David Olsen, CppCon, 17 Sep 2020 BACK TO BASICS: MOVE SEMANTICS2 David Olsen — Back to Basics: Move Semantics — CppCon 2020 #include The Situation It’s 2003… You need to write code that creates Olsen — Back to Basics: Move Semantics — CppCon 2020 #include typedef std::map dictionary_t;4 David Olsen — Back to Basics: Move Semantics — CppCon 2020 #include typedef std::map dictionary_t;5 David Olsen — Back to Basics: Move Semantics — CppCon 2020 #include typedef std::map dictionary_t; dictionary_t

C++ Move Semantics Back to Basics: Move Semantics NICOLAI JOSUTTIS Cppcon 罗 AN Nicolai M. Josuttis C++ Move Semantics C++ Move Semantics Motivation BE josuttis | | eckstein Crpmaae Tcommunication Copy Semantics std: :vector coll1 coll.reserve(3) ; Staclc ，Heap: stad: :string sfgetData() }

Back to Basics: RValue and Move Semantics, Amir Kirsh @ CppCon, 2024About me 2 Back to Basics: RValue and Move Semantics, Amir Kirsh @ CppCon, 2024 Lecturer Academic College of Tel-Aviv-Yaffo Tel-Aviv (C++, but not only)The motivation for Move Semantics 3 Back to Basics: RValue and Move Semantics, Amir Kirsh @ CppCon, 2024The motivation for Move Semantics Godzilla g1 = factory.createFrighteningGodzilla(); RValue and Move Semantics, Amir Kirsh @ CppCon, 2024The motivation for Move Semantics Avoid redundant copying when we can steal from a dead object! 5 Back to Basics: RValue and Move Semantics, Amir Kirsh

know Highly Regarded Principle: std::variant always has a value Rule through Desired Semantics meeting Implementation Reality Exception: valueless_by_exception • std::variant::val b0 } b0 b1 b2 Stack-based objects have outstanding lifecycle symmetry (Guaranteed!) Value Semantics is preferred in Modern C++, due to superior symmetry in object lifecycle Use whenever possible leaked instances • Allocation amortization • Whole-system reset • Resource prioritization and recovery • Resource metrics and runtime monitoring Composable!Charley Bay - charleyb123 at gmail dot com

Goals ▶ Read optimized ▶ Persistence needs no flush or fence after first read ▶ Runtime over recovery ▶ Persist as little as possible ▶ Compact representation and few cache misses ▶ Arrays ▶ Open 25Introduction Persistent Hash Map Persistent Transactional Data Structures Live Demonstration References Recovery ▶ Only persist keys and values ▶ Filesystem data (name, size) infers contents ▶ DAX (Direct Access) mode ▶ Link-and-persist ensures data is consistent ▶ Orphans are possible but discarded during recovery 1024 bits 512 bits 256 bits A Persistent Hash Map for Graph Processing Workloads and a Methodology

30SAFETY ▪ It is a tradeoff… ▪ Performance ▪ Correctness ▪ This is a bitter pill to swallow ▪ Recovery ▪ Getting the system back into a known state ▪ Halting the system not always appropriate 31UNDEFINED environments will have different semantics P2759 DG OPINION ON SAFETY FOR ISO C++P2759 DG OPINION ON SAFETY FOR ISO C++ ▪ Collection of restrictions, requirements, related semantics that define a safety property ban unchecked subscripting, but instead provides runtime checked alternative ▪ Do not change semantics of a valid program ▪ With the exception of UB <==> well-defined behavior ▪ Open (hard) questions:

that the total number of disks in Curve system is N, the number of replicas is R, and the data recovery Time in the case of failed disks is T. The Annual Failure Rate of disks is AFR, and the average 50 other disks will restore the data on that disk at the same time. According to this, the data recovery time T can be estimated. 𝐹𝐼𝑇 = 𝐴𝐹𝑅 24 ∗ 365 The probability of failure of disk within 𝑒"#$%∗('"!)∗(,7 ∗ 1 3 ∗ 50 𝑁 3 The probability of the failure of the third disk in tr during the recovery time is as follows𝑃- = 61 − 𝑒"#$%∗(' " ))∗(,7 ∗ 1 3 ∗ 50 ∗ 3 𝑁 The annual probability of

upgrads supported FULL LIFECYCLE Plan to Support app lifecycle, storage lifecycle(backup, failure, recovery) DEEP INSIGHTS Plan to Support metrics, alerts, log processing and workload analysis AUTO PILOT upgrade automation • Supporting Curve Cluster provisioning in helm chart • metadata backup and recovery • MDS / ChunkServer should respect failure domains of Kubernetes • Support for public cloud

性、容灾要求较⾼。传统的解决⽅案是同城两个机房提供服务、异地⼀个机房提供数据容灾能⼒但不提供服务，此解决⽅案存 在以下缺点：资源利⽤率低、维护成本⾼、RTO (Recovery Time Objective) 及 RPO (Recovery Point Objective) ⽆法真实达到企业所期望的值。TiDB 采⽤多副本 + Multi-Raft 协议 的⽅式将数据调度到不同的机房、机架、机器，

string_view, span, optional, expected etc., raise new questions regarding values types, whole-part semantics, copies, composite objects, ordering and equality. Designing and implementing regular types Regular types Values Objects Concepts Ordering Relations Requirements Equality Whole-part semantics2023 Victor Ciura | @ciura_victor - Regular, Revisited 17 Slide Title This talk is not just Regular types Values Objects Concepts Ordering Relations Requirements Equality Whole-part semantics Lifetimes2023 Victor Ciura | @ciura_victor - Regular, Revisited 17 Slide Title This talk