File: aarch64-sve-inline-asm-datatypes.c

package info (click to toggle)
llvm-toolchain-15 1%3A15.0.6-4
  • links: PTS, VCS
  • area: main
  • in suites: bookworm
  • size: 1,554,644 kB
  • sloc: cpp: 5,922,452; ansic: 1,012,136; asm: 674,362; python: 191,568; objc: 73,855; f90: 42,327; lisp: 31,913; pascal: 11,973; javascript: 10,144; sh: 9,421; perl: 7,447; ml: 5,527; awk: 3,523; makefile: 2,520; xml: 885; cs: 573; fortran: 567
file content (252 lines) | stat: -rw-r--r-- 13,932 bytes parent folder | download | duplicates (9)
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
241
242
243
244
245
246
247
248
249
250
251
252
// RUN: %clang_cc1 -triple aarch64-none-linux-gnu -target-feature +sve -fallow-half-arguments-and-returns \
// RUN:   -target-feature +neon -S -O1 -o - -emit-llvm %s | FileCheck %s

// Tests to check that all sve datatypes can be passed in as input operands
// and passed out as output operands.

#define SVINT_TEST(DT, KIND)\
DT func_int_##DT##KIND(DT in)\
{\
  DT out;\
  asm volatile (\
    "ptrue p0.b\n"\
    "mov %[out]." #KIND ", p0/m, %[in]." #KIND "\n"\
    : [out] "=w" (out)\
    : [in] "w" (in)\
    : "p0"\
    );\
  return out;\
}

SVINT_TEST(__SVUint8_t,b);
// CHECK: call <vscale x 16 x i8> asm sideeffect "ptrue p0.b\0Amov $0.b, p0/m, $1.b\0A", "=w,w,~{p0}"(<vscale x 16 x i8> %in)
SVINT_TEST(__SVUint8_t,h);
// CHECK: call <vscale x 16 x i8> asm sideeffect "ptrue p0.b\0Amov $0.h, p0/m, $1.h\0A", "=w,w,~{p0}"(<vscale x 16 x i8> %in)
SVINT_TEST(__SVUint8_t,s);
// CHECK: call <vscale x 16 x i8> asm sideeffect "ptrue p0.b\0Amov $0.s, p0/m, $1.s\0A", "=w,w,~{p0}"(<vscale x 16 x i8> %in)
SVINT_TEST(__SVUint8_t,d);
// CHECK: call <vscale x 16 x i8> asm sideeffect "ptrue p0.b\0Amov $0.d, p0/m, $1.d\0A", "=w,w,~{p0}"(<vscale x 16 x i8> %in)

SVINT_TEST(__SVUint16_t,b);
// CHECK: call <vscale x 8 x i16> asm sideeffect "ptrue p0.b\0Amov $0.b, p0/m, $1.b\0A", "=w,w,~{p0}"(<vscale x 8 x i16> %in)
SVINT_TEST(__SVUint16_t,h);
// CHECK: call <vscale x 8 x i16> asm sideeffect "ptrue p0.b\0Amov $0.h, p0/m, $1.h\0A", "=w,w,~{p0}"(<vscale x 8 x i16> %in)
SVINT_TEST(__SVUint16_t,s);
// CHECK: call <vscale x 8 x i16> asm sideeffect "ptrue p0.b\0Amov $0.s, p0/m, $1.s\0A", "=w,w,~{p0}"(<vscale x 8 x i16> %in)
SVINT_TEST(__SVUint16_t,d);
// CHECK: call <vscale x 8 x i16> asm sideeffect "ptrue p0.b\0Amov $0.d, p0/m, $1.d\0A", "=w,w,~{p0}"(<vscale x 8 x i16> %in)

