Classes
class	CounterBatch
	Private counters for one worker in a worksharing phase. Nested batches merge into the enclosing worker's batch. The outer batch merges under one lock after the loop; hydraulic arithmetic is unaffected. When profiling is disabled, no counters are initialized or merged. More...

struct	GatedTimer

struct	ProcessPhaseName

class	ProcessTimer

struct	ScopedTimer

struct	SigmaEvaluations

Typedefs
using	FvCounts = std::array<long, n_fv_counter_count>

using	AllocationSnapshot = void ()(std::uint64_t, std::uint64_t*)

Enumerations
enum	FvCounter : std::size_t { n_fv_substep , n_fv_census , n_fv_census_face , n_fv_invert , n_fv_savestate , n_fv_restore , n_fv_structref , n_fv_alg_visit , n_fv_alg_passthru , n_fv_alg_solve , n_fv_alg_resid , n_fv_alg_flux , n_fv_geom_area , n_fv_geom_width , n_fv_geom_i1 , n_fv_geom_hydrad , n_fv_macro_cycles , n_fv_macro_rejected , n_fv_counter_count }

enum class	ProcessPhase : std::size_t { surface_faces , surface_cells , surface_rebuild , surface_boundary , surface_exchange , surface_publish , surface_species , surface_species_cells , surface_reactions , gw_faces , gw_cells , gw_node_exchange , gw_dt , gw_infiltration , gw_species , gw_reactions , gw_sigma_map , heat_columns , surface_quality , runoff , groundwater , lid , lid_nodes , richards , mesh_snow , subcatch_snow , common_source , source_trial , outfall , source_results , engine_source_results , count }

enum class	ProcessCounter : std::size_t { sigma_evaluations , sigma_firings , sigma_le_1 , sigma_le_8 , sigma_le_32 , sigma_le_64 , sigma_gt_64 , sigma_map_refreshes , water_trial_copies , source_stage_copies , rk2_snapshots , mesh_snow_copies , count }

Functions
long	value (FvCounter counter) noexcept
	Read one process-global completed FV counter total.

void	reset_fv () noexcept
	Zero FV phase timers and completed counter totals.

void	dump_fv () noexcept
	One machine-scrapeable line for the FV phase split.

std::chrono::steady_clock::time_point	now () noexcept
	Sample the steady monotonic clock regardless of profiling enablement.

double	since (std::chrono::steady_clock::time_point t0) noexcept
	Measure elapsed wall time from a steady-clock sample.

bool	enabled () noexcept
	Read the process-cached profiling environment gate.

void	reset_load () noexcept
	Zero load/open/initialize/start phase accumulators.

void	dump_load () noexcept
	One machine-scrapeable line per phase on stderr.

void	count (FvCounter counter, long n=1) noexcept
	Accumulate a batch of events into worker-private or serial global counters.

AllocationSnapshot	allocation_snapshot () noexcept

void	count (ProcessCounter counter, unsigned long long n=1) noexcept

unsigned long long	value (ProcessCounter counter) noexcept

void	clear_process_totals () noexcept

void	reset_process () noexcept

void	begin_process_model (const void *model) noexcept

void	dump_process () noexcept

void	end_process_model (const void *model, bool report=true) noexcept

Variables
double	sec_2d_window = 0.0

double	sec_2d_advance = 0.0

double	sec_1d_step = 0.0

double	sec_open_prescan2d = 0.0

double	sec_open_read = 0.0

double	sec_read_dispatch = 0.0

double	sec_open_resolve = 0.0

double	sec_open_validate = 0.0

double	sec_res_extfiles = 0.0

double	sec_res_tables = 0.0

double	sec_res_transects = 0.0

double	sec_res_xsect = 0.0

double	sec_res_shrink = 0.0

double	sec_init_state = 0.0

double	sec_init_hydraulics = 0.0

double	sec_init_hydrology = 0.0

double	sec_init_quality = 0.0

double	sec_init_geometry = 0.0

double	sec_start_iface = 0.0

double	sec_start_plugins = 0.0

double	sec_fv_census = 0.0

double	sec_fv_flux = 0.0

double	sec_fv_nodesolve = 0.0

double	sec_fv_positivity = 0.0

double	sec_fv_cellupdate = 0.0

double	sec_fv_nodeupdate = 0.0

double	sec_fv_refreshdep = 0.0

double	sec_fv_savestate = 0.0

double	sec_fv_restore = 0.0

double	sec_fv_structref = 0.0

double	sec_fv_bndcallback = 0.0

double	sec_fv_rebuild = 0.0

double	sec_fv_reconstruct = 0.0

double	sec_fv_ltsfire = 0.0

double	sec_fv_settle = 0.0

double	sec_fv_tier = 0.0

FvCounts	fv_counts {}

thread_local FvCounts *	local_fv_counts = nullptr

std::mutex	fv_counts_mutex

thread_local double	nested_wall = 0.0
	ScopedTimer that reads the clock only when OPENSWMM_PERF is set.

