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
|
From d3cdd96a300f9003a1cc242541605169aacdc811 Mon Sep 17 00:00:00 2001
From: willschm <willschm@138bc75d-0d04-0410-961f-82ee72b054a4>
Date: Mon, 25 Sep 2017 14:35:02 +0000
Subject: [PATCH] [gcc]
2017-09-25 Will Schmidt <will_schmidt@vnet.ibm.com>
* config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling
for early folding of vector stores (ALTIVEC_BUILTIN_ST_*).
(rs6000_builtin_valid_without_lhs): New helper function.
* config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
Remove obsoleted code for handling ALTIVEC_BUILTIN_VEC_ST.
git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@253152 138bc75d-0d04-0410-961f-82ee72b054a4
Fix internal compiler error for testcase gcc.dg/vmx/7d-02.c
Upstream commit: d3cdd96a300f9003a1cc242541605169aacdc811
Backport by Kaushik Phatak <Kaushik.Phatak@kpit.com>
This patch removes changes to Changelog from the original upstream patch.
This will help us avoid conflicts.
Upstream-Status: Backport
Signed-off-by: Mark Hatle <mark.hatle@windriver.com>
---
gcc/ChangeLog | 8 +++++
gcc/config/rs6000/rs6000-c.c | 72 -------------------------------------------
gcc/config/rs6000/rs6000.c | 73 ++++++++++++++++++++++++++++++++++++++++++--
3 files changed, 78 insertions(+), 75 deletions(-)
diff --git a/gcc/config/rs6000/rs6000-c.c b/gcc/config/rs6000/rs6000-c.c
index a49db97..4a363a1 100644
--- a/gcc/config/rs6000/rs6000-c.c
+++ b/gcc/config/rs6000/rs6000-c.c
@@ -6279,70 +6279,6 @@ altivec_resolve_overloaded_builtin (loca
}
}
- /* Similarly for stvx. */
- if (fcode == ALTIVEC_BUILTIN_VEC_ST
- && (BYTES_BIG_ENDIAN || !VECTOR_ELT_ORDER_BIG)
- && nargs == 3)
- {
- tree arg0 = (*arglist)[0];
- tree arg1 = (*arglist)[1];
- tree arg2 = (*arglist)[2];
-
- /* Construct the masked address. Let existing error handling take
- over if we don't have a constant offset. */
- arg1 = fold (arg1);
-
- if (TREE_CODE (arg1) == INTEGER_CST)
- {
- if (!ptrofftype_p (TREE_TYPE (arg1)))
- arg1 = build1 (NOP_EXPR, sizetype, arg1);
-
- tree arg2_type = TREE_TYPE (arg2);
- if (TREE_CODE (arg2_type) == ARRAY_TYPE && c_dialect_cxx ())
- {
- /* Force array-to-pointer decay for C++. */
- arg2 = default_conversion (arg2);
- arg2_type = TREE_TYPE (arg2);
- }
-
- /* Find the built-in to make sure a compatible one exists; if not
- we fall back to default handling to get the error message. */
- for (desc = altivec_overloaded_builtins;
- desc->code && desc->code != fcode; desc++)
- continue;
-
- for (; desc->code == fcode; desc++)
- if (rs6000_builtin_type_compatible (TREE_TYPE (arg0), desc->op1)
- && rs6000_builtin_type_compatible (TREE_TYPE (arg1), desc->op2)
- && rs6000_builtin_type_compatible (TREE_TYPE (arg2),
- desc->op3))
- {
- tree addr = fold_build2_loc (loc, POINTER_PLUS_EXPR, arg2_type,
- arg2, arg1);
- tree aligned
- = fold_build2_loc (loc, BIT_AND_EXPR, arg2_type,
- addr, build_int_cst (arg2_type, -16));
-
- tree arg0_type = TREE_TYPE (arg0);
- if (TYPE_MODE (arg0_type) == V2DImode)
- /* Type-based aliasing analysis thinks vector long
- and vector long long are different and will put them
- in distinct alias classes. Force our address type
- to be a may-alias type to avoid this. */
- arg0_type
- = build_pointer_type_for_mode (arg0_type, Pmode,
- true/*can_alias_all*/);
- else
- arg0_type = build_pointer_type (arg0_type);
- aligned = build1 (NOP_EXPR, arg0_type, aligned);
- tree stg = build_indirect_ref (loc, aligned, RO_NULL);
- tree retval = build2 (MODIFY_EXPR, TREE_TYPE (stg), stg,
- convert (TREE_TYPE (stg), arg0));
- return retval;
- }
- }
- }
-
for (n = 0;
!VOID_TYPE_P (TREE_VALUE (fnargs)) && n < nargs;
fnargs = TREE_CHAIN (fnargs), n++)
