timemory 3.1.0rc1

Lightweight cross-language instrumentation API for C, C++, Python, Fortran, and CUDA which allows arbitrarily bundling tools together into a single performance analysis handle

timemory

Timing + Memory + Hardware Counter Utilities for C / C++ / CUDA / Python

timemory on GitHub (Source code)

timemory General Documentation (ReadTheDocs)

timemory Source Code Documentation (Doxygen)

timemory Testing Dashboard (CDash)

timemory Tutorials

GitHub	git clone https://github.com/NERSC/timemory.git
PyPi	pip install timemory
Spack	spack install timemory

Purpose

The goal of timemory is to create an open-source performance measurement and analyis package with modular and reusable components which can be used to adapt to any existing C/C++ performance measurement and analysis API and is arbitrarily extendable by users within their application. Timemory is not just another profiling tool, it is a profling toolkit which streamlines building custom profiling tools through modularity and then utilizes the toolkit to provides several pre-built tools.

Versioning

Timemory originated as a very simple tool for recording timing and memory measurements (hence the name) in C, C++, and Python and only supported three modes prior to the 3.0.0 release: a fixed set of timers, a pair of memory measurements, and the combination of the two. Prior to the 3.0.0 release, timemory was almost completely rewritten from scratch with the sole exceptions of some C/C++ macro, e.g. TIMEMORY_AUTO_TIMER, and some Python decorators and context-manager, e.g. timemory.util.auto_timer, whose behavior were able to be fully replicated in the new release. Thus, while it may appear that timemory is a mature project at v3.0+, it is essentially still in it's first major release.

Documentation

The full documentation is available at timemory.readthedocs.io. Detailed source documentation is provided in the doygen section of the full documentation. Tutorials are available in the github.com/NERSC/timemory-tutorials.

Overview

Timemory is designed, first and foremost, to be a portable, modular, and fully customizable toolkit for performance measurement and analysis of serial and parallel programs written in C, C++, Fortran, Python, and CUDA.

Timemory arose out of the need for a universal adapator kit for the various APIs provided several existing tools and a straight-forward and intuitive method for user-defined expression of performance measurements which can easily encapsulated in a generic structure. Performance measurement components written with timemory are arbitrarily scalable up to any number of threads and processes and fully support intermixing different measurements at different locations within the program -- this uniquely enables timemory to be deployed to collect performance data at scale in HPC because highly detailed collection can occur at specific locations within the program where ubiquitous collection would simulatenously degrade performance significantly and require a prohibitive amount of memory.

Timemory can be used as a backend to bundle instrumentation and sampling tools together, support serialization to JSON/XML, and provide statistics among other uses. It can also be utilized as a front-end to invoke custom instrumentation and sampling tools. Timemory uses the abstract term "component" for a structure which encapsulates some performance analysis operation. The structure might encapsulate function calls to another tool, record timestamps for timing, log values provided by the application, provide a operator for replacing a function in the code dynamically, audit the incoming arguments and/or outgoing return value from function, or just provide stubs which can be overloaded by the linker.

Features

• C++ Template API
  • Modular and fully-customizable
  • Adheres to C++ standard template library paradigm of "you don't pay for what you don't use"
  • Simplifies and facilitates creation and implementation of performance measurement tools
    • Create your own instrumentation profiler
    • Create your own instrumentation library
    • Create your own sampling profiler
    • Create your own sampling library
    • Create your own execution wrappers
    • Supplement timemory-provided tools with your own custom component(s)
    • Thread-safe data aggregation
    • Aggregate collection over multiple processes (MPI and UPC++ support)
    • Serialization to text, JSON, XML
  • Components are composable with other components
  • Variadic component bundlers which maintain complete type-safety
    • Components can be bundled together into a single handle without abstractions
  • Components can store data in any valid C++ data type
  • Components can return data in any valid C++ data type
• C / C++ / CUDA / Fortran Library API
  • Straight-forward collection of functions and macros for creating built-in performance analysis to your code
  • Component collection can be arbitrarily inter-mixed
    • E.g. collect "A" and "B" in one region, "A" and "C" in another region
  • Component collection can be dynamically manipulated at runtime
    • E.g. add/remove "A" at any point, on any thread, on any process
• Python API
  • Decorators and context-managers for functions or regions in code
  • Python function profiling
  • Python line-by-line profiling
  • Every component in timemory-avail is provided as a stand-alone Python class
    • Provide low-overhead measurements for building your own Python profiling tools
• Command-line Tools
  • timemory-avail
    • Provides available components, settings, and hardware counters
    • Quick API reference tool
  • timem (UNIX)
    • Extended version of UNIX time command-line tool that includes additional information on memory usage, context switches, and hardware counters
    • Support collecting hardware counters (Linux-only, requires PAPI)
  • timemory-run (Linux)
    • Dynamic instrumentation profiling tool
    • Supports runtime instrumentation and binary re-writing
  • timemory-python-profiler
    • Python function profiler
  • timemory-python-line-profiler
    • Python line-by-line profiler
    • Design based on line-profiler package
    • Extended to use components: cpu-clock, memory-usage, context-switches, etc. (all components which collect scalar values)
• Instrumentation Libraries
  • kokkos-connector: Kokkos Profiling Libraries
  • timemory-mpip: MPI Profiling Library (Linux-only)
  • timemory-ompt: OpenMP Profiling Library

