CppCon 2026

This talk is the 2026 edition of our series that highlights low-latency, networking, parallelism, and concurrency proposals to the C++ Standards Committee, especially those discussed in the Kona 2025, Croydon 2026, and Brno 2026 meetings. This talk from the Concurrency TS2 Editors, DG and SG14/19 chair will describe the following features and show how they can be used. It will also give their motivation, background, and their place within the overall framework of C++ low latency, networking, parallelism and concurrency:

Atomic Compare and Hazard Pointers Reclamation Control

• We discuss a proposal for atomic compare operations that clarify programmer intent and simplify static analysis by avoiding the unnecessary leakage of atomic values. We also explore how these operations complement recent proposals to enhance the expressiveness of atomics, such as the use of tagged pointers.
• We examine techniques for controlling C++26 hazard pointer reclamation, demonstrating how to offload retirement overhead to asynchronous executors to ensure predictable thread latency. We show how this capability can prevent specific deadlock scenarios and assess whether such controllers should be standardized or remain user-space patterns.

Twenty-One Years Without Sockets, The History of Networking in C++ and What We Learned Along the Way

• Go has standard networking. Rust has standard networking. C++ has been trying since 2005. This talk tells the twenty-one-year story in seventeen minutes: the library that shipped but never standardized, the dependency that blocked it, the coroutine revolution that changed the assumptions, the technical discovery that explains why I/O doesn‚Äôt fit the three-channel model cleanly, and the bridge that lets both models win. Three decision points, three lessons, one live question: how should coroutines and senders coexist for C++29? No villains. Just the story and the engineering. Networking is now finally a C++29 priority, so said the Direction Group‚Äôs P5000.

Out-of-Thin-Air (OOTA) Values.

• The memory models for high-level languages, including C++, C, Rust, and even Java, have been prone to out-of-thin-air values. For example, given two atomic integers X and Y, both initially zero, if one thread does a relaxed assignment from X to Y and the other does a relaxed assignment from Y to X, the mathematical C++ memory model does not rule out the outcome where both X and Y have the value 42. However, both communication and computation take time, and so we have proven that OOTA values cannot happen on real-world production-quality computer systems. This result makes compiler-writers‚Äô lives easier, and also eases the jobs of those brave individuals pursuing the full mathematical OOTA problem by relieving them of the need to live within CPU and memory constraints of production-quality compilers.

Discussion

There are other features under discussion, but the features we will present seem close to approval, are particularly interesting, and/or are relatively non-controversial. We will show use cases for each and prognosticate their journeys to C++26 and beyond. This will help programmers in concurrency, lock-free programming, and low-latency applications understand how best to take advantage of each of these important facilities.

Reverse-mode automatic differentiation (AD) computes exact gradients with respect to thousands of inputs in roughly the time of a single function evaluation. If you write numerical C++ for finance, scientific computing, or ML and you care about performance, this talk will show you where AD overhead actually comes from and what to do about it.

A straightforward operator-overloading implementation records every arithmetic operation onto a tape and replays it backward to propagate derivatives. On real numerical code this easily introduces 50-100x overhead, and most of that has nothing to do with calculus. It is heap allocation on every operation. It is cache misses walking the tape backward. It is branch mispredictions at memory chunk boundaries. It is recording tape entries for operations where none of the operands are even being differentiated.

We work through a series of C++ techniques that bring this down to around 4x on production code in finance. The two that matter most: a chunked arena allocator with cache-line alignment and branch-prediction hints that turns the recording hot path into a pointer bump and a store, and expression templates that collapse an entire right-hand side into one tape push with all derivatives accumulated in a compile-time-sized stack buffer. We pull up compiler output to show the abstraction really does compile away.

