`University of Massachusetts Amherst Powered by AI:
 A Cambrian Explosion
 for C++ Software Development Tools Emery BergerCretaceous–Paleogene (K-Pg) extinction eventCretaceous–Paleogene (K-Pg) extinction ALLOCATED MEMORY USAGE GPU UTIL %, PEAK MEMORY (MB/s) MEMORY PYTHON NATIVE AI-powered optimizations!AI-powered optimizations... COMING SOON!evolveevolve profiler that suggests optimizationsevolve

© 2023 Apex.AI, Inc. | All rights reserved. Apex.AI, Apex.OS, Apex.Grace, Apex.Ida, Apex.Alan, Apex.Ray are trademarks of Apex.AI. CppCon 2023 Building bridges: Leveraging C++ and ROS for simulators Divya Aggarwal October 5, 2023© 2023 Apex.AI, Inc. | All rights reserved. Apex.AI, Apex.OS, Apex.Grace, Apex.Ida, Apex.Alan, Apex.Ray are trademarks of Apex.AI. Need for Speed? Asphalt? GT? This is an driving simulator. 2 CppCon© 2023 Apex.AI, Inc. | All rights reserved. Apex.AI, Apex.OS, Apex.Grace, Apex.Ida, Apex.Alan, Apex.Ray are trademarks of Apex.AI. 3 Agenda 1. ROS and simulators 2. What

© 2021 Apex.AI, Inc. The vehicle OS company. Ⓡ © 2021 Apex.AI, Inc. Igor Bogoslavskyi Testing the compilability of the code at runtime© 2021 Apex.AI, Inc. Errors have a high cost, so rigorous that we deliver to our customers© 2021 Apex.AI, Inc. Imagine designing a user-facing API float get_half_of(float smth) { return smth / 2.0F; }© 2021 Apex.AI, Inc. Maybe a bit more generic template< std::enable_if_t>> T get_half_of(T smth) { return smth / T{2}; }© 2021 Apex.AI, Inc. Write a rigorous testing suite #include TEST(TestMyApi, Halving) { EXPECT_FLOAT_EQ(21

2019 Apex.AI, Inc. Safe and certified software for autonomous mobility TM Practical memory pool based allocators for Modern C++ by Misha Shalem misha.shalem@apex.ai © 2020 Apex.AI, Inc.● CppCon one piece of the solution” Today we going to talk about this one piece in (more) depth © 2020 Apex.AI, Inc.Memory allocations in real-time safety-critical environment ● The solution should be safe and “It should comply to Autosar C++ 14 Coding Guidelines regarding memory allocations” © 2020 Apex.AI, Inc.Autosar C++ guidelines Rule A18-5-5 (required, implementation, partially automated) Memory management

Entertainment Computing Research Group • Efficiency and performance of game engines, AI, networking, VR … • Teaching: 3D Graphics, AI, Physics for games, Game Streaming, … • IFIP (International Federation for Vienna Entity Component System + Vienna Type List Library • Vienna Physics Engine + • Vienna Game AI Engine + • GUI • = Vienna Vulkan Game Engine 2.0 https://github.com/hlavacs 20Helmut Hlavacs – count()); //https://gafferongames.com/post/fix_your_timestep/ game_logic.tick(delta_t.count()); AI.tick(delta_t.count()); //… prepare_render_next_frame(); //get idle frame buffer and record command

finally blows the door off its hinges •The two facets of reflection are equally important •Where do AI tools fit within this craze?The Reflection Circularity Problem Without generation, we’re chasing our typename[:^char:] c = '*'; // Same as: char c = '*';Suddenly, an AIWe Can’t Not Talk About This •How is generative AI changing language (features) design? •jippity.pro: 350 lines of working code written by a 9 yo in in the future: • Boilerplate generation? AI could do that • Read all that boilerplate? AI editors can fold that away • Refactor/change large swaths of code? AI could read/modify/write codeQuery (o1): Please

MLIR Bisheng C++ Celerity SYCL MPI Vulkan Sylkan Nvidia HPC SDK MotorSYCL TVM/AKG Ascend AI Chipsets DPC++ fork PIM SDK Samsung PIM-HBM2 PIM emulator DPC++ fork SYCLops NewSYCL ON COMPILER accelerators ▪ Unify the heterogeneous compute ecosystem around open standards ▪ Focus Areas: AI, HPC, Edge AI, Edge Compute ▪ Open source collaborationROLE OF THE FOUNDATION • Host a specification that Interest Groups Working Groups Technical Steering Committee UNIFIED ACCELERATION FOUNDATION (UXL) AI Image Hardware Math Language Open Source Specificatio nUXL FOUNDATION BENEFITS ▪ Single code

NonCoroutineFunction() { auto saveWorld = ApplyImmunization( 3000 ); saveWorld.Wait(); } void __AI_resume(void* ptr) { // ... initial_suspend __sp1: frame->seq_for = sequential_for(frame->saveWorldSteps) coroutine_handle token<> awaitable awaitable::await_suspend coroutine_handle ApplyImmunization void __AI_resume(void* ptr) { // ... initial_suspend __sp1: frame->seq_for = sequential_for(frame->saveWorldSteps) sequential_for isa::await_suspend parent coroutine_handle fsa Job::Job final_suspend void __AI_resume(void* ptr) { // ... initial_suspend __sp1: frame->seq_for = sequential_for(frame->saveWorldSteps)

Scenario • Made up App uses AI to process images to Text • UI App running as medium Integrity Level • Service running as high Integrity Level • Cloud Service to handle the AI OCR processing UI App [Medium Service [High IL] Serialized Data Cloud AI OCR Processing Serialized DataAdversary Perspective UI App [Medium IL] Service [High IL] Serialized Data Cloud AI OCR Processing File Input Validation data • IPC, Parsing, Network connectionsApp Scenario UI App [Medium IL] Service [High IL] Cloud AI OCR ProcessingApp Scenario User Input AppUI.exe Medium IL AppService.exe High IL File Upload

game-engine library as an example, which provides several components such as algorithms, ai, rendering, and network. Both ai and rendering depend on the algorithms component. Please, first clone the sources CMakeLists.txt conanfile.py include ai.h algorithms.h network.h rendering.h src (continues on next page) 169 Conan Documentation, Release 2.0.17 algorithms ai rendering network Fig. 1: components components of the game-engine package (continued from previous page) ai.cpp algorithms.cpp network.cpp rendering.cpp test_package CMakeLists.txt CMakeUserPresets.json conanfile.py src example.cpp As you