Design Goals

• Toolkit for creating new performance analysis tools
• Common instrumentation framework
  • Eliminate need for projects to explicitly support multiple instrumentation frameworks
• High performance during data collection
• Low overhead when dormant (disabled at runtime)
• Zero overhead when disabled at compile time
• Support arbitrarily intermixing components:
  • Instrument measurements of A, B, and C around arbitrary region 1
  • Instrument measurements of A and C around arbitrary region 1.1 (nested with Section 1)
  • Instrument measurements of C around arbitrary region 2
  • Instrument measurements of D around arbitrary region 3
  • No instrumentation around arbitrary region 4
• Intuitive and simple API to use and extend

Component Basics

Timemory components are C++ structs (class which defaults to public instead of private) which define a single collection instance, e.g. the wall_clock component is written as a simple class with two 64-bit integers with start() and stop() member functions.

// This "component" is for conceptual demonstration only
// It is not intended to be copy+pasted
struct wall_clock
{
    int64_t m_value = 0;
    int64_t m_accum = 0;

    void start();
    void stop();
};

The start() member function which records a timestamp and assigns it to one of the integers temporarily, the stop() member function which records another timestamp, computes the difference and then assigns the difference to the first integer and adds the difference to the second integer.

void wall_clock::start()
{
    m_value = get_timestamp();
}

void wall_clock::stop()
{
    // compute difference b/t when start and stop were called
    m_value = (get_timestamp() - m_value);
    // accumulate the difference
    m_accum += m_value;
}

Thus, after start() and stop() is invoked twice on the object:

wall_clock foo;

foo.start();
sleep(1); // sleep for 1 second
foo.stop();

foo.start();
sleep(1); // sleep for 1 second
foo.stop();

The first integer (m_value) represents the most recent timing interval of 1 second and the second integer (m_accum) represents the accumulated timing interval totaling 2 seconds. This design not only encapsulates how to take the measurement, but also provides it's own data storage model. With this design, timemory measurements naturally support asynchronous data collection. Additionally, as part of the design for generating the call-graph, call-graphs are accumulated locally on each thread and on each process and merged at the termination of the thread or process. This allows parallel data to be collection free from synchronization overheads. On the worker threads, there is a concept of being at "sea-level" -- the call-graphs relative position based on the base-line of the primary thread in the application. When a worker thread is at sea-level, it reads the position of the call-graph on the primary thread and creates a copy of that entry in it's call-graph, ensuring that when merged into the primary thread at the end, the accumulated call-graph across all threads is inserted into the appropriate location. This approach has been found to produce the fewest number of artifacts.

