pcod_common::PillarPreprocessCudaContext Class Reference

#include <pillar_preprocess_cuda.hpp>

Classes
struct	Impl

Public Member Functions
	PillarPreprocessCudaContext ()
	~PillarPreprocessCudaContext ()
	PillarPreprocessCudaContext (const PillarPreprocessCudaContext &)=delete
PillarPreprocessCudaContext &	operator= (const PillarPreprocessCudaContext &)=delete
	PillarPreprocessCudaContext (PillarPreprocessCudaContext &&other) noexcept
PillarPreprocessCudaContext &	operator= (PillarPreprocessCudaContext &&other) noexcept
bool	isAvailable () const
bool	run (const PillarPreprocessPoint *points, std::int32_t num_points, const PillarPreprocessCudaConfig &config, const PillarPreprocessCudaOutputs &outputs, std::string *error_message)
bool	runToDevice (const PillarPreprocessPoint *points, std::int32_t num_points, const PillarPreprocessCudaConfig &config, const PillarPreprocessCudaDeviceOutputs &outputs, std::string *error_message)

Detailed Description

Own the reusable CUDA resources required for pillar preprocessing.

Definition at line 67 of file pillar_preprocess_cuda.hpp.

Constructor & Destructor Documentation

PillarPreprocessCudaContext() [1/3]

pcod_common::PillarPreprocessCudaContext::PillarPreprocessCudaContext

(

)

Construct a context, retaining an unavailable state when CUDA initialization fails.

Definition at line 12 of file pillar_preprocess_cuda_stub.cpp.

: impl_(std::make_unique<Impl>()) {}

~PillarPreprocessCudaContext()

pcod_common::PillarPreprocessCudaContext::~PillarPreprocessCudaContext ( )

default

Release all owned CUDA resources.

PillarPreprocessCudaContext() [2/3]

pcod_common::PillarPreprocessCudaContext::PillarPreprocessCudaContext ( const PillarPreprocessCudaContext & )

delete

Copy construction is disabled because the context owns backend resources.

PillarPreprocessCudaContext() [3/3]

pcod_common::PillarPreprocessCudaContext::PillarPreprocessCudaContext ( PillarPreprocessCudaContext && other )

defaultnoexcept

Parameters

other

Context whose resources are transferred.

Member Function Documentation

isAvailable()

bool pcod_common::PillarPreprocessCudaContext::isAvailable

(

)

const

Return whether this context can execute CUDA preprocessing.

Definition at line 20 of file pillar_preprocess_cuda_stub.cpp.

{ return impl_ != nullptr && HasCudaPillarPreprocessSupport(); }

References pcod_common::HasCudaPillarPreprocessSupport().

operator=() [1/2]

PillarPreprocessCudaContext & pcod_common::PillarPreprocessCudaContext::operator= ( const PillarPreprocessCudaContext & )

delete

Copy assignment is disabled because the context owns backend resources.

operator=() [2/2]

PillarPreprocessCudaContext & pcod_common::PillarPreprocessCudaContext::operator= ( PillarPreprocessCudaContext && other )

defaultnoexcept

Parameters

other

Context whose resources are transferred.

Returns

This context.

run()

bool pcod_common::PillarPreprocessCudaContext::run	(	const PillarPreprocessPoint *	points,
		std::int32_t	num_points,
		const PillarPreprocessCudaConfig &	config,
		const PillarPreprocessCudaOutputs &	outputs,
		std::string *	error_message )

Run preprocessing and copy results to host-accessible buffers.

Parameters

points	Host input point array.
num_points	Number of input points.
config	Preprocessing contract.
outputs	Host output buffers.
error_message	Optional failure detail destination.

Returns

Whether preprocessing completed successfully.

Definition at line 22 of file pillar_preprocess_cuda_stub.cpp.

                                                                {
  (void)points;
  (void)num_points;
  (void)config;
  (void)outputs;
  if (error_message != nullptr) {
    *error_message =
        impl_ == nullptr ? "CUDA preprocessing context is not initialized" : "CUDA preprocessing is not available in this build";
  }
  return false;
}

runToDevice()

bool pcod_common::PillarPreprocessCudaContext::runToDevice	(	const PillarPreprocessPoint *	points,
		std::int32_t	num_points,
		const PillarPreprocessCudaConfig &	config,
		const PillarPreprocessCudaDeviceOutputs &	outputs,
		std::string *	error_message )

Run preprocessing directly into device-resident buffers.

Parameters

points	Host input point array.
num_points	Number of input points.
config	Preprocessing contract.
outputs	Device output buffers.
error_message	Optional failure detail destination.

Returns

Whether preprocessing completed successfully.

Definition at line 38 of file pillar_preprocess_cuda_stub.cpp.

                                                                        {
  (void)points;
  (void)num_points;
  (void)config;
  (void)outputs;
  if (error_message != nullptr) {
    *error_message =
        impl_ == nullptr ? "CUDA preprocessing context is not initialized" : "CUDA preprocessing is not available in this build";
  }
  return false;
}

---

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

• include/pcod_common/pillar_preprocess_cuda.hpp
• src/pillar_preprocess_cuda_stub.cpp