This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
GCC 5 Status Report (2015-01-19), Trunk in Stage 4 - an exception for OpenACC 2.0, nvptx and KNL offloading support?
- From: Mark Farnell <mark dot farnell at gmail dot com>
- To: gcc at gcc dot gnu dot org
- Date: Tue, 20 Jan 2015 23:59:29 +1300
- Subject: GCC 5 Status Report (2015-01-19), Trunk in Stage 4 - an exception for OpenACC 2.0, nvptx and KNL offloading support?
- Authentication-results: sourceware.org; auth=none
I know that we are already in stage 4, but features such as OpenACC
2.0, nvptx and KNL (xeon phi) offloading support are sooooo important
for GCC, and if they have to be deferred to GCC 6.0, then it would be
a great loss to GCC, as OpenACC 2.0 makes heterogeneous manycore
programming so much easier, as it relieves users from doing
assembly-like low level stuff as in CUDA and OpenCL.
Currently, OpenACC 2.0, nvptx and KNL (xeon phi) support live in the
gomp-4_0-branch From previous discussions in this list, these
features are mature enough to be merged into the trunk, and the owner
of the trunk is just waiting for the codesourcery to do the job.
So in order to include these features into GCC 5.0, is it possible to
make an exception so that if these features merges into the trunk
within two weeks, they will be allowed to be merged and included into
GCC 5.0?
Otherwise it would be a big blow to both GCC, and OpenACC, as users
would otherwise be forced to use proprietary compilers.