Using the Profiler Utility

The Natural Profiler is used to monitor the internal process flow of a Natural application and to analyze the performance and the code coverage of the application. Profiling of Natural applications is activated by switching on the ACTIVE subparameter of the PROFILER profile parameter. The Natural Profiler writes the collected Profiler event data to a Profiler resource file. See PROFILER – Profile a Natural Session in the Parameter Reference documentation.

Code coverage of Natural applications is activated by switching on the ACTIVE subparameter of the COVERAGE profile parameter (see the Parameter Reference documentation). Natural code coverage writes the collected coverage data to a code coverage resource file.

The Profiler utility runs in batch mode only. It provides functions to control the Profiler and code coverage data collection and to process the resulting data.

1. With the Profiler data collection functions, the data collection can be paused and restarted (see also Starting and Pausing Data Collection).

2. The data processing functions read and process the event data from the Profiler resource file. Unconsolidated event data can be consolidated.

   You can output statistics, a program summary, a line summary and a transaction summary, a program trace, an event trace with the most important data, and reports on program and statement coverage. You can export the resulting data in text or CSV (comma-separated values) format.

The following tools offer also functions to read Profiler resource files and to analyze the data:

• NaturalONE displays the full event trace and provides a performance analysis (hot spots) of the Natural batch application. Coverage data can be inspected in the NaturalONE Coverage view and in the NaturalONE source editor.

• The Profiler Rich GUI, which is started from the NaturalONE Tools and Utilities, offers functions to evaluate or consolidate Profiler data, or to delete a resource file. Interactive pie charts show the distribution of Profiler KPIs for selected criterions like the distribution of the CPU time for programs. For monitored programs, the source is combined with the Profiler data whereby the source is colored according to the value of the selected KPI.

This section covers the following topics:

• Quick Start for Profiling

• Quick Start for Code Coverage

• Prerequisites

• Invoking and Terminating the Profiler Utility

• Syntax and Keywords

• Events and Data Collected

• Starting and Pausing Data Collection

• Consolidating Event Data

• Evaluating Event Data

• Maintaining Profiler Resource Files

• Including Profiler Input from Natural Text Objects

• Event Trace

• Tracing Natural Code Coverage

• Internal Trace

• Profiler Statistics

Quick Start for Profiling

This section briefly describes the steps required for profiling Natural applications and viewing the results. The instructions provided here may serve as a guideline when starting to use the Natural Profiler. Detailed information regarding the steps is provided in the remainder of this chapter.

The steps to take depend on the evaluation you want to perform for your application as illustrated in the following graphic:

1. Check that the prerequisites are met.

2. If you plan to monitor Natural statement events (NS), compile the corresponding Natural objects with the following profile parameter:

   GPGEN=(PROFILER=ON)

   The GPGEN parameter is described in the Parameter Reference documentation.

3. Activate the profiling of the Natural session by switching on the ACTIVE subparameter of the PROFILER profile parameter in the NATPARM parameter file or dynamically when invoking Natural. Example for dynamic parameter specification:

   PROFILER=(ACTIVE=ON,RESNAME=ResName,RESLIB=RESLIB)

   In the example above, the Profiler event data is written to a resource file with the name ResNam.nprf in the library RESLIB. See PROFILER – Profile a Natural Session in the Parameter Reference documentation.

4. Open the NPRF resource in NaturalONE to view the hot spots and the event trace.

5. Submit a Natural batch job with the Profiler utility READ function to print an event trace, a program trace and the Profiler statistics. Example:

   FUNCTION=READ           /* Read Profiler data
     RESOURCE-LIB=RESLIB   /* Resource library
     RESOURCE-TYPE=NPRF    /* Use resource type NPRF
     EVENT=ON              /* Print event trace
     PROGRAM=ON            /* Print program trace
     STATISTICS=ON         /* Print statistics

   See also Profiler Utility READ Function. This section also describes how to generate a transaction summary.

6. Consolidate (aggregate) the event data with the Profiler Rich GUI or submit a Natural batch job with the Profiler utility CONSOLIDATE function. Example:

   FUNCTION=CONSOLIDATE    /* Consolidate Profiler data
     RESOURCE-LIB=RESLIB   /* Resource library
     REPLACE=YES           /* Replace resource

   The consolidated Profiler event data is written to the resource ResNam.nprc in the library RESLIB. See Consolidating Event Data.

7. Open the NPRC resource in NaturalONE to view the hot spots.

8. Open the NPRC resource with the Profiler Rich GUI Data Evaluation or Program Analysis to visualize the Profiler data on a graphical, interactive browser interface.

9. Submit a Natural batch job with the Profiler utility READ function to generate a program summary, a line summary and the Profiler statistics. Example:

   FUNCTION=READ           /* Read Profiler Data
     RESOURCE-LIB=RESLIB   /* Resource library
     RESOURCE-TYPE=NPRC    /* Use resource type NPRC
     PROGRAM=ON            /* Print program summary
     LINE=ON               /* Print line summary
     STATISTICS=ON         /* Print statistics

   See also Profiler Utility READ Function. This section also describes how to generate a transaction summary.

Notes:

1. If the resource name is not explicitly specified in the READ or CONSOLIDATE function of the Profiler utility, the last created NPRF or NPRC resource in the library is used.
2. If you plan to profile a long-running batch application, refer to the section Profiling Long-Running Applications. It covers strategies of how to minimize the number of events to be monitored.
3. The NaturalONE Profiler and the Profiler Rich GUI are described in detail in the NaturalONE documentation.

Quick Start for Code Coverage

This section briefly describes the steps required for performing the code coverage of a Natural batch applications and viewing the results. The instructions provided here may serve as a guideline when starting to use Natural code coverage. Detailed information regarding the steps is provided in the remainder of this chapter.

The steps to take depend on the evaluation you want to perform for your application as illustrated in the following graphic:

1. Check that the prerequisites are met.

2. Compile the Natural objects which are examined with the following profile parameter:

   GPGEN=(COVERAGE=ON)

   The GPGEN parameter is described in the Parameter Reference documentation.

3. Activate code coverage of the Natural session by switching on the ACTIVE subparameter of the COVERAGE profile parameter (see the Parameter Reference documentation) in the NATPARM parameter file or dynamically when invoking Natural. Example for dynamic parameter specification:

   COVERAGE=(ACTIVE=ON,RESNAME=ResName,RESLIB=RESLIB)

   In the example above, the coverage data is written to a resource file with the name ResNam.ncvf in the library RESLIB.

4. Open the NCVF resource in NaturalONE to obtain the Code Coverage view.

5. From the NaturalONE Code Coverage view, you can directly edit the source. The editor shows all lines containing covered statements with a green background.

6. Submit a Natural batch job with the Profiler utility READ function to print the program and statement coverage.

   Example:

   FUNCTION=READ           /* Read Profiler data
     RESOURCE-LIB=RESLIB   /* Resource library  
     RESOURCE-TYPE=NCVF    /* Use resource type 
     EVENT=ON              /* Print statement coverage
     PROGRAM=ON            /* Print program coverage
     EXPORT=ON             /* write to work 7     
     FORMAT=C              /* Semicolon/Comma/Text

   If the EXPORT keyword of the Profiler utility READ function is switched on, the output is written to Work File 7. If FORMAT is specified as C or S, the result is written as comma-separated values (CSV) where a comma or a semicolon is used as a separator, respectively.

