CHEP 2018, Sofia, Bulgaria Optimizing Frameworks’ Performance Using C++ Modules-Aware ROOT Yuka Takahashi, Vassil Vassilev, Raphael Isemann yuka.takahashi, vvasilev, raphael.isemann @cern.ch

1. Introduction PCH files are precompiled header files and ever, with modules, we can automatically We will present our results and challenges work similar to C++ modules. The advan- resolve symbols and cases like those are with C++ modules in ROOT. ROOT was ex- tage of modules over PCH is that they can now correctly handled. be used by experiments. Experiments are tended with experimental support for using 6. Implementation C++ modules during runtime, with aims to still using textual includes as PCH only cov- reduce it’s memory usage and improve its ers ROOT. PCH cannot be exported to ex- correctness. periments because of various technical limi- tations. 2. C++ Modules in a Nutshell • Header information is stored in precom- 5. Results piled PCM files 5.1 Perfomance • No more header parsing during ROOT’s runtime In C/C++, the interface of a library is ac- cessed by including the appropriate header files: # include < s t d i o . h> Figure 5: Visualization of ROOT interpreter These textual includes cause well-known core. problems: As shown in Fig.5, we are developing • Compile-time scalability: #include is and using LLVM/Clang implementation of copying the contents to the includer’s C++ modules, collaborating with develop- code, so the parser has to reparse the ers from Google and Apple. Cling is a C++ same common header files multiple times, interpreter developed by CERN, and root- which is expensive. cling is a dictionary generator for ROOT. We • Fragility: Textual includes are influenced Figure 3: CPU Time required to run se- are implementing runtime modules in these by previously defined macros. For ex- lected tutorials. The first column is display- parts while integrating ROOT with them. ample, macro PI is #defined in Rcpp li- ing ROOT’s time to start into an empty shell. brary. Including this library while using 7. Roadmap local variable PI will end up in a redefini- • Compile ROOT with C++ modules tion error. Status: Completed C++ modules is a mechanism to boost com- • Compile CMSSW with C++ modules pilation time by precompiling headers into Status: Work in progress PCM files, where AST information can be • Use runtime C++ modules in ROOT lazily loaded. Status: Mostly Done 3. From C++ Modules to Runtime C++ • Use runtime C++ modules in experiments Modules Status: Work in progress To summarize, runtime modules are mostly working, but need work to get bet- ter performance. Figure 4: Residential memory used to run 8. Conclusion tutorials. Here we briefly introduced our experimen- Figure 1: Runtime Modules (pcms). Each Fig.3 and Fig.4 are the performance re- tal runtime C++ modules support in ROOT PCM file (E.g. Core.pcm) corresponds to a sults we receive from modules, compared and how it will affect experiments’ software library (E.g. libCore.so). to textual headers. The results are coming stacks. Modules are not yet competent com- from synthetic benchmarks close to the ex- C++ modules are able to reduce compi- pared to PCH, but are more flexible and periment software stacks and in particular lation times. However, the compilation have clear advantage over texual includes. CMSSW. scalability issues in C/C++ becomes run- 9. Future work 5.2 Correctness time issues for an interpretative environ- Runtime modules are still an experimental ment which ROOT provides. They span PCH: feature. Our ultimate goal is to make it de- from slow prompt to slow IO. $ bin/root.exe −l fault in ROOT and in experiments: root [0] gMinuit 4. Advantages over the Status Quo IncrementalExecutor :: executeFunction : • Stabilize modules behavior and tests symbol ’gMinuit’ unresolved while • Adoption by experiments and other ROOT linking [cling interface function]! users Runtime Modules: • Improve performance of loading modules $ bin/root.exe −l root [0] gMinuit Further Information ( TMinuit ) n u l l p t r * • https://clang.llvm.org/docs/Modules.html Runtime Modules are supporting more • https://root.cern.ch/ features than PCH. For example, gMinuit • https://root.cern.ch/cling Figure 2: Precompiled Headers (PCH). In- is an extern variable which cannot be au- formation of the header is stored in one file. toloaded by ROOT at the moment. How-