In general, components do not need to conform to a specific interface. This is relatively unique approach. Most performance analysis which allow user extensions use callbacks and dynamic polymorphism to integrate the user extensions into their workflow. It should be noted that there is nothing preventing a component from creating a similar system but timemory is designed to query the presence of member function names for feature detection and adapts accordingly to the overloads of that function name and it's return type. This is all possible due to the template-based design which makes extensive use of variadic functions to accept any arguments at a high-level and SFINAE to decide at compile-time which function to invoke (if a function is invoked at all). For example:

• component A can contain these member functions:
  • void start()
  • int get()
  • void set_prefix(const char*)
• component B can contains these member functions:
  • void start()
  • void start(cudaStream_t)
  • double get()
• component C can contain these member functions:
  • void start()
  • void set_prefix(const std::string&)

And for a given bundle component_tuple<A, B, C> obj:

• When obj is created, a string identifer, instance of a source_location struct, or a hash is required
  • This is the label for the measurement
  • If a string is passed, obj generates the hash and adds the hash and the string to a hash-map if it didn't previously exist
  • A::set_prefix(const char*) will be invoked with the underlying const char* from the string that the hash maps to in the hash-map
  • C::set_prefix(const std::string&) will be invoked with string that the hash maps to in the hash-map
  • It will be detected that B does not have a member function named set_prefix and no member function will be invoked
• Invoking obj.start() calls the following member functions on instances of A, B, and C:
  • A::start()
  • B::start()
  • C::start()
• Invoking obj.start(cudaStream_t) calls the following member functions on instances of A, B, and C:
  • A::start()
  • B::start(cudaStream_t)
  • C::start()
• Invoking obj.get():
  • Returns std::tuple<int, double> because it detects the two return types from A and B and the lack of get() member function in component C.

This design makes has several benefits and one downside in particular. The benefits are that timemory: (1) makes it extremely each to create a unified interface between two or more components which different interfaces and different capabilities, (2) invoking the different interfaces is very efficient because no logic is required at runtime to determine if a particular feature is implemented by a component, and (3) components get to define their own interface.

With respect to #2, consider the two more traditional implementations. If callbacks are used, a function pointer exists and a component that does not implement this feature will either have a null function pointer (requiring a check at compile time) or the tool will implement an array of function pointers with an unknown size at compile-time. In the latter case, this will require heap allocations (expensive operations) and in both cases, the loop of the function pointers will likely be quiet ineffienct since function pointers have a very high probability of thrashing the instruction cache. If dynamic polymorphism is the approach taken, then the result is multiple virtual table look-ups during every iteration. In the timemory approach, none of these additional overheads are present and there isn't even a for loop -- the bundle expands into a direct call to the member function without any abstractions or nothing.

With respect to #1 and #3, this has some interesting implications with regard to a universal instrumentation interface and is discussed in the following section and the CONTRIBUTING.md documentation.

The aforementioned downside is that the byproduct of all this flexibility and adaption to custom interfaces by each component is that directly using the template interface can take quite a long time to compile.

Support for timemory in external tools

The previous section gave an overview of how components can define their own interfaces. Currently, timemory internally provides compatibility with multiple tools but the end goal is for the majority of this to be maintained by the authors of the tool. This will benefits users by provided a single method for using all of their favorite tools and make it extremely easy for them to try out new tools. This will benefit the authors of the tools because there will be a significantly lower the introduction barrier required for users to try out the new tool -- if the user is familiar with timemory, the tool can be trivially integrated into either their code or into the profiler.

An external tool can easily provide compatibility with timemory and leverage all of its work creating a low-overhead measurement system in parallel environments, Python extensions, and dynamic instrumentation, by simply providing a header in their source code which defines the interface the tool wants to provide and the tools can add/remove support at will without having to maintain any source code in timemory or worry about version compatability with timemory. Versioning issues do not inherently exist because for several reasons which are detailed the CONTRIBUTING.md documentation.

Support for Multiple Instrumentation APIs

• LIKWID
• Caliper
• TAU
• gperftools
• MPI
• OpenMP
• CrayPAT
• Allinea-MAP
• PAPI
• ittnotify (Intel Parallel Studio API)
• CUPTI (NVIDIA performance API)
• NVTX (NVIDIA marker API)

Generic Bundling of Multiple Tools

• CPU and GPU hardware counters via PAPI
• NVIDIA GPU hardware counters via CUPTI
• NVIDIA GPU tracing via CUPTI
• Generating a Roofline for performance-critical sections on the CPU and NVIDIA GPUs
  • Classical Roofline (FLOPs)
  • Instruction Roofline
• Memory usage
• Tool insertion around malloc, calloc, free, cudaMalloc, cudaFree
• Wall-clock, cpu-clock, system-clock timing
• Number of bytes read/written to file-system (and rate)
• Number of context switches
• Trip counts
• CUDA kernel runtime(s)
• Data value tracking

