#include <ginkgo/core/log/profiler_hook.hpp>

Inheritance diagram for gko::log::ProfilerHook:

Classes
struct	nested_summary_entry

class	NestedSummaryWriter

struct	summary_entry

class	SummaryWriter

class	TableSummaryWriter

Public Types
using	hook_function = std::function< void(const char *, profile_event_category)>

Public Types inherited from gko::log::Logger
using	mask_type = gko::uint64

Public Member Functions
void	on_allocation_started (const gko::Executor *exec, const gko::size_type &) const override

void	on_allocation_completed (const gko::Executor *exec, const gko::size_type &, const gko::uintptr &) const override

void	on_free_started (const gko::Executor *exec, const gko::uintptr &) const override

void	on_free_completed (const gko::Executor *exec, const gko::uintptr &) const override

void	on_copy_started (const gko::Executor from, const gko::Executor to, const gko::uintptr &, const gko::uintptr &, const gko::size_type &) const override

void	on_copy_completed (const gko::Executor from, const gko::Executor to, const gko::uintptr &, const gko::uintptr &, const gko::size_type &) const override

void	on_operation_launched (const Executor exec, const Operation operation) const override

void	on_operation_completed (const Executor exec, const Operation operation) const override

void	on_polymorphic_object_copy_started (const Executor exec, const PolymorphicObject from, const PolymorphicObject *to) const override

void	on_polymorphic_object_copy_completed (const Executor exec, const PolymorphicObject from, const PolymorphicObject *to) const override

void	on_polymorphic_object_move_started (const Executor exec, const PolymorphicObject from, const PolymorphicObject *to) const override

void	on_polymorphic_object_move_completed (const Executor exec, const PolymorphicObject from, const PolymorphicObject *to) const override

void	on_linop_apply_started (const LinOp A, const LinOp b, const LinOp *x) const override

void	on_linop_apply_completed (const LinOp A, const LinOp b, const LinOp *x) const override

void	on_linop_advanced_apply_started (const LinOp A, const LinOp alpha, const LinOp b, const LinOp beta, const LinOp *x) const override

void	on_linop_advanced_apply_completed (const LinOp A, const LinOp alpha, const LinOp b, const LinOp beta, const LinOp *x) const override

void	on_linop_factory_generate_started (const LinOpFactory factory, const LinOp input) const override

void	on_linop_factory_generate_completed (const LinOpFactory factory, const LinOp input, const LinOp *output) const override

void	on_criterion_check_started (const stop::Criterion criterion, const size_type &num_iterations, const LinOp residual, const LinOp residual_norm, const LinOp solution, const uint8 &stopping_id, const bool &set_finalized) const override

void	on_criterion_check_completed (const stop::Criterion criterion, const size_type &num_iterations, const LinOp residual, const LinOp residual_norm, const LinOp solution, const uint8 &stopping_id, const bool &set_finalized, const array< stopping_status > *status, const bool &one_changed, const bool &all_stopped) const override

void	on_criterion_check_completed (const stop::Criterion criterion, const size_type &num_iterations, const LinOp residual, const LinOp residual_norm, const LinOp implicit_sq_resnorm, const LinOp solution, const uint8 &stopping_id, const bool &set_finalized, const array< stopping_status > status, const bool &one_changed, const bool &all_stopped) const override

void	on_iteration_complete (const LinOp solver, const LinOp right_hand_side, const LinOp solution, const size_type &num_iterations, const LinOp residual, const LinOp residual_norm, const LinOp implicit_sq_residual_norm, const array< stopping_status > *status, bool stopped) const override

void	on_iteration_complete (const LinOp solver, const size_type &num_iterations, const LinOp residual, const LinOp solution, const LinOp residual_norm) const override

void	on_iteration_complete (const LinOp solver, const size_type &num_iterations, const LinOp residual, const LinOp solution, const LinOp residual_norm, const LinOp *implicit_sq_residual_norm) const override

bool	needs_propagation () const override

void	set_object_name (ptr_param< const PolymorphicObject > obj, std::string name)

void	set_synchronization (bool synchronize)

profiling_scope_guard	user_range (const char *name) const

Public Member Functions inherited from gko::log::Logger
template<size_type Event, typename... Params>
std::enable_if_t< Event==0 &&(0< event_count_max)>	on (Params &&... params) const

