David Lange

Just as data sets from next-generation experiments grow, processing requirements for physics analysis become more computationally demanding, necessitating performance optimizations for RooFit. One possibility to speed-up minimization and add stability is the use of Automatic Differentiation (AD). Unlike for numerical differentiation, the computation cost scales linearly with the number of parameters, making AD particularly appealing for statistical models with many parameters. In this paper, we report on one possible way to implement AD in RooFit. Our approach is to add a facility to generate C++ code for a full RooFit model automatically. Unlike the original RooFit model, this generated code is free of virtual function calls and other RooFit-specific overhead. In particular, this code is then used to produce the gradient automatically with Clad. Clad is a source transformation AD tool implemented as a plugin to the clang compiler, which automatically generates the derivative code for input C++ functions. We show results demonstrating the improvements observed when applying this code generation strategy to HistFactory and other commonly used RooFit models. HistFactory is the subcomponent of RooFit that implements binned likelihood models with probability densities based on histogram templates. These models frequently have a very large number of free parameters and are thus an interesting first target for AD support in RooFit.

Among the well-known methods to approximate derivatives of expectancies computed by Monte-Carlo simulations, averages of pathwise derivatives are often the easiest one to apply. Computing them via algorithmic differentiation typically does not require major manual analysis and rewriting of the code, even for very complex programs like simulations of particle-detector interactions in high-energy physics. However, the pathwise derivative estimator can be biased if there are discontinuities in the program, which may diminish its value for applications. This work integrates algorithmic differentiation into the electromagnetic shower simulation code HepEmShow based on G4HepEm, allowing us to study how well pathwise derivatives approximate derivatives of energy depositions in a sampling calorimeter with respect to parameters of the beam and geometry. We found that when multiple scattering is disabled in the simulation, means of pathwise derivatives converge quickly to their expected values, and these are close to the actual derivatives of the energy deposition. Additionally, we demonstrate the applicability of this novel gradient estimator for stochastic gradient-based optimization in a model example.

RooFit is a toolkit for statistical modeling and fitting used by most experiments in particle physics. Just as data sets from next-generation experiments grow, processing requirements for physics analysis become more computationally demanding, necessitating performance optimizations for RooFit. One possibility to speed-up minimization and add stability is the use of Automatic Differentiation (AD). Unlike for numerical differentiation, the computation cost scales linearly with the number of parameters, making AD particularly appealing for statistical models with many parameters. In this paper, we report on one possible way to implement AD in RooFit. Our approach is to add a facility to generate C++ code for a full RooFit model automatically. Unlike the original RooFit model, this generated code is free of virtual function calls and other RooFit-specific overhead. In particular, this code is then used to produce the gradient automatically with Clad. Clad is a source transformation AD tool implemented as a plugin to the clang compiler, which automatically generates the derivative code for input C++ functions. We show results demonstrating the improvements observed when applying this code generation strategy to HistFactory and other commonly used RooFit models. HistFactory is the subcomponent of RooFit that implements binned likelihood models with probability densities based on histogram templates. These models frequently have a very large number of free parameters and are thus an interesting first target for AD support in RooFit.

File formats for generic data structures, such as ROOT, Avro, and Parquet, pose a problem for deserialization: it must be fast, but its code depends on the type of the data structure, not known at compile-time. Just-in-time compilation can satisfy both constraints, but we propose a more portable solution: specialized virtual machines. AwkwardForth is a Forth-driven virtual machine for deserializing data into Awkward Arrays. As a language, it is not intended for humans to write, but it loosens the coupling between Uproot and Awkward Array. AwkwardForth programs for deserializing record-oriented formats (ROOT and Avro) are about as fast as C++ ROOT and 10–80× faster than fastavro. Columnar formats (simple TTrees, RNTuple, and Parquet) only require specialization to interpret metadata and are therefore faster with precompiled code.

C++ Modules, one of the new features of C++20, aim to fix the long-standing build scalability problems in the language. They provide an IOefficient, on-disk representation capable to reduce build times and peak memory usage. ROOT already employs the C++ modules technology in its dictionary system to improve performance and reduce the memory footprint. ROOT with C++ Modules was released as a technology preview in fall 2018, after intensive development during the previous few years. The current state is ready for production, however, there is still room for performance optimizations. In this talk, we show the road map for making this technology enabled by default in ROOT. We demonstrate a global module indexing optimization which allows reducing the memory footprint dramatically for many workflows. We will report user feedback on the migration to ROOT with C++ Modules.

The Awkward Array library has been an important tool for physics analysis in Python since September 2018. However, some interface and implementation issues have been raised in Awkward Array’s first year that argue for a reimplementation in C++ and Numba. We describe those issues, the new architecture, and present some examples of how the new interface will look to users. Of particular importance is the separation of kernel functions from data structure management, which allows a C++ implementation and a Numba implementation to share kernel functions, and the algorithm that transforms recordoriented data into columnar Awkward Arrays.

C++ Modules come in C++20 to fix the long-standing build scalability problems in the language. They provide an io-efficient, on-disk representation capable to reduce build times and peak memory usage. ROOT employs the C++ modules technology further in the ROOT dictionary system to improve its performance and reduce the memory footprint. ROOT with C++ Modules was released as a technology preview in fall 2018, after intensive development during the last few years. The current state is ready for production, however, there is still room for performance optimizations. In this talk, we show the roadmap for making the technology default in ROOT. We demonstrate a global module indexing optimization which allows reducing the memory footprint dramatically for many workflows. We will report user feedback on the migration to ROOT with C++ Modules.

The Awkward Array library has been an important tool for physics analysis in Python since September 2018. However, some interface and implementation issues have been raised in Awkward Array's first year that argue for a reimplementation in C++ and Numba. We describe those issues, the new architecture, and present some examples of how the new interface will look to users. Of particular importance is the separation of kernel functions from data structure management, which allows a C++ implementation and a Numba implementation to share kernel functions, and the algorithm that transforms record-oriented data into columnar Awkward Arrays.

The high-luminosity program has seen numerous extrapolations of its needed computing resources that each indicate the need for substantial changes if the desired HL-LHC physics program is to be supported within the current level of computing resource budgets. Drivers include large increases in event complexity (leading to increased processing time and analysis data size) and trigger rates needed (5-10 fold increases) for the HL-LHC program. The CMS experiment has recently undertaken an effort to merge the ideas behind short-term and long-term resource models in order to make easier and more reliable extrapolations to future needs. Near term computing resource estimation requirements depend on numerous parameters: LHC uptime and beam intensities; detector and online trigger performance; software performance; analysis data requirements; data access, management, and retention policies; site characteristics; and network performance. Longer term modeling is affected by the same characteristics, but with much larger uncertainties that must be considered to understand the most interesting handles for increasing the "physics per computing dollar" of the HL-LHC. In this presentation, we discuss the current status of long term modeling of the CMS computing resource needs for HL-LHC with emphasis on techniques for extrapolations, uncertainty quantification, and model results. We illustrate potential ways that high-luminosity CMS could accomplish its desired physics program within today's computing budgets.