After the big two, we get into the smaller wins that are still worth knowing about: skipping zero adjoints during the backward sweep to exploit sparsity, picking derivative formulas at compile time that reuse the forward result instead of recomputing it, and demoting active-times-passive operations from binary to unary to halve their tape footprint. We also show how a should-record check propagated through the expression tree lets the tape skip entire statements when no operand is active, and how aggressive inlining lets the compiler eliminate these checks entirely for passive subexpressions, so the cost of asking is zero when the answer is no. We benchmark each one in isolation on four numerical workloads ranging from 8 to 161 sensitivities, so you can see what each technique is actually worth.

A fundamental control structure of each programming language are loops. And we all expect them to be simple and self explanatory. However, C++ would not be C++, if there would be no tricky details to learn and respect about loops in C++.

This talk takes its time to discuss the various ways and approaches to program loops in C++. Beside basic while and for loops, we talk about the range-based for loop, and all the extensions recently added to these control structures.

In addition, we will look at other ways to program loops in C++, such as using algorithms and how to deal with parallel computing in a loop.

As a result you get a deeper understanding of the various ways loops can be programmed in Modern C++ so that you know better how to use them in practice.

Atomic memory ordering is often taught operationally: Use acquire here, release there, and perhaps add a fence “to be safe.” This approach tends to produce cargo-cult synchronization, where atomic operations are selected mechanically without a clear understanding of what information is actually being propagated between threads. In practice, memory ordering is not about memorizing enum values, but about establishing which facts become visible to which observers, and when.

This talk approaches atomic synchronization from first principles. Beginning with the fundamental ideas of publication, visibility, ownership transfer, and synchronization edges, the talk incrementally develops several concurrent structures drawn from real asynchronous systems. Case studies include outstanding-work reference counting, a multi-producer singly-linked publication structure, a concurrently-mutated doubly-linked intrusive list, and the coordination machinery underlying a concurrent operation with multiple concurrent completion modalities. Each structure is used to derive the synchronization requirements imposed by its invariants, rather than selecting memory orderings mechanically.

Along the way, the talk explores the practical meaning of relaxed operations, acquire/release synchronization, and atomic thread fences. Particular attention is paid to understanding what each actually does, when it is necessary, and when it has become a decorative synchronization cargo cult. The goal is not simply to present lock-free algorithms, but to develop a principled way of reasoning about memory visibility and synchronization in real concurrent systems.

According to VGInsights, about 10% of games released on Steam in 2024 used a custom engine, but those games represent 41% of units sold. Most "how an engine works" talks come from the AAA end of that distribution. This one comes from the other end of that distribution: a teaching-scale engine, where every system has to be small enough to understand from first principles. The problems are the same as in a big engine, but the smaller surface is what makes the architecture legible.

In this talk, we'll walk the spine of the engine in modern C++: allocators and handles in the core, the job system, the graphics abstraction question, the asset pipeline split between pipeline-time and runtime, and scene representation including ECS. We finish on serialization, where C++26 reflection (P2996) offers a path past today's reliance on macros and external code generators.

C++20 modules are finally usable end to end, from your own modules to 'import std;' The functional build-system support with real diagnostics. This talk teaches modules the way #include was once taught: with small files, a build, and concrete use cases. Modules stopped being experimental somewhere around 2025. This session builds the mental model from scratch: what a primary module interface unit is, how partitions and implementation units fit together, and, crucially, how import differs from #include in ways that matter day to day. It uses a small library, exposes it as a module, splits it across partitions, consumes import std; and observes the compile-time effect. Then it answers the questions every team hits in week one: how modules interact with macros, with templates in headers, with header-only dependencies, and with a mixed codebase that can't convert everything at once. The examples will be using CMake, clang and recent gcc. You will leave able to structure a small library as a module, explain why a macro didn't cross a module boundary, and plan an incremental adoption that doesn't require converting the whole tree.

Applications that present rapidly changing market data to human users face a practical challenge: updates may arrive thousands of times per second, but humans only benefit from periodic refreshes. Efficiently coalescing and scheduling that work without overwhelming CPU resources becomes a concurrency problem rather than simply a rendering problem.

