Building Caliper

Building and installing Caliper requires cmake, a current C++11-compatible compiler, and a Python interpreter. To build Caliper manually, clone it from the github repository. Next, configure and build Caliper, e.g.:

$ cd <path to caliper root directory>
$ mkdir build && cd build
$ cmake -DCMAKE_INSTALL_PREFIX=<path to install location> ..
$ make
$ make install

This is a minimal configuration, see below for additional CMake configuration flags.

CMake Flags

Configure Caliper using the usual CMake flags to select the C and C++ compiler, build type, and install prefix (CMAKE_C_COMPILER, CMAKE_CXX_COMPILER, CMAKE_BUILD_TYPE, and CMAKE_INSTALL_PREFIX). In addition, Caliper supports the following CMake flags:

BUILD_DOCS

Build documentation.

BUILD_SHARED_LIBS

Build shared or static library. Default: On (build shared library).

BUILD_TESTING

Build unit tests.

WITH_ADIAK

Enable support for recording program metadata with the Adiak library. Point CMake to adiak CMake module, e.g. with -Dadiak_PREFIX=<path-to-adiak>/lib/cmake/adiak.

WITH_CUPTI

Enable support for CUDA performance analysis (wrapping of driver/runtime API calls and CUDA activity tracing).

WITH_FORTRAN

Build the Fortran wrappers.

WITH_GOTCHA

Enable Gotcha support. Allows pthread, IO, and malloc/free tracking, and enables dynamic wrapping of MPI functions. If Gotcha is disabled, MPI wrappers use the PMPI interface. Requires Linux.

WITH_LIBDW

Enables libdw support for DWARF symbol lookup. Required for most sampling-based configurations.

WITH_LIBPFM

Enable Linux perf_event sampling. Set libpfm installation dir in LIBPFM_INSTALL.

WITH_LIBUNWIND

Enables libunwind support for call-path unwinding.

WITH_MPI

Build with MPI support.

WITH_NVTX

Build adapters to forward Caliper annotations to NVidia’s nvtx annotation API. Set CUDA_TOOLKIT_ROOT_DIR to the CUDA installation.

WITH_OMPT

Build with support for the OpenMP tools interface.

WITH_PAPI

Enable PAPI support. Set PAPI installation dir in PAPI_PREFIX.

WITH_ROCTX

Build adapters to forward Caliper annotations to AMD’s roctx annotation API.

WITH_ROCTRACER

Enable support for ROCm/HIP performance analysis (runtime API profiling and GPU activity tracing).

WITH_SAMPLER

Enable time-based sampling on Linux.

WITH_TOOLS

Build Caliper’s tools (i.e, cali-query and mpi-caliquery). Default: On.

WITH_VTUNE

Build adapters to forward Caliper annotations to Intel’s VTune annotation API. Set Intel ITT API installation dir in ITT_PREFIX.

All options are off by default. On Linux, Gotcha is enabled by default.

Linking Caliper programs

Typically, all that is needed to create a Caliper-enabled program is to link it with the Caliper runtime library, which resides in libcaliper.so. An example link command for a C++ program built with g++ could look like this:

CALIPER_DIR = /path/to/caliper/installation

g++ -o target-program $(OBJECTS) -L$(CALIPER_DIR)/lib64 -lcaliper

Using Caliper in CMake projects

Caliper creates a CMake package file (caliper-config.cmake) and installs it in <caliper-installation-dir>/lib64/cmake/caliper. The package file defines Caliper’s include directories and exports targets for the Caliper libraries. Projects using CMake can use find_package() and target_link_libraries() to integrate Caliper as a dependency.

This example CMake code builds a program which depends on Caliper:

find_package(caliper)
add_executable(MyExample MyExample.cpp)
target_link_libraries(MyExample PRIVATE caliper)

When configuring the target program, point CMake to the desired Caliper installation with caliper_DIR:

cmake -Dcaliper_DIR=<caliper-installation-dir>/share/cmake/caliper ..

The CMake package defines the following variables and targets:

${caliper_INCLUDE_DIR}	Caliper include directory (variable)
caliper	The Caliper runtime library (target)
caliper-serial	Caliper runtime library without MPI dependencies (target)
caliper-tools-util	Utilities for caliper tools (target)

In most cases, just link the “caliper” target.