template<size_type Event, typename... Params>
std::enable_if_t< Event==1 &&(1< event_count_max)>	on (Params &&... params) const

template<size_type Event, typename... Params>
std::enable_if_t< Event==2 &&(2< event_count_max)>	on (Params &&... params) const

template<size_type Event, typename... Params>
std::enable_if_t< Event==3 &&(3< event_count_max)>	on (Params &&... params) const

template<size_type Event, typename... Params>
std::enable_if_t< Event==4 &&(4< event_count_max)>	on (Params &&... params) const

template<size_type Event, typename... Params>
std::enable_if_t< Event==5 &&(5< event_count_max)>	on (Params &&... params) const

template<size_type Event, typename... Params>
std::enable_if_t< Event==6 &&(6< event_count_max)>	on (Params &&... params) const

template<size_type Event, typename... Params>
std::enable_if_t< Event==7 &&(7< event_count_max)>	on (Params &&... params) const

template<size_type Event, typename... Params>
std::enable_if_t< Event==8 &&(8< event_count_max)>	on (Params &&... params) const

template<size_type Event, typename... Params>
std::enable_if_t< Event==9 &&(9< event_count_max)>	on (Params &&... params) const

template<size_type Event, typename... Params>
std::enable_if_t< Event==10 &&(10< event_count_max)>	on (Params &&... params) const

template<size_type Event, typename... Params>
std::enable_if_t< Event==11 &&(11< event_count_max)>	on (Params &&... params) const

template<size_type Event, typename... Params>
std::enable_if_t< Event==12 &&(12< event_count_max)>	on (Params &&... params) const

template<size_type Event, typename... Params>
std::enable_if_t< Event==13 &&(13< event_count_max)>	on (Params &&... params) const

template<size_type Event, typename... Params>
std::enable_if_t< Event==14 &&(14< event_count_max)>	on (Params &&... params) const

template<size_type Event, typename... Params>
std::enable_if_t< Event==15 &&(15< event_count_max)>	on (Params &&... params) const

template<size_type Event, typename... Params>
std::enable_if_t< Event==16 &&(16< event_count_max)>	on (Params &&... params) const

template<size_type Event, typename... Params>
std::enable_if_t< Event==17 &&(17< event_count_max)>	on (Params &&... params) const

template<size_type Event, typename... Params>
std::enable_if_t< Event==18 &&(18< event_count_max)>	on (Params &&... params) const

template<size_type Event, typename... Params>
std::enable_if_t< Event==19 &&(19< event_count_max)>	on (Params &&... params) const

template<size_type Event, typename... Params>
std::enable_if_t< Event==20 &&(20< event_count_max)>	on (Params &&... params) const

template<size_type Event, typename... Params>
std::enable_if_t< Event==21 &&(21< event_count_max)>	on (Params &&... params) const

template<size_type Event, typename... Params>
std::enable_if_t< Event==22 &&(22< event_count_max)>	on (Params &&... params) const

template<size_type Event, typename... Params>
std::enable_if_t< Event==23 &&(23< event_count_max)>	on (Params &&... params) const

template<size_type Event, typename... Params>
std::enable_if_t< Event==24 &&(24< event_count_max)>	on (Params &&... params) const

template<size_type Event, typename... Params>
std::enable_if_t< Event==25 &&(25< event_count_max)>	on (Params &&... params) const

template<size_type Event, typename... Params>
std::enable_if_t< Event==26 &&(26< event_count_max)>	on (Params &&... params) const

Static Public Member Functions
static std::shared_ptr< ProfilerHook >	create_tau (bool initialize=true)

static std::shared_ptr< ProfilerHook >	create_vtune ()

static std::shared_ptr< ProfilerHook >	create_nvtx (uint32 color_argb=color_yellow_argb)

static std::shared_ptr< ProfilerHook >	create_roctx ()

static std::shared_ptr< ProfilerHook >	create_for_executor (std::shared_ptr< const Executor > exec)

static std::shared_ptr< ProfilerHook >	create_summary (std::shared_ptr< Timer > timer=std::make_shared< CpuTimer >(), std::unique_ptr< SummaryWriter > writer=std::make_unique< TableSummaryWriter >(), bool debug_check_nesting=false)

static std::shared_ptr< ProfilerHook >	create_nested_summary (std::shared_ptr< Timer > timer=std::make_shared< CpuTimer >(), std::unique_ptr< NestedSummaryWriter > writer=std::make_unique< TableSummaryWriter >(), bool debug_check_nesting=false)

static std::shared_ptr< ProfilerHook >	create_custom (hook_function begin, hook_function end)

