Debian Codegraph @

gcc/testsuite/gcc.target/aarch64/sve2/acle/asm/mlalb_mf8.c

raw
1 /* { dg-do assemble { target aarch64_asm_fp8fma_ok } } */
2 /* { dg-do compile { target { ! aarch64_asm_fp8fma_ok } } } */
3 /* { dg-final { check-function-bodies "**" "" "-DCHECK_ASM" } } */
4
5 #include "test_sve_acle.h"
6
7 #pragma GCC target "+fp8fma"
8 #ifdef STREAMING_COMPATIBLE
9 #pragma GCC target "+ssve-fp8fma"
10 #endif
11
12 /*
13 ** mlalb_f16_mf8_tied1:
14 ** msr fpmr, x0
15 ** fmlalb z0\.h, z4\.b, z5\.b
16 ** ret
17 */
18 TEST_DUAL_Z (mlalb_f16_mf8_tied1, svfloat16_t, svmfloat8_t,
19 z0 = svmlalb_f16_mf8_fpm (z0, z4, z5, fpm0),
20 z0 = svmlalb_fpm (z0, z4, z5, fpm0))
21
22 /*
23 ** mlalb_f16_mf8_tied2:
24 ** msr fpmr, x0
25 ** mov (z[0-9]+)\.d, z0\.d
26 ** movprfx z0, z4
27 ** fmlalb z0\.h, \1\.b, z1\.b
28 ** ret
29 */
30 TEST_DUAL_Z_REV (mlalb_f16_mf8_tied2, svfloat16_t, svmfloat8_t,
31 z0_res = svmlalb_f16_mf8_fpm (z4, z0, z1, fpm0),
32 z0_res = svmlalb_fpm (z4, z0, z1, fpm0))
33
34 /*
35 ** mlalb_f16_mf8_tied3:
36 ** msr fpmr, x0
37 ** mov (z[0-9]+)\.d, z0\.d
38 ** movprfx z0, z4
39 ** fmlalb z0\.h, z1\.b, \1\.b
40 ** ret
41 */
42 TEST_DUAL_Z_REV (mlalb_f16_mf8_tied3, svfloat16_t, svmfloat8_t,
43 z0_res = svmlalb_f16_mf8_fpm (z4, z1, z0, fpm0),
44 z0_res = svmlalb_fpm (z4, z1, z0, fpm0))
45
46 /*
47 ** mlalb_f16_mf8_untied:
48 ** msr fpmr, x0
49 ** movprfx z0, z1
50 ** fmlalb z0\.h, z4\.b, z5\.b
51 ** ret
52 */
53 TEST_DUAL_Z (mlalb_f16_mf8_untied, svfloat16_t, svmfloat8_t,
54 z0 = svmlalb_f16_mf8_fpm (z1, z4, z5, fpm0),
55 z0 = svmlalb_fpm (z1, z4, z5, fpm0))
56
57 /*
58 ** mlalb_h7_f16_tied1:
59 ** msr fpmr, x0
60 ** mov (z[0-9]+\.b), b7
61 ** fmlalb z0\.h, z4\.b, \1
62 ** ret
63 */
64 TEST_DUAL_ZD (mlalb_h7_f16_tied1, svfloat16_t, svmfloat8_t, mfloat8_t,
65 z0 = svmlalb_n_f16_mf8_fpm (z0, z4, d7, fpm0),
66 z0 = svmlalb_fpm (z0, z4, d7, fpm0))
67
68 /*
69 ** mlalb_h7_f16_untied:
70 ** msr fpmr, x0
71 ** mov (z[0-9]+\.b), b7
72 ** movprfx z0, z1
73 ** fmlalb z0\.h, z4\.b, \1
74 ** ret
75 */
76 TEST_DUAL_ZD (mlalb_h7_f16_untied, svfloat16_t, svmfloat8_t, mfloat8_t,
77 z0 = svmlalb_n_f16_mf8_fpm (z1, z4, d7, fpm0),
78 z0 = svmlalb_fpm (z1, z4, d7, fpm0))