Debugging Chapel Programs

This document discusses support for debugging your Chapel program and a set of experimental settings to enable task monitoring and memory tracking.

Running in gdb/lldb

The compiler-generated executable has a --gdb flag that can be used to launch the program within a gdb session. A similar flag, --lldb, exists to launch the program within a lldb session. For best results, you should follow Best Known Configuration to build Chapel and build your application.

Running in gdb/lldb with a launcher

When using almost any launcher, you can launch gdb by setting the environment variable CHPL_COMM_USE_GDB when running the program. This will open up a separate terminal emulator window for each locale, each running the debugger on that locale’s program instance. On the Mac OS X (darwin) platform, you can launch lldb instead, by setting the CHPL_COMM_USE_LLDB environment variable. This works in all of these launchers:

amudprun
aprun
gasnetrun_ibv
gasnetrun_mpi
gasnetrun_ofi
mpirun4ofi
pbs-aprun
smp

The default terminal emulator program is xterm, but by setting the environment variable CHPL_COMM_DBG_TERM=urxvt you can force use of urxvt instead. Whichever terminal emulator is used must be in your PATH on the compute node or an error will result. Note that it is the user’s responsibility to make sure things are set up so the terminal emulator run in the target environment can open its display window in the launch environment.

The Debugger.breakpoint statement

The Debugger module provides a parenless function called breakpoint. When the code is compiled and run with debug symbols, i.e. -g, the attached debugger will automatically stop at calls to this function as a breakpoint. Code that contains breakpoint that is compiled without -g will work as normal with no side effects. Saving the generated code to as temporary directory with --savec DIRECTORY will also allow the debugger to read and display Chapel source code. This works well with either the LLVM or C backends.

Note

Executables will not run as expected if breakpoint is used in code compiled with -g and not run attached to a debugger.

Best Known Configuration

The current best practice for debugging Chapel source code is to use the C backend and use a series of flags to improve the debuggability of the generated executable. This can be done in two steps.

  1. Build the compiler with CHPL_TARGET_COMPILER set to gnu:

    CHPL_TARGET_COMPILER=gnu make
    
  2. Build the executable from Chapel source code:

    chpl -g --target-compiler=gnu --savec <dir> --preserve-inlined-line-numbers --no-munge-user-idents --no-return-by-ref --no-inline <source_file>
    

For more details on these settings, read the rest of this section.

Building the Compiler

For best results while debugging, we recommend building the compiler with CHPL_TARGET_COMPILER set to gnu (or clang if on Mac). See Setting up Your Environment for Chapel for more information on building the compiler.

With two invocations of the build command, both backends can be built. First execute make (which uses the LLVM backend by default) and then execute CHPL_TARGET_COMPILER=gnu make. This will keep the default as LLVM and allow switching to the C backend as needed for debugging. This can be done for a particular invocation of the compiler with chpl --target-compiler=gnu ....

Building the Application

The following flags can be useful for making the generated C more amenable to debugging. Any of them can be omitted as desired.

Flag

Description

-g

Generate debug symbols in the executable

--target-compiler=gnu

Target the C backend

--savec <dir>

Write out the generated C to a given directory

--preserve-inlined-line-numbers

When code gets inlined (e.g. replacing a function call with the function body) maintain the filename and line number information of the original function.

--no-munge-user-idents

Don’t munge user identifiers (e.g. variable or function names). Munging typically prevents conflicts with identifiers in external code but makes debugging harder.

--no-return-by-ref

Don’t use an extra reference argument when compiling a Chapel function that returns a record.

--no-inline

Avoid inlining in many cases.

Notes on munging

The utility of using a debugger with Chapel depends greatly on your familiarity with the Chapel generated code. However, if your program is crashing or running into a runtime error, you can often determine where that is taking place by looking at a stack trace within gdb.

When debugging Chapel, it is useful to know that in generating its code, the Chapel compiler renames user identifiers. By default, the Chapel compiler munges all user identifiers, such that a variable named x would be code generated as x_chpl. This munging can be controlled using the --[no-]munge-user-idents flag (see the chpl man page for more information). In some cases, additional munging may be required or applied that cannot be turned off.

The net effect of this is that Chapel variables can often be inspected using p name_chpl (or p name_chpl<TAB> in cases where the compiler has further renamed the variable). If the --no-munge-user-idents flag is used, p name or p name<TAB> should work in most cases.

See Why are my identifiers renamed in the generated code? for more information on munging.

Over time, we plan to improve our ability to debug the generated C code for a Chapel program. If you find yourself debugging the generated code a lot and need help or have requests for better support, please let us know so that we can prioritize accordingly.

Tracking and Reporting on Tasks

For certain tasking layers, Chapel supports an experimental capability for tracking the status of tasks, primarily designed for use in a single-locale execution. To enable this capability, your program must be compiled with the --task-tracking flag.

The feature itself is enabled at execution time by setting the boolean environment variable CHPL_RT_ENABLE_TASK_REPORTING to any of the values “1”, “yes”, or “true”. If this is done, then when <CTRL-C> is entered while a program is executing, a list of pending and executing tasks will be printed to the console, giving an indication of which tasks are at which source locations. This is only supported with CHPL_TASKS=fifo.

Note that task tracking adds a fair amount of runtime overhead to task-parallel programs.

Configuration Constants for Tracking Memory

Chapel supports a number of configuration constants related to dynamic memory allocation for the compiler-generated executable, currently designed for use primarily by the development team to track memory usage in tests. Please note that our generated code currently contains memory leaks, so you should not be surprised if your program requires more memory than it seems it should.

For full information on these configuration constants consult MemDiagnostics.

A brief synopsis of these configuration constants is as follows:

--memTrack

turn on memory tracking and enable reporting

--memStats

call printMemAllocStats() on normal termination

--memLeaksByType

call printMemAllocsByType() on normal termination

--memLeaks

call printMemAllocs() on normal termination

--memMax=int

set maximum level of allocatable memory

--memThreshold=int

set minimum threshold for memory tracking

--memLog=string

file to contain all memory reporting

--memLeaksLog=string

if set, append final stats and leaks-by-type here