constexpr std::array< ProcessPhaseName, std::size_t(ProcessPhase::count)>	process_names

std::array< std::atomic< double >, std::size_t(ProcessPhase::count)>	process_seconds {}

std::array< std::atomic< unsigned long long >, std::size_t(ProcessPhase::count)>	process_calls {}

std::array< std::atomic< unsigned long long >, std::size_t(ProcessPhase::count)>	process_allocations {}

std::array< std::atomic< unsigned long long >, std::size_t(ProcessPhase::count)>	process_allocated_bytes {}

std::atomic< bool >	process_attribution_valid {true}

std::mutex	process_models_mutex

std::unordered_set< const void * >	process_models

bool	process_tracking_failed = false

constexpr std::array< const char *, std::size_t(ProcessCounter::count)>	process_counter_names

std::array< std::atomic< unsigned long long >, std::size_t(ProcessCounter::count)>	process_counts {}

Typedef Documentation

◆ AllocationSnapshot

using openswmm::perf::AllocationSnapshot = void (*)(std::uint64_t*, std::uint64_t*)

◆ FvCounts

using openswmm::perf::FvCounts = std::array<long, n_fv_counter_count>

Enumeration Type Documentation

◆ FvCounter

enum openswmm::perf::FvCounter : std::size_t

Enumerator
n_fv_substep
n_fv_census
n_fv_census_face
n_fv_invert
n_fv_savestate
n_fv_restore
n_fv_structref
n_fv_alg_visit
n_fv_alg_passthru
n_fv_alg_solve
n_fv_alg_resid
n_fv_alg_flux
n_fv_geom_area
n_fv_geom_width
n_fv_geom_i1
n_fv_geom_hydrad
n_fv_macro_cycles
n_fv_macro_rejected
n_fv_counter_count

◆ ProcessCounter

enum class openswmm::perf::ProcessCounter : std::size_t

strong

Enumerator
sigma_evaluations
sigma_firings
sigma_le_1
sigma_le_8
sigma_le_32
sigma_le_64
sigma_gt_64
sigma_map_refreshes
water_trial_copies
source_stage_copies
rk2_snapshots
mesh_snow_copies
count

◆ ProcessPhase

enum class openswmm::perf::ProcessPhase : std::size_t

strong

Enumerator
surface_faces
surface_cells
surface_rebuild
surface_boundary
surface_exchange
surface_publish
surface_species
surface_species_cells
surface_reactions
gw_faces
gw_cells
gw_node_exchange
gw_dt
gw_infiltration
gw_species
gw_reactions
gw_sigma_map
heat_columns
surface_quality
runoff
groundwater
lid
lid_nodes
richards
mesh_snow
subcatch_snow
common_source
source_trial
outfall
source_results
engine_source_results
count

Function Documentation

◆ allocation_snapshot()

AllocationSnapshot openswmm::perf::allocation_snapshot ( )

inlinenoexcept

Here is the caller graph for this function:

◆ begin_process_model()

void openswmm::perf::begin_process_model ( const void * model )

inlinenoexcept

Here is the call graph for this function:

◆ clear_process_totals()

void openswmm::perf::clear_process_totals ( )

inlinenoexcept

Here is the call graph for this function:

Here is the caller graph for this function:

◆ count() [1/2]

void openswmm::perf::count	(	FvCounter	counter,
		long	n = 1 )

inlinenoexcept

Accumulate a batch of events into worker-private or serial global counters.

No-op when profiling is disabled. Parallel callers must install a CounterBatch; without one updates global fv_counts without a lock.

Parameters

[in]	counter	Valid FvCounter below n_fv_counter_count; no bounds check.
[in]	n	Increment, default one; may be zero/negative. Caller prevents signed-long overflow.

Here is the call graph for this function:

Here is the caller graph for this function:

◆ count() [2/2]

void openswmm::perf::count	(	ProcessCounter	counter,
		unsigned long long	n = 1 )

inlinenoexcept

Here is the call graph for this function:

◆ dump_fv()

void openswmm::perf::dump_fv ( )

inlinenoexcept

One machine-scrapeable line for the FV phase split.

