When we relaxed the vectorizers constraint on loop structure verifying the emptiness of the latch became too lose as can be seen in the case for PR116879 where the latch effectively contains two basic-blocks which one being an unmerged forwarder that's not empty.
Bootstrapped and tested on x86_64-unknown-linux-gnu, pushed. PR tree-optimization/116879 * tree-vect-loop.cc (vect_analyze_loop_form): Scan all blocks that form the latch. * gcc.dg/pr116879.c: New testcase. --- gcc/testsuite/gcc.dg/pr116879.c | 15 +++++++++++++++ gcc/tree-vect-loop.cc | 15 +++++++++++---- 2 files changed, 26 insertions(+), 4 deletions(-) create mode 100644 gcc/testsuite/gcc.dg/pr116879.c diff --git a/gcc/testsuite/gcc.dg/pr116879.c b/gcc/testsuite/gcc.dg/pr116879.c new file mode 100644 index 00000000000..73ddb2b658c --- /dev/null +++ b/gcc/testsuite/gcc.dg/pr116879.c @@ -0,0 +1,15 @@ +/* { dg-do compile } */ +/* { dg-options "-O2 -ftree-vectorize -fallow-store-data-races -fno-tree-ch -ftree-loop-distribution" } */ + +static int b; +int *a, c, *d = &c; +int main() { + int e = 0; + for (; e < 8; e = (char)(e + 1)) { + int *f = &b, g[8], h = 0; + for (; h < 8; h++) + g[h] = 0; + --*f != (*d = g[0] || a); + } + return 0; +} diff --git a/gcc/tree-vect-loop.cc b/gcc/tree-vect-loop.cc index 5cd4bdb32e0..0ce1bf8ebba 100644 --- a/gcc/tree-vect-loop.cc +++ b/gcc/tree-vect-loop.cc @@ -1851,10 +1851,17 @@ vect_analyze_loop_form (class loop *loop, vect_loop_form_info *info) " too many incoming edges.\n"); /* We assume that the latch is empty. */ - if (!empty_block_p (loop->latch) - || !gimple_seq_empty_p (phi_nodes (loop->latch))) - return opt_result::failure_at (vect_location, - "not vectorized: latch block not empty.\n"); + basic_block latch = loop->latch; + do + { + if (!empty_block_p (latch) + || !gimple_seq_empty_p (phi_nodes (latch))) + return opt_result::failure_at (vect_location, + "not vectorized: latch block not " + "empty.\n"); + latch = single_pred (latch); + } + while (single_succ_p (latch)); /* Make sure there is no abnormal exit. */ auto_vec<edge> exits = get_loop_exit_edges (loop); -- 2.43.0