Static Public Attributes
static constexpr uint32	color_yellow_argb = 0xFFFFCB05U

Static Public Attributes inherited from gko::log::Logger
static constexpr size_type	event_count_max = sizeof(mask_type) * byte_size

static constexpr mask_type	all_events_mask = ~mask_type{0}

static constexpr size_type	allocation_started { 0 }

static constexpr mask_type	allocation_started_mask {mask_type{1} << 0 }

static constexpr size_type	allocation_completed { 1 }

static constexpr mask_type	allocation_completed_mask {mask_type{1} << 1 }

static constexpr size_type	free_started { 2 }

static constexpr mask_type	free_started_mask {mask_type{1} << 2 }

static constexpr size_type	free_completed { 3 }

static constexpr mask_type	free_completed_mask {mask_type{1} << 3 }

static constexpr size_type	copy_started { 4 }

static constexpr mask_type	copy_started_mask {mask_type{1} << 4 }

static constexpr size_type	copy_completed { 5 }

static constexpr mask_type	copy_completed_mask {mask_type{1} << 5 }

static constexpr size_type	operation_launched { 6 }

static constexpr mask_type	operation_launched_mask {mask_type{1} << 6 }

static constexpr size_type	operation_completed { 7 }

static constexpr mask_type	operation_completed_mask {mask_type{1} << 7 }

static constexpr size_type	polymorphic_object_create_started { 8 }

static constexpr mask_type	polymorphic_object_create_started_mask {mask_type{1} << 8 }

static constexpr size_type	polymorphic_object_create_completed { 9 }

static constexpr mask_type	polymorphic_object_create_completed_mask {mask_type{1} << 9 }

static constexpr size_type	polymorphic_object_copy_started { 10 }

static constexpr mask_type	polymorphic_object_copy_started_mask {mask_type{1} << 10 }

static constexpr size_type	polymorphic_object_copy_completed { 11 }

static constexpr mask_type	polymorphic_object_copy_completed_mask {mask_type{1} << 11 }

static constexpr size_type	polymorphic_object_deleted { 12 }

static constexpr mask_type	polymorphic_object_deleted_mask {mask_type{1} << 12 }

static constexpr size_type	linop_apply_started { 13 }

static constexpr mask_type	linop_apply_started_mask {mask_type{1} << 13 }

static constexpr size_type	linop_apply_completed { 14 }

static constexpr mask_type	linop_apply_completed_mask {mask_type{1} << 14 }

static constexpr size_type	linop_advanced_apply_started { 15 }

static constexpr mask_type	linop_advanced_apply_started_mask {mask_type{1} << 15 }

static constexpr size_type	linop_advanced_apply_completed { 16 }

static constexpr mask_type	linop_advanced_apply_completed_mask {mask_type{1} << 16 }

static constexpr size_type	linop_factory_generate_started { 17 }

static constexpr mask_type	linop_factory_generate_started_mask {mask_type{1} << 17 }

static constexpr size_type	linop_factory_generate_completed { 18 }

static constexpr mask_type	linop_factory_generate_completed_mask {mask_type{1} << 18 }

static constexpr size_type	criterion_check_started { 19 }

static constexpr mask_type	criterion_check_started_mask {mask_type{1} << 19 }

static constexpr size_type	criterion_check_completed { 20 }

static constexpr mask_type	criterion_check_completed_mask {mask_type{1} << 20 }

static constexpr size_type	iteration_complete {21}

static constexpr mask_type	iteration_complete_mask {mask_type{1} << 21}

static constexpr size_type	polymorphic_object_move_started { 22 }

static constexpr mask_type	polymorphic_object_move_started_mask {mask_type{1} << 22 }

static constexpr size_type	polymorphic_object_move_completed { 23 }

static constexpr mask_type	polymorphic_object_move_completed_mask {mask_type{1} << 23 }

static constexpr size_type	batch_linop_factory_generate_started { 24 }

static constexpr mask_type	batch_linop_factory_generate_started_mask {mask_type{1} << 24 }

