Collect independent graph bindings without mixing timers into numerical restart history.
More...
#include <timing.hpp>
|
| const std::vector< std::shared_ptr< const GraphTimingRun > > & | runs () const noexcept |
| | Inspect completed and active measurements without changing their catalog or indices.
|
| |
| template<typename GraphT> |
| std::shared_ptr< GraphTimingRun > | start_run (std::type_identity< GraphT >) |
| | Allocate names and counters once, before any measured node begins.
|
| |
Collect independent graph bindings without mixing timers into numerical restart history.
- Note
- One cache executes its nodes serially, although each node may contain parallel kernels. Concurrent caches require separate collectors. Returned observers own their run; the collector need only outlive factory invocation. Formatting and sorting belong outside timed execution.
◆ runs()
| const std::vector< std::shared_ptr< const GraphTimingRun > > & serif::utils::dependencies::GraphTiming::runs |
( |
| ) |
const |
|
inlinenodiscardnoexcept |
Inspect completed and active measurements without changing their catalog or indices.
◆ start_run()
template<typename GraphT>
| std::shared_ptr< GraphTimingRun > serif::utils::dependencies::GraphTiming::start_run |
( |
std::type_identity< GraphT > | | ) |
|
|
inlinenodiscard |
Allocate names and counters once, before any measured node begins.
◆ m_runs
| std::vector<std::shared_ptr<const GraphTimingRun> > serif::utils::dependencies::GraphTiming::m_runs |
|
private |
The documentation for this class was generated from the following file: