1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
|
2011-10-06 Ira Rosen <ira.rosen@linaro.org>
gcc/testsuite/
* gcc.dg/vect/bb-slp-26.c: Simplify to make the basic block
vectorizable.
Backport from mainline:
2011-09-25 Ira Rosen <ira.rosen@linaro.org>
gcc/
* tree-vect-slp.c (vect_slp_analyze_bb_1): Split out core part
of vect_analyze_bb here.
(vect_analyze_bb): Loop over vector sizes calling vect_analyze_bb_1.
gcc/testsuite/
* lib/target-supports.exp (check_effective_target_vect64): New.
* gcc.dg/vect/bb-slp-11.c: Expect the error message twice in case
of multiple vector sizes.
* gcc.dg/vect/bb-slp-26.c: New.
=== modified file 'gcc/testsuite/gcc.dg/vect/bb-slp-11.c'
--- old/gcc/testsuite/gcc.dg/vect/bb-slp-11.c 2010-11-22 12:16:52 +0000
+++ new/gcc/testsuite/gcc.dg/vect/bb-slp-11.c 2011-10-02 10:40:34 +0000
@@ -49,6 +49,7 @@
}
/* { dg-final { scan-tree-dump-times "basic block vectorized using SLP" 0 "slp" } } */
-/* { dg-final { scan-tree-dump-times "SLP with multiple types" 1 "slp" } } */
+/* { dg-final { scan-tree-dump-times "SLP with multiple types" 1 "slp" { xfail vect_multiple_sizes } } } */
+/* { dg-final { scan-tree-dump-times "SLP with multiple types" 2 "slp" { target vect_multiple_sizes } } } */
/* { dg-final { cleanup-tree-dump "slp" } } */
=== added file 'gcc/testsuite/gcc.dg/vect/bb-slp-26.c'
--- old/gcc/testsuite/gcc.dg/vect/bb-slp-26.c 1970-01-01 00:00:00 +0000
+++ new/gcc/testsuite/gcc.dg/vect/bb-slp-26.c 2011-10-02 10:40:34 +0000
@@ -0,0 +1,60 @@
+/* { dg-require-effective-target vect_int } */
+
+#include <stdarg.h>
+#include "tree-vect.h"
+
+#define A 3
+#define B 4
+#define N 256
+
+char src[N], dst[N];
+
+void foo (char * __restrict__ dst, char * __restrict__ src, int h,
+ int stride, int dummy)
+{
+ int i;
+ h /= 16;
+ for (i = 0; i < h; i++)
+ {
+ dst[0] += A*src[0];
+ dst[1] += A*src[1];
+ dst[2] += A*src[2];
+ dst[3] += A*src[3];
+ dst[4] += A*src[4];
+ dst[5] += A*src[5];
+ dst[6] += A*src[6];
+ dst[7] += A*src[7];
+ dst += 8;
+ src += 8;
+ if (dummy == 32)
+ abort ();
+ }
+}
+
+
+int main (void)
+{
+ int i;
+
+ check_vect ();
+
+ for (i = 0; i < N; i++)
+ {
+ dst[i] = 0;
+ src[i] = i/8;
+ }
+
+ foo (dst, src, N, 8, 0);
+
+ for (i = 0; i < N/2; i++)
+ {
+ if (dst[i] != A * src[i])
+ abort ();
+ }
+
+ return 0;
+}
+
+/* { dg-final { scan-tree-dump-times "basic block vectorized using SLP" 1 "slp" { target vect64 } } } */
+/* { dg-final { cleanup-tree-dump "slp" } } */
+
=== modified file 'gcc/testsuite/lib/target-supports.exp'
--- old/gcc/testsuite/lib/target-supports.exp 2011-09-20 07:54:28 +0000
+++ new/gcc/testsuite/lib/target-supports.exp 2011-10-02 10:40:34 +0000
@@ -3283,6 +3283,24 @@
return $et_vect_multiple_sizes_saved
}
+# Return 1 if the target supports vectors of 64 bits.