Same [PERF-FV] key=value contract as dump_load(); emitted from close() when OPENSWMM_PERF is set and the FV solver actually ran. Each phase is SELF time: a phase that calls another timed phase is charged only for its own work (see nested_wall), so the phases partition the time rather than overlapping it. total is their sum, NOT the router step — the difference between the two is unattributed time and is itself the useful signal when a phase is missing. It cannot be negative; if it ever is, a timer is nested inside a scope that does not go through GatedTimer.

Note: Emits to stderr whenever called; the caller checks enabled() and serializes access to totals. Printing failures are not returned.

Here is the call graph for this function:

◆ dump_load()

void openswmm::perf::dump_load ( )

inlinenoexcept

One machine-scrapeable line per phase on stderr.

Emitted from close() so all three benchmark cuts (open only, open+initialize, full) report. Format is [PERF-LOAD] key=value pairs in seconds; tests/benchmarks/scripts/ parses it.

Note: Emits to stderr whenever called; the caller checks enabled() and serializes access to totals. Printing failures are not returned.

◆ dump_process()

void openswmm::perf::dump_process ( )

inlinenoexcept

Here is the call graph for this function:

Here is the caller graph for this function:

◆ enabled()

bool openswmm::perf::enabled ( )

inlinenoexcept

Read the process-cached profiling environment gate.

Subsequent environment changes are ignored. C++ function-local static initialization is thread-safe.

Returns: True if OPENSWMM_PERF existed on the first call, including an empty or zero string.

Here is the caller graph for this function:

◆ end_process_model()

void openswmm::perf::end_process_model	(	const void *	model,
		bool	report = true )

inlinenoexcept

Here is the call graph for this function:

◆ now()

std::chrono::steady_clock::time_point openswmm::perf::now ( )

inlinenoexcept

Sample the steady monotonic clock regardless of profiling enablement.

Returns: Current steady_clock time point for use with since().

Here is the caller graph for this function:

◆ reset_fv()

void openswmm::perf::reset_fv ( )

inlinenoexcept

Zero FV phase timers and completed counter totals.

Does not reset sec_1d_step or active thread-local batches. Caller invokes only between worksharing regions with no concurrent counter/timer use.

◆ reset_load()

void openswmm::perf::reset_load ( )

inlinenoexcept

Zero load/open/initialize/start phase accumulators.

Does not reset runtime FV/1D/2D totals or environment gate; caller serializes reset with timer updates.

◆ reset_process()

void openswmm::perf::reset_process ( )

inlinenoexcept

Here is the call graph for this function:

◆ since()

double openswmm::perf::since ( std::chrono::steady_clock::time_point t0 )

inlinenoexcept

Measure elapsed wall time from a steady-clock sample.

Parameters

[in] t0 Earlier steady-clock time point; a future value yields a negative duration.

Returns: now()-t0 in fractional seconds, regardless of profiling enablement.

Here is the call graph for this function:

Here is the caller graph for this function:

◆ value() [1/2]

long openswmm::perf::value ( FvCounter counter )

inlinenoexcept

Read one process-global completed FV counter total.

Read only between worksharing regions or under external synchronization.

Parameters

[in] counter Valid FvCounter below n_fv_counter_count; no bounds check.

Returns: Stored count, with no locking or merge of currently active thread-local batches.

Here is the caller graph for this function:

◆ value() [2/2]

unsigned long long openswmm::perf::value ( ProcessCounter counter )

inlinenoexcept

Variable Documentation

◆ fv_counts

FvCounts openswmm::perf::fv_counts {}

inline

◆ fv_counts_mutex

std::mutex openswmm::perf::fv_counts_mutex

inline

◆ local_fv_counts

thread_local FvCounts* openswmm::perf::local_fv_counts = nullptr

inline

◆ nested_wall

thread_local double openswmm::perf::nested_wall = 0.0

inline

ScopedTimer that reads the clock only when OPENSWMM_PERF is set.

ScopedTimer is used at per-routing-step and per-load-phase call sites, where two unconditional clock reads are free. The FV phase timers are not in that regime: a 30 h / 5 s run is ~21,600 routing steps, each of which can carry hundreds of substeps, each carrying a dozen phases. At ~20 ns a read that is seconds of pure instrument on a run that has nothing to do with profiling. enabled() is a cached static bool, so the gated form costs a predicted branch instead. Wall time booked by GatedTimers that opened and closed inside the one currently running on this thread. Read and reset by each timer so that what lands in the accumulator is SELF time.