SVINT_TEST(__SVUint32_t,b);
// CHECK: call <vscale x 4 x i32> asm sideeffect "ptrue p0.b\0Amov $0.b, p0/m, $1.b\0A", "=w,w,~{p0}"(<vscale x 4 x i32> %in)
SVINT_TEST(__SVUint32_t,h);
// CHECK: call <vscale x 4 x i32> asm sideeffect "ptrue p0.b\0Amov $0.h, p0/m, $1.h\0A", "=w,w,~{p0}"(<vscale x 4 x i32> %in)
SVINT_TEST(__SVUint32_t,s);
// CHECK: call <vscale x 4 x i32> asm sideeffect "ptrue p0.b\0Amov $0.s, p0/m, $1.s\0A", "=w,w,~{p0}"(<vscale x 4 x i32> %in)
SVINT_TEST(__SVUint32_t,d);
// CHECK: call <vscale x 4 x i32> asm sideeffect "ptrue p0.b\0Amov $0.d, p0/m, $1.d\0A", "=w,w,~{p0}"(<vscale x 4 x i32> %in)

SVINT_TEST(__SVUint64_t,b);
// CHECK: call <vscale x 2 x i64> asm sideeffect "ptrue p0.b\0Amov $0.b, p0/m, $1.b\0A", "=w,w,~{p0}"(<vscale x 2 x i64> %in)
SVINT_TEST(__SVUint64_t,h);
// CHECK: call <vscale x 2 x i64> asm sideeffect "ptrue p0.b\0Amov $0.h, p0/m, $1.h\0A", "=w,w,~{p0}"(<vscale x 2 x i64> %in)
SVINT_TEST(__SVUint64_t,s);
// CHECK: call <vscale x 2 x i64> asm sideeffect "ptrue p0.b\0Amov $0.s, p0/m, $1.s\0A", "=w,w,~{p0}"(<vscale x 2 x i64> %in)
SVINT_TEST(__SVUint64_t,d);
// CHECK: call <vscale x 2 x i64> asm sideeffect "ptrue p0.b\0Amov $0.d, p0/m, $1.d\0A", "=w,w,~{p0}"(<vscale x 2 x i64> %in)

SVINT_TEST(__SVInt8_t,b);
// CHECK: call <vscale x 16 x i8> asm sideeffect "ptrue p0.b\0Amov $0.b, p0/m, $1.b\0A", "=w,w,~{p0}"(<vscale x 16 x i8> %in)
SVINT_TEST(__SVInt8_t,h);
// CHECK: call <vscale x 16 x i8> asm sideeffect "ptrue p0.b\0Amov $0.h, p0/m, $1.h\0A", "=w,w,~{p0}"(<vscale x 16 x i8> %in)
SVINT_TEST(__SVInt8_t,s);
// CHECK: call <vscale x 16 x i8> asm sideeffect "ptrue p0.b\0Amov $0.s, p0/m, $1.s\0A", "=w,w,~{p0}"(<vscale x 16 x i8> %in)
SVINT_TEST(__SVInt8_t,d);
// CHECK: call <vscale x 16 x i8> asm sideeffect "ptrue p0.b\0Amov $0.d, p0/m, $1.d\0A", "=w,w,~{p0}"(<vscale x 16 x i8> %in)

SVINT_TEST(__SVInt16_t,b);
// CHECK: call <vscale x 8 x i16> asm sideeffect "ptrue p0.b\0Amov $0.b, p0/m, $1.b\0A", "=w,w,~{p0}"(<vscale x 8 x i16> %in)
SVINT_TEST(__SVInt16_t,h);
// CHECK: call <vscale x 8 x i16> asm sideeffect "ptrue p0.b\0Amov $0.h, p0/m, $1.h\0A", "=w,w,~{p0}"(<vscale x 8 x i16> %in)
SVINT_TEST(__SVInt16_t,s);
// CHECK: call <vscale x 8 x i16> asm sideeffect "ptrue p0.b\0Amov $0.s, p0/m, $1.s\0A", "=w,w,~{p0}"(<vscale x 8 x i16> %in)
SVINT_TEST(__SVInt16_t,d);
// CHECK: call <vscale x 8 x i16> asm sideeffect "ptrue p0.b\0Amov $0.d, p0/m, $1.d\0A", "=w,w,~{p0}"(<vscale x 8 x i16> %in)