7. Export the data of Work File 7 with any tool (such as FTP) as a CSV-formatted file to a Windows environment if you want to process it further in Microsoft Excel.

Notes:

1. If the resource name is not explicitly specified in the READ function of the Profiler utility, the NCVF resource created last in the library is used.
2. The NaturalONE Code Coverage view and editor are described in the NaturalONE documentation.

Prerequisites

The following prerequisite must be met before you can use the Profiler utility:

• Natural Parameter Settings

• Activate Data Processing

Natural Parameter Settings

The Profiler utility data processing functions (CONSOLIDATE, READ, and LIST) cannot be executed if profiling is active. Deactivate the profiling infrastructure with the following (default) parameter setting:

PROFILER=(ACTIVE=OFF)

For details regarding the PROFILER parameter, see PROFILER – Profile a Natural Session in the Parameter Reference documentation.

Activate Data Processing

If NaturalONE is installed at your site, you can activate the Profiler utility data processing functions (CONSOLIDATE, READ, and LIST) with the following steps:

1. Start NaturalONE.

2. In the Natural Server view, map to the environment where the Profiler resources reside.

3. Add the program ACTIVATE contained in the system library SYSPRFLR to a new or existing project in NaturalONE.

4. Profile the program ACTIVATE with the context menu function Profile As > Natural Application.

5. Verify that the user-defined event data on the Event Trace page of the NaturalONE Profiler contains the activation success message.

Alternatively, you can activate the Profiler utility data processing functions from the NaturalONE Tools and Utilities:

1. Start NaturalONE.

2. In your workspace, select a Natural project which is mapped to the environment where the Profiler resources reside.

3. Invoke the context menu and choose NaturalONE > Tools and Utilities to start the SYSUTIL utility.

4. Select Natural Profiler Activation in the tree and click on Execute.

When you perform the steps described above, a NaturalONE Profiler key is generated and written to the Natural resource NaturalONEProfilerKey.nprk in the system library SYSPRFLR. Each Profiler data processing function reads this resource and checks the key. If the key is valid, the function is performed. A newly generated key is valid for one year. It can always be regenerated.

The Profiler data processing function starts issuing a warning 9 days before the key expires, and returns an error message if no key is found or if the key is not valid.

Notes:

1. Natural statement events (NS) are only generated for profiling if the corresponding Natural object was compiled with the profile parameter GPGEN set to (PROFILER=ON).GPGEN is described in the Parameter Reference documentation.
2. Code coverage of a Natural application can only be performed if the corresponding Natural objects were compiled with the profile parameter GPGEN set to (COVERAGE=ON).

Invoking and Terminating the Profiler Utility

This section provides instructions for invoking and terminating the Profiler utility in batch mode.

To invoke the Profiler utility

• Enter the following system command into the primary command input data set CMSYNIN:

  PROFILER

  Note:
  After the PROFILER system command, the Profiler expects one or more lines with Profiler keyword entries.

To terminate the Profiler utility

• Enter the following Profiler keyword into the primary command input data set CMSYNIN:

  END-PROFILER

  Or:

  END

  Or:

  .

Syntax and Keywords

The Profiler utility in batch mode reads the Profiler keywords that control the profiling from the primary command input data set CMSYNIN. The Profiler reads the input lines until it reaches the END-PROFILER keyword (or END or .).

This section covers the following topics:

• Profiler Utility Syntax

• Profiler Utility Keywords

Profiler Utility Syntax

The symbols used in the syntax diagrams shown in this section are explained in System Command Syntax in the System Commands documentation.

You enter a Profiler utility command using either of the following syntax formats:

Or:

Notes:

1. If a value is associated with a keyword but no equal sign is found, the Profiler expects the value in a separate input line without any other keyword (second syntax format).
2. The first syntax format expects input in delimiter mode (IM=D).
3. The second syntax format can be used if the Profiler is to be executed with the Natural STACK profile parameter or if the data is entered in forms mode (IM=F).

The following rules apply:

• Empty lines and lines starting with an asterisk (*) are ignored.

• All characters in a line from /* to */ or to the end of the line are ignored.

• Some keywords have no associated value.

• Blanks can be added before or after the keyword or value.

• Multiple keywords in a line are separated by commas (applies to the first syntax format only).

• A value can be enclosed in apostrophes ('value').

• A value must not contain a comma.

• Keywords and values can be specified in upper or lower case.

• The maximum input line length is 78 characters.

The Profiler utility can be executed multiple times in one Natural session. For example, it is first executed with the START function, and then, after the execution of a user program, it is executed with the PAUSE function.

Example

The following Natural batch example runs the Profiler utility READ function:

natural BATCHMODE CMSYNIN=cmd.txt CMOBJIN=data.txt CMPRINT=out.txt CMWRK07=wrk07.txt

The content of the batch input file cmd.txt which contains the PROFILER command is shown below:

PROFILER
FIN

The content of the input file data.txt which contains the input for the Profiler utility is shown below:

*********************************************************************
* Read Profiler NPRF Resource                                        
*********************************************************************
TRACE=3                  /* Set Profiler trace level
FUNCTION=READ            /* Read Profiler resource
  RESOURCE-NAME='Demo01' /* Resource name   
  RESOURCE-LIB=PRFDATA   /* Resource library
  RESOURCE-TYPE=NPRF     /* Resource type   
  EVENT=ON               /* List events     
  STATISTICS=ON          /* List properties and statistics
  PROGRAM=ON             /* Program trace           
  PRINT=ON               /* Write to standard output
  EXPORT=ON              /* Write to work file 7   
  FORMAT=TEXT            /* Use text format   
END-PROFILER             /* End profiler input

After execution, out.txt contains the Profiler utility output and the internal trace (TRACE=3). wrk07.txt contains the Profiler utility output in text format (FORMAT=TEXT).

The following Natural example demonstrates how the Profiler utility PAUSE function is to be executed with the Natural STACK profile parameter:

natural PROFILER=(ACTIVE=ON,RESNAME=ResName,RESLIB=RESLIB)
STACK=(
PROFILER FUNCTION:PAUSE:END-PROFILER;
LOGON PRFDEMO
)

After execution, profiling of the Natural session is activated but the data collection is paused. The data collection can be started later with the Profiler utility START function.

Profiler Utility Keywords

The main keywords used in the syntax of the Profiler utility in batch mode are described in the following table. Any additional (subordinate) keywords available for a main keyword are described in the sections referenced in the table. In general, a subordinate keyword value must follow the main keyword value, for example:

FUNCTION=READ
  PRINT=ON

A subordinate keyword specified before the first FUNCTION or FILTER keyword is treated as a subordinate keyword of the first FUNCTION or FILTER keyword.

The following main keywords are available:

Events and Data Collected

This section describes the events and data processed by the Profiler utility.

• Events

• Data Collected

Events

During a Natural session, different types of events can occur (for example, a program start) where the Profiler collects data specific to the event in a trace record. Each event is associated with an event type, that is, a one or two letter code. Related event types are combined into an event group which is denoted by a one letter code.

The following events, event types and event groups are available:

With each collected event, a CPU and an event timestamp are recorded. In general, a timestamp is taken at the beginning of an event. The duration of an event therefore equals the time that elapses between the timestamp of the event and the timestamp of the event that follows.