Powerful GOTCHA Extensions

• GOTCHA is an API wrapping function calls similar to the use of LD_PRELOAD
  • Significantly simplify existing implementations
• Scoped GOTCHA
  • Enables temporary wrapping over regions
• Use gotcha component to replace external function calls with custom replacements
  • E.g. replace the C math function exp with custom exp implementation
• Use gotcha component to wrap external library calls with custom instrumentation

Multi-language Support

• Variadic interface to all the utilities from C code
• Variadic interface to all the utilities from C++ code
• Variadic interface to all the utilities from Python code
  • Includes context-managers and decorators

Profiling and timemory

Timemory includes the timemory-run as a full profiler for Linux systems. This executable supports dynamic instrumentation (instrumenting at the target applicaiton's runtime), attaching to a running process, and binary re-writing (creating a new instrumented binary). The instrumented applications support flat-profiling, call-stack profiling, and timeline profiling and can be configured to use any of the components timemory provides or, with a little work, can also be used to instrument custom components defined by the user. It is highly recommended for custom tools targetting specific functions to use the combination of GOTCHA and the dynamic instrumentation. Using the GOTCHA extensions for profiling specific functions enables creating components which replace the function or audit the incoming arguments and return values for the functions and the dynamic instrumentation makes it easy to inject using the GOTCHA wrappers into an executable or library.

Interface Basics

Timemory can be quite useful as the backend for creating your own profiling interface, but a frontend interface is also provided for those who want to do quick performance analysis. The following is an example in many languages for collecting the total cache misses in the L1, L2, and L3 cache levels. The particular hardware counters can be set in the environment or directly in C++ or Python. See timemory-avail documentation for more settings and descriptions.

• Environment:
  • export TIMEMORY_PAPI_EVENTS="PAPI_L1_TCM, PAPI_L2_TCM, PAPI_L3_TCM"
• C++:
  • tim::settings::papi_events() = "PAPI_L1_TCM, PAPI_L2_TCM, PAPI_L3_TCM"
• Python:
  • timemory.settings.papi_events = "PAPI_L1_TCM, PAPI_L2_TCM, PAPI_L3_TCM"

C / C++ Library Interface

timemory_push_components("papi_vector");
timemory_push_region("MY_REGION_OF_INTEREST");
//
// do something in region of interest...
//
timemory_pop_region("MY_REGION_OF_INTEREST");

Fortran

call timemory_push_components("papi_vector")
call timemory_push_region("MY_REGION_OF_INTEREST")
!
! do something in region of interest...
!
call timemory_pop_region("MY_REGION_OF_INTEREST")

C++ Template Interface

using hwcounters_t = tim::component_tuple<tim::component::papi_vector>;

hwcounters_t roi("MY_REGION_OF_INTEREST");
roi.start();
//
// do something in region of interest...
//
roi.stop();
// access to data
auto hwc = roi.get();
// print
std::cout << hwc << '\n';

Or encoding the PAPI enumeration types explicitly:

using hwcounters_t = tim::component_tuple<
    tim::component::papi_tuple<PAPI_L1_TCM, PAPI_L2_TCM, PAPI_L3_TCM>>;

hwcounters_t roi("MY_REGION_OF_INTEREST");
roi.start();
//
// do something in region of interest...
//
roi.stop();
// access to data
auto hwc = roi.get();
// print
std::cout << hwc << '\n';

Python Context Manager

from timemory.util import marker

with marker(["papi_vector"], key="MY_REGION_OF_INTEREST"):
    #
    # do something in region of interest...
    #

Python Component Class

from timemory.component import PapiVector

hwc = PapiVector("MY_REGION_OF_INTEREST")
hwc.start()
//
// do something in region of interest...
//
hwc.stop()
// get values
l1_tcm, l2_tcm, l3_tcm = hwc.get()
// print as string
print("{}".format(hwc))

C Enumeration Interface

void* roi = TIMEMORY_BLANK_MARKER("MY_REGION_OF_INTEREST", PAPI_VECTOR);
//
// do something in region of interest...
//
FREE_TIMEMORY_MARKER(roi)

Create Your Own Tools/Components

• Written in C++
• Direct access to performance analysis data in Python and C++
• Create your own components: any one-time measurement or start/stop paradigm can be wrapped with timemory
  • Flexible and easily extensible interface: no data type restrictions in custom components

