[atomics] Is the read-modify-write operation a single indivisible operation?

[xmh0511]

Full name of submitter (unless configured in github; will be published with the issue): Jim X

As mentioned in this SO question https://stackoverflow.com/questions/65568185/for-purposes-of-ordering-is-atomic-read-modify-write-one-operation-or-two

It is unclear whether read-modify-write is a single operation or three operations implied by its name. As discussed in that question, certain people think read-modify-write comprises two operations, and certain people think it is a single operation, however, it is not explicitly specified in the document

[atomics.order] p10 says

Atomic read-modify-write operations shall always read the last value (in the modification order) written before the write associated with the read-modify-write operation.

Note the emphasized wording, read and write seem to imply that a read-modify-write operation comprises two operations: read and write.

Suggested Resolution

Accurately define the read-modify-write operation

[t3nsor]

Can you explain the normative implications of the issue that you believe exists?

[jensmaurer]

The concern you have is about a library clause, and was already filed as a library isuse: LWG3980

[xmh0511]

The concern you have is about a library clause, and was already filed as a library isuse: LWG3980

I think they are different issues.

[t3nsor]

But you haven't really explained the issue.