static constexpr size_type	batch_linop_factory_generate_completed { 25 }

static constexpr mask_type	batch_linop_factory_generate_completed_mask {mask_type{1} << 25 }

static constexpr size_type	batch_solver_completed {26}

static constexpr mask_type	batch_solver_completed_mask {mask_type{1} << 26}

static constexpr mask_type	executor_events_mask

static constexpr mask_type	operation_events_mask

static constexpr mask_type	polymorphic_object_events_mask

static constexpr mask_type	linop_events_mask

static constexpr mask_type	linop_factory_events_mask

static constexpr mask_type	batch_linop_factory_events_mask

static constexpr mask_type	criterion_events_mask

Additional Inherited Members
Protected Member Functions inherited from gko::log::Logger
virtual void	on_polymorphic_object_create_started (const Executor exec, const PolymorphicObject po) const

virtual void	on_polymorphic_object_create_completed (const Executor exec, const PolymorphicObject input, const PolymorphicObject *output) const

virtual void	on_polymorphic_object_deleted (const Executor exec, const PolymorphicObject po) const

virtual void	on_batch_linop_factory_generate_started (const batch::BatchLinOpFactory factory, const batch::BatchLinOp input) const

virtual void	on_batch_linop_factory_generate_completed (const batch::BatchLinOpFactory factory, const batch::BatchLinOp input, const batch::BatchLinOp *output) const

virtual void	on_batch_solver_completed (const array< int > &iters, const array< double > &residual_norms) const

virtual void	on_batch_solver_completed (const array< int > &iters, const array< float > &residual_norms) const

virtual void	on_batch_solver_completed (const array< int > &iters, const array< gko::half > &residual_norms) const

	Logger (std::shared_ptr< const gko::Executor > exec, const mask_type &enabled_events=all_events_mask)

	Logger (const mask_type &enabled_events=all_events_mask)

Detailed Description

This Logger can be used to annotate the execution of Ginkgo functionality with profiler-specific ranges. It currently supports TAU, VTune, NSightSystems (NVTX) and rocPROF(ROCTX) and custom profiler hooks.

The Logger should be attached to the Executor that is being used to run the application for a full, program-wide annotation, or to individual objects to only highlight events caused directly by them (not operations and memory allocations though)

Member Function Documentation

◆ create_custom()

static std::shared_ptr< ProfilerHook > gko::log::ProfilerHook::create_custom	(	hook_function	begin,
		hook_function	end
	)

static

Creates a logger annotating Ginkgo events with a custom set of functions for range begin and end.

◆ create_for_executor()

static std::shared_ptr< ProfilerHook > gko::log::ProfilerHook::create_for_executor ( std::shared_ptr< const Executor > exec )

static

Creates a logger annotating Ginkgo events with the most suitable backend for the given executor: NVTX for NSight Systems in CUDA, ROCTX for rocprof in HIP, TAU for everything else.

◆ create_nested_summary()

static std::shared_ptr< ProfilerHook > gko::log::ProfilerHook::create_nested_summary	(	std::shared_ptr< Timer >	timer = `std::make_shared< CpuTimer >()`,
		std::unique_ptr< NestedSummaryWriter >	writer = `std::make_unique< TableSummaryWriter >()`,
		bool	debug_check_nesting = `false`
	)

static

Creates a logger measuring the runtime of Ginkgo events in a nested fashion and printing a summary when it is destroyed.

Parameters

timer	The timer used to record time points.
writer	The NestedSummaryWriter to receive the performance results.
debug_check_nesting	Enable this flag if the output looks like it might contain incorrect nesting. This increases the overhead slightly, but recognizes mismatching push/pop pairs on the range stack.

Note: For this logger to provide reliable GPU timings, either use Timer::create_for_executor or enable synchronization via set_synchronization(true).

◆ create_nvtx()

static std::shared_ptr< ProfilerHook > gko::log::ProfilerHook::create_nvtx ( uint32 color_argb = color_yellow_argb )

static

Creates a logger annotating Ginkgo events with NVTX ranges for CUDA.

Parameters

color_argb The color of the NVTX ranges in the NSight Systems output. It has to be a 32 bit packed ARGB value.

◆ create_roctx()

static std::shared_ptr< ProfilerHook > gko::log::ProfilerHook::create_roctx ( )

static

Creates a logger annotating Ginkgo events with ROCTX ranges for HIP.