SVINT_TEST(__SVInt32_t,b);
// CHECK: call <vscale x 4 x i32> asm sideeffect "ptrue p0.b\0Amov $0.b, p0/m, $1.b\0A", "=w,w,~{p0}"(<vscale x 4 x i32> %in)
SVINT_TEST(__SVInt32_t,h);
// CHECK: call <vscale x 4 x i32> asm sideeffect "ptrue p0.b\0Amov $0.h, p0/m, $1.h\0A", "=w,w,~{p0}"(<vscale x 4 x i32> %in)
SVINT_TEST(__SVInt32_t,s);
// CHECK: call <vscale x 4 x i32> asm sideeffect "ptrue p0.b\0Amov $0.s, p0/m, $1.s\0A", "=w,w,~{p0}"(<vscale x 4 x i32> %in)
SVINT_TEST(__SVInt32_t,d);
// CHECK: call <vscale x 4 x i32> asm sideeffect "ptrue p0.b\0Amov $0.d, p0/m, $1.d\0A", "=w,w,~{p0}"(<vscale x 4 x i32> %in)

SVINT_TEST(__SVInt64_t,b);
// CHECK: call <vscale x 2 x i64> asm sideeffect "ptrue p0.b\0Amov $0.b, p0/m, $1.b\0A", "=w,w,~{p0}"(<vscale x 2 x i64> %in)
SVINT_TEST(__SVInt64_t,h);
// CHECK: call <vscale x 2 x i64> asm sideeffect "ptrue p0.b\0Amov $0.h, p0/m, $1.h\0A", "=w,w,~{p0}"(<vscale x 2 x i64> %in)
SVINT_TEST(__SVInt64_t,s);
// CHECK: call <vscale x 2 x i64> asm sideeffect "ptrue p0.b\0Amov $0.s, p0/m, $1.s\0A", "=w,w,~{p0}"(<vscale x 2 x i64> %in)
SVINT_TEST(__SVInt64_t,d);
// CHECK: call <vscale x 2 x i64> asm sideeffect "ptrue p0.b\0Amov $0.d, p0/m, $1.d\0A", "=w,w,~{p0}"(<vscale x 2 x i64> %in)


//Test that floats can also be used as datatypes for integer instructions
//and check all the variants which would not be possible with a float
//instruction
SVINT_TEST(__SVFloat16_t,b);
// CHECK: call <vscale x 8 x half> asm sideeffect "ptrue p0.b\0Amov $0.b, p0/m, $1.b\0A", "=w,w,~{p0}"(<vscale x 8 x half> %in)
SVINT_TEST(__SVFloat16_t,h);
// CHECK: call <vscale x 8 x half> asm sideeffect "ptrue p0.b\0Amov $0.h, p0/m, $1.h\0A", "=w,w,~{p0}"(<vscale x 8 x half> %in)
SVINT_TEST(__SVFloat16_t,s);
// CHECK: call <vscale x 8 x half> asm sideeffect "ptrue p0.b\0Amov $0.s, p0/m, $1.s\0A", "=w,w,~{p0}"(<vscale x 8 x half> %in)
SVINT_TEST(__SVFloat16_t,d);
// CHECK: call <vscale x 8 x half> asm sideeffect "ptrue p0.b\0Amov $0.d, p0/m, $1.d\0A", "=w,w,~{p0}"(<vscale x 8 x half> %in)

SVINT_TEST(__SVFloat32_t,b);
// CHECK: call <vscale x 4 x float> asm sideeffect "ptrue p0.b\0Amov $0.b, p0/m, $1.b\0A", "=w,w,~{p0}"(<vscale x 4 x float> %in)
SVINT_TEST(__SVFloat32_t,h);
// CHECK: call <vscale x 4 x float> asm sideeffect "ptrue p0.b\0Amov $0.h, p0/m, $1.h\0A", "=w,w,~{p0}"(<vscale x 4 x float> %in)
SVINT_TEST(__SVFloat32_t,s);
// CHECK: call <vscale x 4 x float> asm sideeffect "ptrue p0.b\0Amov $0.s, p0/m, $1.s\0A", "=w,w,~{p0}"(<vscale x 4 x float> %in)
SVINT_TEST(__SVFloat32_t,d);
// CHECK: call <vscale x 4 x float> asm sideeffect "ptrue p0.b\0Amov $0.d, p0/m, $1.d\0A", "=w,w,~{p0}"(<vscale x 4 x float> %in)

