diff options
| author | Richard Guenther <rguenther@suse.de> | 2010-10-07 14:36:38 +0000 |
|---|---|---|
| committer | Richard Biener <rguenth@gcc.gnu.org> | 2010-10-07 14:36:38 +0000 |
| commit | bb67d9c7cef80c80e817cafdd9ee2001cb29c9f7 (patch) | |
| tree | 2b9ac40e2ddfd76e458ab91dc73fbb371682b0f1 /gcc/tree-vect-slp.c | |
| parent | 767f865fb828fba54459911661d15289e1a42930 (diff) | |
| download | gcc-bb67d9c7cef80c80e817cafdd9ee2001cb29c9f7.tar.gz | |
machmode.h (mode_for_vector): Declare.
2010-10-07 Richard Guenther <rguenther@suse.de>
* machmode.h (mode_for_vector): Declare.
* stor-layout.c (mode_for_vector): New function, split out from ...
(layout_type): ... here.
* tree-vectorizer.h (current_vector_size): Declare.
* tree-vect-stmts.c (perm_mask_for_reverse): Check if the
mask vector type is available.
(get_vectype_for_scalar_type): Rename to ...
(get_vectype_for_scalar_type_and_size): ... this. Get a vector
size argument.
(get_vectype_for_scalar_type): New wrapper around
get_vectype_for_scalar_type_and_size using current_vector_size.
(get_same_sized_vectype): Use get_vectype_for_scalar_type_and_size.
* tree-vect-loop.c (vect_analyze_loop_2): Split out core part
of vect_analyze_loop here.
(vect_analyze_loop): Loop over vector sizes calling vect_analyze_loop_3.
* tree-vect-slp.c (vect_slp_analyze_bb): Set current_vector_size
to autodetect.
* config/i386/i386.c (ix86_vectorize_builtin_conversion): Fix
V8SF to V8SI conversion builtin.
From-SVN: r165116
Diffstat (limited to 'gcc/tree-vect-slp.c')
| -rw-r--r-- | gcc/tree-vect-slp.c | 1 |
1 files changed, 1 insertions, 0 deletions
diff --git a/gcc/tree-vect-slp.c b/gcc/tree-vect-slp.c index 4739538348c..2dbe328ba25 100644 --- a/gcc/tree-vect-slp.c +++ b/gcc/tree-vect-slp.c @@ -1643,6 +1643,7 @@ vect_slp_analyze_bb (basic_block bb) int max_vf = MAX_VECTORIZATION_FACTOR; bool data_dependence_in_bb = false; + current_vector_size = 0; if (vect_print_dump_info (REPORT_DETAILS)) fprintf (vect_dump, "===vect_slp_analyze_bb===\n"); |