◆ create_summary()

static std::shared_ptr< ProfilerHook > gko::log::ProfilerHook::create_summary	(	std::shared_ptr< Timer >	timer = `std::make_shared< CpuTimer >()`,
		std::unique_ptr< SummaryWriter >	writer = `std::make_unique< TableSummaryWriter >()`,
		bool	debug_check_nesting = `false`
	)

static

Creates a logger measuring the runtime of Ginkgo events and printing a summary when it is destroyed.

Parameters

timer	The timer used to record time points.
writer	The SummaryWriter to receive the performance results.
debug_check_nesting	Enable this flag if the output looks like it might contain incorrect nesting. This increases the overhead slightly, but recognizes mismatching push/pop pairs on the range stack.

Note: For this logger to provide reliable GPU timings, either use Timer::create_for_executor or enable synchronization via set_synchronization(true).

◆ create_tau()

static std::shared_ptr< ProfilerHook > gko::log::ProfilerHook::create_tau ( bool initialize = true )

static

Creates a logger annotating Ginkgo events with TAU ranges via PerfStubs.

Parameters

initialize Should we call TAU's initialization and finalization functions, or does the application take care of it? The initialization will happen immediately, the finalization at program exit.

◆ create_vtune()

static std::shared_ptr< ProfilerHook > gko::log::ProfilerHook::create_vtune ( )

static

Creates a logger annotating Ginkgo events with VTune ITT ranges.

◆ needs_propagation()

bool gko::log::ProfilerHook::needs_propagation ( ) const

overridevirtual

Returns true if this logger, when attached to an Executor, needs to be forwarded all events from objects on this executor.

Reimplemented from gko::log::Logger.

◆ on_allocation_completed()

void gko::log::ProfilerHook::on_allocation_completed	(	const gko::Executor *	exec,
		const gko::size_type &	num_bytes,
		const gko::uintptr &	location
	)		const

overridevirtual

Executor's allocation completed event.

Parameters

exec	the executor used
num_bytes	the number of bytes allocated
location	the address at which the data was allocated

Reimplemented from gko::log::Logger.

◆ on_allocation_started()

void gko::log::ProfilerHook::on_allocation_started	(	const gko::Executor *	exec,
		const gko::size_type &	num_bytes
	)		const

overridevirtual

Executor's allocation started event.

Parameters

exec	the executor used
num_bytes	the number of bytes to allocate

Reimplemented from gko::log::Logger.

◆ on_copy_completed()

void gko::log::ProfilerHook::on_copy_completed	(	const gko::Executor *	exec_from,
		const gko::Executor *	exec_to,
		const gko::uintptr &	loc_from,
		const gko::uintptr &	loc_to,
		const gko::size_type &	num_bytes
	)		const

overridevirtual

Executor's copy completed event.

Parameters

exec_from	the executor copied from
exec_to	the executor copied to
loc_from	the address at which the data was copied from
loc_to	the address at which the data was copied to
num_bytes	the number of bytes copied

Reimplemented from gko::log::Logger.

◆ on_copy_started()

void gko::log::ProfilerHook::on_copy_started	(	const gko::Executor *	exec_from,
		const gko::Executor *	exec_to,
		const gko::uintptr &	loc_from,
		const gko::uintptr &	loc_to,
		const gko::size_type &	num_bytes
	)		const

overridevirtual

Executor's copy started event.

Parameters

exec_from	the executor to be copied from
exec_to	the executor to be copied to
loc_from	the address at which the data will be copied from
loc_to	the address at which the data will be copied to
num_bytes	the number of bytes to be copied

Reimplemented from gko::log::Logger.

◆ on_criterion_check_completed() [1/2]

void gko::log::ProfilerHook::on_criterion_check_completed	(	const stop::Criterion *	criterion,
		const size_type &	it,
		const LinOp *	r,
		const LinOp *	tau,
		const LinOp *	implicit_tau_sq,
		const LinOp *	x,
		const uint8 &	stopping_id,
		const bool &	set_finalized,
		const array< stopping_status > *	status,
		const bool &	one_changed,
		const bool &	all_converged
	)		const

overridevirtual

stop::Criterion's check completed event. Parameters are the Criterion, the stoppingId, the finalized boolean, the stopping status, plus the output one_changed boolean and output all_converged boolean.

