https://gcc.gnu.org/bugzilla/show_bug.cgi?id=100408
Bug ID: 100408
Summary: [nvptx][OpenMP] Enable SIMT for user-defined reduction
Product: gcc
Version: 12.0
Status: UNCONFIRMED
Keywords: missed-optimization, openmp
Severity: normal
Priority: P3
Component: target
Assignee: unassigned at gcc dot gnu.org
Reporter: burnus at gcc dot gnu.org
Target Milestone: ---
Target: nvptx-none
Follow-up to PR100321 - which is about the same issue but solved it by
disabling SIMT
>From the patch:
The test-case included in this patch contains this target region:
...
for (int i0 = 0 ; i0 < N0 ; i0++ )
counter_N0.i += 1;
...
When running with nvptx accelerator, the counter variable is expected to
be N0 after the region, but instead is N0 / 32. The problem is that rather
than getting the result for all warp lanes, we get it for just one lane.
This is caused by the implementation of SIMT being incomplete. It handles
regular reductions, but appearantly not user-defined reductions.
For now, handle this by disabling SIMT in this case, specifically by
setting
sctx->max_vf to 1.
For details (code location, longer testcases, etc.) see PR100321
