int tail_flag = ...

// Return the warp-wide sums to each lane0

int aggregate = WarpReduce(temp_storage).TailSegmentedSum(

thread_data, tail_flag);

: Suppose the set of input thread_data and tail_flag across the block of threads is {0, 1, 2, 3, ..., 31 and is {0, 0, 0, 1, 0, 0, 0, 1, ..., 0, 0, 0, 1, respectively. The corresponding output aggregate in threads 0, 4, 8, etc. will be 6, 22, 38, etc. (and is undefined in other threads).

Template Parameters

ReductionOp [inferred] Binary reduction operator type having member T operator()(const T &a, const T &b)

Parameters

[in]	input	Calling thread's input
[in]	tail_flag	Head flag denoting whether or not `input` is the start of a new segment

template<typename T , int LOGICAL_WARP_THREADS = CUB_PTX_WARP_THREADS, int PTX_ARCH = CUB_PTX_ARCH>

template<typename ReductionOp >

__device__ __forceinline__ T cub::WarpReduce< T, LOGICAL_WARP_THREADS, PTX_ARCH >::Reduce	(	T	input,
		ReductionOp	reduction_op
	)

inline

Computes a warp-wide reduction in the calling warp using the specified binary reduction functor. The output is valid in warp lane₀.

Supports non-commutative reduction operators

A subsequent __syncthreads() threadblock barrier should be invoked after calling this method if the collective's temporary storage (e.g., temp_storage) is to be reused or repurposed.

Snippet: The code snippet below illustrates four concurrent warp max reductions within a block of 128 threads (one per each of the 32-thread warps).

: #include <cub/cub.cuh>

__global__ void ExampleKernel(...)

{

// Specialize WarpReduce for type int

typedef cub::WarpReduce<int> WarpReduce;

// Allocate WarpReduce shared memory for 4 warps

__shared__ typename WarpReduce::TempStorage temp_storage[4];

// Obtain one input item per thread

int thread_data = ...

// Return the warp-wide reductions to each lane0

int warp_id = threadIdx.x / 32;

int aggregate = WarpReduce(temp_storage[warp_id]).Reduce(

thread_data, cub::Max());

: Suppose the set of input thread_data across the block of threads is {0, 1, 2, 3, ..., 127}. The corresponding output aggregate in threads 0, 32, 64, and 96 will 31, 63, 95, and 127, respectively (and is undefined in other threads).

Template Parameters

ReductionOp [inferred] Binary reduction operator type having member T operator()(const T &a, const T &b)

Parameters

[in]	input	Calling thread's input
[in]	reduction_op	Binary reduction operator

template<typename T , int LOGICAL_WARP_THREADS = CUB_PTX_WARP_THREADS, int PTX_ARCH = CUB_PTX_ARCH>

template<typename ReductionOp >

__device__ __forceinline__ T cub::WarpReduce< T, LOGICAL_WARP_THREADS, PTX_ARCH >::Reduce	(	T	input,
		ReductionOp	reduction_op,
		int	valid_items
	)

inline

Computes a partially-full warp-wide reduction in the calling warp using the specified binary reduction functor. The output is valid in warp lane₀.

All threads across the calling warp must agree on the same value for valid_items. Otherwise the result is undefined.

Supports non-commutative reduction operators

A subsequent __syncthreads() threadblock barrier should be invoked after calling this method if the collective's temporary storage (e.g., temp_storage) is to be reused or repurposed.

Snippet: The code snippet below illustrates a max reduction within a single, partially-full block of 32 threads (one warp).

: #include <cub/cub.cuh>

__global__ void ExampleKernel(int *d_data, int valid_items)

{

// Specialize WarpReduce for type int

typedef cub::WarpReduce<int> WarpReduce;

// Allocate WarpReduce shared memory for one warp

__shared__ typename WarpReduce::TempStorage temp_storage;

// Obtain one input item per thread if in range

int thread_data;

if (threadIdx.x < valid_items)

thread_data = d_data[threadIdx.x];

// Return the warp-wide reductions to each lane0

int aggregate = WarpReduce(temp_storage).Reduce(

thread_data, cub::Max(), valid_items);

: Suppose the input d_data is {0, 1, 2, 3, 4, ... and valid_items is 4. The corresponding output aggregate in thread0 is 3 (and is undefined in other threads).