Data Collected

This section describes the data collected by the Natural Profiler:

General Data

The following data elements are collected at every event:

• Event counter

• Event type

• Event time in units of microseconds

• Session CPU time in units of microseconds

• Trace session ID

• Natural Security user group ID

• Natural user ID

• Natural application name

• Program library

• Program name

• Program level

• Copycode library

• Copycode name

• Statement line number

• Statement op-code

• Coverage flag (for Natural code coverage)

Notes:

1. The Natural Profiler for Linux and Windows does not yet collect RPC-related events.
2. Natural code coverage only collects NS and PI events.
3. Natural code coverage does not collect time values.
4. A PI event is collected for each object accessed and for all copycodes included in the object (recursively).

Event-Specific Data

The following data is only collected at the following events:

Event	Data Elements
Session Initialization	None
Session Termination	Termination return code Natural termination message code NAT99nn Name of back-end program Monitor CPU time in units of microseconds
Program Load	Name of program to be loaded Name of load library Invocation type
Program Resume	None
Program Start/Termination	Program type Database ID of program library File number of program library
Program Information	Program type Number of statements in the program or copycode First statement item INCLUDE line number Parent copycode ID
Database Call	Database type Command code Command ID Database ID File number Response code (event type DA) Error subcode (event type DA) Adabas command time (event type DA)
Terminal I/O	Number of bytes sent Number of bytes read Total session storage allocated Compressed session storage length
External Program Call	Name of program called Calling mode such as dynamic or static mode Program link location Parameter type such as reference or value Response code (event type CA)
Runtime Error	Natural system error message code External abend code Name of error handling program
Natural Statement	Profiling: None Natural code coverage: Statement item identifier (GP offset)
Start of RPC Request Execution	Environment (C = client, S = server) Subprogram name Adabas user ID (ETID) Conversation status Logon indicator (Y = logon performed) Impersonation indicator of RPC request (Y = impersonation performed)
Outbound/Inbound RPC Message / RPC Wait for Client	Environment (C = client, S = server) Transport protocol RPC function Type of client user ID Length of message RPC return code External conversation ID Client user ID Server node (event types RO and RW) Server name (event types RO and RW)
User-Defined Event	Subtype of the user-defined event Up to 249 bytes of user-defined information
Monitor Pause	Type of monitor pause Possible values:
	R	Monitor pause requested. This value is also set when the session is initialized with the Pause option.
	F	Start of a block of filtered-out events. Block filters are: library, program, line, FNAT, event count, or time filter.
	W	Trace session waits because of a data pool full situation.

Starting and Pausing Data Collection

When a Natural session is profiled, all event data of the session is collected by default.

You can start and pause data collection with the following methods:

• Using Profiler Utility Functions

• Using Profiler Utility Programs

• Using the Application Programming Interface

Using Profiler Utility Functions

The Profiler utility START and PAUSE functions are used to start and pause data collection. The following syntax applies:

Syntax Description:

Using Profiler Utility Programs

The following Natural programs in the system library SYSPRFLR are supplied to perform Profiler utility functions:

To use Profiler utility programs

• Logon to the library SYSPRFLR or copy the programs to the library SYSTEM, to the appropriate steplib library, or to the required library.

  If PRFFCT is used, the application programming interface USR8210N must be copied as well (see the following section).

  If PRFFCT is used in a client/server environment, copy PRFFCT to the client library and USR8210N to the server library.

  Note:
  PRFFCT expects as input the value START, PAUSE or STATE to perform the corresponding function.

To start the data collection

• Execute the following program:

  PRFSTART

  Or:

  PRFFCT
  START

To pause the data collection

• Execute the following program:

  PRFPAUSE

  Or:

  PRFFCT
  PAUSE

To retrieve the current state of the data collection

• Execute the following program:

  PRFSTATE

  Or:

  PRFFCT
  STATE

Using the Application Programming Interface

The data collection can be started and paused from the profiled Natural application by calling the application programming interface (API) USR8210N. The API can also be used to get the current state of the monitoring process. The API is delivered in the SYSEXT library. For more information, see SYSEXT Utility - Natural Application Programming Interfaces.

To use the API

• Copy the subprogram USR8210N to the library SYSTEM, to the appropriate steplib library, or to the required library.

  Note:
  USR8210N expects as the first parameter the value START, PAUSE or STATE to perform the corresponding function. The parameter values can be specified in uppercase or lowercase. On return, P-RETURN contains the return code and P-MESSAGE the success or error message.

To start the data collection

• Use the interface with the CALLNAT statement:

  CALLNAT 'USR8210N' 'START' P-RETURN P-MESSAGE /* Start Profiler

To pause the data collection

• Use the interface with the CALLNAT statement:

  CALLNAT 'USR8210N' 'PAUSE' P-RETURN P-MESSAGE /* Pause Profiler

To retrieve the current state of the data collection

• Use the interface with the CALLNAT statement:

  CALLNAT 'USR8210N' 'STATE' P-RETURN P-MESSAGE /* Get Profiler state

  The state is coded in the field P-RETURN:

  P-RETURN	Description
  0	Natural Profiler data collection is started.
  1	Natural Profiler data collection is paused.

Consolidating Event Data

The Profiler utility CONSOLIDATE function consolidates event data.

Note that the Natural Profiler Rich GUI also offers a function to consolidate Profiler event data.

For general information regarding data consolidation, see Data Consolidation in the section Basic Concepts of the Profiler Utility.

Syntax of CONSOLIDATE:

Syntax Description:

Example of a Consolidation

The following example consolidates the Profiler resource Test.nprf in the library PRFDATA and writes the consolidated data to the Profiler resource Test.nprc. I/O and Natural RPC client time are included in the consolidated data.

In addition, the consolidated data is written in CSV (semicolon-separated values) format to Work File 7.

The event and consolidation traces are switched off.

FUNCTION=CONSOLIDATE    /* Consolidate Profiler data
  RESOURCE=ON           /* Write to resource
  RESOURCE-NAME='Test'  /* Resource name
  RESOURCE-LIB=PRFDATA  /* Resource library
  REPLACE=YES           /* Replace resource
  TRANSACTION=OFF       /* Do not add transaction identifiers 
  IO-TIME=ON            /* Include I/O and RPC client times
  EXPORT=ON             /* Write to Work File 7       
  FORMAT=S              /* CSV format with semicolon separator 
  TRACE-EVENT=OFF       /* No event trace
  TRACE-CONSOLIDATE=OFF /* No consolidation trace

Evaluating Event Data

When a Natural application is profiled, the Natural Profiler utility writes the event data to an NPRF resource file. Consolidated data is stored in an NPRC resource file and coverage data is stored in an NCVF resource file. The Profiler utility READ function reads and evaluates the Profiler resource data and writes the results to standard output or to a Natural work file. The evaluations performed depend on the type of the resource file read as described in the following table:

This section covers the following topics:

• Profiler Utility READ Function

• Example of READ

• Event Trace

• Consolidation Trace

• Program Trace

• Program Summary

• Line Summary

• Transaction Summary

• Program Coverage

• Statement Coverage

• Using a Microsoft Excel Template to Visualize Coverage Results

• Statistics

Profiler Utility READ Function

The Profiler utility READ function reads and evaluates the resource data.

