RunTimeControl< CLOCK > Class Template Reference

Collaboration diagram for RunTimeControl< CLOCK >:

Public Member Functions
	RunTimeControl (RunTimeManager< CLOCK > &rm, int maxCPUSecs, const std::string &stop_file_prefix, bool cleanup=true)
	constructor More...
bool	checkStop (LoopTimer< CLOCK > &loop_timer)
	check if the run needs to stop because of walltime or stop control file. More...
std::string	generateProgressMessage (const std::string &driverName, int block, int num_blocks) const
	generate terse progress messages More...
std::string	generateStopMessage (const std::string &driverName, int block) const
	generate stop message explaining why More...
void	runtime_padding (int runtime_padding)
void	loop_margin (int loopMargin)

Private Types
enum	StopStatus { CONTINUE, MAX_SECONDS_PASSED, NOT_ENOUGH_TIME, STOP_FILE }

Private Member Functions
bool	enough_time_for_next_iteration (LoopTimer< CLOCK > &loop_timer)
bool	stop_file_requested ()

Private Attributes
const int	MaxCPUSecs
double	m_runtime_safety_padding
double	m_loop_margin
double	m_loop_time
double	m_elapsed
double	m_remaining
RunTimeManager< CLOCK > &	runtimeManager
const std::string	stop_filename_
	the prefix of the stop file (stop_file_prefix + ".STOP") More...
enum qmcplusplus::RunTimeControl::StopStatus	stop_status_

Detailed Description

template<class CLOCK = ChronoClock>
class qmcplusplus::RunTimeControl< CLOCK >

Definition at line 70 of file RunTimeManager.h.

Member Enumeration Documentation

StopStatus

enum StopStatus

strongprivate

Enumerator
CONTINUE
MAX_SECONDS_PASSED
NOT_ENOUGH_TIME
STOP_FILE

Definition at line 82 of file RunTimeManager.h.

  {
    CONTINUE,           // keep running
    MAX_SECONDS_PASSED, // all already passed max_seconds
    NOT_ENOUGH_TIME,    // not enough time for next iteration
    STOP_FILE,          // reqsuted stop from a file
  } stop_status_;

Constructor & Destructor Documentation

RunTimeControl()

RunTimeControl	(	RunTimeManager< CLOCK > &	rm,
		int	maxCPUSecs,
		const std::string &	stop_file_prefix,
		bool	cleanup = true
	)

constructor

Parameters

rm	the RunTimeManager attached to
maxCPUSecs	maxmimal allowed seconds from the beginning of the simulation
stop_file_prefix	the prefix of the stop file
cleanup	if true, clean up stop files left from previous simulations. Rank 0 handle this.

Definition at line 67 of file RunTimeManager.cpp.

References RunTimeControl< CLOCK >::m_loop_margin, RunTimeControl< CLOCK >::m_runtime_safety_padding, and RunTimeControl< CLOCK >::stop_filename_.

    : MaxCPUSecs(maxCPUSecs),
      runtimeManager(rm),
      stop_filename_(stop_file_prefix + ".STOP"),
      stop_status_(StopStatus::CONTINUE)
{
  if (stop_file_prefix.empty())
    throw std::runtime_error("Stop control filename prefix must not be empty!");

  if (cleanup)
  {
    std::remove(stop_filename_.c_str());
    if (std::ifstream(stop_filename_.c_str()))
      throw std::runtime_error("Failed to delete the existing stop control file \"" + stop_filename_ +
                               "\", cannot continue!");
  }

  m_runtime_safety_padding = 30.0; // generous 30 seconds to allow for shut down?
  m_loop_margin            = 1.1;  // 10% margin on average loop time?
}

Member Function Documentation

checkStop()

bool checkStop

(

LoopTimer< CLOCK > &

loop_timer

)

check if the run needs to stop because of walltime or stop control file.

it should be called at the end of each block in a driver.

Definition at line 125 of file RunTimeManager.cpp.

Referenced by qmcplusplus::TEST_CASE().

{
  bool need_to_stop = false;
  need_to_stop |= !enough_time_for_next_iteration(loop_timer);
  need_to_stop |= stop_file_requested();
  return need_to_stop;
}

enough_time_for_next_iteration()

bool enough_time_for_next_iteration ( LoopTimer< CLOCK > &  loop_timer )

private

Definition at line 92 of file RunTimeManager.cpp.

References LoopTimer< CLOCK >::get_time_per_iteration().

