diff options
| author | Alejandro Martinez <alejandro.martinezvicente@arm.com> | 2019-05-28 13:48:44 +0000 |
|---|---|---|
| committer | Alejandro Martinez <alejandro@gcc.gnu.org> | 2019-05-28 13:48:44 +0000 |
| commit | 997636716c5dde7d59d026726a6f58918069f122 (patch) | |
| tree | d18d1d7f5388bc0bb743583ce506e284c8eb60d3 /gcc/tree-vectorizer.c | |
| parent | 8b4e7143550cd1f3f4b1dca005a5e656506979d9 (diff) | |
| download | gcc-997636716c5dde7d59d026726a6f58918069f122.zip gcc-997636716c5dde7d59d026726a6f58918069f122.tar.gz gcc-997636716c5dde7d59d026726a6f58918069f122.tar.bz2 | |
Current vectoriser doesn't support masked loads for SLP.
Current vectoriser doesn't support masked loads for SLP. We should add that, to
allow things like:
void
f (int *restrict x, int *restrict y, int *restrict z, int n)
{
for (int i = 0; i < n; i += 2)
{
x[i] = y[i] ? z[i] : 1;
x[i + 1] = y[i + 1] ? z[i + 1] : 2;
}
}
to be vectorized using contiguous loads rather than LD2 and ST2.
This patch was motivated by SVE, but it is completely generic and should apply
to any architecture with masked loads.
From-SVN: r271704
Diffstat (limited to 'gcc/tree-vectorizer.c')
| -rw-r--r-- | gcc/tree-vectorizer.c | 1 |
1 files changed, 1 insertions, 0 deletions
diff --git a/gcc/tree-vectorizer.c b/gcc/tree-vectorizer.c index d271049..4f6c65f 100644 --- a/gcc/tree-vectorizer.c +++ b/gcc/tree-vectorizer.c @@ -641,6 +641,7 @@ vec_info::new_stmt_vec_info (gimple *stmt) STMT_VINFO_VECTORIZABLE (res) = true; STMT_VINFO_VEC_REDUCTION_TYPE (res) = TREE_CODE_REDUCTION; STMT_VINFO_VEC_CONST_COND_REDUC_CODE (res) = ERROR_MARK; + STMT_VINFO_SLP_VECT_ONLY (res) = false; if (gimple_code (stmt) == GIMPLE_PHI && is_loop_header_bb_p (gimple_bb (stmt))) |