SVINT_TEST(__SVFloat64_t,b);
// CHECK: call <vscale x 2 x double> asm sideeffect "ptrue p0.b\0Amov $0.b, p0/m, $1.b\0A", "=w,w,~{p0}"(<vscale x 2 x double> %in)
SVINT_TEST(__SVFloat64_t,h);
// CHECK: call <vscale x 2 x double> asm sideeffect "ptrue p0.b\0Amov $0.h, p0/m, $1.h\0A", "=w,w,~{p0}"(<vscale x 2 x double> %in)
SVINT_TEST(__SVFloat64_t,s);
// CHECK: call <vscale x 2 x double> asm sideeffect "ptrue p0.b\0Amov $0.s, p0/m, $1.s\0A", "=w,w,~{p0}"(<vscale x 2 x double> %in)
SVINT_TEST(__SVFloat64_t,d);
// CHECK: call <vscale x 2 x double> asm sideeffect "ptrue p0.b\0Amov $0.d, p0/m, $1.d\0A", "=w,w,~{p0}"(<vscale x 2 x double> %in)


#define SVBOOL_TEST(KIND)\
__SVBool_t func_bool_##KIND(__SVBool_t in1, __SVBool_t in2)\
{\
  __SVBool_t out;\
  asm volatile (\
    "zip1 %[out]." #KIND ", %[in1]." #KIND ", %[in2]." #KIND "\n"\
    : [out] "=Upa" (out)\
    :  [in1] "Upa" (in1),\
      [in2] "Upa" (in2)\
    :);\
  return out;\
}

SVBOOL_TEST(b) ;
// CHECK: call <vscale x 16 x i1> asm sideeffect "zip1 $0.b, $1.b, $2.b\0A", "=@3Upa,@3Upa,@3Upa"(<vscale x 16 x i1> %in1, <vscale x 16 x i1> %in2)
SVBOOL_TEST(h) ;
// CHECK: call <vscale x 16 x i1> asm sideeffect "zip1 $0.h, $1.h, $2.h\0A", "=@3Upa,@3Upa,@3Upa"(<vscale x 16 x i1> %in1, <vscale x 16 x i1> %in2)
SVBOOL_TEST(s) ;
// CHECK: call <vscale x 16 x i1> asm sideeffect "zip1 $0.s, $1.s, $2.s\0A", "=@3Upa,@3Upa,@3Upa"(<vscale x 16 x i1> %in1, <vscale x 16 x i1> %in2)
SVBOOL_TEST(d) ;
// CHECK: call <vscale x 16 x i1> asm sideeffect "zip1 $0.d, $1.d, $2.d\0A", "=@3Upa,@3Upa,@3Upa"(<vscale x 16 x i1> %in1, <vscale x 16 x i1> %in2)


#define SVBOOL_TEST_UPL(DT, KIND)\
__SVBool_t func_bool_upl_##KIND(__SVBool_t in1, DT in2, DT in3)\
{\
  __SVBool_t out;\
  asm volatile (\
    "fadd %[out]." #KIND ", %[in1]." #KIND ", %[in2]." #KIND ", %[in3]." #KIND "\n"\
    : [out] "=w" (out)\
    :  [in1] "Upl" (in1),\
      [in2] "w" (in2),\
      [in3] "w" (in3)\
    :);\
  return out;\
}

