]> git.ipfire.org Git - thirdparty/gcc.git/commit
tree-optimization/107865 - ICE with outlining of loops
authorRichard Biener <rguenther@suse.de>
Fri, 25 Nov 2022 07:27:42 +0000 (08:27 +0100)
committerRichard Biener <rguenther@suse.de>
Fri, 25 Nov 2022 07:27:42 +0000 (08:27 +0100)
commitbcc2449384f2092cbdf5d6ac2357aeabe3212b2e
tree273c3818c49048483e3731089711dc9f556020a6
parentf120196382ac5ac49ec4a60f8abad42f22d45a91
tree-optimization/107865 - ICE with outlining of loops

The following makes sure to clear loops number of iterations when
outlining them as part of a SESE region as can happen with
auto-parallelization.  The referenced SSA names become stale otherwise.

PR tree-optimization/107865
* tree-cfg.cc (move_sese_region_to_fn): Free the number of
iterations of moved loops.

* gfortran.dg/graphite/pr107865.f90: New testcase.
gcc/testsuite/gfortran.dg/graphite/pr107865.f90 [new file with mode: 0644]
gcc/tree-cfg.cc