Syntax of READ:

Syntax Description:

Example of READ

The following example reads the Natural Profiler resource Test.nprf in the library PRFDATA and writes the event trace, program trace and the Profiler statistics to standard output and to Work File 7 in text format.

FUNCTION=READ           /* Read Profiler Data
  RESOURCE-NAME='Test'  /* Resource name
  RESOURCE-LIB=PRFDATA  /* Resource library
  RESOURCE-TYPE=NPRF    /* Use resource type NPRF
  EVENT=ON              /* Print event trace
  PROGRAM=ON            /* Print program trace
  STATISTICS=ON         /* Print statistics
  PRINT=ON              /* Write to standard output
  EXPORT=ON             /* Write to Work File 7
  FORMAT=TEXT           /* Export in text format

Event Trace

If EVENT=ON is specified for an NPRF resource file, the Profiler event trace is generated.

The event trace shows the data of each Natural event which occurred while the application executed. The trace can be referenced if detailed information of an event is required. For example, if a Natural error occurred during application execution, the event trace shows the corresponding error number and message.

If the event trace is written to standard output (PRINT=ON) or exported in text format (EXPORT=ON, FORMAT=TEXT), it is similar to the event trace written by the Profiler monitor session while the application was profiled (see Event Trace). If the data is exported in CSV (comma-separated values) format, it contains all data fields provided by the Profiler (see Data Collected).

Example of an Event Trace

The following example shows an extract of an event trace:

Natural Profiler Event Trace ---------------------------- Count Time CPU-Time (ms) Ev Lev Library Program Line CC-Lib CC-Name Statement Local-Data 0 10:20:58.219911 63.318 MP 003 SYSPRFD PRBINIT 8350 Call Monitor pause requested 102 10:20:58.277586 76.106 PL 000 0000 Execute SYSEDMD/MENU 103 10:20:58.277591 76.139 PS 001 SYSEDMD MENU 0000 PgmStart 00010/02430 Type: P 103 10:20:58.277594 76.151 NS 001 SYSEDMD MENU 0250 Compute Assign/Compute/Move 103 10:20:58.277596 76.155 NS 001 SYSEDMD MENU 0270 Fetch Fetch 104 10:20:58.277598 76.169 DB 001 SYSEDMD MENU 0270 Fetch 00010/02430 S1 ...

Explanations:

• The Count column shows the number of the event.

• The Time and CPU-Time columns show the event time and the CPU timestamp of the event execution, respectively.

• The event with the number 104 is a Database Before (DB) event caused by an Adabas S1 command issued against the file 00010/02430 which was triggered by a FETCH statement in the line 0270 of the Natural object MENU.

For further explanations of the trace columns and event types, see the sections Event Trace and Events and Data Collected.

Consolidation Trace

If EVENT=ON is specified for an NPRC resource file, the Natural Profiler consolidation trace is generated. The consolidation trace is also generated if TRACE-CONSOLIDATE=ON is set for the Profiler utility CONSOLIDATE function.

The consolidation trace shows general event data, summarized values of the elapsed time and CPU time and the hit count of the consolidated record. If two trace entries show the same general event data, they have different event-specific data which is not displayed in the consolidation trace.

The consolidated records are used as the basis for further evaluations like the NaturalONE hot spots or the Natural Profiler Rich GUI data evaluation and program analysis. The consolidation trace can be used to validate the consolidated data.

If the consolidation trace is written to standard output (PRINT=ON), it is similar to the consolidation trace written by the Profiler data consolidation (see Consolidating Event Data). If the data is exported, it contains all consolidated data fields provided by the Profiler.

Example of a Consolidation Trace

The following example shows an extract of a consolidation trace:

Natural Profiler Consolidation Trace ------------------------------------ Count Transact Ev User Lev Library Program Line CC-Lib CC-Name Statement Hit-Count Elapsed(ms) CPU(ms) 1 DA PRFO82D 000 0000 1 75.692 0.870 2 DA PRFO82D 000 0000 1 0.002 0.004 3 DA PRFO82D 000 0000 1 0.006 0.025 4 NS PRFO82D 006 SYSLIBS A82CLS 0010 SYSAOSSU C-COPYRT Reset 43 0.043 0.118 5 NS PRFO82D 006 SYSTEM NOMSTCS 4360 End 1 0.000 0.003 6 PL PRFO82D 006 SYSTEM NOMSTCS 0970 Callnat 1 0.008 0.058 7 PL PRFO82D 006 SYSTEM NOMSTCS 1020 Perform 1 0.004 0.017 ...

Explanations:

• The Count column shows the number of the consolidated record.

• The Transact column shows the transaction identifier.

  The transaction identifier starts with 1 and is increased with every IA (after terminal I/O) or RI (inbound RPC message) event. Transaction identifiers are only available if data is consolidated with the option TRANSACTION=ON.

• The consolidated record 4 shows that the RESET statement in the line 0010 of the copycode C-COPYRT (included in the Natural object A82CLS) executed 43 times spending a total elapsed time of 0.043 milliseconds (ms) and a total CPU time of 0.118 ms.

For further explanations of the trace columns and event types, see the sections Event Trace and Events and Data Collected.

Program Trace

If PROGRAM=ON is specified for an NPRF resource file, the Profiler program trace is generated. The program trace shows the program flow of the profiled application. In general, the program trace exclusively shows program and session events (see Events and Data Collected for a list of possible event types).

If the program trace is written to standard output (PRINT=ON) or exported in text format (EXPORT=ON, FORMAT=TEXT), the program names are indented (see the example below) according to the program level to provide a quick overview of the application calling structure.

If the data is exported in CSV (comma-separated values) format, the program names are not indented. In addition to the output in text format, the exported data contains the CPU timestamp and the summarized Adabas time.

Example of a Program Trace

The following example shows an extract of a program trace and the totals of the application run:

Natural Profiler Program Trace ------------------------------ Time Ev Library CC-Name Line Lev Program Events 10:20:58.309812 PL 0000 000 10:20:58.309817 PS SYSEDMD 0000 001 .OPTTEST D=4 N=2 10:20:58.357694 PL SYSEDMD 5620 001 .OPTTEST 10:20:58.357704 PS SYSEDMD 0000 002 ..CALLMON3 N=3 10:20:58.385263 PL SYSEDMD 0980 002 ..CALLMON3 10:20:58.385274 PS SYSEDMD 0000 003 ...OP3DISC D=3 N=4 10:20:58.412207 PL SYSEDMD 1670 003 ...OP3DISC 10:20:58.412221 PS SYSEDMD 0000 004 ....OPTINFO N=57 10:20:58.443203 PL SYSEDMD 5830 004 ....OPTINFO 10:20:58.443210 PS SYSEDMD 0000 005 ....:OPTPARM1 D=3 N=19 10:20:58.449549 PL SYSEDMD 1960 005 ....:OPTPARM1 10:20:58.449555 PS SYSEDMD 0000 006 ....:.OPTPARM2 D=3 N=10 10:20:58.458286 PL SYSEDMD 0560 006 ....:.OPTPARM2 10:20:58.458300 PS SYSEDMD 0000 007 ....:..OPTPARM3 N=16 10:20:58.458390 PL SYSEDMD 1530 007 ....:..OPTPARM3 10:20:58.458408 PS SYSLIBS 0000 008 ....:...NAT41004 D=10 C=6 N=7345 10:20:58.471017 PT SYSLIBS 5235 008 ....:...NAT41004 10:20:58.471017 PR SYSEDMD 1530 007 ....:..OPTPARM3 N=2898 10:20:58.473293 PL SYSEDMD 1530 007 ....:..OPTPARM3 10:20:58.473297 PS SYSLIBS 0000 008 ....:...NAT41004 D=5 C=6 N=1416 10:20:58.475581 PT SYSLIBS 5235 008 ....:...NAT41004 10:20:58.475581 PR SYSEDMD 1530 007 ....:..OPTPARM3 N=466 10:20:58.475957 PT SYSEDMD 2190 007 ....:..OPTPARM3 10:20:58.475957 PR SYSEDMD 0560 006 ....:.OPTPARM2 N=283 10:20:58.476187 PT SYSEDMD 0860 006 ....:.OPTPARM2 10:20:58.476187 PR SYSEDMD 1960 005 ....:OPTPARM1 N=42 10:20:58.476222 PT SYSEDMD 7510 005 ....:OPTPARM1 10:20:58.476222 PR SYSEDMD 5830 004 ....OPTINFO D=3 N=10 10:20:58.497926 PL SYSEDMD 6080 004 ....OPTINFO 10:20:58.521954 PR SYSEDMD 1670 003 ...OP3DISC N=241 10:21:11.205102 PR SYSEDMD 0980 002 ..CALLMON3 D=7 N=6070 10:21:41.704996 PR SYSEDMD 5620 001 .OPTTEST D=8 I=3 N=26 10:21:41.731229 PT SYSEDMD 7370 001 .OPTTEST 10:21:41.731229 PR 0000 000 D=14 I=1 10:21:42.248348 ST 0000 000 Totals ------ Ev Event Count S Session ................ 1 P Program ................ 5297 D Database Call .......... 2140 I Terminal I/O ........... 12 C External Program Call .. 6510 E Runtime Error .......... 43 N Natural Statement ...... 857384 R RPC Request............. 0 U User-Defined Event ..... 0 M Monitor Pause .......... 2

Explanations:

• For each event listed, the time when the event occurred, the active library, program (Natural object), copycode, line number and program level is displayed.

• The program name is followed by the number of events that occurred from one program event to the next program event.

• Events which belong to one event group are combined into one count using the maximum count of the corresponding event types. Example: One Database Before (DB) and one Database After (DA) event are combined into one Database event (D=1).

• In the example above, the Natural object OPTTEST was started at the level 1. This program calls the subprogram CALLMON3 which calls further subprograms. The highest Level 8 is reached when the subprogram NAT41004 executes. During the first execution, this subprogram performs 10 database calls (D=10), 6 external program calls (C=6) and 7345 Natural statements (N=7345).

• The Totals section at the end of the program trace shows the maximum count of each event group. For example: a total of 2140 database calls corresponds to 2140 Database Before (DB) and 2140 Database After (DA) events.

• The totals of the Session (S) and Program (P) event groups are only listed under Totals; they are not listed next to the program name.

For further explanations of the trace columns, see the section Event Trace.

For explanations of event types and associated event groups, see the section Events.

Program Summary

If PROGRAM=ON is specified for an NPRC resource file, the Profiler program summary is generated.

The program summary shows for each Natural object how many Natural events have occurred, the total CPU time (in milliseconds) and the percentage of the CPU time spent by the Natural object with respect to the total CPU time.

Monitor Pause events and events at Level 0 are not considered for the program summary. Events which belong to one event group are combined into one count: see Events.

Program starts and load requests are listed separately.

If the data is exported in CSV (comma-separated values) format, the count of each event type is listed. Additionally, the elapsed time and the Adabas times (absolute and percentage values) are displayed. The exported time values are indicated in microseconds.

Example of a Program Summary

The following example shows the extract of a program summary:

Natural Profiler Program Summary -------------------------------- Library Program Start Load Database I/O External Error Statement User CPU-Time (ms) CPU % SYSEDMD ADA-CL 41 0 40 0 41 0 621 0 3.785 0.14 SYSEDMD ADA-RC 45 0 44 0 45 0 545 0 4.704 0.17 SYSEDMD AOS-CL 115 97 15 0 0 0 2507 0 42.890 1.63 SYSEDMD AOS-OP 169 154 22 0 0 0 6975 0 70.286 2.68 SYSEDMD BYTE 1 0 0 0 0 0 11 0 0.034 0.00 SYSEDMD CALLMON3 1 5 23 0 0 0 7089 0 20.001 0.76 SYSEDMD CALLNOM 6 6 19 0 0 0 18 0 1.342 0.05 SYSEDMD CALLNOPM 2 2 4 0 0 0 16 0 0.395 0.01 SYSEDMD CALLNOPN 1 1 4 0 0 0 8 0 0.244 0.00 SYSEDMD CALLNOPS 3 4 23 0 0 1 31 0 1.841 0.07 SYSEDMD DISNOP 1 7 6 0 0 0 515 0 2.260 0.08 SYSEDMD DISNO4I 1 47 3 0 1 0 8075 0 25.516 0.97 SYSEDMD DISNO4IS 57 0 0 0 624 0 36877 0 105.650 4.03 SYSEDMD DISNRS 1 0 0 0 44 0 511 0 3.343 0.12 SYSEDMD DISNSP 1 18 15 0 0 0 1850 0 6.074 0.23 SYSEDMD DISNTMZ 1 4 11 0 0 0 324 0 2.309 0.08 SYSEDMD MENU 1 1 3 0 0 0 2 0 0.235 0.00 SYSEDMD MONACSH 1 6 6 0 0 0 1217 0 3.470 0.13 SYSEDMD MONADA 1 3176 71 0 0 0 272180 0 680.214 25.98 SYSEDMD MONAREP 1 9 28 0 0 0 1964 0 6.378 0.24 ... Total 5294 5293 2122 7 6510 43 857384 0 2617.326 100.00

Explanations:

• The Natural object MONADA consumed the most CPU time: 680.214 ms which corresponds to 25.98 percent of the total CPU time.

• MONADA was started once, it loaded 3176 other Natural objects, performed 71 database calls and 272180 Natural statements. There was no I/O, no external call and no error in the program.

• At the end of the program summary, the Total counts of the profiling are listed.

Line Summary

If LINE=ON is specified for an NPRC resource file, the Profiler line summary is generated.

The line summary shows for each source line in a Natural object, the number of Natural events that occurred (hit count), the CPU and elapsed time (in milliseconds and percent) spent by the line. The percentage of times is calculated in relation to the total times of the application.

The line summary does not count Monitor Pause events and events at Level 0.

If the data is exported in CSV (comma-separated values) format, the count of each event type is listed. Additionally, the Adabas times (absolute and percentage values) are displayed. The exported time values are indicated in microseconds.

Example of a Line Summary

The following example shows the extract of a line summary:

Natural Profiler Line Summary -------------------------------- Library Program Line CC-Lib CC-Name Hit-Count CPU (ms) CPU % Elapsed (ms) Ela % PRFTEST XINT 0000 1 0.016 0.46 0.003 0.01 PRFTEST XINT 0140 1 0.005 0.14 0.001 0.00 PRFTEST XINT 0150 1 0.006 0.17 0.002 0.01 PRFTEST XINT 0160 1 0.004 0.11 0.001 0.00 PRFTEST XINT 0170 23 0.128 3.75 0.029 0.18 PRFTEST XINT 0180 10 0.049 1.43 0.012 0.07 PRFTEST XINT 0190 10 0.054 1.58 0.010 0.06 ... Total 371 3.408 100.00 15.992 100.00

Explanations:

• Line 0170 in the Natural object XINT consumed 0.128 ms of the CPU time and 0.029 ms of the elapsed time. This corresponds to 3.75 percent of the total CPU time and 0.18 percent of the total elapsed time. 23 events (hit count) were executed in the line.