+
+proc check_effective_target_vect64 { } {
+ global et_vect64
+
+ if [info exists et_vect64_saved] {
+ verbose "check_effective_target_vect64: using cached result" 2
+ } else {
+ set et_vect64_saved 0
+ if { ([istarget arm*-*-*] && [check_effective_target_arm_neon_ok]) } {
+ set et_vect64_saved 1
+ }
+ }
+
+ verbose "check_effective_target_vect64: returning $et_vect64_saved" 2
+ return $et_vect64_saved
+}
+
# Return 1 if the target supports section-anchors
proc check_effective_target_section_anchors { } {
=== modified file 'gcc/tree-vect-slp.c'
--- old/gcc/tree-vect-slp.c 2011-07-06 12:04:10 +0000
+++ new/gcc/tree-vect-slp.c 2011-10-02 10:40:34 +0000
@@ -1664,42 +1664,18 @@
/* Check if the basic block can be vectorized. */
-bb_vec_info
-vect_slp_analyze_bb (basic_block bb)
+static bb_vec_info
+vect_slp_analyze_bb_1 (basic_block bb)
{
bb_vec_info bb_vinfo;
VEC (ddr_p, heap) *ddrs;
VEC (slp_instance, heap) *slp_instances;
slp_instance instance;
- int i, insns = 0;
- gimple_stmt_iterator gsi;
+ int i;
int min_vf = 2;
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");
-
- for (gsi = gsi_start_bb (bb); !gsi_end_p (gsi); gsi_next (&gsi))
- {
- gimple stmt = gsi_stmt (gsi);
- if (!is_gimple_debug (stmt)
- && !gimple_nop_p (stmt)
- && gimple_code (stmt) != GIMPLE_LABEL)
- insns++;
- }
-
- if (insns > PARAM_VALUE (PARAM_SLP_MAX_INSNS_IN_BB))
- {
- if (vect_print_dump_info (REPORT_UNVECTORIZED_LOCATIONS))
- fprintf (vect_dump, "not vectorized: too many instructions in basic "
- "block.\n");
-
- return NULL;
- }
-
bb_vinfo = new_bb_vec_info (bb);
if (!bb_vinfo)
return NULL;
@@ -1819,6 +1795,61 @@
}
+bb_vec_info
+vect_slp_analyze_bb (basic_block bb)
+{
+ bb_vec_info bb_vinfo;
+ int insns = 0;
+ gimple_stmt_iterator gsi;
+ unsigned int vector_sizes;
+
+ if (vect_print_dump_info (REPORT_DETAILS))
+ fprintf (vect_dump, "===vect_slp_analyze_bb===\n");
+
+ for (gsi = gsi_start_bb (bb); !gsi_end_p (gsi); gsi_next (&gsi))
+ {
+ gimple stmt = gsi_stmt (gsi);
+ if (!is_gimple_debug (stmt)
+ && !gimple_nop_p (stmt)
+ && gimple_code (stmt) != GIMPLE_LABEL)
+ insns++;
+ }
+
+ if (insns > PARAM_VALUE (PARAM_SLP_MAX_INSNS_IN_BB))
+ {
+ if (vect_print_dump_info (REPORT_UNVECTORIZED_LOCATIONS))
+ fprintf (vect_dump, "not vectorized: too many instructions in basic "
+ "block.\n");
+
+ return NULL;
+ }
+
+ /* Autodetect first vector size we try. */
+ current_vector_size = 0;
+ vector_sizes = targetm.vectorize.autovectorize_vector_sizes ();
+
+ while (1)
+ {
+ bb_vinfo = vect_slp_analyze_bb_1 (bb);
+ if (bb_vinfo)
+ return bb_vinfo;
+
+ destroy_bb_vec_info (bb_vinfo);
+
+ vector_sizes &= ~current_vector_size;
+ if (vector_sizes == 0
+ || current_vector_size == 0)
+ return NULL;
+
+ /* Try the next biggest vector size. */
+ current_vector_size = 1 << floor_log2 (vector_sizes);
+ if (vect_print_dump_info (REPORT_DETAILS))
+ fprintf (vect_dump, "***** Re-trying analysis with "
+ "vector size %d\n", current_vector_size);
+ }
+}
+
+
/* SLP costs are calculated according to SLP instance unrolling factor (i.e.,
the number of created vector stmts depends on the unrolling factor).
However, the actual number of vector stmts for every SLP node depends on
|