Composable Components Example

Building a brand-new component is simple and straight-forward. In fact, new components can simply be composites of existing components. For example, if a component for measuring the FLOP-rate (floating point operations per second) is desired, it is arbitrarily easy to create and this new component will have all the features of wall_clock and papi_vector component:

// This "component" is for conceptual demonstration only
// It is not intended to be copy+pasted
struct flop_rate : base<flop_rate, double>
{
private:
    wall_clock  wc;
    papi_vector hw;

public:
    static void global_init()
    {
        papi_vector::add_event(PAPI_DP_OPS);
    }

    void start()
    {
        wc.start();
        hw.start();
    }

    void stop()
    {
        wc.stop();
        hw.stop();
    }

    auto get() const
    {
        return hw.get() / wc.get();
    }
};

Extended Example

The simplicity of creating a custom component that inherits category-based formatting properties (is_timing_category) and timing unit conversion (uses_timing_units) can be easily demonstrated with the wall_clock component and the simplicity and adaptability of forwarding timemory markers to external instrumentation is easily demonstrated with the tau_marker component:

TIMEMORY_DECLARE_COMPONENT(wall_clock)
TIMEMORY_DECLARE_COMPONENT(tau_marker)

// type-traits for wall-clock
TIMEMORY_DEFINE_CONCRETE_TRAIT(is_timing_category, component::wall_clock, true_type)
TIMEMORY_DEFINE_CONCRETE_TRAIT(uses_timing_units, component::wall_clock, true_type)
TIMEMORY_STATISTICS_TYPE(component::wall_clock, double)

namespace tim
{
namespace component
{
//
// the system's real time (i.e. wall time) clock, expressed as the
// amount of time since the epoch.
//
// NOTE: 'value', 'accum', 'get_units()', etc. are provided by base class
//
struct wall_clock : public base<wall_clock, int64_t>
{
    using ratio_t    = std::nano;
    using value_type = int64_t;
    using base_type  = base<wall_clock, value_type>;

    static std::string label() { return "wall"; }
    static std::string description() { return "wall-clock timer"; }

    static value_type  record()
    {
        // use STL steady_clock to get time-stamp in nanoseconds
        using clock_type    = std::chrono::steady_clock;
        using duration_type = std::chrono::duration<clock_type::rep, ratio_t>;
        return std::chrono::duration_cast<duration_type>(
            clock_type::now().time_since_epoch()).count();
    }

    double get_display() const { return get(); }

    double get() const
    {
        // get_unit() provided by base_clock via uses_timing_units type-trait
        auto val = (is_transient) ? accum : value;
        return static_cast<double>(val) / ratio_t::den * get_unit();
    }

    void start()
    {
        value = record();
    }

    void stop()
    {
        value = (record() - value);
        accum += value;
    }
};

//
// forwards timemory instrumentation to TAU instrumentation.
//
struct tau_marker : public base<tau_marker, void>
{
    // timemory component api
    using value_type = void;
    using this_type  = tau_marker;
    using base_type  = base<this_type, value_type>;

    static std::string label() { return "tau"; }
    static std::string description() { return "TAU_start and TAU_stop instrumentation"; }

    static void global_init(storage_type*) { Tau_set_node(dmp::rank()); }
    static void thread_init(storage_type*) { TAU_REGISTER_THREAD();     }

    tau_marker() = default;
    tau_marker(const std::string& _prefix) : m_prefix(_prefix) {}

    void start() { Tau_start(m_prefix.c_str()); }
    void stop()  { Tau_stop(m_prefix.c_str());  }

    void set_prefix(const std::string& _prefix) { m_prefix = _prefix; }
    // This 'set_prefix(...)' member function is a great example of the template
    // meta-programming provided by timemory: at compile-time, timemory checks
    // whether components have this member function and, if and only if it exists,
    // timemory will call this member function for the component and provide the
    // marker label.

private:
    std::string m_prefix = "";
};

}  // namespace component
}  // namespace tim

Using the two tools together in C++ is as easy as the following:

