holoscan::utils::transmit_data_per_model

gxf_result_t holoscan::utils::transmit_data_per_model(
    gxf_context_t &cont,
    const HoloInfer::MultiMappings &model_to_tensor_map,
    HoloInfer::DataMap &input_data_map,
    OutputContext &op_output,
    std::vector<std::string> &out_tensors,
    HoloInfer::DimType &tensor_out_dims_map,
    bool cuda_buffer_in,
    bool cuda_buffer_out,
    const nvidia::gxf::Handle<nvidia::gxf::Allocator> &allocator_,
    const std::string &module,
    CudaStreamHandler &cuda_stream_handler
)

Transmits multiple buffers via GXF Transmitters.

This version of transmit_data_per_model uses the legacy CudaStreamHandler utility. Use the variant without a cuda_stream_handler argument to use the built-in CudaObjectHandler instead.

Returns: GXF result code

Parameters

cont

gxf_context_t &

GXF context for transmission

model_to_tensor_map

const HoloInfer::MultiMappings &

Map of model name as key, mapped to a vector of tensor names

input_data_map

HoloInfer::DataMap &

Map of tensor name as key, mapped to the data buffer as a vector

op_output

OutputContext &

Output context. Assume that the output port’s name is “transmitter”.

out_tensors

std::vector<std::string> &

Output tensor names

tensor_out_dims_map

HoloInfer::DimType &

Map is updated with model name as key mapped to dimension of output tensor as a vector

cuda_buffer_in

bool

Flag to demonstrate if memory storage of input buffers is on CUDA

cuda_buffer_out

bool

Flag to demonstrate if memory storage of output message is on CUDA

allocator_

const nvidia::gxf::Handle<nvidia::gxf::Allocator> &

GXF Memory allocator

module

const std::string &

Module that called for data transmission

cuda_stream_handler

CudaStreamHandler &

Cuda stream handler

holoscan::utils::transmit_data_per_model

Transmits multiple buffers via GXF Transmitters (with cuda stream handler)

Transmits multiple buffers via GXF Transmitters (with cstream)

Overload 3

Transmits multiple buffers via GXF Transmitters (with cuda stream handler)

Transmits multiple buffers via GXF Transmitters (with cstream)

Overload 3