> Yes, we can. C++20 Modules are usable in a Linux + Clang environment. There are also examples showing that C++20 Modules are usable in a Windows environment with MSVC. I have not yet heard of GCC’s C++20 Modules being used in non-trivial projects.
People keep saying this and yet I do not know of a good example from a real life project which did this which I can test. This seems very much still an experimental thing.
> C++ 26 reflections have now been voted in. This would get rid of moc entirely, but I really do not see how this will become widely available in the next 5-10 Years+. This would require Qt to move to C++ 26, but only if compiler support is complete for all 3 compilers AND older Linux distros that ship these compilers. For example, MSVC still has no native C++ 23 flag (In CMake does get internally altered to C++ latest aka. C++ 26) , because they told me that they will only enable it is considered 100% stable. So I guess we need to add modules support into moc now, waiting another 10 years is not an option for me .
It is just beyond experimental now, and finally in the early adopter phase. Those early adopters are trying things are trying to develop best practices - which is to say as always: they will be trying things that future us will laugh at how stupid it was to do.
There are still some features that are missing from compilers, but enough is there that you can target all 3 major compilers and still get most of modules and benefit from them. However if you do this remember you are an early adopter and you need to be prepared to figure out the right way to do things - including fixing things that you get wrong once you figure out what is right.
Also, if you are writing a library you cannot benefit from modules unless you are willing to force all your consumers to adopt modules. This is not reasonable for major libraries used by many so they will be waiting until more projects adopt modules.
Still modules need early adopters and they show great promise. If you write C++ you should spend a little time playing with them in your current project even if you can't commit anything.
I think it still is in a "well technically its possible" state. And I fear it'll remain that way for a bit longer.
A while ago I made a small example to test how it would work in an actual project and that uses cmake (https://codeberg.org/JulianGmp/cpp-modules-cmake-example).
And while it works™, you can't use any compiler provided modules or header modules. Which means that
1) so you'll need includes for anything from the standard library, no import std
2) you'll also need includes for any third party library you want to use
When I started a new project recently I was considering going with modules, but in the end I chose against it because I dont want to mix modules and includes in one project.
We are building a document rendering tool using them. It’s a pretty large project, and there have been some really good improvements in Clang’s implementation of C++20 modules in the past few versions.
> The data I have obtained from practice ranges from 25% to 45%, excluding the build time of third-party libraries, including the standard library.
> Online, this number varies widely. The most exaggerated figure I recall is a 26x improvement in project compilation speed after a module-based refactoring.
> Furthermore, if a project uses extensive template metaprogramming and stores constexpr variable values in Modules, the compilation speed can easily increase by thousands of times, though we generally do not discuss such cases.
> Apart from these more extreme claims, most reports on C++20 Modules compilation speed improvements are between 10% and 50%.
I'd like to see references to those claims and experiments, size of the codebase etc. I find it hard to believe the figures since the bottleneck in large codebases is not a compute, e.g. headers preprocessing, but it's a memory bandwidth.
> I find it hard to believe the figures since the bottleneck in large codebases is not a compute, e.g. headers preprocessing, but it's a memory bandwidth.
source? language? what exactly does memory bandwidth have to do with compilation times in your example?
> I'd like to see references to those claims and experiments, size of the codebase etc. I find it hard to believe the figures since the bottleneck in large codebases is not a compute, e.g. headers preprocessing, but it's a memory bandwidth.
Edit: I think I misunderstood what you meant by memory bandwidth at first?
Modules reduce the amount of work being done by the compiler in parsing and interpreting C++ code (think constexpr). Even if your compilation infrastructure is constrained by RAM access, modules replace a compute+RAM heavy part with a trivial amount of loading a module into compiler memory so it's a win.
Why is something which shall makes things easy and secure so complicated?
I'm used to:
g++ -o hello hello.cpp
It can use headers. Or doesn't use headers. I doesn't matter. That's the decision of the source file. To be fair, the option -std=c++20 probably isn't necessary in future.
We use c++ modules at Waymo, inside the google monorepo. The Google toolchain team did all the hard work, but we applied it more aggressively than any team I know of. The results have been fantastic, with our largest compilation units getting a 30-40% speedup. It doesn't make a huge difference in a clean build, as that's massively distributed. But it makes an enormous difference for iterative compilation. It also has the benefit of avoiding recompilation entirely in some cases.