#include <timemory/timemory.hpp>

using namespace tim::component;
using comp_bundle_t = tim::component_tuple_t <wall_clock, tau_marker>;
using auto_bundle_t = tim::auto_tuple_t      <wall_clock, tau_marker>;
// "auto" types automatically start/stop based on scope

void foo()
{
    comp_bundle_t t("foo");
    t.start();
    // do something
    t.stop();
}

void bar()
{
    auto_bundle_t t("foo");
    // do something
}

int main(int argc, char** argv)
{
    tim::init(argc, argv);
    foo();
    bar();
    tim::finalize();
}

Using the pure template interface will cause longer compile-times and is only available in C++ so a library interface for C, C++, and Fortran is also available:

#include <timemory/library.h>

void foo()
{
    uint64_t idx = timemory_get_begin_record("foo");
    // do something
    timemory_end_record(idx);
}

void bar()
{
    timemory_push_region("bar");
    // do something
    timemory_pop_region("bar");
}

int main(int argc, char** argv)
{
    timemory_init_library(argc, argv);
    timemory_push_components("wall_clock,tau_marker");
    foo();
    bar();
    timemory_pop_components();
    timemory_finalize_library();
}

In Python:

import timemory
from timemory.profiler import profile
from timemory.util import auto_tuple

def get_config(items=["wall_clock", "tau_marker"]):
    """
    Converts strings to enumerations
    """
    return [getattr(timemory.component, x) for x in items]

@profile(["wall_clock", "tau_marker"])
def foo():
    """
    @profile (also available as context-manager) enables full python instrumentation
    of every subsequent python call
    """
    # ...

@auto_tuple(get_config())
def bar():
    """
    @auto_tuple (also available as context-manager) enables instrumentation
    of only this function
    """
    # ...

if __name__ == "__main__":
    foo()
    bar()
    timemory.finalize()

GOTCHA and timemory

C++ codes running on the Linux operating system can take advantage of the built-in GOTCHA functionality to insert timemory markers around external function calls. GOTCHA is similar to LD_PRELOAD but operates via a programmable API. This include limited support for C++ function mangling (in general, mangling template functions are not supported -- yet).

Writing a GOTCHA hook in timemory is greatly simplified and applications using timemory can specify their own GOTCHA hooks in a few lines of code instead of being restricted to a pre-defined set of GOTCHA hooks.

Example GOTCHA

If an application wanted to insert tim::auto_timer around (unmangled) MPI_Allreduce and (mangled) ext::do_work in the following executable:

#include <mpi.h>
#include <vector>

int main(int argc, char** argv)
{
    init();

    MPI_Init(&argc, &argv);

    int sizebuf = 100;
    std::vector<double> sendbuf(sizebuf, 1.0);
    // ... do some stuff
    std::vector<double> recvbuf(sizebuf, 0.0);

    MPI_Allreduce(sendbuf.data(), recvbuf.data(), sizebuf, MPI_DOUBLE, MPI_SUM, MPI_COMM_WORLD);
    // ... etc.

    int64_t nitr = 10;
    std::pair<float, double> settings{ 1.25f, 2.5 };
    std::tuple<float, double> result = ext::do_work(nitr, settings);
    // ... etc.

    return 0;
}

This would be the required specification using the TIMEMORY_C_GOTCHA macro for unmangled functions and TIMEMORY_CXX_GOTCHA macro for mangled functions:

#include <timemory/timemory.hpp>

static constexpr size_t NUM_FUNCS = 2;
using gotcha_t = tim::component::gotcha<NUM_FUNCS, tim::auto_timer_t>;
void init()
{
    TIMEMORY_C_GOTCHA(gotcha_t, 0, MPI_Allreduce);
    TIMEMORY_CXX_GOTCHA(gotcha_t, 1, ext::do_work);
}
// uses comma operator to call init() during static construction of boolean
static bool is_initialized = (init(), true);

Compilation with the Template Interface

It was noted above that direct use of the template interface can introduce long compile-times. However, this interface is extremely powerful and one might be tempted to use it directly. The 2011 standard of C++ introduced the concept of an extern template and it is highly recommended to use this feature if the template interface is used. In general, a project using the template interface should have a header which declares the component bundle as an extern template at the end. Here is example of what this might look like:

#include <timemory/variadic/component_bundle.hpp>
#include <timemory/variadic/auto_bundle.hpp>
#include <timemory/components/types.hpp>
#include <timemory/macros.hpp>

