* [PATCH] tree-optimization/97769 - fix assert in peeling for alignment
@ 2020-11-10 12:38 Richard Biener
0 siblings, 0 replies; only message in thread
From: Richard Biener @ 2020-11-10 12:38 UTC (permalink / raw)
To: gcc-patches
The following removes an assert that can not easily be adjusted to
cover the additional cases we now handle after the removal of
the same-align DRs vector.
Tested on x86_64-unknown-linux and aarch64, pushed.
2020-11-10 Richard Biener <rguenther@suse.de>
PR tree-optimization/97769
* tree-vect-data-refs.c (vect_update_misalignment_for_peel):
Remove assert.
* gcc.dg/vect/pr97769.c: New testcase.
---
gcc/testsuite/gcc.dg/vect/pr97769.c | 32 +++++++++++++++++++++++++++++
gcc/tree-vect-data-refs.c | 7 +------
2 files changed, 33 insertions(+), 6 deletions(-)
create mode 100644 gcc/testsuite/gcc.dg/vect/pr97769.c
diff --git a/gcc/testsuite/gcc.dg/vect/pr97769.c b/gcc/testsuite/gcc.dg/vect/pr97769.c
new file mode 100644
index 00000000000..127f91aa8fd
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/vect/pr97769.c
@@ -0,0 +1,32 @@
+/* { dg-do compile } */
+/* { dg-additional-options "-O3" } */
+
+typedef struct {
+ int alg;
+ int h1[8];
+ unsigned d1[1];
+} tmp;
+typedef struct {
+ tmp itmp;
+ tmp otmp;
+} h1;
+h1 c;
+
+static void
+fn1(char *p1, int p2)
+{
+ int i = 0;
+ for (; i < 4; i++)
+ *p1++ = p2;
+}
+
+static void
+fn2(tmp *p1)
+{
+ char *d = (char *)p1->d1;
+ int *b = p1->h1;
+ for (int a; a; a++, d += 4)
+ fn1(d, *b++);
+}
+
+void fn3() { fn2(&(&c)->otmp); }
diff --git a/gcc/tree-vect-data-refs.c b/gcc/tree-vect-data-refs.c
index 8afd3044461..0efab495407 100644
--- a/gcc/tree-vect-data-refs.c
+++ b/gcc/tree-vect-data-refs.c
@@ -1186,14 +1186,9 @@ static void
vect_update_misalignment_for_peel (dr_vec_info *dr_info,
dr_vec_info *dr_peel_info, int npeel)
{
- /* It can be assumed that if dr_info has the same alignment as dr_peel,
- it is aligned in the vector loop. */
+ /* If dr_info is aligned of dr_peel_info is, then mark it so. */
if (vect_dr_aligned_if_peeled_dr_is (dr_info, dr_peel_info))
{
- gcc_assert (!known_alignment_for_access_p (dr_info)
- || !known_alignment_for_access_p (dr_peel_info)
- || (DR_MISALIGNMENT (dr_info)
- == DR_MISALIGNMENT (dr_peel_info)));
SET_DR_MISALIGNMENT (dr_info, 0);
return;
}
--
2.26.2
^ permalink raw reply [flat|nested] only message in thread
only message in thread, other threads:[~2020-11-10 12:38 UTC | newest]
Thread overview: (only message) (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2020-11-10 12:38 [PATCH] tree-optimization/97769 - fix assert in peeling for alignment Richard Biener
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).