SVBOOL_TEST_UPL(__SVInt8_t, b) ;
// CHECK: call <vscale x 16 x i1> asm sideeffect "fadd $0.b, $1.b, $2.b, $3.b\0A", "=w,@3Upl,w,w"(<vscale x 16 x i1> %in1, <vscale x 16 x i8> %in2, <vscale x 16 x i8> %in3)
SVBOOL_TEST_UPL(__SVInt16_t, h) ;
// CHECK: call <vscale x 16 x i1> asm sideeffect "fadd $0.h, $1.h, $2.h, $3.h\0A", "=w,@3Upl,w,w"(<vscale x 16 x i1> %in1, <vscale x 8 x i16> %in2, <vscale x 8 x i16> %in3)
SVBOOL_TEST_UPL(__SVInt32_t, s) ;
// CHECK: call <vscale x 16 x i1> asm sideeffect "fadd $0.s, $1.s, $2.s, $3.s\0A", "=w,@3Upl,w,w"(<vscale x 16 x i1> %in1, <vscale x 4 x i32> %in2, <vscale x 4 x i32> %in3)
SVBOOL_TEST_UPL(__SVInt64_t, d) ;
// CHECK: call <vscale x 16 x i1> asm sideeffect "fadd $0.d, $1.d, $2.d, $3.d\0A", "=w,@3Upl,w,w"(<vscale x 16 x i1> %in1, <vscale x 2 x i64> %in2, <vscale x 2 x i64> %in3)

#define SVFLOAT_TEST(DT,KIND)\
DT func_float_##DT##KIND(DT inout1, DT in2)\
{\
  asm volatile (\
    "ptrue p0." #KIND ", #1 \n"\
    "fsub %[inout1]." #KIND ", p0/m, %[inout1]." #KIND ", %[in2]." #KIND "\n"\
    : [inout1] "=w" (inout1)\
    : "[inout1]" (inout1),\
      [in2] "w" (in2)\
    : "p0");\
  return inout1 ;\
}\

SVFLOAT_TEST(__SVFloat16_t,s);
// CHECK: call <vscale x 8 x half> asm sideeffect "ptrue p0.s, #1 \0Afsub $0.s, p0/m, $0.s, $2.s\0A", "=w,0,w,~{p0}"(<vscale x 8 x half> %inout1, <vscale x 8 x half> %in2)
SVFLOAT_TEST(__SVFloat16_t,d);
// CHECK: call <vscale x 8 x half> asm sideeffect "ptrue p0.d, #1 \0Afsub $0.d, p0/m, $0.d, $2.d\0A", "=w,0,w,~{p0}"(<vscale x 8 x half> %inout1, <vscale x 8 x half> %in2)

SVFLOAT_TEST(__SVFloat32_t,s);
// CHECK: call <vscale x 4 x float> asm sideeffect "ptrue p0.s, #1 \0Afsub $0.s, p0/m, $0.s, $2.s\0A", "=w,0,w,~{p0}"(<vscale x 4 x float> %inout1, <vscale x 4 x float> %in2)
SVFLOAT_TEST(__SVFloat32_t,d);
// CHECK: call <vscale x 4 x float> asm sideeffect "ptrue p0.d, #1 \0Afsub $0.d, p0/m, $0.d, $2.d\0A", "=w,0,w,~{p0}"(<vscale x 4 x float> %inout1, <vscale x 4 x float> %in2)

SVFLOAT_TEST(__SVFloat64_t,s);
// CHECK: call <vscale x 2 x double> asm sideeffect "ptrue p0.s, #1 \0Afsub $0.s, p0/m, $0.s, $2.s\0A", "=w,0,w,~{p0}"(<vscale x 2 x double> %inout1, <vscale x 2 x double> %in2)
SVFLOAT_TEST(__SVFloat64_t,d);
// CHECK: call <vscale x 2 x double> asm sideeffect "ptrue p0.d, #1 \0Afsub $0.d, p0/m, $0.d, $2.d\0A", "=w,0,w,~{p0}"(<vscale x 2 x double> %inout1, <vscale x 2 x double> %in2)

#define SVFLOAT_TEST_Y(DT, KIND)\
__SVBool_t func_float_y_##KIND(DT in1, DT in2)\
{\
  __SVBool_t out;\
  asm volatile (\
    "fmul %[out]." #KIND ", %[in1]." #KIND ", %[in2]." #KIND "\n"\
    : [out] "=w" (out)\
    :  [in1] "w" (in1),\
      [in2] "y" (in2)\
    :);\
  return out;\
}

