ProLaza and ProLaza64 - Features

32 bit version:
ProLaza supports 32 bit applications developed with Lazarus 1.0 or better for 32 and 64 bit processors.

64 bit version:
ProLaza64 supports 64 bit applications developed with Lazarus 1.0 or better for 64 bit processors.

Action controlled measurement:
Measurement can be switched on and off dynamically by entering and finishing a procedure or a function. This can be activated user friendly by a selection form (for fast changing the activation points), by special comments and by API-calls in the source files (if permanently the same points shall be used for a long period).

All-in-one-profiling:
With one profiling session the number of calls for a method (call count profiling ), the average runtime of a method (function profiling),  function coverage profiling and emulation profiling can be performed. When developing ProLaza, the philosophy was (and is) to develop a high precision profiling tool (thats why it is neither a sampling  nor a machine code instrumenting profiler) and to produce a minimum of overhead (that's why there is no line profiling mode). See also Profiler type .

Call graphs:

A graph shows by who a procedure is called and which procedure it is calling. It also shows how often this happened an the time consumed when processing these calls. By clicking on the procedure symbol a quick following the execution path of the application.

Command line:
Profiling from command line is possible.

Compiler versions:
ProLaza supports Lazarus 1.x.

Conditional compilation:
Fully supported. Compiler symbols and switches are read from the compilers project file (LPI).

Coverage profiling:
A browser can display all methods which were not called.

Cyclic storage of measurement results:

Measurements can be stored cyclically (activated by online operation window). With the viewer it can be scrolled through the different measurements.

Documentation:
ProLaza comes with a real User Guide (and not with slim online help).

Emulation profiling:
The measured results can be recalculated for a faster or slower PC. So it's e.g. not necessary to have a customers PC for the development process. By entering a speed constant and the number of MHz'es the other PC can be emulated. Of course the results are not as accurate as they are with measuring on the original equipment.

Exclusion of parts of the program from measurement:

Directories:

Units or Include files: Parts of a unit or an include file: Automatic exclusion: Freeware Versions:
The Freeware Versions can be downloaded from the ProDelphi homepage and Torry's .
The Professional Versions can be ordered at ShareIt software registration service and is sent as e-mail download link.
(See also'Differences between Freeware- and Professional version' ).

Granularity:
ProLaza measures in CPU-cycles. The smallest measurable unit is 1 CPU-cycle. This means that on a Pentium with 1000 MHz the smallest measurable duration is 0.001 s = 1ns.

Handling of idle times:
Idle times produced by some Windows-API calls are not measured, also idle times of some Lazarus calls. See list of handled functions in the manual.

Hardware:
ProLaza supports Pentium and Pentium compatible processors.

History function:
ProLaza comes with a built-in history function. The viewer shows by a colored grid, which functions got faster and which got slower since the last storing of results into the history file. By a mouse click on a result line of the viewers grid, the time from the history file is displayed for the selected procedure. Multiple history files can be stored.

History of ProLaza:
ProLaza is ProDelphi ported for Lazarus developpment.

Integration into the Lazarus IDE:
ProLaza can be integrated into the Lazarus tools menu. The setup program displays a window with instructions.

Libraries:
In the professional mode the user part of libraries can be profiled.

Limitations:
There are no limitations regarding size or accuracy: Even the smallest procedures are measured accurate, there is no minimum size required for a procedure to be measured.

Line profiling:
Line profiling is not supported. This has two reasons: Counting runtimes for lines changes the instruction processing of the CPU so much, that no useful results can be expected. Counting runtime for lines also slows down a program too much.

Local Procedures:
Can optionally be excluded from measurement..

Measuring parts of procedures:
By special comments extra measurement points inside procedures can be defined.

Measuring runtimes in DLL's:
Fully supported. Program and DLL can be measured simultaneously.

Measurement accuracy:
ProLaza measures very accurately. The measurement overhead is automatically removed from the measured times. A sophisticated algorithm calculates the runtime used for measurement at initialization time. Nested function calls are no problem for ProLaza. 

Measurement results:
The measured durations are displayed either in CPU-Cycles, in a fixed or in a variable time format. Following time formats are automatically selected:

For all methods the time for a single call and the runtime sum is displayed (also additionally with included child times).

Multiple profiling directories:
Fully supported. ProLaza reads the project file (LPI) of Lazarus and automatically scans all directories in the search path + the directory of the DPR-file.

Object orientation:
Specially supported by the optional feature of adding the runtimes of all inherited methods with the same name.

