In stand-alone mode, profiling options are specified in a file.
The following describes the file format and the valid profiling options that can be used in the file.
Option key |
Option value |
- MONITOR_MODE
- This option controls whether monitor events appear in the trace. Monitor events
are of two kinds: monitor contended events, and monitor wait events. Monitor
contended events occur when a thread tries acquiring a monitor that is already
acquired by another thread. Monitor wait events occur when a thread waits on a
monitor (for example, by executing the wait() method).
|
- all
- All monitor events will appear in the trace.
- none
- No monitor events will appear in the trace
|
- FILTERS
- Used to specify whether the used filters are printed out at the beginning of the
trace.
|
- true
- Print out the filters at the beginning of the trace.
- false
- Do not print out the filters at the beginning of the trace.
|
- TRACK_GC_EVENT_TYPES
- Used to indicate what type of garbage collection (GC) events should be printed
to the trace. The two types of GC events are object free events and object move
events.
|
- none
- Do not print out any GC events.
- deletes
- frees
- Print out only object free events
- moves
- Print out only object move events
- deletesAndMoves
- movesAndFrees
- Print out both object frees and object moves.
|
- ID_STYLE
- Used to specify the types of object ID's that are printed. The possible types
are Static ID and Relocatable ID.
A Static ID is essentially a one-time id assigned to an object in the trace the
first time it is seen. Furthermore, the id doesn't change during the duration
of execution (even if the JVM moves the object internally). Static id's are the type that are
typically used by the profiler.
A Relocatable ID is the id the the JVM associates with the objects and can
change during execution (if, for example, during garbage collection, an object
move event occurs).
|
- static
- Print only Static IDs
- relocatable
- Print only Relocatable IDs
- staticAndRelocatable
- Print both Static and Relocatable IDs
|
- OPTIONS
- Used to specify whether the profiling options should be printed to the trace.
|
- true
- Print the profiling options to the trace.
- false
- Do not print the profiling options to the trace.
|
- TIMESTAMPS
- Used to specify whether timestamps should be printed in the trace.
|
- true
- Print timestamps
- false
- Do not print timestamps
|
- METHOD_COUNTS
- Used to specify whether the method counts should be printed at the end of the trace.
|
- true
- For the methods that were traced, print a total of the number of times they were
called at the end of the trace just before JVM shutdown.
- false
- Do not print method counts.
|
- OBJ_ALLOC_IS_ARRAY
- Used to specify whether to print out the isArray attribute for object
allocations. (i.e., when an object is allocated, it is either an array or it is
not; the isArray attribute contains that information.)
|
- true
- Print the isArray attribute.
- false
- Do not print the isArray attribute.
|
- STACK_INFORMATION
- This option is used to specify what type of method tracing should be done.
|
- none
- Do not do any method tracing. This is the equivalent of NOT selecting Execution
Time Analysis in the Profile configuration window in the workbench.
- normal
- Trace all non-filtered methods.
- boundary
- Trace all non-filtered methods PLUS all boundary methods. This is the
equivalent of having Profile Configuration > Execution Time Analysis > Collect
boundary classes excluded by the filter set selected in the workbench.
- contiguous
- Trace all non-filtered methods PLUS all filtered methods that occur on the
method invocation stack that are between two non-filtered methods. For example,
suppose methodA calls methodB, which calls methodC, which in turn calls methodD.
Further suppose methodA and methodD are not filtered, but methodB and methodC
are. Since methodB and methodC are between two non-filtered methods, they also
get traced.
- boundaryAndContiguous
- Same behavior as having both the contiguous and boundary option values set for STACK_INFORMATION.
|
- BOUNDARY_DEPTH
- Used to specify the boundary tracing depth when the STACK_INFORMATION=boundary
or boundaryAndContiguous.
|
- Integer
- Integer value specifying the boundary tracing depth.
|
- TICKET
- Used to specify whether the ticket attribute should appear in the trace.
|
- true
- Ticket attribute should appear in the trace.
- false
- Ticket attribute should not appear in the trace.
|
- TRACE_MODE
- Used to specify the tracing mode. Basically the tracing mode specifies whether
object (heap) information is kept track of or not. (The 'none' value has
special meaning.)
|
- full
- Trace all possible object information. That is, trace object allocations and
trace the objects that are associated with method invocations (if
STACK_INFORMATION is set to do method tracing).
- none
- Do not do any tracing whatsoever. (If this value is set, then the
STACK_INFORMATION option gets overridden even if it is set to trace method
invocations.)
- noObjectCorrelation
- Do not trace object information. This is different from the 'none' value in that
method invocation tracing can still take place (if STACK_INFORMATION is
appropriately set), but the objects associated with method invocations won't be
kept track of, nor will any object allocation information be kept track of.
|
- TRACE_ID_REFS
- Used to specify whether to print out the trace id reference attribute. The trace
reference id is an identifier that is unique to the trace.
|
- true
- Print out the trace id reference attribute.
- false
- Do not print out the trace id reference attribute.
|
(C) Copyright IBM Corporation 2000, 2006. All Rights Reserved.