Especially within the world of multi-threaded programming does atomic access become a crucial topic, as multiple execution contexts may seek to access the same memory locations at the same time. Yet the exact meaning of the word ‘atomic’ is also essential here, as there is in fact not just a single meaning of the word within the world of computer science. One type of atomic access refers merely to whether a single value can be written or read atomically (e.g. reading or writing a 32-bit integer on a 32-bit system versus a 16-bit system), whereas atomic operations are a whole other kettle of atomic fish.

Until quite recently very few programming languages offered direct support for the latter, whereas the former has been generally something that either Just Worked™ if you know the platform you are on, or could often be checked fairly trivially using the programming language’s platform support headers. For C and C++ atomic operations didn’t become supported by the language itself until C11 and C++11 respectively, previously requiring built-in functions provided by the toolchain (e.g. GCC intrinsics).

In the case of Ada there has been a reluctance among the language designers to add support for atomic operations to the language, with the (GNU) toolchain offering the same intrinsics as a fallback. With the Ada 2022 standard there is now direct support in the System.Atomic_Operations library, however.

Continue reading “Programming Ada: Atomics And Other Low-Level Details” →