Hi!
On Wed, 19 Oct 2016 12:28:39 +0200, I wrote:
> [PR tree-optimization/78024] Clear basic block flags before using
> BB_VISITED for OpenACC loops processing
To fix the same problem, committed the following to gcc-6-branch in
r244264:
commit 82f25a52380b366a99af8045fb615d79d4ff94b6
Date: Wed Oct 19 10:19:24 2016 +
[PR tree-optimization/78024] Clear basic block flags before using
BB_VISITED for OpenACC loops processing
gcc/
* omp-low.c (oacc_loop_discovery): Call clear_bb_flags before, and
don't clear BB_VISITED after processin
(apart from performance regression, obviously); it just caused problems
>> > > in a few libgomp test cases that actually at run time test for
>> > > parallelized execution -- which will/did trigger only with nvptx
>> > > offloading enabled, which not too many peop
part from performance regression, obviously); it just caused problems
> > > in a few libgomp test cases that actually at run time test for
> > > parallelized execution -- which will/did trigger only with nvptx
> > > offloading enabled, which not too many people are testin
clear that this pattern will need to be adjusted if changing the
OpenACC loop processing.
> > This won't generally cause any problem
> > (apart from performance regression, obviously); it just caused problems
> > in a few libgomp test cases that actually at run time test for
gt; This won't generally cause any problem
> (apart from performance regression, obviously); it just caused problems
> in a few libgomp test cases that actually at run time test for
> parallelized execution -- which will/did trigger only with nvptx
> off
that ought to be parallelized. This won't generally cause any problem
(apart from performance regression, obviously); it just caused problems
in a few libgomp test cases that actually at run time test for
parallelized execution -- which will/did trigger only with nvptx
offloading enabled,
tree-dump line?
Ugh. Not worse to what we do in various dwarf scanning I guess.
Doesn't failure lead to a miscompile eventually? So you could formulate
this as a dg-do run test with a check for the desired outcome?
Richard.
> commit 4e8abdfd25aa08abbad0c3fe2e9ec6182308f78c
&g
ase appropriate (which would
have caught this issue right away), in particular the dg-final
scan-tree-dump line?
commit 4e8abdfd25aa08abbad0c3fe2e9ec6182308f78c
Author: Thomas Schwinge
Date: Mon Oct 17 11:29:43 2016 +0200
Clear basic block flags before using BB_VISITED for OpenACC loops processin
