Date: Thu, 25 Jul 2013 15:10:19 -0700
On 7/25/13, Gabriel Dos Reis <gdr_at_[hidden]> wrote:
> Lawrence Crowl <Lawrence_at_[hidden]> writes:
>
> | On 7/25/13, Gabriel Dos Reis <gdr_at_[hidden]> wrote:
> | > Lawrence Crowl <Lawrence_at_[hidden]> writes:
> | >
> | > | On 7/25/13, Howard Hinnant <howard.hinnant_at_[hidden]> wrote:
> | > | > It was this SO question that started this thread:
> | > | >
> | > | > http://stackoverflow.com/q/17789928/576911
> | > | >
> | > | > I'm curious: The accepted answer uses memcpy and the claim is that
> this
> | > is
> | > | > a correct answer to the question. That is it does not exhibit
> | > undefined
> | > | > behavior. My current understanding is that I agree with this
> answer.
> | > But I
> | > | > wanted to check here. Do people here agree that:
> | > | >
> | > | > http://stackoverflow.com/a/17790026/576911
> | > | >
> | > | > does not break the aliasing rules, or otherwise invoke undefined
> | > behavior?
> | > |
> | > | I believe so.
> | >
> | > I can't find the rule (in the standards) that the memcpy() into i
> | > constitutes an initialization.
> |
> | It is trivial assignment, and initialization is unnecessary because it
> | is trivial. We could probably be clearer.
>
> I can't find the rule for that interpretation. The closest I can come
> to is 3.9/3, but that paragraph requires copying from an object of the
> same type:
>
> # For any trivially copyable type T, if two pointers to T point to
> # distinct T objects obj1 and obj2, where neither obj1 nor obj2 is a
> # base-class subobject, if the underlying bytes (1.7) making up obj1 are
> # copied into obj2,41 obj2 shall subsequently hold the same value as
> obj1.
>
> | > One also needs something for the
> | > memcpy() back into x, saying that does not yield a trap representation,
> | > which would otherwise lead to undefined behaviour in the update:
> | >
> | > x = x*(1.5f - xhalf*x*x);
> |
> | But at that point aren't we already platform dependent?
> | I.e. dealing with the representation?
>
> I can't find rules that cover that, or that would make it platform
> dependent.
I was trying to get a sound intuition. Once we have that, I think we
can make the necessary wording.
> Lawrence Crowl <Lawrence_at_[hidden]> writes:
>
> | On 7/25/13, Gabriel Dos Reis <gdr_at_[hidden]> wrote:
> | > Lawrence Crowl <Lawrence_at_[hidden]> writes:
> | >
> | > | On 7/25/13, Howard Hinnant <howard.hinnant_at_[hidden]> wrote:
> | > | > It was this SO question that started this thread:
> | > | >
> | > | > http://stackoverflow.com/q/17789928/576911
> | > | >
> | > | > I'm curious: The accepted answer uses memcpy and the claim is that
> this
> | > is
> | > | > a correct answer to the question. That is it does not exhibit
> | > undefined
> | > | > behavior. My current understanding is that I agree with this
> answer.
> | > But I
> | > | > wanted to check here. Do people here agree that:
> | > | >
> | > | > http://stackoverflow.com/a/17790026/576911
> | > | >
> | > | > does not break the aliasing rules, or otherwise invoke undefined
> | > behavior?
> | > |
> | > | I believe so.
> | >
> | > I can't find the rule (in the standards) that the memcpy() into i
> | > constitutes an initialization.
> |
> | It is trivial assignment, and initialization is unnecessary because it
> | is trivial. We could probably be clearer.
>
> I can't find the rule for that interpretation. The closest I can come
> to is 3.9/3, but that paragraph requires copying from an object of the
> same type:
>
> # For any trivially copyable type T, if two pointers to T point to
> # distinct T objects obj1 and obj2, where neither obj1 nor obj2 is a
> # base-class subobject, if the underlying bytes (1.7) making up obj1 are
> # copied into obj2,41 obj2 shall subsequently hold the same value as
> obj1.
>
> | > One also needs something for the
> | > memcpy() back into x, saying that does not yield a trap representation,
> | > which would otherwise lead to undefined behaviour in the update:
> | >
> | > x = x*(1.5f - xhalf*x*x);
> |
> | But at that point aren't we already platform dependent?
> | I.e. dealing with the representation?
>
> I can't find rules that cover that, or that would make it platform
> dependent.
I was trying to get a sound intuition. Once we have that, I think we
can make the necessary wording.
-- Lawrence Crowl
Received on 2013-07-26 00:10:35