Parameters

criterion	the criterion used
it	the current iteration count
r	the residual
tau	the residual norm
implicit_tau_sq	the implicit residual norm squared
x	the solution
stopping_id	the id of the stopping criterion
set_finalized	whether this finalizes the iteration
status	the stopping status of the right hand sides
one_changed	whether at least one right hand side converged or not
all_converged	whether all right hand sides are converged

Reimplemented from gko::log::Logger.

◆ on_criterion_check_completed() [2/2]

void gko::log::ProfilerHook::on_criterion_check_completed	(	const stop::Criterion *	criterion,
		const size_type &	it,
		const LinOp *	r,
		const LinOp *	tau,
		const LinOp *	x,
		const uint8 &	stopping_id,
		const bool &	set_finalized,
		const array< stopping_status > *	status,
		const bool &	one_changed,
		const bool &	all_converged
	)		const

overridevirtual

stop::Criterion's check completed event. Parameters are the Criterion, the stoppingId, the finalized boolean, the stopping status, plus the output one_changed boolean and output all_converged boolean.

Parameters

criterion	the criterion used
it	the current iteration count
r	the residual
tau	the residual norm
x	the solution
stopping_id	the id of the stopping criterion
set_finalized	whether this finalizes the iteration
status	the stopping status of the right hand sides
one_changed	whether at least one right hand side converged or not
all_converged	whether all right hand sides

Note: The on_criterion_check_completed function that this macro declares is deprecated. Please use the one with the additional implicit_tau_sq parameter as below.

Reimplemented from gko::log::Logger.

◆ on_criterion_check_started()

void gko::log::ProfilerHook::on_criterion_check_started	(	const stop::Criterion *	criterion,
		const size_type &	it,
		const LinOp *	r,
		const LinOp *	tau,
		const LinOp *	x,
		const uint8 &	stopping_id,
		const bool &	set_finalized
	)		const

overridevirtual

stop::Criterion's check started event.

Parameters

criterion	the criterion used
it	the current iteration count
r	the residual
tau	the residual norm
x	the solution
stopping_id	the id of the stopping criterion
set_finalized	whether this finalizes the iteration

Reimplemented from gko::log::Logger.

◆ on_free_completed()

void gko::log::ProfilerHook::on_free_completed	(	const gko::Executor *	exec,
		const gko::uintptr &	location
	)		const

overridevirtual

Executor's free completed event.

Parameters

exec	the executor used
location	the address at which the data was freed

Reimplemented from gko::log::Logger.

◆ on_free_started()

void gko::log::ProfilerHook::on_free_started	(	const gko::Executor *	exec,
		const gko::uintptr &	location
	)		const

overridevirtual

Executor's free started event.

Parameters

exec	the executor used
location	the address at which the data will be freed

Reimplemented from gko::log::Logger.

◆ on_iteration_complete() [1/3]

void gko::log::ProfilerHook::on_iteration_complete	(	const LinOp *	solver,
		const LinOp *	b,
		const LinOp *	x,
		const size_type &	it,
		const LinOp *	r,
		const LinOp *	tau,
		const LinOp *	implicit_tau_sq,
		const array< stopping_status > *	status,
		bool	stopped
	)		const

overridevirtual

Register the iteration_complete event which logs every completed iterations.

Parameters

solver	the solver executing the iteration
b	the right-hand-side vector
x	the solution vector
it	the current iteration count
r	the residual (optional)
tau	the implicit residual norm squared (optional)
implicit_tau_sq	the residual norm (optional)
status	the stopping status of the right hand sides (optional)
stopped	whether all right hand sides have stopped (invalid if status is not provided)

Reimplemented from gko::log::Logger.

◆ on_iteration_complete() [2/3]

void gko::log::ProfilerHook::on_iteration_complete	(	const LinOp *	solver,
		const size_type &	it,
		const LinOp *	r,
		const LinOp *	x,
		const LinOp *	tau
	)		const

overridevirtual

Register the iteration_complete event which logs every completed iterations.

Parameters

it	the current iteration count
r	the residual
x	the solution vector (optional)
tau	the residual norm (optional)

Warning: This on_iteration_complete function that this macro declares is deprecated. Please use the version with the stopping information.

Reimplemented from gko::log::Logger.