diff --git a/gcc/config/rs6000/rs6000.c b/gcc/config/rs6000/rs6000.c
index a49db97..4a363a1 100644
--- a/gcc/config/rs6000/rs6000.c
+++ b/gcc/config/rs6000/rs6000.c
@@ -55,6 +55,7 @@
#include "reload.h"
#include "sched-int.h"
#include "gimplify.h"
+#include "gimple-fold.h"
#include "gimple-iterator.h"
#include "gimple-ssa.h"
#include "gimple-walk.h"
@@ -17089,6 +17090,25 @@ rs6000_fold_builtin (tree fndecl, int n_
#endif
}
+/* Helper function to sort out which built-ins may be valid without having
+ a LHS. */
+bool
+rs6000_builtin_valid_without_lhs (enum rs6000_builtins fn_code)
+{
+ switch (fn_code)
+ {
+ case ALTIVEC_BUILTIN_STVX_V16QI:
+ case ALTIVEC_BUILTIN_STVX_V8HI:
+ case ALTIVEC_BUILTIN_STVX_V4SI:
+ case ALTIVEC_BUILTIN_STVX_V4SF:
+ case ALTIVEC_BUILTIN_STVX_V2DI:
+ case ALTIVEC_BUILTIN_STVX_V2DF:
+ return true;
+ default:
+ return false;
+ }
+}
+
/* Fold a machine-dependent built-in in GIMPLE. (For folding into
a constant, use rs6000_fold_builtin.) */
@@ -17102,6 +17122,10 @@ rs6000_gimple_fold_builtin (gimple_stmt_
= (enum rs6000_builtins) DECL_FUNCTION_CODE (fndecl);
tree arg0, arg1, lhs;
+ /* Prevent gimple folding for code that does not have a LHS, unless it is
+ allowed per the rs6000_builtin_valid_without_lhs helper function. */
+ if (!gimple_call_lhs (stmt) && !rs6000_builtin_valid_without_lhs (fn_code))
+ return false;
switch (fn_code)
{
/* Flavors of vec_add. We deliberately don't expand
@@ -17169,6 +17193,54 @@ rs6000_gimple_fold_builtin (gimple_stmt_
gsi_replace (gsi, g, true);
return true;
}
+ /* Vector stores. */
+ case ALTIVEC_BUILTIN_STVX_V16QI:
+ case ALTIVEC_BUILTIN_STVX_V8HI:
+ case ALTIVEC_BUILTIN_STVX_V4SI:
+ case ALTIVEC_BUILTIN_STVX_V4SF:
+ case ALTIVEC_BUILTIN_STVX_V2DI:
+ case ALTIVEC_BUILTIN_STVX_V2DF:
+ {
+ /* Do not fold for -maltivec=be on LE targets. */
+ if (VECTOR_ELT_ORDER_BIG && !BYTES_BIG_ENDIAN)
+ return false;
+ arg0 = gimple_call_arg (stmt, 0); /* Value to be stored. */
+ arg1 = gimple_call_arg (stmt, 1); /* Offset. */
+ tree arg2 = gimple_call_arg (stmt, 2); /* Store-to address. */
+ location_t loc = gimple_location (stmt);
+ tree arg0_type = TREE_TYPE (arg0);
+ /* Use ptr_type_node (no TBAA) for the arg2_type.
+ FIXME: (Richard) "A proper fix would be to transition this type as
+ seen from the frontend to GIMPLE, for example in a similar way we
+ do for MEM_REFs by piggy-backing that on an extra argument, a
+ constant zero pointer of the alias pointer type to use (which would
+ also serve as a type indicator of the store itself). I'd use a
+ target specific internal function for this (not sure if we can have
+ those target specific, but I guess if it's folded away then that's
+ fine) and get away with the overload set."
+ */
+ tree arg2_type = ptr_type_node;
+ /* POINTER_PLUS_EXPR wants the offset to be of type 'sizetype'. Create
+ the tree using the value from arg0. The resulting type will match
+ the type of arg2. */
+ gimple_seq stmts = NULL;
+ tree temp_offset = gimple_convert (&stmts, loc, sizetype, arg1);
+ tree temp_addr = gimple_build (&stmts, loc, POINTER_PLUS_EXPR,
+ arg2_type, arg2, temp_offset);
+ /* Mask off any lower bits from the address. */
+ tree aligned_addr = gimple_build (&stmts, loc, BIT_AND_EXPR,
+ arg2_type, temp_addr,
+ build_int_cst (arg2_type, -16));
+ gsi_insert_seq_before (gsi, stmts, GSI_SAME_STMT);
+ /* The desired gimple result should be similar to:
+ MEM[(__vector floatD.1407 *)_1] = vf1D.2697; */
+ gimple *g;
+ g = gimple_build_assign (build2 (MEM_REF, arg0_type, aligned_addr,
+ build_int_cst (arg2_type, 0)), arg0);
+ gimple_set_location (g, loc);
+ gsi_replace (gsi, g, true);
+ return true;
+ }
default:
break;
|