This talk demonstrates how modern C++ atomics can be used to build a lock-free timer scheduler optimized for extremely high update rates. The scheduler follows a single-producer, multiple-consumer design in which work items represent recurring tasks that must execute periodically. Consumers process scheduled work and re-schedule it for future execution without relying on traditional locks or centralized coordination.

The most unusual aspect of the design is its “reverse work stealing” behavior: consumers can proactively give away work to other consumers and become idle themselves. Rather than distributing work evenly across all threads, the scheduler attempts to pack work onto as few cores as possible, reducing idle spinning, lowering CPU utilization, and improving cache locality. Because tasks may have highly variable execution times, the scheduler must also prevent pathological cases where work is endlessly redistributed between consumers.

Attendees will learn how atomics, lock-free forward lists, and memory ordering can be combined to implement high-throughput schedulers with predictable latency characteristics. The talk also discusses practical tradeoffs, implementation challenges, and performance measurements from production-inspired workloads, including producer throughput of approximately 50 million scheduled tasks per second and consumer processing throughput approaching 700 million tasks per second, while keeping migrated work items below 1% in typical workloads.

Your system supports multiple strategies, with the user selecting one at startup, but each strategy must handle different semantics. Traditional runtime polymorphism can require a virtual override for every combination, making code harder to extend and easier to get wrong

This talk shows how OpenJDK solves this problem by using a layered architecture of compile-time patterns with a runtime trampoline. We'll build them up through OpenJDK's GC barrier system, where a single templated API stands in for dozens of differently-shaped memory accesses across the runtime: plain stores, raw/unsafe access, native calls, weak references, and more, each needing different treatment depending on the semantics and which garbage collector was chosen by the user.

We'll also discuss trade-offs against std::variant, function pointers, and plain virtual dispatch. No JVM knowledge required.

Practical HPC — Forcing the Compiler's Hand with Modern C++

Achieving near-peak FLOPs on modern CPUs requires exploiting SIMD, register files, and cache hierarchies — yet the levers that matter most (loop unrolling, vectorization width, tile sizes, register tiling) are largely outside the programmer's direct control. Autovectorization is fragile, #pragma unroll is advisory and non-portable, and the information the compiler needs most — loop bounds, problem shapes, working-set sizes — is typically only available at runtime. The result is a familiar gap between theoretical peak and delivered performance, bridged in practice only by hand-written intrinsics or inline assembly, fragile macro forests, or external code generators that sacrifice maintainability and type safety.

Modern C++ is, sometimes surprisingly, an excellent language for closing that gap without leaving the host language. We walk through the concrete features that make this possible and how to use them in practice — templates and constexpr to promote runtime values into compile-time constants, if constexpr and parameter packs for shape-specialized kernels, generic and template lambdas for loop bodies that are unrolled and inlined by construction, and concepts to turn silent performance cliffs into compile-time errors. Together, these give the programmer precise control over what the compiler emits: compile-time dispatch that specializes kernels per shape and target, guaranteed static unrolling that unlocks ILP and register tiling without relying on optimizer heuristics, and hardware-aware specialization parameterized on register count, SIMD width, and cache sizes.

Looking forward, C++26 and static reflection push this approach considerably further, turning today's disciplined metaprogramming into something closer to first-class, in-language code generation — and squarely aligning the language with the goal of making inline assembly unnecessary, even at the highest performance tiers.

These patterns are powerful but verbose and reappear across kernels, so we briefly introduce POET (Performance Optimized Excessive Templates), a header-only library usable from C++17 onward that packages the most common cases as ready-to-use building blocks, alongside xsimd for portable SIMD.

As a concrete case study, we walk through a real scientific-computing kernel — a workload of the kind that has historically demanded hand-tuned assembly — and show how these abstractions get within a small fraction of theoretical peak throughput while keeping the source readable, portable, and maintainable, with no inline assembly, no compiler builtins, and no external code-generation step. Scientific computing does not have to choose between performance and engineering quality, and with the right abstractions, development becomes dramatically more effective.