SVFLOAT_TEST_Y(__SVFloat16_t,h);
// CHECK: call <vscale x 16 x i1> asm sideeffect "fmul $0.h, $1.h, $2.h\0A", "=w,w,y"(<vscale x 8 x half> %in1, <vscale x 8 x half> %in2)
SVFLOAT_TEST_Y(__SVFloat32_t,s);
// CHECK: call <vscale x 16 x i1> asm sideeffect "fmul $0.s, $1.s, $2.s\0A", "=w,w,y"(<vscale x 4 x float> %in1, <vscale x 4 x float> %in2)
SVFLOAT_TEST_Y(__SVFloat64_t,d);
// CHECK: call <vscale x 16 x i1> asm sideeffect "fmul $0.d, $1.d, $2.d\0A", "=w,w,y"(<vscale x 2 x double> %in1, <vscale x 2 x double> %in2)


// Another test for floats to include h suffix

#define SVFLOAT_CVT_TEST(DT1,KIND1,DT2,KIND2)\
DT1 func_float_cvt_##DT1##KIND1##DT2##KIND2(DT2 in1)\
{\
  DT1 out1 ;\
  asm volatile (\
    "ptrue p0." #KIND2 ", #1 \n"\
    "fcvt %[out1]." #KIND1 ", p0/m, %[in1]." #KIND2 "\n"\
    : [out1] "=w" (out1)\
    : [in1] "w" (in1)\
    : "p0");\
  return out1 ;\
}\

SVFLOAT_CVT_TEST(__SVFloat64_t,d,__SVFloat32_t,s);
// CHECK: call <vscale x 2 x double> asm sideeffect "ptrue p0.s, #1 \0Afcvt $0.d, p0/m, $1.s\0A", "=w,w,~{p0}"(<vscale x 4 x float> %in1)
SVFLOAT_CVT_TEST(__SVFloat64_t,d,__SVFloat16_t,h);
// CHECK: call <vscale x 2 x double> asm sideeffect "ptrue p0.h, #1 \0Afcvt $0.d, p0/m, $1.h\0A", "=w,w,~{p0}"(<vscale x 8 x half> %in1)
SVFLOAT_CVT_TEST(__SVFloat32_t,s,__SVFloat16_t,h);
// CHECK: call <vscale x 4 x float> asm sideeffect "ptrue p0.h, #1 \0Afcvt $0.s, p0/m, $1.h\0A", "=w,w,~{p0}"(<vscale x 8 x half> %in1)
SVFLOAT_CVT_TEST(__SVFloat32_t,s,__SVFloat64_t,d);
// CHECK: call <vscale x 4 x float> asm sideeffect "ptrue p0.d, #1 \0Afcvt $0.s, p0/m, $1.d\0A", "=w,w,~{p0}"(<vscale x 2 x double> %in1)
SVFLOAT_CVT_TEST(__SVFloat16_t,h,__SVFloat64_t,d);
// CHECK: call <vscale x 8 x half> asm sideeffect "ptrue p0.d, #1 \0Afcvt $0.h, p0/m, $1.d\0A", "=w,w,~{p0}"(<vscale x 2 x double> %in1)
SVFLOAT_CVT_TEST(__SVFloat16_t,h,__SVFloat32_t,s);
// CHECK: call <vscale x 8 x half> asm sideeffect "ptrue p0.s, #1 \0Afcvt $0.h, p0/m, $1.s\0A", "=w,w,~{p0}"(<vscale x 4 x float> %in1)

//Test a mix of float and ints
SVFLOAT_CVT_TEST(__SVInt16_t,h,__SVFloat32_t,s);
// CHECK: call <vscale x 8 x i16> asm sideeffect "ptrue p0.s, #1 \0Afcvt $0.h, p0/m, $1.s\0A", "=w,w,~{p0}"(<vscale x 4 x float> %in1)
SVFLOAT_CVT_TEST(__SVFloat16_t,s,__SVUint32_t,d);
// CHECK: call <vscale x 8 x half> asm sideeffect "ptrue p0.d, #1 \0Afcvt $0.s, p0/m, $1.d\0A", "=w,w,~{p0}"(<vscale x 4 x i32> %in1)