Thread: Int64 (long long) Supporting Compiler Requirement Status?

There was a discussion once about using 64 bit long long compiler support to
increase the size of the transaction ids to solve the wrap around problem. I
understand that there is a different solution for this now.

However, my question is:  Are we to the point where int64's can be used in
mainstream code yet, or are there supported platforms that this will not work
with?  And if not, when (if ever) will such capability be standardized?

The reason why I ask is I would like to experiment with a variable length
base-(2^32) numeric type that I hope might be accepted someday, and
base-(2^32) operations need long long support to implement in a
straightforward fashion.

-  Mark Butler

Mark Butler <butlerm@middle.net> writes:
> However, my question is:  Are we to the point where int64's can be used in
> mainstream code yet, or are there supported platforms that this will not work
> with?  And if not, when (if ever) will such capability be standardized?

I don't foresee ever being willing to *require* int64 support.  It'll
always be optional.

> The reason why I ask is I would like to experiment with a variable length
> base-(2^32) numeric type that I hope might be accepted someday, and
> base-(2^32) operations need long long support to implement in a
> straightforward fashion.

I really doubt that base 2^32 would be enough faster than base 10000 to
be worth taking any portability risks for.
        regards, tom lane