re PR middle-end/36286 (ICE with -ftree-loop-linear and -O1 and above)
authorSebastian Pop <sebastian.pop@amd.com>
Wed, 21 May 2008 23:13:00 +0000 (23:13 +0000)
committerSebastian Pop <spop@gcc.gnu.org>
Wed, 21 May 2008 23:13:00 +0000 (23:13 +0000)
commit4e8d71770098bd2fc4ddf78c68fa72a2df205308
treed94b5ef5e17a03628d716a1445a4b91366182b82
parent5d6342ebc23344e8168b1589589b68f7b6781f66
re PR middle-end/36286 (ICE with -ftree-loop-linear and -O1 and above)

2008-05-21  Sebastian Pop  <sebastian.pop@amd.com>

PR tree-optimization/36286
PR tree-optimization/36287
* lambda-code.c (build_access_matrix): Do not use the loop->num
for computing the number of induction variables: use the loop depth
instead.

* testsuite/gcc.dg/tree-ssa/pr36287.c: New.
* testsuite/gfortran.dg/pr36286.f90: New.

From-SVN: r135741
gcc/ChangeLog
gcc/lambda-code.c
gcc/testsuite/ChangeLog
gcc/testsuite/gcc.dg/tree-ssa/pr36287.c [new file with mode: 0644]
gcc/testsuite/gfortran.dg/pr36286.f90 [new file with mode: 0644]