• At the end of the line summary, the Total counts of the profiling are listed.

Transaction Summary

A transaction is the code executed between two consecutive terminal I/O operations. The elapsed time spent on executing a transaction is called response time.

The transaction summary generated by the Profiler shows how many Natural events have occurred for each transaction, the response time (in milliseconds) and the percentage of the response time spent by the transaction with respect to the total response time.

Prerequisites Required for Transaction Evaluations

The Profiler data must be consolidated with the CONSOLIDATE function set to TRANSACTION=ON so that the consolidated data records in the NPRC resource file contain transaction identifiers required for evaluating the transaction response time.

TRANSACTION=ON is specified for the READ function.

Event Data Evaluated for Transactions

In general, one terminal I/O event relates to one transaction. The number of terminal I/O events is not listed in standard output (for PRINT=ON).

Data exported in CSV (comma-separated values) format contains the count of each event type, including terminal I/O. Additionally, the data list contains the CPU time and the Adabas times (absolute and percentage values). Exported time values are indicated in microseconds.

The transaction summary does not consider Before Terminal I/O (IB) events, RPC Wait for Client (RW) events, Monitor Pause (MP) events and events at Level 0. Events that belong to one event group are combined into one count.

For explanations of event types and associated event groups, see the section Events.

Modified Line and Program Summary

If TRANSACTION=ON is specified for the READ function of an NPRC resource file, the program summary and the line summary also consider the transaction identifier and aggregate the program and line values for each transaction separately. In the transaction line summary, the ID column indicates the program line in which the transaction was started.

Example of a Transaction Summary

The following example shows a transaction summary:

Natural Profiler Transaction Summary ------------------------------------ Transact TA-Lib TA-Prog TA-CC TA-Line Program Database External Error Statement User Elapsed (ms) Time% 1 1 0 1 0 0 0 0.829 0.50 2 SYSPRFLR PRFMENM 0020 9 13 8 0 0 0 17.618 10.65 3 SYSEDM MOPTTEST 0020 4 0 0 0 0 0 6.167 3.72 4 SYSEDM OPTWLS80 0470 0 0 0 0 0 0 1.108 0.66 5 SYSEDM OPTWLS80 0470 2 0 0 0 0 0 1.133 0.68 6 SYSEDM OPTTEST 1750 1 0 0 0 0 0 1.180 0.71 7 SYSEDM MOPTTEST 0020 27 33 30 0 0 0 94.236 56.96 8 SYSEDM OPTTEST 5590 1 0 0 0 0 0 1.185 0.71 9 SYSEDM MOPTTEST 0020 37 33 56 0 0 0 37.311 22.55 10 SYSEDM OPTTEST 5760 1 0 0 0 0 0 1.096 0.66 11 SYSEDM MOPTTEST 0020 6 0 6 0 0 0 2.164 1.30 12 SYSPRFLR PRFMENM 0020 2 0 3 0 0 0 1.394 0.84 Total 86 79 103 0 0 0 165.421 100.00

Explanations using the example of Transaction 7:

• Transaction 7 spent the longest response (elapsed) time: 94.236 milliseconds (ms) which correspond to 56.96 percent of the total response time.

• Transaction 7 started an I/O operation in the library SYSEDM (TA-Lib column), program MOPTTEST (TA-Prog), program line 0020 (TA-Line).

• Transaction 7 accessed 27 programs and issued 33 database calls and 30 external program calls.

• In the example above, all counters for statement events are 0 because statement events were not collected during this Profiler run.

• At the end of the transaction summary, the Total counts of the profiled transactions are listed.

Program Coverage

The program coverage table is generated if PROGRAM=ON is specified for an NCVF resource file.

The program coverage table shows the code coverage results for each accessed Natural object. If the table is given in text format, only the GP coverage results (copycodes included) are displayed. In CSV (comma-separated values) format, the table shows lines containing copycode values, additional columns with source counters (copycodes not included) and information regarding INCLUDE statements.

In text format, the table provides the coverage count for each accessed library and for the whole application.

The table contains the following columns:

Example of Program Coverage

The following example shows the result of program coverage in text format:

Program Coverage
----------------
Library   Object   Ty Coverage%  Covered    Missed    Total
COVDEMO   TESTCOVN N      78.5%       11         3       14
COVDEMO   TESTCOVP P      57.1%        4         3        7
COVDEMO   -------- --     71.4%       15         6       21
Totals    -------- --     71.4%       15         6       21

Explanations:

• The application accesses two objects, the TESTCOVN subprogram (N) and the TESTCOVP program (P).

• In TESTCOVN, there are 14 executable statements from which 11 were covered (executed) and 3 missed (not executed), giving a total coverage of 78.5%.

• The summarized values of the two objects accessed in the library COVDEMO show coverage of 71.4%.

• Total coverage is also 71.4% because only one library is accessed by the objects.

Statement Coverage

Statement coverage is generated if EVENT=ON is specified for an NCVF resource file.

For statement coverage, the Profiler utility reads the source of the monitored objects. First it searches the source in the library containing the Natural Profiler resource. If the source is not found in this library, it is searched in the library given with the Profiler data. If the source is still not found or if the source does not match the collected data, source lines are not printed in the statement coverage report. The Profiler utility resolves INCLUDE statements and merges the source of the corresponding copycode into the including program. If the INCLUDE structure cannot be resolved, the copycodes are printed separately.

If a source has been modified after the coverage run, the coverage data does no longer fit to the source, and the Profiler utility can no longer provide the full information. To prevent this, copy the resource file together with the related program sources into another library. If the Profiler utility reads the resource, the sources from the resource library are taken and a modification of the original sources does not affect the statement coverage.

Statement coverage shows the percentage of statements covered for each source line of the accessed programs. If the result is written in text format, for each object listed in the statistics, the object coverage values are shown before the statement coverage data. If the result is written in CSV (comma-separated values) format, additional information regarding statement coverage is provided.