Without this the phase split double-counts wherever one timed phase calls another, which it does in three places today: restoreState and settleAccumulators both call refreshDepths, and runMacroCycle's boundary callback sits inside the settle window. The visible symptom was total exceeding the routing step it is a breakdown of — the LTS rows of the phase table reported between -7.9 % and -18.1 % "unattributed" time, which is not a quantity that can be negative. A breakdown whose parts can exceed the whole cannot rank optimization targets, which is the only reason these timers exist.

Thread-local because the accumulators are serial-path today but the flux sweep is slated to go parallel; a shared counter would then attribute one thread's children to another thread's parent. Nothing here is atomic and nothing needs to be: a timer only ever touches its own thread's tally.

◆ process_allocated_bytes

std::array<std::atomic<unsigned long long>, std::size_t(ProcessPhase::count)> openswmm::perf::process_allocated_bytes {}

inline

◆ process_allocations

std::array<std::atomic<unsigned long long>, std::size_t(ProcessPhase::count)> openswmm::perf::process_allocations {}

inline

◆ process_attribution_valid

std::atomic<bool> openswmm::perf::process_attribution_valid {true}

inline

◆ process_calls

std::array<std::atomic<unsigned long long>, std::size_t(ProcessPhase::count)> openswmm::perf::process_calls {}

inline

◆ process_counter_names

std::array<const char*, std::size_t(ProcessCounter::count)> openswmm::perf::process_counter_names

inlineconstexpr

Initial value:

{{
    "sigma.evaluate", "sigma.firings", "sigma.hist.le1", "sigma.hist.2_8",
    "sigma.hist.9_32", "sigma.hist.33_64", "sigma.hist.gt64", "sigma_map.refresh",
    "copy.water_trial", "copy.source_stage", "copy.rk2_snapshot", "copy.mesh_snow"
}}

◆ process_counts

std::array<std::atomic<unsigned long long>, std::size_t(ProcessCounter::count)> openswmm::perf::process_counts {}

inline

◆ process_models

std::unordered_set<const void*> openswmm::perf::process_models

inline

◆ process_models_mutex

std::mutex openswmm::perf::process_models_mutex

inline

◆ process_names

std::array<ProcessPhaseName, std::size_t(ProcessPhase::count)> openswmm::perf::process_names

inlineconstexpr

Initial value:

{{
    {"2D", "faces"}, {"2D", "cells"}, {"2D", "rebuild"}, {"2D", "boundary"},
    {"2D", "exchange"}, {"2D", "publish"}, {"2D", "species_faces"},
    {"2D", "species_cells"}, {"2D", "reactions"},
    {"GW2D", "faces"}, {"GW2D", "cells"}, {"GW2D", "node_exchange"},
    {"GW2D", "refresh_dt"}, {"GW2D", "infiltration"}, {"GW2D", "species"},
    {"GW2D", "reactions"}, {"GW2D", "sigma_map"}, {"GW2D", "heat_columns"},
    {"2D", "quality"}, {"HYD", "runoff"}, {"HYD", "groundwater"}, {"HYD", "lid"},
    {"HYD", "lid_nodes"}, {"HYD", "richards"}, {"HYD", "mesh_snow"},
    {"HYD", "subcatch_snow"}, {"HYD", "common_source"}, {"HYD", "source_trial"},
    {"HYD", "outfall"}, {"HYD", "source_results"}, {"HYD", "engine_source_results"}
}}

◆ process_seconds

std::array<std::atomic<double>, std::size_t(ProcessPhase::count)> openswmm::perf::process_seconds {}

