2 offline days

Anton will talk about new C++ tricks and techniques. You will learn about: LazyPrvalue; ODR violation detection from C++ code; FixedArray; "the funniest bimap". The latest trick with "the funniest bimap"  has already been discussed in the C++ Russia conference chat. The talk details both the background and the final results with benchmarks. Ready-to-use code from the talk is already available at https://userver.tech/

The talk describes the interacting semantic processes in C++, both from the point of view of the compiler internals and from the point of view of the language standard.

Aleksandr will propose a distributed C++ compiler — nocc. It's like distcc, but faster. It's like ccache, but remote.

Many of us are engaged in software development, being in a kind of project process. Alexander will talk about the variants of such projects and the role of the developer in these processes. He will touch upon the problem of technical debt: where it comes from and why it doesn't disappear. The talk will discuss what a developer can influence in the production process and what he cannot usually reach. We will definitely consider the peculiarities of C++ development in project activities.

Marcel will talk about the standard ways of working in multi-threaded environment (mutexes, compare and swap, hazard pointer) and the algorithm of taking a snapshot of two atomic "registers". The talk will be useful for newcomers to multithreading and for experienced developers who want to dust off already rusty knowledge.

The speaker will compare flat_map with existing standard associative containers, look at popular existing implementations of flat_map and what is proposed to be added to the standard C++ library, and finally share conclusions about what kind of flat_map we want to have in C++. After the talk, you will have a good understanding of what flat_map is and in what situations it should be used.

Yandex presents the library that allows to implement JSON serializable structures in C++. In his talk Ivan will speak about high-level design of the library; technical difficulties in implementation (behavior of virtual methods, linking of template classes, etc.); experience in implementing the library in a large code base; attempts to make a good interface.

A talk about the CUDA C++ tool, which allows you to get the most out of the hardware you use.

Not so long ago, the National Security Agency (NSA) published a guide declaring C/C++ "dangerous" and requiring a transition to "safe" C#, Go, Java, Ruby, and Swift. Sergey will analyze this document in detail and tell if the security situation in C++ is really that bad, and what the modern industry offers to solve this issue.

The talk discusses various mistakes when writing comparison functions (comparators) in API like std::sort — their consequences and the means of detecting them provided by modern toolchains.

Inheritance, virtual functions, v-tables, virtual inheritance — things that are painfully familiar to every C++ developer. These things directly affect performance, the size of our binaries, and sometimes the security of our code. So let's go back to the beginning and understand the details of this machine. Let's look at what's new in this area and how it works under the hood.

Maxim will tell you a lot of practical things about performance optimization. What tools are worth using to find places for optimizations. How not to make a mistake when testing performance. How to correctly select and test algorithms and data structures. In the talk there will be a lot of examples from practice.

Elbrus as an architecture is very different from mainstream processor architectures. As a consequence, it requires some special optimizations from the compiler. Victor will briefly describe the peculiarities of Elbrus and its special optimizations in the LCC compiler. In more detail, he will talk about the code merge optimization: what it is, why it is needed, and how it works.

A huge amount of C++ code in Yandex can be described by the words "distributed computing". The talk is devoted to the C++ framework, which is actively being developed now, inspired by std::ranges and Apache Beam libraries. This framework allows rather compact and simple description of such processes and running them in MapReduce mode to process data tables or in real-time mode to process endless streams (queues) of data.

Despite the Committee's work on the new language standards, it seems that some invariants cannot be revised, and activists are working in creating their own languages compatible with C++. You will dive back into the design in order to try to answer the question: why? What is so new proposed and what might be missing in C++?

Vladimir will speak about the work of conventional™ indexes in databases and will tell: how the index speeds up the search; whether to index conditions where; whether to index conditions in order by; whether to index foreign keys; what to do if there are several search criteria; in what order to specify columns in the index; when the index is slow and how to reduce the impact of the index on the application.

ASIO allows you to use std::future out of the box to handle asynchronous events. But what if you want to use your own implementation of future? Ilya will tell and show you how to do it. At the same time, he will talk about customization below, on the example of io_uring.

Evgenii will share his experience of using Bazel for two projects. In the first project, he used it as a C++/Python developer. In the second project, at Kaspersky Lab, he's part of the team that migrates the code base from CMake to Bazel. He will talk about the killer features of this build system — its advantages and disadvantages.

A sequel to the talk on programming automation in the USSR. This time we will talk about the Soviet translators created in the 60-70s. The presentation concentrates on simple and actual even today algorithms. The central theme of the talk is the problem of register allocation. What do the developments of the Soviet compiler school and LLVM with SSA and instruction scheduling for modern processors have in common?