◆ on_iteration_complete() [3/3]

void gko::log::ProfilerHook::on_iteration_complete	(	const LinOp *	solver,
		const size_type &	it,
		const LinOp *	r,
		const LinOp *	x,
		const LinOp *	tau,
		const LinOp *	implicit_tau_sq
	)		const

overridevirtual

Register the iteration_complete event which logs every completed iterations.

Parameters

it	the current iteration count
r	the residual
x	the solution vector (optional)
tau	the residual norm (optional)
implicit_tau_sq	the implicit residual norm squared (optional)

Warning: This on_iteration_complete function that this macro declares is deprecated. Please use the version with the stopping information.

Reimplemented from gko::log::Logger.

◆ on_linop_advanced_apply_completed()

void gko::log::ProfilerHook::on_linop_advanced_apply_completed	(	const LinOp *	A,
		const LinOp *	alpha,
		const LinOp *	b,
		const LinOp *	beta,
		const LinOp *	x
	)		const

overridevirtual

LinOp's advanced apply completed event.

Parameters

A	the system matrix
alpha	scaling of the result of op(b)
b	the input vector(s)
beta	scaling of the input x
x	the output vector(s)

Reimplemented from gko::log::Logger.

◆ on_linop_advanced_apply_started()

void gko::log::ProfilerHook::on_linop_advanced_apply_started	(	const LinOp *	A,
		const LinOp *	alpha,
		const LinOp *	b,
		const LinOp *	beta,
		const LinOp *	x
	)		const

overridevirtual

LinOp's advanced apply started event.

Parameters

A	the system matrix
alpha	scaling of the result of op(b)
b	the input vector(s)
beta	scaling of the input x
x	the output vector(s)

Reimplemented from gko::log::Logger.

◆ on_linop_apply_completed()

void gko::log::ProfilerHook::on_linop_apply_completed	(	const LinOp *	A,
		const LinOp *	b,
		const LinOp *	x
	)		const

overridevirtual

LinOp's apply completed event.

Parameters

A	the system matrix
b	the input vector(s)
x	the output vector(s)

Reimplemented from gko::log::Logger.

◆ on_linop_apply_started()

void gko::log::ProfilerHook::on_linop_apply_started	(	const LinOp *	A,
		const LinOp *	b,
		const LinOp *	x
	)		const

overridevirtual

LinOp's apply started event.

Parameters

A	the system matrix
b	the input vector(s)
x	the output vector(s)

Reimplemented from gko::log::Logger.

◆ on_linop_factory_generate_completed()

void gko::log::ProfilerHook::on_linop_factory_generate_completed	(	const LinOpFactory *	factory,
		const LinOp *	input,
		const LinOp *	output
	)		const

overridevirtual

LinOp Factory's generate completed event.

Parameters

factory	the factory used
input	the LinOp object used as input for the generation (usually a system matrix)
output	the generated LinOp object

Reimplemented from gko::log::Logger.

◆ on_linop_factory_generate_started()

void gko::log::ProfilerHook::on_linop_factory_generate_started	(	const LinOpFactory *	factory,
		const LinOp *	input
	)		const

overridevirtual

LinOp Factory's generate started event.

Parameters

factory	the factory used
input	the LinOp object used as input for the generation (usually a system matrix)

Reimplemented from gko::log::Logger.

◆ on_operation_completed()

void gko::log::ProfilerHook::on_operation_completed	(	const Executor *	exec,
		const Operation *	op
	)		const

overridevirtual

Executor's operation completed event (method run).

Parameters

exec	the executor used
op	the completed operation

Note: For the GPU, to be certain that the operation completed it is required to call synchronize. This burden falls on the logger. Most of the loggers will do lightweight logging, and therefore this operation for the GPU just notes that the Operation has been sent to the GPU.

Reimplemented from gko::log::Logger.

◆ on_operation_launched()

void gko::log::ProfilerHook::on_operation_launched	(	const Executor *	exec,
		const Operation *	op
	)		const

overridevirtual

Executor's operation launched event (method run).

Parameters

exec	the executor used
op	the operation launched

Reimplemented from gko::log::Logger.

◆ on_polymorphic_object_copy_completed()

void gko::log::ProfilerHook::on_polymorphic_object_copy_completed	(	const Executor *	exec,
		const PolymorphicObject *	input,
		const PolymorphicObject *	output
	)		const

overridevirtual

PolymorphicObject's copy completed event.

