Scientific Unit Testing Dave Steffen, Ph.D. Software Lead in Physics which is relevant dsteffen@scitec.com www.scitec.com 1If I have seen further than others, it is by standing upon the shoulders Limit tests to the code in question 26Reproducibility (handling nondeterministic results) Most scientific experiments have measurement error or noise. Interference from the environment: noisy signals correct Unit tests attempt to show !C Confidence in C tracks thoroughness of tests A.K.A "The Scientific Method" 42Acknowledgements Kris Jusiak, Quantlab Financial Check out his C++20 macro-free unit

Poster submission: Modern C++ for Parallelism in High Performance Computing Victor Eijkhout CppCon 2024 Introduction This poster reports on ‘D2D’, a benchmark that explores elegance of expression and and perfor- mance in the context of a High Performance Computing ‘mini-application’. The same code has been implemented using a number of different approaches to parallelism. Implementations are discussed discussed with performance results. Relevance C++ is making inroads into HPC / Scientific Computing, a field traditionally dominated by C and Fortran. With all the developments in modern C++ such as range

Constraints Architecture Substitution Construction Conclusion Symbolic Calculus for High-Performance Computing from Scratch using C++23 Vincent Reverdy Laboratoire d’Annecy de Physique des Particules, France Symbolic Computation, Joël Falcou and Vincent Reverdy, CppCon 2019 Hypothesis This is the Scientific Computing Track so you all know about optimization, performance, parallelism, . . . What this talk is not about Complicated maths (you are smart people, you can do it yourself) High-performance computing (you all know about it + see the 2019 talk for that) Benchmarks, assembly, and optimization (see

connected and accessible information being supercharged by artificial intelligence, accelerating computing power, and semi-borderless capital…all driving massive change. Sport provides a good analogy for Enterprise Impact’ via Morgan Stanley (10/23) Enabling Infrastructure CPUs Big Data / Cloud GPUs Computing Cycles Over Time – 1960s-2020s, per Morgan Stanley Note: Axis is logarithmic; i.e., there are Estimates how much progress comes from bigger models versus smarter algorithms, based on how much computing power you'd need to reach top performance without any improvements. Source: Epoch AI (3/24) Impact

from developing across multiple architectures ▪ Develop with open standards for accelerator computing ▪ Standards and industry defined libraries For Software Developers For Processor Developers ▪ optimized toolchain Free and based on open standards“this work supports the productivity of scientific application developers and users through performance portability of applications between Aurora nliber@anl.govWHO AM I? ▪ Argonne National Laboratory ▪ Computer Scientist ▪ Argonne Leadership Computing Facility ▪ C++, SYCL, Kokkos ▪ Aurora ▪ WG21 - ISO C++ Committee ▪ Vice Chair, Library Evolution

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 308 23 Parallel Computing 310 24 Asynchronous Programming 311 24.1 Basic Task operations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 327 26 Multi-processing and Distributed Computing 329 26.1 Code Availability and Loading Packages . . . . . . . . . . . . . . . . . . . . . . . Arrays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 509 38.14 Computing cluster . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 510 38.15 Julia Releases

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 308 23 Parallel Computing 310 24 Asynchronous Programming 311 24.1 Basic Task operations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 327 26 Multi-processing and Distributed Computing 329 26.1 Code Availability and Loading Packages . . . . . . . . . . . . . . . . . . . . . . . Arrays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 509 38.14 Computing cluster . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 510 38.15 Julia Releases

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 308 23 Parallel Computing 310 24 Asynchronous Programming 311 24.1 Basic Task operations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 327 26 Multi-processing and Distributed Computing 329 26.1 Code Availability and Loading Packages . . . . . . . . . . . . . . . . . . . . . . . Arrays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 509 38.14 Computing cluster . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 510 38.15 Julia Releases

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 311 23 Parallel Computing 313 24 Asynchronous Programming 314 24.1 Basic Task operations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 331 26 Multi-processing and Distributed Computing 332 26.1 Code Availability and Loading Packages . . . . . . . . . . . . . . . . . . . . . . . Arrays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 520 38.14 Computing cluster . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 521 38.15 Julia Releases

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 311 23 Parallel Computing 313 24 Asynchronous Programming 314 24.1 Basic Task operations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 331 26 Multi-processing and Distributed Computing 333 26.1 Code Availability and Loading Packages . . . . . . . . . . . . . . . . . . . . . . . Arrays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 521 38.14 Computing cluster . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 522 38.15 Julia Releases