This fixes the FAIL of PR50058 - I tightened the struct copy matching
too much and somehow the FAIL got lost during testing.
Bootstrapped and tested on x86_64-unknown-linux-gnu, applied to trunk.
Richard.
2011-08-15 Richard Guenther <rguent...@suse.de>
PR tree-optimization/50058
* tree-ssa-sccvn.c (vn_reference_lookup_3): Relax aggregate
copy matching.
Index: gcc/tree-ssa-sccvn.c
===================================================================
*** gcc/tree-ssa-sccvn.c (revision 177757)
--- gcc/tree-ssa-sccvn.c (working copy)
*************** vn_reference_lookup_3 (ao_ref *ref, tree
*** 1485,1496 ****
may fail when comparing types for compatibility. But we really
don't care here - further lookups with the rewritten operands
will simply fail if we messed up types too badly. */
! if (j == 0 && i == 0
&& VEC_index (vn_reference_op_s, lhs_ops, 0)->opcode == MEM_REF
! && VEC_index (vn_reference_op_s, vr->operands, i)->opcode == MEM_REF
! && tree_int_cst_equal
! (VEC_index (vn_reference_op_s, lhs_ops, 0)->op0,
! VEC_index (vn_reference_op_s, vr->operands, i)->op0))
i--, j--;
/* i now points to the first additional op.
--- 1485,1495 ----
may fail when comparing types for compatibility. But we really
don't care here - further lookups with the rewritten operands
will simply fail if we messed up types too badly. */
! if (j == 0 && i >= 0
&& VEC_index (vn_reference_op_s, lhs_ops, 0)->opcode == MEM_REF
! && VEC_index (vn_reference_op_s, lhs_ops, 0)->off != -1
! && (VEC_index (vn_reference_op_s, lhs_ops, 0)->off
! == VEC_index (vn_reference_op_s, vr->operands, i)->off))
i--, j--;
/* i now points to the first additional op.
