The test is for loop vectorization producing non-canonical multiplications. We can now BB vectorize the whole function when the target supports .REDUC_PLUS for V2SImode but we don't have a dejagnu selector for that. Disable BB vectorization like we disabled epilogue vectorization.
Tested on aarch64-linux, pushed. PR testsuite/111125 * gcc.dg/vect/pr53773.c: Disable BB vectorization. --- gcc/testsuite/gcc.dg/vect/pr53773.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) diff --git a/gcc/testsuite/gcc.dg/vect/pr53773.c b/gcc/testsuite/gcc.dg/vect/pr53773.c index 7f8229571ec..213e74e0b9d 100644 --- a/gcc/testsuite/gcc.dg/vect/pr53773.c +++ b/gcc/testsuite/gcc.dg/vect/pr53773.c @@ -1,7 +1,10 @@ /* Disabling epilogues until we find a better way to deal with scans. */ /* { dg-additional-options "--param vect-epilogues-nomask=0" } */ /* { dg-do compile } */ -/* { dg-additional-options "-fdump-tree-optimized" } */ +/* Disable BB vectorization, we now can cover the whole loop with that + when the target supports .REDUC_PLUS for integers, we don't have a + good target selector for this. */ +/* { dg-additional-options "-fdump-tree-optimized -fno-tree-slp-vectorize" } */ int foo (int integral, int decimal, int power_ten) -- 2.35.3