Does it mean that if we use transactional memory, we need to put the atomic code in the same code block? If it is the case, I think it's not a good feature for modern software engineering..
bwf
I'm not sure I'd agree with you on that statement. Even without transactional memory, programmers still need to be aware of these kinds of issues. If anything, I think it simplifies that matter by allowing us to simplify the way we think about our code since we can ignore locks since we know the code that is written in a block will execute atomically.
lament
@zhanpenf My impression is that all operations that a "tread of control" performs prior to the end of the atomic region are perform atomically, so functions that are called would be run atomically. It would be great to get some confirmation on the subject, though.
Does it mean that if we use transactional memory, we need to put the atomic code in the same code block? If it is the case, I think it's not a good feature for modern software engineering..
I'm not sure I'd agree with you on that statement. Even without transactional memory, programmers still need to be aware of these kinds of issues. If anything, I think it simplifies that matter by allowing us to simplify the way we think about our code since we can ignore locks since we know the code that is written in a block will execute atomically.
@zhanpenf My impression is that all operations that a "tread of control" performs prior to the end of the atomic region are perform atomically, so functions that are called would be run atomically. It would be great to get some confirmation on the subject, though.