NDDS::Utility Class Reference

Unsupported utility APIs. More...

#include <managed_config_dotnet.h>

Static Public Member Functions
static void	spin (System::UInt64 spinCount)
	Performs the spin operation as many times as indicated. More...
static System::UInt64	get_spin_per_microsecond ()
	Returns the number of spin operations to perform to wait 1 microsecond. More...
static System::Boolean	enable_heap_monitoring ()
	Starts monitoring the heap memory used by RTI Connext. More...
static void	disable_heap_monitoring ()
	Stops monitoring the heap memory used by RTI Connext. More...
static System::Boolean	pause_heap_monitoring ()
	Pauses heap monitoring. More...
static System::Boolean	resume_heap_monitoring ()
	Resumes heap monitoring. More...
static System::Boolean	take_heap_snapshot (System::String^ filename, System::Boolean print_details)
	Saves the current heap memory usage in a file. More...

Detailed Description

Unsupported utility APIs.

Member Function Documentation

spin()

static void NDDS::Utility::spin ( System::UInt64  spinCount )

static

Performs the spin operation as many times as indicated.

Spinning is the action of performing useless operations in a for loop in order to actively wait some time without yielding the CPU. Given that the resolution of sleep is in the order of ms, you can use this utility to wait times in the order of microseconds. To properly use this functionality, it is useful to measure previously the number of spin operations needed to wait the equivalent to 1 microsecond (using the utility get_spin_per_microsecond) and then compute the corresponding spin count desired.

Parameters

spinCount

<<in>> Number of spin operations to perform.

get_spin_per_microsecond()

static System::UInt64 NDDS::Utility::get_spin_per_microsecond ( )

static

Returns the number of spin operations to perform to wait 1 microsecond.

This utility can be used to measure how many spin operations must be performed to wait 1 microsecond. Since the time that it takes the CPU to perform 1 spin operation depends on the CPU frequency, it is recommended to use this utility before using spin().

Returns

Number of spin operations to wait 1 microsecond.

enable_heap_monitoring()

static System::Boolean NDDS::Utility::enable_heap_monitoring ( )

static

Starts monitoring the heap memory used by RTI Connext.

This function must be called before any other function in the RTI Connext library is called.

Once heap monitoring is enabled, you can take heap snapshots by using NDDS::Utility::take_heap_snapshot.

Use this method only for debugging purposes, as it may introduce a significant performance impact.

MT Safety:

UNSAFE. It is not safe to call this method while another thread may be simultaneously calling another heap-related method, including this one.

Returns

true if success. Otherwise, false

See also

NDDS::Utility::disable_heap_monitoring

disable_heap_monitoring()

static void NDDS::Utility::disable_heap_monitoring ( )

static

Stops monitoring the heap memory used by RTI Connext.

This method must be the last method called from RTI Connext.

See also

NDDS::Utility::enable_heap_monitoring

pause_heap_monitoring()

static System::Boolean NDDS::Utility::pause_heap_monitoring ( )

static

Pauses heap monitoring.

New memory allocations will not be monitored and they will not appear in the snapshot generated by NDDS::Utility::take_heap_snapshot.

Returns

true if success. Otherwise, false

See also

NDDS::Utility::resume_heap_monitoring

resume_heap_monitoring()

static System::Boolean NDDS::Utility::resume_heap_monitoring ( )

static

Resumes heap monitoring.

Returns

true if success. Otherwise, false

See also

NDDS::Utility::pause_heap_monitoring

take_heap_snapshot()

static System::Boolean NDDS::Utility::take_heap_snapshot ( System::String^  filename, System::Boolean  print_details  )

static

Saves the current heap memory usage in a file.

After NDDS::Utility::enable_heap_monitoring is called, you may invoke this method periodically to save the current heap memory usage to a file.

By comparing two snapshots, you can tell if new memory has been allocated and in many cases where. This is why this operation can be used to debug unexpected memory growth.

The format of a snapshot is as follows:

First, there is a memory usage summary like this:

Process virtual memory: 2552352768
Process physical memory: 16187392
Current heap usage: 10532131
High watermark: 10532131
Alloc count: 17634
Free count: 3518

• Process virtual memory: The amount of virtual memory in bytes taken by the process. This memory includes RTI Connext and not RTI Connext memory.
• Process virtual memory: The amount of physical memory in bytes taken by the process.
• Current heap usage: The amount of heap memory in bytes used by the middleware. For Java and .NET APIs, this memory only accounts for unmanaged RTI Connext memory, not memory living in the managed heap.
• High watermark: The maximum amount of heap usage by RTI Connext since NDDS::Utility::enable_heap_monitoring was invoked.
• Alloc count: The number of invocations to malloc, realloc, or calloc operations done by RTI Connext.
• Free count: The number of invocations to the free operation done by RTI Connext.

After the previous summary, and only if you set the parameter print_details to true, the method will print the details of every single outstanding heap allocation done by RTI Connext. For example:

block_id, timestamp, block_size, pool_alloc, pool_buffer_size, pool_buffer_count, topic_name, activity, alloc_method_name, type_name

12830, 1492838970, 104, MALLOC, 0, 0, PRESServiceRequest, PRESCstReaderCollator_new, RTIOsapiHeap_allocateStructure, struct REDAFastBufferPool

• block_id: Block ID of the allocation. This number increases with every allocation.
• timestamp: Timestamp in UTC seconds corresponding to the time where the allocation was done.
• block_size: The number of bytes allocated.
• pool_alloc: Indicates if the heap allocation is a RTI Connext pool allocation (POOL) or a regular allocation (MALLOC).
• pool_buffer_size: For pool allocations, this number indicates the size of the elements in the pool in number of bytes. block_size is equal to (pool_buffer_size * pool_buffer_count).
• pool_buffer_count: For pool allocations, this number indicates the number of buffers allocated for the pool. block_size is equal to (pool_buffer_size * pool_buffer_count).
• topic_name: The topic name associated with the allocation or 'n/a' if it is not available.
• activity: Activity associated with the allocation or 'n/a' if it is not available.
• alloc_method_name: The allocation RTI Connext method name.
• type_name: The allocation typename.

Parameters

filename	<<in>>. Name of file in which to store the snapshot.
print_details	<<in>>. Indicates if the snapshot will contain only the memory usage summary or the details of the individual allocations.

Returns

true if success. Otherwise, false