1 /* { dg-do compile } */
2 /* { dg-options "-O2 -ftree-vectorize -msve-vector-bits=scalable" } */
3
4 #include <stdint.h>
5
6 #define LOOP(TYPE, NAME, STEP) \
7 __attribute__((noinline, noclone)) \
8 void \
9 test_##TYPE##_##NAME (TYPE *dst, TYPE base, int count) \
10 { \
11 for (int i = 0; i < count; ++i, base += STEP) \
12 dst[i] += base; \
13 }
14
15 #define TEST_TYPE(T, TYPE) \
16 T (TYPE, m17, -17) \
17 T (TYPE, m16, -16) \
18 T (TYPE, m15, -15) \
19 T (TYPE, m1, -1) \
20 T (TYPE, 1, 1) \
21 T (TYPE, 15, 15) \
22 T (TYPE, 16, 16) \
23 T (TYPE, 17, 17)
24
25 #define TEST_ALL(T) \
26 TEST_TYPE (T, int8_t) \
27 TEST_TYPE (T, int16_t) \
28 TEST_TYPE (T, int32_t) \
29 TEST_TYPE (T, int64_t)
30
31 TEST_ALL (LOOP)
32
33 /* { dg-final { scan-assembler-times {\tindex\tz[0-9]+\.b, w[0-9]+, #-16\n} 1 } } */
34 /* { dg-final { scan-assembler-times {\tindex\tz[0-9]+\.b, w[0-9]+, #-15\n} 1 } } */
35 /* { dg-final { scan-assembler-times {\tindex\tz[0-9]+\.b, w[0-9]+, #1\n} 1 } } */
36 /* { dg-final { scan-assembler-times {\tindex\tz[0-9]+\.b, w[0-9]+, #15\n} 1 } } */
37 /* { dg-final { scan-assembler-times {\tindex\tz[0-9]+\.b, w[0-9]+, w[0-9]+\n} 3 } } */
38 /* { dg-final { scan-assembler-times {\tld1b\tz[0-9]+\.b, p[0-7]+/z, \[x[0-9]+, x[0-9]+\]} 8 } } */
39 /* { dg-final { scan-assembler-times {\tst1b\tz[0-9]+\.b, p[0-7]+, \[x[0-9]+, x[0-9]+\]} 8 } } */
40 /* { dg-final { scan-assembler-times {\tincb\tx[0-9]+\n} 8 } } */
41
42 /* { dg-final { scan-assembler-not {\tdecb\tz[0-9]+\.b} } } */
43 /* We don't need to increment the vector IV for steps -16 and 16, since the
44 increment is always a multiple of 256. */
45 /* { dg-final { scan-assembler-times {\tadd\tz[0-9]+\.b, z[0-9]+\.b, z[0-9]+\.b\n} 14 } } */
46
47 /* { dg-final { scan-assembler-times {\tindex\tz[0-9]+\.h, w[0-9]+, #-16\n} 1 } } */
48 /* { dg-final { scan-assembler-times {\tindex\tz[0-9]+\.h, w[0-9]+, #-15\n} 1 } } */
49 /* { dg-final { scan-assembler-times {\tindex\tz[0-9]+\.h, w[0-9]+, #1\n} 1 } } */
50 /* { dg-final { scan-assembler-times {\tindex\tz[0-9]+\.h, w[0-9]+, #15\n} 1 } } */
51 /* { dg-final { scan-assembler-times {\tindex\tz[0-9]+\.h, w[0-9]+, w[0-9]+\n} 3 } } */
52 /* { dg-final { scan-assembler-times {\tld1h\tz[0-9]+\.h, p[0-7]+/z, \[x[0-9]+, x[0-9]+, lsl 1\]} 8 } } */
53 /* { dg-final { scan-assembler-times {\tst1h\tz[0-9]+\.h, p[0-7]+, \[x[0-9]+, x[0-9]+, lsl 1\]} 8 } } */
54 /* { dg-final { scan-assembler-times {\tincb\tx[0-9]+\n} 8 } } */
55
56 /* { dg-final { scan-assembler-times {\tdech\tz[0-9]+\.h, all, mul #16\n} 1 } } */
57 /* { dg-final { scan-assembler-times {\tdech\tz[0-9]+\.h, all, mul #15\n} 1 } } */
58 /* { dg-final { scan-assembler-times {\tdech\tz[0-9]+\.h\n} 1 } } */
59 /* { dg-final { scan-assembler-times {\tinch\tz[0-9]+\.h\n} 1 } } */
60 /* { dg-final { scan-assembler-times {\tinch\tz[0-9]+\.h, all, mul #15\n} 1 } } */