The table contains the following columns:

Example of Statement Coverage

The following example assumes that the development has delivered a new version of the TESTCOVN subprogram to the quality engineering. After running the test programs, statement coverage of the subprogram shows the following result (text format):

M Cov% CC-Lib   CC-Name  Line Source
*                        0010 * Test function Coverage 
*                        0020 * Subprogram TESTCOVN 
+                        0030 DEFINE DATA 
+                        0040 PARAMETER 
+                        0050 1 FUNC     (I2)  /* function 
+                        0060 1 RET-CODE (I4)  /* Return code 
+                        0070 END-DEFINE 
*                        0080 * 
*                        0090 /* Return 0 by default 
C 100%                   0100 RESET RET-CODE 
*                        0110 * 
C 100%                   0120 DECIDE ON FIRST VALUE OF FUNC 
+                        0130   VALUE 0 
+                        0140     PRINT 'Test function 0' 
+                        0150   VALUE 1 
C 100%                   0160     PRINT 'Test function 1' 
+                        0170   VALUE 2 
C 100%                   0180     PRINT 'Test function 2' 
+                        0190   VALUE 3 
C 100%                   0200     PRINT 'Test function 3' 
+                        0210   VALUE 4 
C 100%                   0220     PRINT 'Test function 4' 
+                        0230   VALUE 5 
C 100%                   0240     PRINT 'Test function 5' 
+                        0250   VALUE 6 
C 100%                   0260     PRINT 'Test function 6' 
+                        0270   VALUE 7 
C 100%                   0280     PRINT 'Test function 7' 
+                        0290   VALUE 8 
C 100%                   0300     PRINT 'Test function 8' 
+                        0310   VALUE 9 
+                        0320     PRINT 'New test function 9' 
+                        0330   NONE VALUE 
+                        0340     RET-CODE := 1  /* Unsupported function 
+                        0350 END-DECIDE 
*                        0360 * 
C 100%                   0370 END

Explanations:

• The Mark (M) column shows whether a line is covered (C).

• No test cases cover the functions Test function 0 and New test function 9. The NONE VALUE case is also not covered.

• All other test cases are covered (denoted with C and 100% coverage).

As a consequence of this coverage analysis, the test cases have to be adjusted so that Test function 0 and Test function 9 (and, perhaps, the error case with an unsupported function code) are also covered.

Using a Microsoft Excel Template to Visualize Coverage Results

Prerequisites: Microsoft Excel and Natural for Windows or Natural for Linux and Cloud.

If you want to analyze the coverage result with Microsoft Excel, you can use the Microsoft Excel template delivered with Natural for Windows and Natural for Linux and Cloud. Perform the following steps:

1. Perform the Profiler READ function and write the output data in CSV (comma-separated values) format to Work File 7. For example:

   FUNCTION=READ           /* Read Profiler Data
     RESOURCE-NAME='Test'  /* Resource name
     RESOURCE-LIB=PRFDATA  /* Resource library
     RESOURCE-TYPE=NCVF    /* Use resource type NCVF
     EVENT=ON              /* Print statement coverage
     PROGRAM=ON            /* Print program coverage
     STATISTICS=ON         /* Print statistics
     PRINT=ON              /* Write to standard output
     EXPORT=ON             /* Write to Work File 7
     FORMAT=COMMA          /* Export in CSV format

2. If your Microsoft Excel requires semicolons as separators, specify the following:

   FORMAT=SEMICOLON      /* Export in CSV format

3. Export the data of Work File 7 with any tool (such as FTP) as a CSV-formatted file to a Windows environment.

4. Open the CSV file with Microsoft Excel.

5. Rearrange the data so that each evaluation type (program, event, statistics) is on its own worksheet in the Microsoft Excel file.

6. Open the delivered template TESTCOV.XLSX with Microsoft Excel. The template is contained in the RES (Resources) subdirectory of the Natural SYSPRFLR system library.

7. For each worksheet, copy the format from the template to your Microsoft Excel:

   • Click on the upper left corner of the table in the template to mark all data in the table.

   • Click on the Microsoft Excel Copy format function.

   • Click on the upper left corner of the table in your worksheet to copy the format.

     Now, all entries are formatted as in the template. The source lines are colored and marked as follows:

     Color	Mark	Description
     Green	C	All statements in the line are covered.
     Yellow	P	The statements in the line are partly covered.
     Pink	M	All statements in the line are missed.
     Gray	*	A comment or an empty line.
     Red	E	Error encountered. For example, if the coverage analysis has collected a line number but the corresponding source line is not found.
     None (white)	+	All other lines such as continuation lines of a statement.

Example of a Microsoft Excel Worksheet

The following example shows a worksheet extract of code coverage for the TESTCOVP program with included TESTCOVC copycode without the columns that contain the object name and library:

Explanations:

• The source lines of the TESTCOVC copycode are included in the source of the TESTCOVP program and placed right after the corresponding INCLUDE statement.

• The lines 40 through 70 of the copycode contain statements which were not executed in the test run.

• All other lines (in green) containing executable statements are covered.

Statistics

If STATISTICS=ON is specified, the Profiler statistics are listed.

If the data is exported in CSV (comma-separated values) format, the properties and values of the Profiler statistics are added as separate columns to the event or consolidation trace. If coverage data is exported in CSV format, the statistics values are added in additional lines indicated by the value Statistics in the Evaluation column.

Example of Statistics

The following example shows an extract of the statistics of an NPRC resource file:

*************************************************************************** * 17:35:59 ***** NATURAL PROFILER UTILITY ***** 2016-01-11 * User SAGTEST1 - Statistics - RESDATA ... * * Profiler Resource File * Resource name ...................... EDM-MONITOR.nprc * Resource type ...................... Natural Profiler Resource Consolidated * Resource allocation date ........... 2015-07-27 10:36:19.6 * Resource size (bytes) .............. 565160 ... * * Data Processing * Number of events ................... 895936 ... * * Data Consolidation * Consolidation ...................... ON * Consolidation records .............. 21624 * Consolidation elapsed time (sec) ... 15.643516 * Consolidation factor ............... 41.4 * Consolidation records/block ........ 191.3 * Bytes/consolidation record ......... 25.8 * ... ***************************************************************************

Explanations:

• The EDM-MONITOR.nprc resource was allocated on 2015-07-27 at 10:36:19 a.m. and has a size of 565160 bytes.

• The profiled application generated a total of 895936 Natural events. The data consolidation took 15.6 seconds and reduced the number of records to 21624 which corresponds to a consolidation factor of 41.4.

All statistics information provided is explained in the section Profiler Statistics.

Maintaining Profiler Resource Files

In general, Profiler resources are listed as NPRF, NPRC or NCVF files by using the Natural SYSMAIN utility, NaturalONE or Natural Studio. These tools also provide functions to copy, rename and delete resource files.

The Natural Profiler Rich GUI also lists the Profiler resources and offers a function to delete a Profiler resource file.

In addition, you can use Profiler utility functions to list Profiler resource files.

This section covers the following topic:

• Listing Profiler Resource Files

Listing Profiler Resource Files

The Profiler utility LIST function lists the Profiler resource files of a given Natural library and the date and time when the resource files were allocated.

Syntax of LIST:

Syntax Description:

Example of LIST

The following example lists the NPRF Profiler resource files of library PRFDATA. The list is written to standard output and to Work File 7 in text format.

FUNCTION=LIST           /* List Profiler resource files
  RESOURCE-LIB=PRFDATA  /* Resource library
  RESOURCE-TYPE=NPRF    /* List NPRF resource files
  PRINT=ON              /* Write to standard output
  EXPORT=ON             /* Write to Work File 7
  FORMAT=TEXT           /* Export in text format

Output:

Natural Profiler Resources
--------------------------
Library: PRFDATA
Resource type: nprf
 
Count Date       Time     Name
    1 2015-06-15 14:32:18 Hello1.nprf
    2 2015-06-26 18:39:57 QDTest1.nprf
    3 2015-06-24 22:00:35 QETest1.nprf
    4 2015-06-30 14:32:42 Studio.nprf
    5 2015-07-02 15:02:32 Test.nprf
 
Number of nprf resources in library PRFDATA: 5

Including Profiler Input from Natural Text Objects

The Profiler can read input data from a Natural text object. The syntax of the data in the Natural text object is the same as for the primary command input data set CMSYNIN (see Syntax and Keywords).

