This is the mail archive of the
gcc-bugs@gcc.gnu.org
mailing list for the GCC project.
[Bug target/80547] [6/7/8 Regression] nvptx back end ICE with OpenACC "reduction(OP:x)", "x = [...]"
- From: "tschwinge at gcc dot gnu.org" <gcc-bugzilla at gcc dot gnu dot org>
- To: gcc-bugs at gcc dot gnu dot org
- Date: Thu, 27 Apr 2017 17:22:11 +0000
- Subject: [Bug target/80547] [6/7/8 Regression] nvptx back end ICE with OpenACC "reduction(OP:x)", "x = [...]"
- Auto-submitted: auto-generated
- References: <bug-80547-4@http.gcc.gnu.org/bugzilla/>
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=80547
--- Comment #1 from Thomas Schwinge <tschwinge at gcc dot gnu.org> ---
I used "#pragma omp target teams num_teams(1) map(x) reduction(+: x)" to test
this with OpenMP "nvptx" offloading, which works.
An additional observation: for host fallback (adding an "if(0)" clause to both
the OpenMP and OpenACC directives, for example), the OpenMP code still works,
but for OpenACC, we then get a result of "x = 23", so apparently the reduction
computation/finalization code is being skipped in this case. (Let's please
make this a separate PR if it's not directly related to this one here.)