RSDK-13439 remove std::memory_order usages

[oliviamiller]

The explicit memory_order_relaxed and memory_order_release/acquire usages added unnecessary complexity without benefit. memory_order_relaxed leaves the CPU and compiler free to reorder the operations, making correctness dependent on the non-obvious happens-before chain through the release on total_samples_written counter. The default memory_order_seq_cst already performs acquire on loads and release on stores, making the explicit arguments redundant.

Switching to using the default memory ordering everywhere to make the code easier to reason about.

ty @dgottlieb for pointing this out to me

1 approval needed.

[seanavery]

Thanks for the explainer!

So there is no perf tradeoff swapping the relaxed loads in the read_samples loop for seq_cst?

[hexbabe]

[optional] performance profiling to get an idea of how performant this opt is

[oliviamiller]

Thanks for the explainer!

So there is no perf tradeoff swapping the relaxed loads in the read_samples loop for seq_cst?

as long as the samples are read before the next portaudio callback theres no performance difference, the overhead from seq_cst is negligible.