{
  m_loop_time = loop_timer.get_time_per_iteration();
  m_elapsed   = runtimeManager.elapsed();

  if (m_elapsed >= MaxCPUSecs)
  {
    stop_status_ = StopStatus::MAX_SECONDS_PASSED;
    return false;
  }

  m_remaining      = MaxCPUSecs - m_elapsed;
  bool enough_time = true;
  if ((m_loop_margin * m_loop_time + m_runtime_safety_padding) > m_remaining)
    enough_time = false;

  stop_status_ = StopStatus::NOT_ENOUGH_TIME;
  return enough_time;
}

generateProgressMessage()

std::string generateProgressMessage	(	const std::string &	driverName,
		int	block,
		int	num_blocks
	)		const

generate terse progress messages

Definition at line 134 of file RunTimeManager.cpp.

References qmcplusplus::log(), and qmcplusplus::hdf::num_blocks.

{
  std::stringstream log;
  if (block == 0 || block + 1 == num_blocks / 4 || block + 1 == num_blocks / 2 || block + 1 == (num_blocks * 3) / 4 ||
      block + 1 == num_blocks)
  {
    log << "Completed block " << std::setw(4) << block + 1 << " of " << num_blocks << " average "
        << std::setprecision(4) << m_loop_time << " secs/block after " << std::setprecision(4) << m_elapsed << " secs"
        << std::endl;
  }
  return log.str();
}

generateStopMessage()

std::string generateStopMessage	(	const std::string &	driverName,
		int	block
	)		const

generate stop message explaining why

Definition at line 150 of file RunTimeManager.cpp.

References qmcplusplus::log().

Referenced by qmcplusplus::TEST_CASE().

{
  std::stringstream log;
  log << "RunTimeControl takes action in " << driverName << " driver." << std::endl;
  if (stop_status_ == StopStatus::MAX_SECONDS_PASSED)
    log << "Time limit reached. Stopping after block " << block << std::endl
        << "Hard limit (seconds) " << MaxCPUSecs << ", elapsed (seconds) " << m_elapsed << std::endl;
  else if (stop_status_ == StopStatus::NOT_ENOUGH_TIME)
  {
    log << "Insufficient time for next block. Stopping after block " << block << std::endl;
    log << "  Iteration time per " << driverName << " block (seconds) = " << m_loop_time << std::endl;
    log << "  Elapsed   time (seconds) = " << m_elapsed << std::endl;
    log << "  Remaining time (seconds) = " << m_remaining << std::endl;
  }
  else if (stop_status_ == StopStatus::STOP_FILE)
    log << "Stop requested from the control file \"" + stop_filename_ + "\", stopping after block " << block
        << std::endl;
  else
    throw std::runtime_error("Unidentified stop status!");

  return log.str();
}

loop_margin()

void loop_margin ( int  loopMargin )

inline

Definition at line 114 of file RunTimeManager.h.

{ m_loop_margin = loopMargin; }

runtime_padding()

void runtime_padding ( int  runtime_padding )

inline

Definition at line 113 of file RunTimeManager.h.

References RunTimeControl< CLOCK >::runtime_padding().

Referenced by RunTimeControl< CLOCK >::runtime_padding(), and qmcplusplus::TEST_CASE().

{ m_runtime_safety_padding = runtime_padding; }

stop_file_requested()

bool stop_file_requested ( )

private

Definition at line 113 of file RunTimeManager.cpp.

{
  if (std::ifstream(stop_filename_.c_str()))
  {
    stop_status_ = StopStatus::STOP_FILE;
    return true;
  }
  else
    return false;
}

Member Data Documentation

m_elapsed

double m_elapsed

private

Definition at line 76 of file RunTimeManager.h.

m_loop_margin

double m_loop_margin

private

Definition at line 74 of file RunTimeManager.h.

Referenced by RunTimeControl< CLOCK >::RunTimeControl().

m_loop_time

double m_loop_time

private

Definition at line 75 of file RunTimeManager.h.

m_remaining

double m_remaining

private

Definition at line 77 of file RunTimeManager.h.

m_runtime_safety_padding

double m_runtime_safety_padding

private

Definition at line 73 of file RunTimeManager.h.

Referenced by RunTimeControl< CLOCK >::RunTimeControl().

MaxCPUSecs

const int MaxCPUSecs

private

Definition at line 72 of file RunTimeManager.h.

runtimeManager

RunTimeManager<CLOCK>& runtimeManager

private

Definition at line 78 of file RunTimeManager.h.

stop_filename_

const std::string stop_filename_

private

the prefix of the stop file (stop_file_prefix + ".STOP")

Definition at line 80 of file RunTimeManager.h.

Referenced by RunTimeControl< CLOCK >::RunTimeControl().

stop_status_

enum qmcplusplus::RunTimeControl::StopStatus stop_status_

private

---

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

• /home/pk7/projects/qmc/for_cron_doxygen/qmcpack/src/Utilities/RunTimeManager.h
• /home/pk7/projects/qmc/for_cron_doxygen/qmcpack/src/Utilities/RunTimeManager.cpp