Tobias Burnus <tbur...@baylibre.com> writes:
> Alex wrote:
>> Tested on x86_64-pc-linux-gnu, this is only a library addition (and a
>> few tests) so it shouldn't cause any major impacts. I also tested
>> libgomp C to ensure the conditional compile was working.
>
> Namely, the change is only to omp.h(.in) - effective for
> __cplusplus >= 201103L.
>
> Note that the following is an OpenMP 5.0 feature that for some reason
> was missed when implementing omp_alloc / omp_free support.
>
> omp::allocator:: ... <T>
>
> where ... is the name of a predefined allocator (with omp_ and _alloc
> stripped).
> [Support for omp::allocator::null_allocator is a (semi-accidental)
> OpenMP 6.0 feature, where omp_null_allocator implies that the allocator
> of default-allocator-var ICV is used.]
>
> The main use case of this feature is to make it easy to use those
> allocators with containers from the STL like:
>
> std::vector<int, omp::allocator::cgroup_mem<int>> var;
>
> where cgroup_mem uses low latency memory on AMD and Nvidia GPU devices,
> which is faster than the normal allocator.
> (→ https://gcc.gnu.org/onlinedocs/libgomp/Offload-Target-Specifics.html
> for cgroup_mem )
As reported in PR libgomp/119849, those new tests FAIL on Solaris (and
most likely any non-Linux targets).
I couldn't find a Bugzilla account for Alex, so he isn't Cc'ed (and BZ
is intolerably slow ATM).
Rainer
--
-----------------------------------------------------------------------------
Rainer Orth, Center for Biotechnology, Bielefeld University
