Debian Codegraph @

gcc/testsuite/gcc.target/aarch64/sme2/acle-asm/clamp_bf16_x2.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 - z25\.h}, z0\.h, z5\.h
12 ** ret
13 */
14 TEST_XN_SINGLE (clamp_z24_z24_z0_z5, svbfloat16x2_t, svbfloat16_t, z24,
15 svclamp_single_bf16_x2 (z24, z0, z5),
16 svclamp (z24, z0, z5))
17
18 /*
19 ** clamp_z24_z28_z5_z7:
20 ** (
21 ** mov [^\n]+
22 ** mov [^\n]+
23 ** bfclamp {z24\.h - z25\.h}, z5\.h, z7\.h
24 ** |
25 ** bfclamp {z28\.h - z29\.h}, z5\.h, z7\.h
26 ** mov [^\n]+
27 ** mov [^\n]+
28 ** )
29 ** ret
30 */
31 TEST_XN_SINGLE (clamp_z24_z28_z5_z7, svbfloat16x2_t, svbfloat16_t, z24,
32 svclamp_single_bf16_x2 (z28, z5, z7),
33 svclamp (z28, z5, z7))
34
35 /*
36 ** clamp_z24_z1_z7_z16:
37 ** (
38 ** mov z24\.d, z1\.d
39 ** mov z25\.d, z2\.d
40 ** |
41 ** mov z25\.d, z2\.d
42 ** mov z24\.d, z1\.d
43 ** )
44 ** bfclamp {z24\.h - z25\.h}, z7\.h, z16\.h
45 ** ret
46 */
47 TEST_XN_SINGLE (clamp_z24_z1_z7_z16, svbfloat16x2_t, svbfloat16_t, z24,
48 svclamp_single_bf16_x2 (z1, z7, z16),
49 svclamp (z1, z7, z16))
50
51 /*
52 ** clamp_z1_z24_z16_z23:
53 ** bfclamp {z24\.h - z25\.h}, z16\.h, z23\.h
54 ** (
55 ** mov z1\.d, z24\.d
56 ** mov z2\.d, z25\.d
57 ** |
58 ** mov z2\.d, z25\.d
59 ** mov z1\.d, z24\.d
60 ** )
61 ** ret
62 */
63 TEST_XN_SINGLE (clamp_z1_z24_z16_z23, svbfloat16x2_t, svbfloat16_t, z1,
64 svclamp_single_bf16_x2 (z24, z16, z23),
65 svclamp (z24, z16, z23))
66
67 /*
68 ** clamp_z1_z1_z23_z0:
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 ** ret
75 */
76 TEST_XN_SINGLE (clamp_z1_z1_z23_z0, svbfloat16x2_t, svbfloat16_t, z1,
77 svclamp_single_bf16_x2 (z1, z23, z0),
78 svclamp (z1, z23, z0))
79
80 /*
81 ** clamp_z18_z18_z0_z23:
82 ** bfclamp {z18\.h - z19\.h}, z0\.h, z23\.h
83 ** ret
84 */
85 TEST_XN_SINGLE (clamp_z18_z18_z0_z23, svbfloat16x2_t, svbfloat16_t, z18,
86 svclamp_single_bf16_x2 (z18, z0, z23),
87 svclamp (z18, z0, z23))
88
89 /*
90 ** clamp_awkward:
91 ** ...
92 ** bfclamp {z[0-9]+\.h - z[0-9]+\.h}, z[0-9]+\.h, z3\.h
93 ** ...
94 ** ret
95 */
96 TEST_XN_SINGLE_AWKWARD (clamp_awkward, svbfloat16x2_t, svbfloat16_t,
97 z0_res = svclamp_single_bf16_x2 (z1, z0, zn),
98 z0_res = svclamp (z1, z0, zn))