Every once in a while something breaks, usually around exotic use of templates. But on the whole we love it, and we'd have to do so much ongoing refactoring to keep things workable without them.
Update: I now recall those numbers are from a partial experiment, and the full deployment was even faster, but I can't recall the exact number. Maybe a 2/3 speedup?
12 comments
[ 4.7 ms ] story [ 36.3 ms ] threadPeople keep saying this and yet I do not know of a good example from a real life project which did this which I can test. This seems very much still an experimental thing.
> C++ 26 reflections have now been voted in. This would get rid of moc entirely, but I really do not see how this will become widely available in the next 5-10 Years+. This would require Qt to move to C++ 26, but only if compiler support is complete for all 3 compilers AND older Linux distros that ship these compilers. For example, MSVC still has no native C++ 23 flag (In CMake does get internally altered to C++ latest aka. C++ 26) , because they told me that they will only enable it is considered 100% stable. So I guess we need to add modules support into moc now, waiting another 10 years is not an option for me .
https://github.com/pjmlp/RaytracingWeekend-CPP
Also shows how to use static libraries alongside modules.
There are still some features that are missing from compilers, but enough is there that you can target all 3 major compilers and still get most of modules and benefit from them. However if you do this remember you are an early adopter and you need to be prepared to figure out the right way to do things - including fixing things that you get wrong once you figure out what is right.
Also, if you are writing a library you cannot benefit from modules unless you are willing to force all your consumers to adopt modules. This is not reasonable for major libraries used by many so they will be waiting until more projects adopt modules.
Still modules need early adopters and they show great promise. If you write C++ you should spend a little time playing with them in your current project even if you can't commit anything.
A while ago I made a small example to test how it would work in an actual project and that uses cmake (https://codeberg.org/JulianGmp/cpp-modules-cmake-example). And while it works™, you can't use any compiler provided modules or header modules. Which means that 1) so you'll need includes for anything from the standard library, no import std 2) you'll also need includes for any third party library you want to use
When I started a new project recently I was considering going with modules, but in the end I chose against it because I dont want to mix modules and includes in one project.
https://github.com/odoo/paper-muncher/blob/main/src/main.cpp
> Online, this number varies widely. The most exaggerated figure I recall is a 26x improvement in project compilation speed after a module-based refactoring.
> Furthermore, if a project uses extensive template metaprogramming and stores constexpr variable values in Modules, the compilation speed can easily increase by thousands of times, though we generally do not discuss such cases.
> Apart from these more extreme claims, most reports on C++20 Modules compilation speed improvements are between 10% and 50%.
I'd like to see references to those claims and experiments, size of the codebase etc. I find it hard to believe the figures since the bottleneck in large codebases is not a compute, e.g. headers preprocessing, but it's a memory bandwidth.
source? language? what exactly does memory bandwidth have to do with compilation times in your example?
Edit: I think I misunderstood what you meant by memory bandwidth at first? Modules reduce the amount of work being done by the compiler in parsing and interpreting C++ code (think constexpr). Even if your compilation infrastructure is constrained by RAM access, modules replace a compute+RAM heavy part with a trivial amount of loading a module into compiler memory so it's a win.
SSD bandwidth: 4-10GB/s RAM bandwidth: 5-10x that, say 40GB/s.
If compute was not a bottleneck, the entire linux kernel should compile in less than 1 second.
I'm used to:
It can use headers. Or doesn't use headers. I doesn't matter. That's the decision of the source file. To be fair, the option -std=c++20 probably isn't necessary in future.I recommend skimming over this issue from Meson:
https://github.com/mesonbuild/meson/issues/5024
Reading the last few blog posts from a developer of Meson, providing some insights why Meson doesn't support modules until now:
https://nibblestew.blogspot.com/
Every once in a while something breaks, usually around exotic use of templates. But on the whole we love it, and we'd have to do so much ongoing refactoring to keep things workable without them.
Update: I now recall those numbers are from a partial experiment, and the full deployment was even faster, but I can't recall the exact number. Maybe a 2/3 speedup?