Message ID | ord0uklwq2.fsf@lxoliva.fsfla.org |
---|---|
State | New |
Headers | show |
Series | fix reassoc cut&pasto | expand |
On 08/14/2018 10:49 AM, Alexandre Oliva wrote: > I guess this was a cut&pasto. I caught it during code review. It > doesn't seem to matter much: I haven't been able to get any codegen > difference from this change, not in a bootstrap, not in test runs. I > even put a gcc_unreachable() in case the test on rhs2 were to return > false (which would never happen after the test on rhs1 passed, when rhs1 > and rhs2 were the same), and still no hits in stage3. Anyway... > > Regstrapped on x86_64-linux-gnu. Ok to install? > > From: Alexandre Oliva <aoliva@redhat.com> > > for gcc/ChangeLog > > * tree-ssa-reassoc.c (is_reassociable_op): Fix cut&pasto. OK. jeff
diff --git a/gcc/tree-ssa-reassoc.c b/gcc/tree-ssa-reassoc.c index 41a37ab29899..6b0bf5c03548 100644 --- a/gcc/tree-ssa-reassoc.c +++ b/gcc/tree-ssa-reassoc.c @@ -618,7 +618,7 @@ is_reassociable_op (gimple *stmt, enum tree_code code, struct loop *loop) && has_single_use (gimple_assign_lhs (stmt))) { tree rhs1 = gimple_assign_rhs1 (stmt); - tree rhs2 = gimple_assign_rhs1 (stmt); + tree rhs2 = gimple_assign_rhs2 (stmt); if (TREE_CODE (rhs1) == SSA_NAME && SSA_NAME_OCCURS_IN_ABNORMAL_PHI (rhs1)) return false;
I guess this was a cut&pasto. I caught it during code review. It doesn't seem to matter much: I haven't been able to get any codegen difference from this change, not in a bootstrap, not in test runs. I even put a gcc_unreachable() in case the test on rhs2 were to return false (which would never happen after the test on rhs1 passed, when rhs1 and rhs2 were the same), and still no hits in stage3. Anyway... Regstrapped on x86_64-linux-gnu. Ok to install? From: Alexandre Oliva <aoliva@redhat.com> for gcc/ChangeLog * tree-ssa-reassoc.c (is_reassociable_op): Fix cut&pasto. --- gcc/tree-ssa-reassoc.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)