Example: A method 'LoadFromDisk' calls the method 'LoadFromDisk' of the inherited class. In this case the runtimes of both classes will be added.

Online (de)activation of measurement:
Measurement can be switched on and off at the programs runtime, also intermediate results can be stored online.

Operating systems:
ProLaza is compatible to Windows 2000/XP/Server 2003/Vista/7/8 and 10.
ProLaza64 is compatible to Windows XP 64 bit / Vista 64 bit / 7 64 bit / 8 64 bit / 10 64 bit.

Optimization option:
The optimization of the Delphi compiler is automatically deactivated if the user forgets this. This guaranties high measurement accuracy.

Overhead:
Measurement of runtime costs measurement overhead. This is another point (like accuracy) where ProLaza shows its outstanding quality: by a very low measurement overhead. All measurement routines are extremely optimized.

Packages:
Package profiling is not supported yet.

Post Mortem Review:
Instead of vaccinating (instrumenting) the sources with statements for runtime measurement, it can be vaccinated with statements for exception trapping. In case of an exception occuring the call stack is stored into a file inclusive the name of the exception. Optionally the program can be traced, see Tracing.
The program can be delivered in an instrumented form to a customer. This could be done in case of sporadic exceptions. The source code does not have to be given to the customer (together with the Lazarus-IDE).

Professional version:

See upgrade to professional version.
(See also 'Differences between Freeware- and Professional Version' ).

Profiler type:
ProLaza is a source code instrumenting profiler. Source instrumenting versus machine code instrumenting has the big advantage, that at the best position possible a time stamp can be taken: at the start and the end of the procedure body. No original code of the measured procedure has to be replaced by a branch to the measurement procedure and stored and executed there. The normal instruction flow (nearly identical as without profiling) is kept. This is one of the reasons for ProLaza's outstanding accuaracy.

Another advantage of a source instrumenting profiler is, that idle times (caused e.g. by opening a message box) can be handled. Before calling the Windows- or LCL - function, the measurement is deactivated.
The only disadvantage of a source instrumenting profiler is that the source files are changed by the vaccination (instrumentation) process. To avoid the risk of impacting the sources, they should be saved before.

Sampling profilers: with a low sampling rate small procedures can not be measured, with a high sampling rate the profiled program has to be interrupted very often, what means, that the runtime behaviour of the tested program is changed very much. Sampling profilers also cannot handle idle times.

See also comparison between the three profiler types .

Programming-API:
Measurement can be switched on and off at runtime. Intermediate results can be stored on disk.

Security:
ProLaza inserts statements into the source files. To prevent a loss of code in case of power failure, the sources is saved before instrumentation (filename_sav). When the instrumentation is processed sucessfully, the saved file is deleted.
Because 'shit happens' all source code files should be saved before profiling (e.g. with WinZip).

Setup:
ProLaza is delivered with an easy to use setup program. It copies all files into the appropriate directories and creates an entry in the Windows list of programs (Start menu / Programs).

Supported Lazarus versions:
Lazarus 1.x.

Supported libraries:
LCL and FPC.

Switching off of all measurement:
Simply done by deleting a compiler symbol and recompilation.

Threads:
ProLaza has special settings for single threaded applications (low overhead, extremly accurate) and for multi threaded applications (medium overhead, medium accurate). A special option allows to exclusively measure the main thread only (Professional version).

Tracing:
An option of Post Mortem Review is tracing. ProLaza can insert trace statement at procedure entry and exit points. For using this function DebugDelphi needs to be installed and started. The version of DebugDelphi for 32 bit applications is free and can be downloaded from this site.

Upgrading to Professional Version:
If you need ProLaza for a larger project and 20 measured procedures are not enough for you, you can order the professional version via ShareIt registration service.

Updates of Professional Version:
Updates of the Professional Version due to bugfixes are free.
Upgrades
of the Professional Version due to new versions of Lazarus or new features of ProLaza are available for 15.-

Viewing of measurement results:
Results can be either displayed in CPU-cycles, in a fixed or in a variable time format. On a pentium with 1000 MHz the smallest time unit is 0.001s (= 1 ns).

    With the built-in viewer and browser
(navigator):

The complete results can be sorted using different criteria.
Procedures which consume the most CPU-cycles or are called most often are additionally prefiltered and sorted.
In professional mode besides average runtimes also minimum and maximum times can be viewed.
The browser can be used to quickly find a certain method, or to find out which method was not called.
Text search for units, classes or methods is available.
Caller / Call graph (see screen shot).

      The viewer can automatically exclude methods with a very low runtime from displaying in the measurement results.