<br>

Configurable nxn crossbar datapath. More...

Functions
template<typename DataType , unsigned NumInputLanes, unsigned NumOutputLanes>
void	crossbar (DataType data_in[NumInputLanes], bool valid_in[NumInputLanes], NVUINTW(nvhls::index_width< NumInputLanes >::val) source[NumOutputLanes], bool valid_source[NumOutputLanes], DataType data_out[NumOutputLanes], bool valid_out[NumOutputLanes])
	Main entry point for crossbar - most generic implementation.
template<typename DataType , unsigned NumInputLanes, unsigned NumOutputLanes>
void	crossbar (DataType data_in[NumInputLanes], bool valid_in[NumInputLanes], NVUINTW(nvhls::index_width< NumInputLanes >::val) source[NumOutputLanes], DataType data_out[NumOutputLanes], bool valid_out[NumOutputLanes])
	Simplified specialization with no valid_source parameter. All source selections are assumed to be valid.
template<typename DataType , unsigned NumInputLanes, unsigned NumOutputLanes>
void	crossbar (DataType data_in[NumInputLanes], NVUINTW(nvhls::index_width< NumInputLanes >::val) source[NumOutputLanes], DataType data_out[NumOutputLanes])
	Simplified specialization with no valid parameter at all. Inputs are assumed to be valid. Validity of outputs is not reported.

Detailed Description

Configurable nxn crossbar datapath.

Function Documentation

crossbar() [1/3]

template<typename DataType , unsigned NumInputLanes, unsigned NumOutputLanes>

void crossbar	(	DataType	data_in[NumInputLanes],
		bool	valid_in[NumInputLanes],
		NVUINTW(nvhls::index_width< NumInputLanes >::val)	source[NumOutputLanes],
		bool	valid_source[NumOutputLanes],
		DataType	data_out[NumOutputLanes],
		bool	valid_out[NumOutputLanes]
	)

Main entry point for crossbar - most generic implementation.

Template Parameters

DataType	Datatype of input and output of each lane
NumInputLanes	Number of input lanes
NumOutputLanes	Number of output lanes

Parameters

[in]	data_in	Array of data inputs for each lane, indexed by input lane id.
[in]	valid_in	Array of boolean valid bits, indexed by input lane id.. False means data can be ignored.
[in]	source	Array indicating the origin lane for each output lane id. Indexed by output lane id.
[in]	valid_source	Array of boolean valid bits, indexed by output lane id. Indicated validity of source selection for given output.
[out]	data_out	Array of outputs, indexed by output lane id.
[out]	valid_out	Array of outputs, indexed by output lane id.

Overview

• Routing input data into output based on source information.
• Output will be invalid if either source information for that output was invalid or selected input lane is invalid.
• Invalid output will have 0 on data bits.

A Simple Example

#include <crossbar.h>
 
...
Src_t   src_in[NUM_OUTPUTS]; // src_in[x]=y means lane x output comes from lane y
Word_t  data_out[NUM_OUTPUTS];
bool    valid_in[NUM_INPUTS];
bool    valid_src[NUM_OUTPUTS];
bool    valid_out[NUM_OUTPUTS];
 
...           
crossbar<DATA_TYPE, NUM_INPUTS, NUM_OUTPUTS>(
    data_in, valid_in,
    source,  valid_source,
    data_out,valid_out);
...

Definition at line 78 of file crossbar.h.

crossbar() [2/3]

template<typename DataType , unsigned NumInputLanes, unsigned NumOutputLanes>

void crossbar	(	DataType	data_in[NumInputLanes],
		bool	valid_in[NumInputLanes],
		NVUINTW(nvhls::index_width< NumInputLanes >::val)	source[NumOutputLanes],
		DataType	data_out[NumOutputLanes],
		bool	valid_out[NumOutputLanes]
	)

Simplified specialization with no valid_source parameter. All source selections are assumed to be valid.

Definition at line 112 of file crossbar.h.

crossbar() [3/3]

template<typename DataType , unsigned NumInputLanes, unsigned NumOutputLanes>

void crossbar	(	DataType	data_in[NumInputLanes],
		NVUINTW(nvhls::index_width< NumInputLanes >::val)	source[NumOutputLanes],
		DataType	data_out[NumOutputLanes]
	)

Simplified specialization with no valid parameter at all. Inputs are assumed to be valid. Validity of outputs is not reported.

Definition at line 130 of file crossbar.h.