public inbox for gcc-cvs@sourceware.org
help / color / mirror / Atom feed
* [gcc r11-1972] fixup BIT_FIELD_REF detection in SLP discovery
@ 2020-07-09 17:54 Richard Biener
0 siblings, 0 replies; only message in thread
From: Richard Biener @ 2020-07-09 17:54 UTC (permalink / raw)
To: gcc-cvs
https://gcc.gnu.org/g:9ddea9306251b7d4e4fd1d67a5941ef7448b2e66
commit r11-1972-g9ddea9306251b7d4e4fd1d67a5941ef7448b2e66
Author: Richard Biener <rguenther@suse.de>
Date: Thu Jul 9 16:06:04 2020 +0200
fixup BIT_FIELD_REF detection in SLP discovery
This fixes a thinko where we end up combining a BIT_FIELD_REF
and a memory access, fixed by checking all stmts are a load or
none.
2020-07-09 Richard Biener <rguenther@suse.de>
PR tree-optimization/96133
* tree-vect-slp.c (vect_build_slp_tree_1): Compare load_p
status between stmts.
Diff:
---
gcc/tree-vect-slp.c | 40 +++++++++++++++++++++-------------------
1 file changed, 21 insertions(+), 19 deletions(-)
diff --git a/gcc/tree-vect-slp.c b/gcc/tree-vect-slp.c
index 35ae6984593..b3645b0a820 100644
--- a/gcc/tree-vect-slp.c
+++ b/gcc/tree-vect-slp.c
@@ -760,7 +760,7 @@ vect_build_slp_tree_1 (vec_info *vinfo, unsigned char *swap,
machine_mode optab_op2_mode;
machine_mode vec_mode;
stmt_vec_info first_load = NULL, prev_first_load = NULL;
- bool load_p = false;
+ bool first_stmt_load_p = false, load_p = false;
/* For every stmt in NODE find its def stmt/s. */
stmt_vec_info stmt_info;
@@ -850,6 +850,7 @@ vect_build_slp_tree_1 (vec_info *vinfo, unsigned char *swap,
{
*node_vectype = vectype;
first_stmt_code = rhs_code;
+ first_stmt_load_p = load_p;
/* Shift arguments should be equal in all the packed stmts for a
vector shift with scalar shift operand. */
@@ -931,24 +932,25 @@ vect_build_slp_tree_1 (vec_info *vinfo, unsigned char *swap,
if (first_stmt_code != rhs_code
&& alt_stmt_code == ERROR_MARK)
alt_stmt_code = rhs_code;
- if (first_stmt_code != rhs_code
- && (first_stmt_code != IMAGPART_EXPR
- || rhs_code != REALPART_EXPR)
- && (first_stmt_code != REALPART_EXPR
- || rhs_code != IMAGPART_EXPR)
- /* Handle mismatches in plus/minus by computing both
- and merging the results. */
- && !((first_stmt_code == PLUS_EXPR
- || first_stmt_code == MINUS_EXPR)
- && (alt_stmt_code == PLUS_EXPR
- || alt_stmt_code == MINUS_EXPR)
- && rhs_code == alt_stmt_code)
- && !(STMT_VINFO_GROUPED_ACCESS (stmt_info)
- && (first_stmt_code == ARRAY_REF
- || first_stmt_code == BIT_FIELD_REF
- || first_stmt_code == INDIRECT_REF
- || first_stmt_code == COMPONENT_REF
- || first_stmt_code == MEM_REF)))
+ if ((first_stmt_code != rhs_code
+ && (first_stmt_code != IMAGPART_EXPR
+ || rhs_code != REALPART_EXPR)
+ && (first_stmt_code != REALPART_EXPR
+ || rhs_code != IMAGPART_EXPR)
+ /* Handle mismatches in plus/minus by computing both
+ and merging the results. */
+ && !((first_stmt_code == PLUS_EXPR
+ || first_stmt_code == MINUS_EXPR)
+ && (alt_stmt_code == PLUS_EXPR
+ || alt_stmt_code == MINUS_EXPR)
+ && rhs_code == alt_stmt_code)
+ && !(STMT_VINFO_GROUPED_ACCESS (stmt_info)
+ && (first_stmt_code == ARRAY_REF
+ || first_stmt_code == BIT_FIELD_REF
+ || first_stmt_code == INDIRECT_REF
+ || first_stmt_code == COMPONENT_REF
+ || first_stmt_code == MEM_REF)))
+ || first_stmt_load_p != load_p)
{
if (dump_enabled_p ())
{
^ permalink raw reply [flat|nested] only message in thread
only message in thread, other threads:[~2020-07-09 17:54 UTC | newest]
Thread overview: (only message) (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2020-07-09 17:54 [gcc r11-1972] fixup BIT_FIELD_REF detection in SLP discovery 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).