To include Profiler input data from a Natural text object

• Enter the following Profiler keywords:

  INCLUDE-LIB=library-name
  INCLUDE=object-name

  The keyword syntax is explained in Profiler Utility Keywords.

The data in the Natural text object is added to the Profiler input data in the line after the INCLUDE keyword. The Profiler input data can contain multiple INCLUDE keywords, and the related Natural text objects can also contain INCLUDE keywords. If a Natural text object contains an END-PROFILER keyword, the Profiler utility terminates and any remaining data in the Natural text object(s) is ignored.

The Natural system library SYSPRFLR supplies text object whose names begin with X which can be used as Profiler input. The individual Profiler functions they perform are described in the sources of these objects.

We recommend that you do not modify any objects in the system library SYSPRFLR because they can be overwritten or removed when a new Natural version is installed. Copy the required object(s) to a user library before you edit it.

Examples of INCLUDE

The following example adds the contents of the Natural MYPROF text object from the library MYLIB to the Profiler input data:

INCLUDE-LIB=MYLIB
INCLUDE=MYPROF

The following example adds the content of the Natural text member XCONS from the library SYSPRFLR to the Profiler input data. The object consolidates Profiler event data. Additionally, it terminates the Profiler utility so that no further Profiler input is expected after the INCLUDE keyword.

INCLUDE=XCONS

Event Trace

The Natural Profiler collects detailed information of each Natural event that occurs while a Natural application executes. This data can be viewed in the event trace.

The traces written for Natural code coverage are described in the section Tracing Natural Code Coverage.

The Profiler utility provides the following options to write a Profiler event trace:

• Write the trace to standard output while the NPRF data is consolidated. In this case, the event trace shows the delta values of the elapsed time and the CPU time instead of event-specific data.

• Write the trace when reading a Profiler NPRF resource file with the Profiler utility READ function.

To enable the event trace

• Enter the following subordinate keyword of the Profiler utility CONSOLIDATE function:

  TRACE-EVENT=ON

  Enter the following subordinate keyword of the Profiler utility READ function:

  EVENT=ON

The Profiler event trace contains the following columns:

Example of an Event Trace

In the following example, the Profiler utility READ function prints the event trace:

FUNCTION=READ               /* Read event data
  EVENT=ON                  /* Write event trace

The event trace is written to standard output:

Count Time              CPU-Time (ms) Ev Lev Library  Program  Line CC-Lib   CC-Name  Statement  Local-Data
    0 17:38:17.200951          42.324 MP 003 SYSPRFLR PRBINIT  8370                   Call       Monitor pause requested
    0 17:38:17.204508          43.471 MP 003 SYSPRFLR PRBSTART 1760                   Call       Start of block filter
   11 17:38:17.218379          48.874 DB 000                   0000                              00010/00032 S1
   12 17:38:17.218941          48.897 DA 000                   0000                              00010/00032 S1   Rsp: 0
   13 17:38:17.218944          48.910 PL 000                   0000                              Execute PRFDEMO/XPROF
   14 17:38:17.218945          48.916 PS 001 PRFDEMO  XPROF    0000                   PgmStart   00010/00032 Type: P
   15 17:38:17.218956          48.979 IB 001 PRFDEMO  XPROF    0300                   Input      Out: 133 In: 0
   16 17:38:17.219235          49.046 IA 001 PRFDEMO  XPROF    0300                   Input      Out: 133 In: 80
   17 17:38:17.219258          49.182 DB 001 PRFDEMO  XPROF    0370                   Callnat    00010/00032 S1
   18 17:38:17.220426          49.211 DA 001 PRFDEMO  XPROF    0370                   Callnat    00010/00032 S1   Rsp: 0
   19 17:38:17.220427          49.216 DB 001 PRFDEMO  XPROF    0370                   Callnat    00010/00032 S1
   ...

Tracing Natural Code Coverage

When the coverage resource is read with the Profiler utility READ function, the coverage data can be traced with the internal data trace.

To enable tracing for code coverage

• Enable the internal trace by specifying the following subordinate keyword of the Profiler utility READ function:

  TRACE=9

The table below describes the properties listed in the trace:

Internal Trace

The Profiler internal trace writes Profiler messages such as errors or warnings.

The internal trace can be activated for the following:

• The Profiler data processing functions. The data is written to standard output.

• The Profiler Rich GUI offers a drop-down box for the internal trace activation. The data is written to standard output.

To activate the internal trace for the Profiler data processing functions

• Enter the following Profiler keyword:

  TRACE=n

  where n is the trace level (see Trace Levels).

  Notes:

  1. By default (if TRACE is not specified), Trace Level 2 (warnings) is used.
  2. The trace is activated as soon as the TRACE keyword is specified. It is therefore recommended to specify the TRACE keyword as soon as possible.
  3. If you execute the Profiler utility multiple times in the job, you need to specify the TRACE keyword with each execution.

Trace Levels

The trace levels used by the Profiler trace and monitor sessions and by the Profiler data processing functions are listed in the following table. In general, a higher trace level also contains the information of the lower trace levels. For example, if you select Trace Level 3 (statistics), error messages and warnings are also logged.

We recommend that you use at least Trace Level 2 (warnings) so that error messages and warnings are logged.

Example

In the following example, the Profiler internal trace is set to 4 (function):

* Set Profiler internal trace
TRACE=4                     /* Trace level

Profiler Statistics

The Natural Profiler on Linux and Windows does not collect properties and statistics during the profiling. If you consolidate an NPRF file on Linux or Windows, statistics which can be determined from the collected trace data, are generated. The statistical data is written to the Profiler NPRC resource file in addition to the event data.

The Profiler utility provides the following options to write and view Profiler statistics:

• Write the statistics to standard output while the data is consolidated.

• Write the statistics when reading a Profiler resource file with the Profiler utility READ function.

• Select a resource in the Natural Profiler Rich GUI to display the corresponding properties and statistics.

To write Profiler statistics, perform one of the following steps

• Enter the following keyword before you start the Profiler utility CONSOLIDATE function:

  TRACE=3

  or a higher trace level (see Trace Levels).

• Enter the following subordinate keyword of the Profiler utility READ function:

  STATISTICS=ON

The Profiler statistical data is displayed in categories combining properties of a similar type. The following categories are available:

• General Information

• Profiler Resource File

• Monitor Session

• Trace Session

• Data Processing

• Event Type Statistics

• Monitor Pause Statistics

• Data Consolidation

• Coverage

• Transaction

General Information

Display environment and Natural Profiler related information.

Profiler Resource File

Display Profiler resource file related information.

Monitor Session

Display statistics of the Profiler monitor session.

Trace Session

Display statistics of the Profiler trace session. The Profiler trace session includes also the application execution.

Data Processing

Display statistics of the data processing, compression and transfer.

The compression rate is calculated by the following formula:

Event Type Statistics

Display statistics of the event types.

Monitor Pause Statistics

Display statistics of the types of Monitor Pause events.

Data Consolidation

Display statistics of the data consolidation.

ConsolidationFactor := NumberOfEvents / ConsolidationRecords

Coverage

Display statistics of Natural code coverage.

Transaction

Display statistics of executed transactions.

For more information on evaluating transactions, see Transaction Summary.