On Jan 25, 2018, Alexandre Oliva <aol...@redhat.com> wrote: > Thanks, I'll retest with the simplified test (just in case; for I can't > recall why I ended up with all those redundant conditions),
As suspected, removing the redundant tests didn't regress anything. I suppose they mattered in some earlier experimental version of the patch (I vaguely recall having a long sequence of tests within the loop at some point), and I just failed to further simplify the final form. Thanks for catching the further simplification opportunities! FTR, here's the patch I'm installing, while awaiting a review from someone else on the rtl auto-inc patch (the second and last patch in https://gcc.gnu.org/ml/gcc-patches/2018-01/msg01994.html). [PR81611] accept copies in simple_iv_increment_p If there are copies between the GIMPLE_PHI at the loop body and the increment that reaches it (presumably through a back edge), still regard it as a simple_iv_increment, so that we won't consider the value in the back edge eligible for forwprop. Doing so would risk making the phi node and the incremented conflicting value live within the loop, and the phi node to be preserved for propagated uses after the loop. for gcc/ChangeLog PR tree-optimization/81611 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening copies. --- gcc/tree-ssa-dom.c | 18 ++++++++++++++---- 1 file changed, 14 insertions(+), 4 deletions(-) diff --git a/gcc/tree-ssa-dom.c b/gcc/tree-ssa-dom.c index 2b371667253a..a6f176c5def0 100644 --- a/gcc/tree-ssa-dom.c +++ b/gcc/tree-ssa-dom.c @@ -1276,8 +1276,11 @@ record_equality (tree x, tree y, class const_and_copies *const_and_copies) /* Returns true when STMT is a simple iv increment. It detects the following situation: - i_1 = phi (..., i_2) - i_2 = i_1 +/- ... */ + i_1 = phi (..., i_k) + [...] + i_j = i_{j-1} for each j : 2 <= j <= k-1 + [...] + i_k = i_{k-1} +/- ... */ bool simple_iv_increment_p (gimple *stmt) @@ -1305,8 +1308,15 @@ simple_iv_increment_p (gimple *stmt) return false; phi = SSA_NAME_DEF_STMT (preinc); - if (gimple_code (phi) != GIMPLE_PHI) - return false; + while (gimple_code (phi) != GIMPLE_PHI) + { + /* Follow trivial copies, but not the DEF used in a back edge, + so that we don't prevent coalescing. */ + if (!gimple_assign_ssa_name_copy_p (phi)) + return false; + preinc = gimple_assign_rhs1 (phi); + phi = SSA_NAME_DEF_STMT (preinc); + } for (i = 0; i < gimple_phi_num_args (phi); i++) if (gimple_phi_arg_def (phi, i) == lhs) -- Alexandre Oliva, freedom fighter http://FSFLA.org/~lxoliva/ You must be the change you wish to see in the world. -- Gandhi Be Free! -- http://FSFLA.org/ FSF Latin America board member Free Software Evangelist|Red Hat Brasil GNU Toolchain Engineer
