openswmm::threadinfo Namespace Reference

Functions
int	logicalCpus ()
	Query OS-visible logical CPU concurrency.
int	ompMaxThreads ()
	Query the current calling-context OpenMP maximum.
int	ompAvailable ()
	Report whether engine OpenMP support was compiled.
int	perfCores ()
	Query macOS performance-cluster logical CPU count.
int	kokkosOmpThreads ()
	Read the engine-recorded Kokkos OpenMP team hint.
void	setKokkosOmpThreads (int n)
	Store the engine-recorded Kokkos OpenMP team hint.
int	hostReservedThreads ()
	Read host thread reservation from the current environment.
bool	isOversubscribed (int threads)
	Compare a team plus host reservation with known logical concurrency.
int	resolveRequested (int requested, const char *what, std::vector< std::string > *warnings)
	Resolve a thread request and optionally append oversubscription diagnostics.
int	dwThreads (int requested, int n_conduits, std::vector< std::string > *warnings)
	Resolve a dynamic-wave team request with model-size and automatic core caps.
int	twoDThreads (int requested, int n_triangles, std::vector< std::string > *warnings)
	Resolve a 2D marcher team request using core and triangle-count heuristics.

Variables
constexpr int	kMinConduitsPerThread = 100

Function Documentation

dwThreads()

int openswmm::threadinfo::dwThreads	(	int	requested,
		int	n_conduits,
		std::vector< std::string > *	warnings )

Resolve a dynamic-wave team request with model-size and automatic core caps.

Applies model-size cap to explicit and automatic requests, warning only for explicit reductions. Exactly requested=0 additionally caps to performance cores minus two on macOS when that count exceeds two. Negative requests use OpenMP maximum without that auto-only clamp.

Parameters

[in]	requested	Requested team size; <=0 resolves from OpenMP maximum. Exactly zero additionally enables automatic core-clamp heuristics.
[in]	n_conduits	Conduit count used in floor(count/100), normalized to a minimum one-worker cap.
[in,out]	warnings	Optional caller-owned sink; new messages append without clearing existing entries.

Returns

Positive recommended team size; does not set OpenMP or honor SWMM_DW_THREADS directly.

Here is the call graph for this function:

Here is the caller graph for this function:

hostReservedThreads()

int openswmm::threadinfo::hostReservedThreads

(

)

Read host thread reservation from the current environment.

Value is read on every call; trailing nonnumeric bytes accepted by atoi are not separately rejected. Caller avoids environment mutation races and out-of-int-range tokens.

Returns

Positive atoi prefix of OPENSWMM_HOST_RESERVED_THREADS, or zero when absent/empty/nonpositive.

Here is the caller graph for this function:

isOversubscribed()

bool openswmm::threadinfo::isOversubscribed

(

int

threads

)

Compare a team plus host reservation with known logical concurrency.

Parameters

[in]

threads

Team count; caller keeps threads+hostReservedThreads() representable as int.

Returns

True only when logical CPU count is known and the sum exceeds it; false when hardware count is unknown.

Here is the call graph for this function:

kokkosOmpThreads()

int openswmm::threadinfo::kokkosOmpThreads

(

)

Read the engine-recorded Kokkos OpenMP team hint.

Does not query or initialize the plugin/runtime.

Returns

Relaxed atomic stored hint; zero until set or after a nonpositive reset.

logicalCpus()

int openswmm::threadinfo::logicalCpus

(

)

Query OS-visible logical CPU concurrency.

Returns

std::thread::hardware_concurrency() converted to int; zero if unavailable.

Here is the caller graph for this function:

ompAvailable()

int openswmm::threadinfo::ompAvailable

(

)

Report whether engine OpenMP support was compiled.

Returns

One with SWMM_USE_OPENMP, zero otherwise; no runtime/plugin qualification.

ompMaxThreads()

int openswmm::threadinfo::ompMaxThreads

(

)

Query the current calling-context OpenMP maximum.

Does not create a team or guarantee that the runtime will provide this many workers.

Returns

Positive omp_get_max_threads() result, normalized to at least one; one without SWMM_USE_OPENMP.

Here is the call graph for this function:

Here is the caller graph for this function:

perfCores()

int openswmm::threadinfo::perfCores

(

)

Query macOS performance-cluster logical CPU count.

Returns

Positive hw.perflevel0.logicalcpu value, or zero for sysctl failure/unsupported platform.

Here is the caller graph for this function:

resolveRequested()

int openswmm::threadinfo::resolveRequested	(	int	requested,
		const char *	what,
		std::vector< std::string > *	warnings )

Resolve a thread request and optionally append oversubscription diagnostics.

For positive requests, warns when request+host reservation exceeds known logical CPUs; otherwise warns above the current OpenMP maximum. On macOS can additionally warn above performance-cluster size. No actual runtime team is changed. Caller keeps request+reservation representable as int.

Parameters

[in]	requested	Requested team size; <=0 resolves from OpenMP maximum. Exactly zero additionally enables automatic core-clamp heuristics.
[in]	what	Borrowed non-null NUL-terminated activity label used in formatted warnings.
[in,out]	warnings	Optional caller-owned sink; new messages append without clearing existing entries.

Returns

OpenMP maximum for requested<=0; otherwise the positive request is returned unchanged.

Here is the call graph for this function:

Here is the caller graph for this function:

setKokkosOmpThreads()

void openswmm::threadinfo::setKokkosOmpThreads

(

int

n

)

Store the engine-recorded Kokkos OpenMP team hint.

Uses relaxed atomic storage; provides no synchronization for plugin initialization or other engine state.

Parameters

[in]

n

Team count; nonpositive values store zero.

twoDThreads()

int openswmm::threadinfo::twoDThreads	(	int	requested,
		int	n_triangles,
		std::vector< std::string > *	warnings )

Resolve a 2D marcher team request using core and triangle-count heuristics.

Exactly requested=0 applies the macOS performance-core-minus-two cap before the mesh test. Positive requests reduced by the mesh test append a diagnostic. Negative requests skip the auto-only core cap.

Parameters

[in]	requested	Requested team size; <=0 resolves from OpenMP maximum. Exactly zero additionally enables automatic core-clamp heuristics.
[in]	n_triangles	Triangle count; recommends one worker when n_triangles<4*resolved team. Caller keeps that product representable as int.
[in,out]	warnings	Optional caller-owned sink; new messages append without clearing existing entries.

Returns

Positive recommended team size; does not create or alter a runtime team.

Here is the call graph for this function:

Variable Documentation

kMinConduitsPerThread

int openswmm::threadinfo::kMinConduitsPerThread = 100

constexpr

Minimum conduits of momentum work per dynamic-wave team thread before an extra thread pays for its barriers (PERFORMANCE-ONLY; bit-identical).