// create an API for your project
TIMEMORY_DEFINE_API(FooBenchmarking)

#if defined(DISABLE_BENCHMARKING)
// this will elimiate all components from the component_bundle or auto_bundle
// with 'api::FooBenchmarking' as the first template parameter
// e.g. bundle<Foo, ...> turns into bundle<Foo> (no components)
TIMEMORY_DEFINE_CONCRETE_TRAIT(is_available, api::FooBenchmarking, false_type)
#endif

// this structure will:
//  - Always record:
//      - wall-clock timer
//      - cpu-clock timer
//      - cpu utilization
//      - Any tools which downstream users inject into the user_global_bundle
//          - E.g. 'user_global_bundle::configure<peak_rss>()'
//  - Optionally enable activating (at runtime):
//      - PAPI hardware counters
//      - GPU kernel tracing
//      - GPU hardware counters
//      - The '*' at the end is what designates the component as optional
#if !defined(FOO_TOOLSET)
#define FOO_TOOLSET                             \
    tim::component_bundle<                      \
        tim::api::FooBenchmarking,              \
        tim::component::wall_clock,             \
        tim::component::cpu_clock,              \
        tim::component::cpu_util,               \
        tim::component::user_global_bundle,     \
        tim::component::papi_vector*,           \
        tim::component::cupti_activity*,        \
        tim::component::cupti_counters*>
#endif

namespace foo
{
namespace benchmark
{
using bundle_t = FOO_TOOLSET;
using auto_bundle_t = typename FOO_TOOLSET::auto_type;
}
}

//  THIS WILL MAKE SURE THE TEMPLATE NEVER GETS INSTANTIATED
//  LEADING TO SIGNIFICANTLY REDUCED COMPILE TIMES
#if !defined(FOO_BENCHMARKING_SOURCE)
extern template class FOO_TOOLSET;
#endif

And then in the one source file:

// avoid the extern template declaration
// make sure this is defined before inclusing the header
#define FOO_BENCHMARKING_SOURCE

// include the header with the code from the previous block
#include "/path/to/header/file"

// pull in all the definitions required to instantiate the template
#include <timemory/timemory.hpp>

// provide an instantiation
template class FOO_TOOLSET;

A similar scheme to the above is used extensively internally by timemory -- the source code contains many almost empty .cpp files which contain only a single line of code: #include "timemory/<some-path>/extern.hpp. These source files are part of the scheme for pre-compiling many of the expensive template instantiations (the templated storage class, in particular), not junk files that were accidentally committed. In this scheme, when the .cpp file is compiled a macro is used to transform the statement in the header into a template instantiation but when included from other headers, the macro transforms the statement into an extern template declaration. In general, this is how it is implemented:

#
# source/timemory/components/foo/CMakeLists.txt
#
add_library(foo SHARED <OTHER_FILES> extern.cpp)
target_compile_definitions(foo
    #  extern.cpp will be compiled with -DTIMEMORY_FOO_SOURCE
    PRIVATE     TIMEMORY_FOO_SOURCE
    #  When the "foo" target part of a 'target_link_libraries(...)'
    #  command by another target downstream, CMake will add
    #  -DTIMEMORY_USE_FOO_EXTERN to the compile definitions
    INTERFACE   TIMEMORY_USE_FOO_EXTERN)

//
// source/timemory/components/foo/extern.hpp
//
#if defined(TIMEMORY_FOO_SOURCE)
#   define FOO_EXTERN_TEMPLATE(...) template __VA_ARGS__;
#elif defined(TIMEMORY_USE_FOO_EXTERN)
#   define FOO_EXTERN_TEMPLATE(...) extern template __VA_ARGS__;
#else
#   define FOO_EXTERN_TEMPLATE(...)
#endif

// in header-only mode, the macro makes the code disappear
FOO_EXTERN_TEMPLATE(tim::component::base<Foo>)
FOO_EXTERN_TEMPLATE(tim::operation::start<Foo>)
FOO_EXTERN_TEMPLATE(tim::operation::stop<Foo>)
FOO_EXTERN_TEMPLATE(tim::storage<Foo>)

//
// source/timemory/components/foo/extern.cpp
//
#include "timemory/components/foo/extern.hpp"

Additional Information

For more information, refer to the documentation.