Case Study: FFI Thread Safety

[kassane]

This case to be studied does not affect or influence the progress of the Zig project and its features. Everything that is or will be tested has an experimental purpose.

References

• https://github.com/ziglang/zig/issues/8224
• https://github.com/ziglang/zig/issues/5962
• https://github.com/ziglang/zig/issues/9922
• https://github.com/ziglang/zig/issues/5311
• https://github.com/ziglang/zig/issues/12251
• https://github.com/ziglang/zig/issues/7769

Related

• https://github.com/kassane/tigerbeetle-cpp/issues/4

asio - FFI

• https://github.com/Heerozh/LuaAsio
• Emilua - internal - @vinipsmaker project

Others

• Vale & C - FFI safe
• Asio and Threads
  • Strand
• How strands work and why you should use them
  • https://github.com/chriskohlhoff/asio/issues/886
• https://github.com/chriskohlhoff/asio/issues/1408#issuecomment-1907081721

[kassane]

There are some threadpool solutions written in Zig, one of which is officially supported in the standard library, but requires memory allocation and another heapless proposal.

Allocator - std

• https://github.com/ziglang/zig/blob/master/lib/std/Thread/Pool.zig

Note: When Observing the std version git history, it can be seen that it is based on the refactored heapless version.

Heapless

• https://github.com/mitchellh/libxev/blob/main/src/ThreadPool.zig
• https://zig.news/kprotty/resource-efficient-thread-pools-with-zig-3291
• https://github.com/mitchellh/libxev/blob/main/docs/xev_threadpool.3.scd

[kassane]

Based on the reference below.

• https://github.com/chriskohlhoff/asio/issues/1054

Something similar occurs in the current CI test. The zig test prints Hello task using posted strand however it is not displayed along with threadpool.