C++ Modules and Large-Scale Development (Part 2)
summary
Argues that allocator-aware infrastructure yields major real-world wins: measured examples include a companywide ~20% speedup and large reductions in memory diffusion by using polymorphic memory_resource (PMR), monotonic and multipool allocators, and mmap arenas; costs mostly appear as upfront library work, ~<20% more boilerplate source and training, and tricky cases like shared_ptr and pair; proposes tooling and language support (BTverify, test allocators, compiler-generated plumbing) to make adoption cheap and safe.
AI-written summary · about
chapters (10)
- 0:00Introduction and motivation
- 4:02The economics of allocators
- 6:17Business case and utility
- 16:54Allocator interface styles
- 21:12Language support roadmap
- 22:59Performance benefits
- 29:28Addressing adoption costs
- 37:39Risks and misuse
- 45:20Collateral benefits
- 54:22Addressing concerns and conclusion
related talks
01h 00m“Deploying C++ modules to 100s of millions of lines of code"
01h 00mSo, you want to use C++ Modules ... cross-platform?
01h 02mModern CMake Modules
01h 04mC++ Modules - Getting started today
01h 04mCommon-sense acceleration of your MLOC build
01h 09mC++ Modules: Getting Started Today
01h 01m10 Problems Large Companies Have Managing C++ Dependencies and How to Solve Them
19mOn-Demand Container Partitioning for Distributed ML