Template Parameters

ReductionOp [inferred] Binary reduction operator type having member T operator()(const T &a, const T &b)

Parameters

[in]	input	Calling thread's input
[in]	reduction_op	Binary reduction operator
[in]	valid_items	Total number of valid items in the calling thread's logical warp (may be less than `LOGICAL_WARP_THREADS`)

template<typename T , int LOGICAL_WARP_THREADS = CUB_PTX_WARP_THREADS, int PTX_ARCH = CUB_PTX_ARCH>

template<typename ReductionOp , typename FlagT >

__device__ __forceinline__ T cub::WarpReduce< T, LOGICAL_WARP_THREADS, PTX_ARCH >::HeadSegmentedReduce	(	T	input,
		FlagT	head_flag,
		ReductionOp	reduction_op
	)

inline

Computes a segmented reduction in the calling warp where segments are defined by head-flags. The reduction of each segment is returned to the first lane in that segment (which always includes lane₀).

Supports non-commutative reduction operators

A subsequent __syncthreads() threadblock barrier should be invoked after calling this method if the collective's temporary storage (e.g., temp_storage) is to be reused or repurposed.

Snippet: The code snippet below illustrates a head-segmented warp max reduction within a block of 32 threads (one warp).

: #include <cub/cub.cuh>

__global__ void ExampleKernel(...)

{

// Specialize WarpReduce for type int

typedef cub::WarpReduce<int> WarpReduce;

// Allocate WarpReduce shared memory for one warp

__shared__ typename WarpReduce::TempStorage temp_storage;

// Obtain one input item and flag per thread

int thread_data = ...

int head_flag = ...

// Return the warp-wide reductions to each lane0

int aggregate = WarpReduce(temp_storage).HeadSegmentedReduce(

thread_data, head_flag, cub::Max());

: Suppose the set of input thread_data and head_flag across the block of threads is {0, 1, 2, 3, ..., 31 and is {1, 0, 0, 0, 1, 0, 0, 0, ..., 1, 0, 0, 0, respectively. The corresponding output aggregate in threads 0, 4, 8, etc. will be 3, 7, 11, etc. (and is undefined in other threads).

Template Parameters

ReductionOp [inferred] Binary reduction operator type having member T operator()(const T &a, const T &b)

Parameters

[in]	input	Calling thread's input
[in]	head_flag	Head flag denoting whether or not `input` is the start of a new segment
[in]	reduction_op	Reduction operator

template<typename T , int LOGICAL_WARP_THREADS = CUB_PTX_WARP_THREADS, int PTX_ARCH = CUB_PTX_ARCH>

template<typename ReductionOp , typename FlagT >

__device__ __forceinline__ T cub::WarpReduce< T, LOGICAL_WARP_THREADS, PTX_ARCH >::TailSegmentedReduce	(	T	input,
		FlagT	tail_flag,
		ReductionOp	reduction_op
	)

inline

Computes a segmented reduction in the calling warp where segments are defined by tail-flags. The reduction of each segment is returned to the first lane in that segment (which always includes lane₀).

Supports non-commutative reduction operators

A subsequent __syncthreads() threadblock barrier should be invoked after calling this method if the collective's temporary storage (e.g., temp_storage) is to be reused or repurposed.

Snippet: The code snippet below illustrates a tail-segmented warp max reduction within a block of 32 threads (one warp).

: #include <cub/cub.cuh>

__global__ void ExampleKernel(...)

{

// Specialize WarpReduce for type int

typedef cub::WarpReduce<int> WarpReduce;

// Allocate WarpReduce shared memory for one warp

__shared__ typename WarpReduce::TempStorage temp_storage;

// Obtain one input item and flag per thread

int thread_data = ...

int tail_flag = ...

// Return the warp-wide reductions to each lane0

int aggregate = WarpReduce(temp_storage).TailSegmentedReduce(

thread_data, tail_flag, cub::Max());

: Suppose the set of input thread_data and tail_flag across the block of threads is {0, 1, 2, 3, ..., 31 and is {0, 0, 0, 1, 0, 0, 0, 1, ..., 0, 0, 0, 1, respectively. The corresponding output aggregate in threads 0, 4, 8, etc. will be 3, 7, 11, etc. (and is undefined in other threads).

Template Parameters

ReductionOp [inferred] Binary reduction operator type having member T operator()(const T &a, const T &b)

Parameters

[in]	input	Calling thread's input
[in]	tail_flag	Tail flag denoting whether or not `input` is the end of the current segment
[in]	reduction_op	Reduction operator

The documentation for this class was generated from the following file:

warp_reduce.cuh

Detailed description

template< typename T, int LOGICAL_WARP_THREADS = CUB_PTX_WARP_THREADS, int PTX_ARCH = CUB_PTX_ARCH> class cub::WarpReduce< T, LOGICAL_WARP_THREADS, PTX_ARCH >

Classes

Public Methods

Constructor & Destructor Documentation

Member Function Documentation

template< typename T, int LOGICAL_WARP_THREADS = CUB_PTX_WARP_THREADS, int PTX_ARCH = CUB_PTX_ARCH>
class cub::WarpReduce< T, LOGICAL_WARP_THREADS, PTX_ARCH >