Remote Execution Caching Compiler (RECC) CppCon 2024 September 19, 2024 Shivam Bairoliya Software Engineer© 2024 Bloomberg Finance L.P. All rights reserved. What is RECC? ● Remote Execution Caching source build tool that wraps compiler commands and optionally forwards them to a remote build execution service ○ Encompasses the capabilities of both ccache and distcc ○ Supports remote linking and CC) ○ Supports multiple operating systems (Linux, macOS, Solaris) ● Compatible with any remote execution API server supported by Bazel ○ Single Host Server/Proxy: BuildBox-CASD ○ Distributed Server:

rights reserved. Presenter: Haichen Shen, Yao Wang Amazon SageMaker Neo, Deep Engine Science Dynamic Model in TVM AWS AI© 2019, Amazon Web Services, Inc. or its Affiliates. All rights reserved. Models with models© 2019, Amazon Web Services, Inc. or its Affiliates. All rights reserved. Support dynamic model in TVM ● Support Any-dim in typing ● Use shape function to compute the type at runtime ● Virtual input_name = "data" input_shape = [tvm.relay.Any(), 3, 224, 224] dtype = "float32" block = get_model('resnet50_v1', pretrained=True) mod, params = relay.frontend.from_mxnet(block, shape={input_name:

manually split up data amongst GPUs / nodes - High-level mechanisms for data distribution / execution necessary. CPU NIC GPU GPU GPU GPU Xe LinkMulti-GPU Systems - NUMA regions: - 4+ GPUs distributed algorithms - Achieve high performance for both multi-GPU, NUMA, and multi-node execution float dot_product(vector& x, vector& y) { auto z = views::zip(x Operate on and modify data using namespace std; using namespace std::ranges; using namespace std::execution; template auto dot_product(R&& x, R&& y) { using T = range_value_t; auto z =

Memory Model C++11 – C++23About Me: alex.dathskovsky@speedata.io www.linkedin.com/in/alexdathskovsky https://www.cppnext.comAlex Dathskovsky | alex.dathskovsky@speedata.io | www.linkedin.com/in/a ExecutionAlex Dathskovsky | alex.dathskovsky@speedata.io | www.linkedin.com/in/alexdathskovsky INO Execution • instruction fetch • if operands available execute it if not fetch them • The instruction is ExecutionAlex Dathskovsky | alex.dathskovsky@speedata.io | www.linkedin.com/in/alexdathskovsky OOO Execution Source:: https://www.semanticscholar.org/paper/RISC-V-Reward:-Building-Out-of-Order-Processo

including some standard ones: an event loop, portable access to the system execution context, nursery for spawned work4 P2300: STD::EXECUTION Proposes: • A set of concepts that represent: • A handle to a compute Example 1: Launching concurrent work6 EXAMPLE: LAUNCHING CONCURRENT WORK namespace ex = std::execution; int compute_intensive(int); int main() { unifex::static_thread_pool pool{8}; ex::scheduler concurrently on a custom execution context libunifex: https://github.com/facebookexperimental/libunifex7 EXAMPLE: LAUNCHING CONCURRENT WORK namespace ex = std::execution; int compute_intensive(int);

perspective: > I understand C++, and I kinda get assembly because of compiler explorer. Our typical model of AoT is “what C and C++ do”, and I want to expand the understanding for what other computation models JiT compilation includes any translation performed dynamically, after a program has started execution. We examine the motivation behind JiT compilation and constraints imposed on JiT compilation systems JiT compilation includes any translation performed dynamically, after a program has started execution. We examine the motivation behind JiT compilation and constraints imposed on JiT compilation systems

or generating a report. Applications that integrate LLMs but don’t use them to control workflow execution—think simple chatbots, single-turn LLMs, or sentiment classifiers—are not agents. More concretely manage workflow execution and make decisions. It recognizes when a workflow is complete and can proactively correct its actions if needed. In case 
 of failure, it can halt execution and transfer control design foundations In its most fundamental form, an agent consists of three core components: 01 Model The LLM powering the agent’s reasoning and decision-making 02 Tools External functions or APIs the

somewhat reflects how we write software (1/3) 11 ● Typically we have one main, sequential thread of execution ● One CPU core executes code sequentially ○ i.e. One instruction after the other. Instruction somewhat reflects how we write software (2/3) 12 ● Typically we have one main, sequential thread of execution ● One CPU core executes code sequentially ○ i.e. One instruction after the other. ○ We can abstract Foo() Bar() Baz() Execute main execution pathThis somewhat reflects how we write software (3/3) 13 ● Typically we have one main, sequential thread of execution ● One CPU core executes code serially

best practices and lessons learned by these early Agile adopters. Broader, more successful Agile execution will take time, trial and error, and shaping of processes, policies, and culture, but with support locked. During sprint execution, the development team runs through the full development cycle for each user story in the sprint backlog as shown in Figure 4. Figure 4 Sprint Execution Cycle The integrated culture often run counter to those in the long-established defense acquisition enterprise. The Agile model represents a change in the way DoD conducts business, and programs must rethink how they are staffed

your code on a simulator or emulator • Use the transactional/behavioral/functional definition to model the HW using SW elementsIntel Confidential Department or Event Name 5 © 2023 Intel Corporation and ready • How do we shift the production of the SW earlier? • Answer: We need a model! Transaction trigger Side Effects Model State Transaction inspect observeIntel Confidential Department or Event Name Detail – Time quanta and synchronization • How often do we really care to know the state of the model? • Consider a Public Key Cryptography Accelerator Complex State Machine Doing Things like a Montgomery