xpu::block_scan< T, BlockSize, Impl > Class Template Reference

Parallel scan inside a block. More...

Classes
struct	storage_t
	Temporary storage for the block scan. Should be allocated in shared memory. More...

Public Member Functions
template<typename ContextT >
XPU_D	block_scan (ContextT &ctx, storage_t &storage)
	Construct a block scan object. More...
XPU_D	block_scan (tpos &pos, storage_t &storage)
	Construct a block scan object. More...
XPU_D void	exclusive_sum (T input, T &output)
template<typename ScanOp >
XPU_D void	exclusive_sum (T input, T &output, T initial_value, ScanOp scan_op)
XPU_D void	inclusive_sum (T input, T &output)
template<typename ScanOp >
XPU_D void	inclusive_sum (T input, T &output, T initial_value, ScanOp scan_op)

Detailed Description

template<typename T, int BlockSize, xpu::driver_t Impl = XPU_COMPILATION_TARGET>
class xpu::block_scan< T, BlockSize, Impl >

Parallel scan inside a block.

Constructor & Destructor Documentation

block_scan() [1/2]

template<typename T , int BlockSize, xpu::driver_t Impl = XPU_COMPILATION_TARGET>

template<typename ContextT >

XPU_D xpu::block_scan< T, BlockSize, Impl >::block_scan	(	ContextT &	ctx,
		storage_t &	storage
	)

Construct a block scan object.

Parameters

ctx	Kernel context.
storage	Temporary storage for the block scan.

Note

This is a shortcut for block_scan(ctx.pos(), storage).

See also

block_scan::storage_t

block_scan() [2/2]

template<typename T , int BlockSize, xpu::driver_t Impl = XPU_COMPILATION_TARGET>

XPU_D xpu::block_scan< T, BlockSize, Impl >::block_scan	(	tpos &	pos,
		storage_t &	storage
	)

Construct a block scan object.

Parameters

pos	Thread position.
storage	Temporary storage for the block scan.

See also

block_scan::storage_t

Member Function Documentation

exclusive_sum() [1/2]

template<typename T , int BlockSize, xpu::driver_t Impl = XPU_COMPILATION_TARGET>

XPU_D void xpu::block_scan< T, BlockSize, Impl >::exclusive_sum	(	T	input,
		T &	output
	)

exclusive_sum() [2/2]

template<typename T , int BlockSize, xpu::driver_t Impl = XPU_COMPILATION_TARGET>

template<typename ScanOp >

XPU_D void xpu::block_scan< T, BlockSize, Impl >::exclusive_sum	(	T	input,
		T &	output,
		T	initial_value,
		ScanOp	scan_op
	)

inclusive_sum() [1/2]

template<typename T , int BlockSize, xpu::driver_t Impl = XPU_COMPILATION_TARGET>

XPU_D void xpu::block_scan< T, BlockSize, Impl >::inclusive_sum	(	T	input,
		T &	output
	)

inclusive_sum() [2/2]

template<typename T , int BlockSize, xpu::driver_t Impl = XPU_COMPILATION_TARGET>

template<typename ScanOp >

XPU_D void xpu::block_scan< T, BlockSize, Impl >::inclusive_sum	(	T	input,
		T &	output,
		T	initial_value,
		ScanOp	scan_op
	)

---

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

• device.h