Debian Codegraph @

gcc/testsuite/gcc.target/aarch64/sme2/acle-asm/clamp_bf16_x4.c

raw
1 /* { dg-do assemble { target aarch64_asm_sve-b16b16_ok } } */
2 /* { dg-do compile { target { ! aarch64_asm_sve-b16b16_ok } } } */
3 /* { dg-final { check-function-bodies "**" "" "-DCHECK_ASM" } } */
4
5 #include "test_sme2_acle.h"
6
7 #pragma GCC target "+sve-b16b16"
8
9 /*
10 ** clamp_z24_z24_z0_z5:
11 ** bfclamp {z24\.h - z27\.h}, z0\.h, z5\.h
12 ** ret
13 */
14 TEST_XN_SINGLE (clamp_z24_z24_z0_z5, svbfloat16x4_t, svbfloat16_t, z24,
15 svclamp_single_bf16_x4 (z24, z0, z5),
16 svclamp (z24, z0, z5))
17
18 /*
19 ** clamp_z24_z28_z5_z7:
20 ** (
21 ** mov [^\n]+
22 ** mov [^\n]+
23 ** mov [^\n]+
24 ** mov [^\n]+
25 ** bfclamp {z24\.h - z27\.h}, z5\.h, z7\.h
26 ** |
27 ** bfclamp {z28\.h - z31\.h}, z5\.h, z7\.h
28 ** mov [^\n]+
29 ** mov [^\n]+
30 ** mov [^\n]+
31 ** mov [^\n]+
32 ** )
33 ** ret
34 */
35 TEST_XN_SINGLE (clamp_z24_z28_z5_z7, svbfloat16x4_t, svbfloat16_t, z24,
36 svclamp_single_bf16_x4 (z28, z5, z7),
37 svclamp (z28, z5, z7))
38
39 /*
40 ** clamp_z24_z1_z7_z16:
41 ** mov [^\n]+
42 ** mov [^\n]+
43 ** mov [^\n]+
44 ** mov [^\n]+
45 ** bfclamp {z24\.h - z27\.h}, z7\.h, z16\.h
46 ** ret
47 */
48 TEST_XN_SINGLE (clamp_z24_z1_z7_z16, svbfloat16x4_t, svbfloat16_t, z24,
49 svclamp_single_bf16_x4 (z1, z7, z16),
50 svclamp (z1, z7, z16))
51
52 /*
53 ** clamp_z1_z24_z16_z23:
54 ** bfclamp {z24\.h - z27\.h}, z16\.h, z23\.h
55 ** mov [^\n]+
56 ** mov [^\n]+
57 ** mov [^\n]+
58 ** mov [^\n]+
59 ** ret
60 */
61 TEST_XN_SINGLE (clamp_z1_z24_z16_z23, svbfloat16x4_t, svbfloat16_t, z1,
62 svclamp_single_bf16_x4 (z24, z16, z23),
63 svclamp (z24, z16, z23))
64
65 /*
66 ** clamp_z1_z1_z23_z0:
67 ** mov [^\n]+
68 ** mov [^\n]+
69 ** mov [^\n]+
70 ** mov [^\n]+
71 ** bfclamp {z[0-9]+\.h - z[0-9]+\.h}, z23\.h, z0\.h
72 ** mov [^\n]+
73 ** mov [^\n]+
74 ** mov [^\n]+
75 ** mov [^\n]+
76 ** ret
77 */
78 TEST_XN_SINGLE (clamp_z1_z1_z23_z0, svbfloat16x4_t, svbfloat16_t, z1,
79 svclamp_single_bf16_x4 (z1, z23, z0),
80 svclamp (z1, z23, z0))
81
82 /*
83 ** clamp_z18_z18_z16_z5:
84 ** mov [^\n]+
85 ** mov [^\n]+
86 ** mov [^\n]+
87 ** mov [^\n]+
88 ** bfclamp {z[0-9]+\.h - z[0-9]+\.h}, z16\.h, z5\.h
89 ** mov [^\n]+
90 ** mov [^\n]+
91 ** mov [^\n]+
92 ** mov [^\n]+
93 ** ret
94 */
95 TEST_XN_SINGLE (clamp_z18_z18_z16_z5, svbfloat16x4_t, svbfloat16_t, z18,
96 svclamp_single_bf16_x4 (z18, z16, z5),
97 svclamp (z18, z16, z5))
98
99 /*
100 ** clamp_awkward:
101 ** ...
102 ** bfclamp {z[0-9]+\.h - z[0-9]+\.h}, z[0-9]+\.h, z5\.h
103 ** ...
104 ** ret
105 */
106 TEST_XN_SINGLE_AWKWARD (clamp_awkward, svbfloat16x4_t, svbfloat16_t,
107 z0_res = svclamp_single_bf16_x4 (z1, z0, zn),
108 z0_res = svclamp (z1, z0, zn))