61 /* { dg-final { scan-assembler-times {\tinch\tz[0-9]+\.h, all, mul #16\n} 1 } } */
62 /* { dg-final { scan-assembler-times {\tadd\tz[0-9]+\.h, z[0-9]+\.h, z[0-9]+\.h\n} 10 } } */
63
64 /* { dg-final { scan-assembler-times {\tindex\tz[0-9]+\.s, w[0-9]+, #-16\n} 1 } } */
65 /* { dg-final { scan-assembler-times {\tindex\tz[0-9]+\.s, w[0-9]+, #-15\n} 1 } } */
66 /* { dg-final { scan-assembler-times {\tindex\tz[0-9]+\.s, w[0-9]+, #1\n} 1 } } */
67 /* { dg-final { scan-assembler-times {\tindex\tz[0-9]+\.s, w[0-9]+, #15\n} 1 } } */
68 /* { dg-final { scan-assembler-times {\tindex\tz[0-9]+\.s, w[0-9]+, w[0-9]+\n} 3 } } */
69 /* { dg-final { scan-assembler-times {\tld1w\tz[0-9]+\.s, p[0-7]+/z, \[x[0-9]+, x[0-9]+, lsl 2\]} 8 } } */
70 /* { dg-final { scan-assembler-times {\tst1w\tz[0-9]+\.s, p[0-7]+, \[x[0-9]+, x[0-9]+, lsl 2\]} 8 } } */
71 /* 2 for the calculations of -17 and 17. */
72 /* { dg-final { scan-assembler-times {\tincw\tx[0-9]+\n} 10 } } */
73
74 /* { dg-final { scan-assembler-times {\tdecw\tz[0-9]+\.s, all, mul #16\n} 1 } } */
75 /* { dg-final { scan-assembler-times {\tdecw\tz[0-9]+\.s, all, mul #15\n} 1 } } */
76 /* { dg-final { scan-assembler-times {\tdecw\tz[0-9]+\.s\n} 1 } } */
77 /* { dg-final { scan-assembler-times {\tincw\tz[0-9]+\.s\n} 1 } } */
78 /* { dg-final { scan-assembler-times {\tincw\tz[0-9]+\.s, all, mul #15\n} 1 } } */
79 /* { dg-final { scan-assembler-times {\tincw\tz[0-9]+\.s, all, mul #16\n} 1 } } */
80 /* { dg-final { scan-assembler-times {\tadd\tz[0-9]+\.s, z[0-9]+\.s, z[0-9]+\.s\n} 10 } } */
81
82 /* { dg-final { scan-assembler-times {\tindex\tz[0-9]+\.d, x[0-9]+, #-16\n} 1 } } */
83 /* { dg-final { scan-assembler-times {\tindex\tz[0-9]+\.d, x[0-9]+, #-15\n} 1 } } */
84 /* { dg-final { scan-assembler-times {\tindex\tz[0-9]+\.d, x[0-9]+, #1\n} 1 } } */
85 /* { dg-final { scan-assembler-times {\tindex\tz[0-9]+\.d, x[0-9]+, #15\n} 1 } } */
86 /* { dg-final { scan-assembler-times {\tindex\tz[0-9]+\.d, x[0-9]+, x[0-9]+\n} 3 } } */
87 /* { dg-final { scan-assembler-times {\tld1d\tz[0-9]+\.d, p[0-7]+/z, \[x[0-9]+, x[0-9]+, lsl 3\]} 8 } } */
88 /* { dg-final { scan-assembler-times {\tst1d\tz[0-9]+\.d, p[0-7]+, \[x[0-9]+, x[0-9]+, lsl 3\]} 8 } } */
89 /* 2 for the calculations of -17 and 17. */
90 /* { dg-final { scan-assembler-times {\tincd\tx[0-9]+\n} 10 } } */
91
92 /* { dg-final { scan-assembler-times {\tdecd\tz[0-9]+\.d, all, mul #16\n} 1 } } */
93 /* { dg-final { scan-assembler-times {\tdecd\tz[0-9]+\.d, all, mul #15\n} 1 } } */
94 /* { dg-final { scan-assembler-times {\tdecd\tz[0-9]+\.d\n} 1 } } */
95 /* { dg-final { scan-assembler-times {\tincd\tz[0-9]+\.d\n} 1 } } */
96 /* { dg-final { scan-assembler-times {\tincd\tz[0-9]+\.d, all, mul #15\n} 1 } } */
97 /* { dg-final { scan-assembler-times {\tincd\tz[0-9]+\.d, all, mul #16\n} 1 } } */
98 /* { dg-final { scan-assembler-times {\tadd\tz[0-9]+\.d, z[0-9]+\.d, z[0-9]+\.d\n} 10 } } */