Parameters

exec	the executor used
input	the PolymorphicObject to be copied from
output	the PolymorphicObject to be copied to

Reimplemented from gko::log::Logger.

◆ on_polymorphic_object_copy_started()

void gko::log::ProfilerHook::on_polymorphic_object_copy_started	(	const Executor *	exec,
		const PolymorphicObject *	input,
		const PolymorphicObject *	output
	)		const

overridevirtual

PolymorphicObject's copy started event.

Parameters

exec	the executor used
input	the PolymorphicObject to be copied from
output	the PolymorphicObject to be copied to

Reimplemented from gko::log::Logger.

◆ on_polymorphic_object_move_completed()

void gko::log::ProfilerHook::on_polymorphic_object_move_completed	(	const Executor *	exec,
		const PolymorphicObject *	input,
		const PolymorphicObject *	output
	)		const

overridevirtual

PolymorphicObject's move completed event.

Parameters

exec	the executor used
input	the PolymorphicObject to be move from
output	the PolymorphicObject to be move into

Reimplemented from gko::log::Logger.

◆ on_polymorphic_object_move_started()

void gko::log::ProfilerHook::on_polymorphic_object_move_started	(	const Executor *	exec,
		const PolymorphicObject *	input,
		const PolymorphicObject *	output
	)		const

overridevirtual

PolymorphicObject's move started event.

Parameters

exec	the executor used
input	the PolymorphicObject to be move from
output	the PolymorphicObject to be move into

Reimplemented from gko::log::Logger.

◆ set_object_name()

void gko::log::ProfilerHook::set_object_name	(	ptr_param< const PolymorphicObject >	obj,
		std::string	name
	)

Sets the name for an object to be profiled. Every instance of that object in the profile will be replaced by the name instead of its runtime type.

Parameters

obj	the object
name	its name

◆ set_synchronization()

void gko::log::ProfilerHook::set_synchronization ( bool synchronize )

Should the events call executor->synchronize on operations and copy/allocation? This leads to a certain overhead, but makes the execution timeline of kernels synchronous.

◆ user_range()

profiling_scope_guard gko::log::ProfilerHook::user_range ( const char * name ) const

Creates a scope guard for a user-defined range to be included in the profile.

Parameters

name	the name of the range

Returns: the scope guard. It will begin a range immediately and end it at the end of its scope.

Member Data Documentation

◆ color_yellow_argb

constexpr uint32 gko::log::ProfilerHook::color_yellow_argb = 0xFFFFCB05U

staticconstexpr

The Ginkgo yellow background color as packed 32 bit ARGB value.

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

ginkgo/core/log/profiler_hook.hpp

ProfilerHook Class Reference

ProfilerHook Class Reference#

Classes

Public Types

Public Member Functions

Static Public Member Functions

Static Public Attributes

Additional Inherited Members

Detailed Description

Member Function Documentation

◆ create_custom()

◆ create_for_executor()

◆ create_nested_summary()

◆ create_nvtx()

◆ create_roctx()

◆ create_summary()

◆ create_tau()

◆ create_vtune()

◆ needs_propagation()

◆ on_allocation_completed()

◆ on_allocation_started()

◆ on_copy_completed()

◆ on_copy_started()

◆ on_criterion_check_completed() [1/2]

◆ on_criterion_check_completed() [2/2]

◆ on_criterion_check_started()

◆ on_free_completed()

◆ on_free_started()

◆ on_iteration_complete() [1/3]

◆ on_iteration_complete() [2/3]

◆ on_iteration_complete() [3/3]

◆ on_linop_advanced_apply_completed()

◆ on_linop_advanced_apply_started()

◆ on_linop_apply_completed()

◆ on_linop_apply_started()

◆ on_linop_factory_generate_completed()

◆ on_linop_factory_generate_started()

◆ on_operation_completed()

◆ on_operation_launched()

◆ on_polymorphic_object_copy_completed()

◆ on_polymorphic_object_copy_started()

◆ on_polymorphic_object_move_completed()

◆ on_polymorphic_object_move_started()

◆ set_object_name()

◆ set_synchronization()

◆ user_range()

Member Data Documentation

◆ color_yellow_argb