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
|
// RUN: %clang_cc1 -triple mips-unknown-linux -o - -emit-llvm %s | FileCheck %s -check-prefixes=ALL,O32 -enable-var-scope
// RUN: %clang_cc1 -triple mipsel-unknown-linux -o - -emit-llvm %s | FileCheck %s -check-prefixes=ALL,O32 -enable-var-scope
// RUN: %clang_cc1 -triple mips64-unknown-linux -o - -emit-llvm -target-abi n32 %s | FileCheck %s -check-prefixes=ALL,N32,NEW -enable-var-scope
// RUN: %clang_cc1 -triple mips64-unknown-linux -o - -emit-llvm -target-abi n32 %s | FileCheck %s -check-prefixes=ALL,N32,NEW -enable-var-scope
// RUN: %clang_cc1 -triple mips64-unknown-linux -o - -emit-llvm %s | FileCheck %s -check-prefixes=ALL,N64,NEW -enable-var-scope
// RUN: %clang_cc1 -triple mips64el-unknown-linux -o - -emit-llvm %s | FileCheck %s -check-prefixes=ALL,N64,NEW -enable-var-scope
#include <stdarg.h>
typedef int v4i32 __attribute__ ((__vector_size__ (16)));
int test_i32(char *fmt, ...) {
va_list va;
va_start(va, fmt);
int v = va_arg(va, int);
va_end(va);
return v;
}
// O32-LABEL: define i32 @test_i32(i8*{{.*}} %fmt, ...)
// N32-LABEL: define signext i32 @test_i32(i8*{{.*}} %fmt, ...)
// N64-LABEL: define signext i32 @test_i32(i8*{{.*}} %fmt, ...)
//
// O32: %va = alloca i8*, align [[$PTRALIGN:4]]
// N32: %va = alloca i8*, align [[$PTRALIGN:4]]
// N64: %va = alloca i8*, align [[$PTRALIGN:8]]
// ALL: [[V:%.*]] = alloca i32, align 4
// NEW: [[PROMOTION_TEMP:%.*]] = alloca i32, align 4
//
// ALL: [[VA:%.+]] = bitcast i8** %va to i8*
// ALL: call void @llvm.va_start(i8* [[VA]])
// ALL: [[AP_CUR:%.+]] = load i8*, i8** %va, align [[$PTRALIGN]]
// O32: [[AP_NEXT:%.+]] = getelementptr inbounds i8, i8* [[AP_CUR]], [[$INTPTR_T:i32]] [[$CHUNKSIZE:4]]
// NEW: [[AP_NEXT:%.+]] = getelementptr inbounds i8, i8* [[AP_CUR]], [[$INTPTR_T:i32|i64]] [[$CHUNKSIZE:8]]
//
// ALL: store i8* [[AP_NEXT]], i8** %va, align [[$PTRALIGN]]
//
// O32: [[AP_CAST:%.+]] = bitcast i8* [[AP_CUR]] to [[CHUNK_T:i32]]*
// O32: [[ARG:%.+]] = load i32, i32* [[AP_CAST]], align [[CHUNKALIGN:4]]
//
// N32: [[AP_CAST:%.+]] = bitcast i8* [[AP_CUR]] to [[CHUNK_T:i64]]*
// N32: [[TMP:%.+]] = load i64, i64* [[AP_CAST]], align [[CHUNKALIGN:8]]
// N64: [[AP_CAST:%.+]] = bitcast i8* [[AP_CUR]] to [[CHUNK_T:i64]]*
// N64: [[TMP:%.+]] = load i64, i64* [[AP_CAST]], align [[CHUNKALIGN:8]]
// NEW: [[TMP2:%.+]] = trunc i64 [[TMP]] to i32
// NEW: store i32 [[TMP2]], i32* [[PROMOTION_TEMP]], align 4
// NEW: [[ARG:%.+]] = load i32, i32* [[PROMOTION_TEMP]], align 4
// ALL: store i32 [[ARG]], i32* [[V]], align 4
//
// ALL: [[VA1:%.+]] = bitcast i8** %va to i8*
// ALL: call void @llvm.va_end(i8* [[VA1]])
// ALL: }
long long test_i64(char *fmt, ...) {
va_list va;
va_start(va, fmt);
long long v = va_arg(va, long long);
va_end(va);
return v;
}
// ALL-LABEL: define i64 @test_i64(i8*{{.*}} %fmt, ...)
//
// ALL: %va = alloca i8*, align [[$PTRALIGN]]
// ALL: [[VA:%.+]] = bitcast i8** %va to i8*
// ALL: call void @llvm.va_start(i8* [[VA]])
// ALL: [[AP_CUR:%.+]] = load i8*, i8** %va, align [[$PTRALIGN]]
//
// i64 is 8-byte aligned, while this is within O32's stack alignment there's no
// guarantee that the offset is still 8-byte aligned after earlier reads.
// O32: [[TMP1:%.+]] = ptrtoint i8* [[AP_CUR]] to i32
// O32: [[TMP2:%.+]] = add i32 [[TMP1]], 7
// O32: [[TMP3:%.+]] = and i32 [[TMP2]], -8
// O32: [[AP_CUR:%.+]] = inttoptr i32 [[TMP3]] to i8*
//
// ALL: [[AP_NEXT:%.+]] = getelementptr inbounds i8, i8* [[AP_CUR]], [[$INTPTR_T]] 8
// ALL: store i8* [[AP_NEXT]], i8** %va, align [[$PTRALIGN]]
//
// ALL: [[AP_CAST:%.*]] = bitcast i8* [[AP_CUR]] to i64*
// ALL: [[ARG:%.+]] = load i64, i64* [[AP_CAST]], align 8
//
// ALL: [[VA1:%.+]] = bitcast i8** %va to i8*
// ALL: call void @llvm.va_end(i8* [[VA1]])
// ALL: }
char *test_ptr(char *fmt, ...) {
va_list va;
va_start(va, fmt);
char *v = va_arg(va, char *);
va_end(va);
return v;
}
// ALL-LABEL: define i8* @test_ptr(i8*{{.*}} %fmt, ...)