inline

◆ process_tracking_failed

bool openswmm::perf::process_tracking_failed = false

inline

◆ sec_1d_step

double openswmm::perf::sec_1d_step = 0.0

inline

◆ sec_2d_advance

double openswmm::perf::sec_2d_advance = 0.0

inline

◆ sec_2d_window

double openswmm::perf::sec_2d_window = 0.0

inline

◆ sec_fv_bndcallback

double openswmm::perf::sec_fv_bndcallback = 0.0

inline

◆ sec_fv_cellupdate

double openswmm::perf::sec_fv_cellupdate = 0.0

inline

◆ sec_fv_census

double openswmm::perf::sec_fv_census = 0.0

inline

◆ sec_fv_flux

double openswmm::perf::sec_fv_flux = 0.0

inline

◆ sec_fv_ltsfire

double openswmm::perf::sec_fv_ltsfire = 0.0

inline

◆ sec_fv_nodesolve

double openswmm::perf::sec_fv_nodesolve = 0.0

inline

◆ sec_fv_nodeupdate

double openswmm::perf::sec_fv_nodeupdate = 0.0

inline

◆ sec_fv_positivity

double openswmm::perf::sec_fv_positivity = 0.0

inline

◆ sec_fv_rebuild

double openswmm::perf::sec_fv_rebuild = 0.0

inline

◆ sec_fv_reconstruct

double openswmm::perf::sec_fv_reconstruct = 0.0

inline

◆ sec_fv_refreshdep

double openswmm::perf::sec_fv_refreshdep = 0.0

inline

◆ sec_fv_restore

double openswmm::perf::sec_fv_restore = 0.0

inline

◆ sec_fv_savestate

double openswmm::perf::sec_fv_savestate = 0.0

inline

◆ sec_fv_settle

double openswmm::perf::sec_fv_settle = 0.0

inline

◆ sec_fv_structref

double openswmm::perf::sec_fv_structref = 0.0

inline

◆ sec_fv_tier

double openswmm::perf::sec_fv_tier = 0.0

inline

◆ sec_init_geometry

double openswmm::perf::sec_init_geometry = 0.0

inline

◆ sec_init_hydraulics

double openswmm::perf::sec_init_hydraulics = 0.0

inline

◆ sec_init_hydrology

double openswmm::perf::sec_init_hydrology = 0.0

inline

◆ sec_init_quality

double openswmm::perf::sec_init_quality = 0.0

inline

◆ sec_init_state

double openswmm::perf::sec_init_state = 0.0

inline

◆ sec_open_prescan2d

double openswmm::perf::sec_open_prescan2d = 0.0

inline

◆ sec_open_read

double openswmm::perf::sec_open_read = 0.0

inline

◆ sec_open_resolve

double openswmm::perf::sec_open_resolve = 0.0

inline

◆ sec_open_validate

double openswmm::perf::sec_open_validate = 0.0

inline

◆ sec_read_dispatch

double openswmm::perf::sec_read_dispatch = 0.0

inline

◆ sec_res_extfiles

double openswmm::perf::sec_res_extfiles = 0.0

inline

◆ sec_res_shrink

double openswmm::perf::sec_res_shrink = 0.0

inline

◆ sec_res_tables

double openswmm::perf::sec_res_tables = 0.0

inline

◆ sec_res_transects

double openswmm::perf::sec_res_transects = 0.0

inline

◆ sec_res_xsect

double openswmm::perf::sec_res_xsect = 0.0

inline

◆ sec_start_iface

double openswmm::perf::sec_start_iface = 0.0

inline

◆ sec_start_plugins

double openswmm::perf::sec_start_plugins = 0.0

inline

Classes

Typedefs

Enumerations

Functions

Variables

Typedef Documentation

◆ AllocationSnapshot

◆ FvCounts

Enumeration Type Documentation

◆ FvCounter

◆ ProcessCounter

◆ ProcessPhase

Function Documentation

◆ allocation_snapshot()

◆ begin_process_model()

◆ clear_process_totals()

◆ count() [1/2]

◆ count() [2/2]

◆ dump_fv()

◆ dump_load()

◆ dump_process()

◆ enabled()

◆ end_process_model()

◆ now()

◆ reset_fv()

◆ reset_load()

◆ reset_process()

◆ since()

◆ value() [1/2]

◆ value() [2/2]

Variable Documentation

◆ fv_counts

◆ fv_counts_mutex

◆ local_fv_counts

◆ nested_wall

◆ process_allocated_bytes

◆ process_allocations

◆ process_attribution_valid

◆ process_calls

◆ process_counter_names

◆ process_counts

◆ process_models

◆ process_models_mutex

◆ process_names

◆ process_seconds

◆ process_tracking_failed

◆ sec_1d_step

◆ sec_2d_advance

◆ sec_2d_window

◆ sec_fv_bndcallback

◆ sec_fv_cellupdate

◆ sec_fv_census

◆ sec_fv_flux

◆ sec_fv_ltsfire

◆ sec_fv_nodesolve

◆ sec_fv_nodeupdate

◆ sec_fv_positivity

◆ sec_fv_rebuild

◆ sec_fv_reconstruct

◆ sec_fv_refreshdep

◆ sec_fv_restore

◆ sec_fv_savestate

◆ sec_fv_settle

◆ sec_fv_structref

◆ sec_fv_tier

◆ sec_init_geometry

◆ sec_init_hydraulics

◆ sec_init_hydrology

◆ sec_init_quality

◆ sec_init_state

◆ sec_open_prescan2d

◆ sec_open_read

◆ sec_open_resolve

◆ sec_open_validate

◆ sec_read_dispatch

◆ sec_res_extfiles

◆ sec_res_shrink

◆ sec_res_tables

◆ sec_res_transects

◆ sec_res_xsect