//
// ALL: %va = alloca i8*, align [[$PTRALIGN]]
// ALL: [[V:%.*]] = alloca i8*, align [[$PTRALIGN]]
// N32: [[AP_CAST:%.+]] = alloca i8*, align 4
// ALL: [[VA:%.+]] = bitcast i8** %va to i8*
// ALL: call void @llvm.va_start(i8* [[VA]])
// ALL: [[AP_CUR:%.+]] = load i8*, i8** %va, align [[$PTRALIGN]]
// ALL: [[AP_NEXT:%.+]] = getelementptr inbounds i8, i8* [[AP_CUR]], [[$INTPTR_T]] [[$CHUNKSIZE]]
// ALL: store i8* [[AP_NEXT]], i8** %va, align [[$PTRALIGN]]
//
// When the chunk size matches the pointer size, this is easy.
// O32: [[AP_CAST:%.+]] = bitcast i8* [[AP_CUR]] to i8**
// N64: [[AP_CAST:%.+]] = bitcast i8* [[AP_CUR]] to i8**
// Otherwise we need a promotion temporary.
// N32: [[TMP1:%.+]] = bitcast i8* [[AP_CUR]] to i64*
// N32: [[TMP2:%.+]] = load i64, i64* [[TMP1]], align 8
// N32: [[TMP3:%.+]] = trunc i64 [[TMP2]] to i32
// N32: [[PTR:%.+]] = inttoptr i32 [[TMP3]] to i8*
// N32: store i8* [[PTR]], i8** [[AP_CAST]], align 4
//
// ALL: [[ARG:%.+]] = load i8*, i8** [[AP_CAST]], align [[$PTRALIGN]]
// ALL: store i8* [[ARG]], i8** [[V]], align [[$PTRALIGN]]
//
// ALL: [[VA1:%.+]] = bitcast i8** %va to i8*
// ALL: call void @llvm.va_end(i8* [[VA1]])
// ALL: }
int test_v4i32(char *fmt, ...) {
va_list va;
va_start(va, fmt);
v4i32 v = va_arg(va, v4i32);
va_end(va);
return v[0];
}
// O32-LABEL: define i32 @test_v4i32(i8*{{.*}} %fmt, ...)
// N32-LABEL: define signext i32 @test_v4i32(i8*{{.*}} %fmt, ...)
// N64-LABEL: define signext i32 @test_v4i32(i8*{{.*}} %fmt, ...)
//
// ALL: %va = alloca i8*, align [[$PTRALIGN]]
// ALL: [[V:%.+]] = alloca <4 x i32>, align 16
// ALL: [[VA1:%.+]] = bitcast i8** %va to i8*
// ALL: call void @llvm.va_start(i8* [[VA1]])
// ALL: [[AP_CUR:%.+]] = load i8*, i8** %va, align [[$PTRALIGN]]
//
// Vectors are 16-byte aligned, however the O32 ABI has a maximum alignment of
// 8-bytes since the base of the stack is 8-byte aligned.
// O32: [[TMP1:%.+]] = ptrtoint i8* [[AP_CUR]] to i32
// O32: [[TMP2:%.+]] = add i32 [[TMP1]], 7
// O32: [[TMP3:%.+]] = and i32 [[TMP2]], -8
// O32: [[AP_CUR:%.+]] = inttoptr i32 [[TMP3]] to i8*
//
// NEW: [[TMP1:%.+]] = ptrtoint i8* [[AP_CUR]] to [[$INTPTR_T]]
// NEW: [[TMP2:%.+]] = add [[$INTPTR_T]] [[TMP1]], 15
// NEW: [[TMP3:%.+]] = and [[$INTPTR_T]] [[TMP2]], -16
// NEW: [[AP_CUR:%.+]] = inttoptr [[$INTPTR_T]] [[TMP3]] to i8*
//
// ALL: [[AP_NEXT:%.+]] = getelementptr inbounds i8, i8* [[AP_CUR]], [[$INTPTR_T]] 16
// ALL: store i8* [[AP_NEXT]], i8** %va, align [[$PTRALIGN]]
//
// ALL: [[AP_CAST:%.+]] = bitcast i8* [[AP_CUR]] to <4 x i32>*
// O32: [[ARG:%.+]] = load <4 x i32>, <4 x i32>* [[AP_CAST]], align 8
// N64: [[ARG:%.+]] = load <4 x i32>, <4 x i32>* [[AP_CAST]], align 16
// N32: [[ARG:%.+]] = load <4 x i32>, <4 x i32>* [[AP_CAST]], align 16
// ALL: store <4 x i32> [[ARG]], <4 x i32>* [[V]], align 16
//
// ALL: [[VA1:%.+]] = bitcast i8** %va to i8*
// ALL: call void @llvm.va_end(i8* [[VA1]])
// ALL: [[VECEXT:%.+]] = extractelement <4 x i32> {{.*}}, i32 0
// ALL: ret i32 [[VECEXT]]
// ALL: }
|
