| 1 |
/* { dg-do assemble { target aarch64_asm_sme-b16b16_ok } } */ |
| 2 |
/* { dg-do compile { target { ! aarch64_asm_sme-b16b16_ok } } } */ |
| 3 |
/* { dg-final { check-function-bodies "**" "" "-DCHECK_ASM" } } */ |
| 4 |
|
| 5 |
#include "test_sme2_acle.h" |
| 6 |
|
| 7 |
#pragma GCC target "+sme-b16b16" |
| 8 |
|
| 9 |
/* |
| 10 |
** sub_0_z0: |
| 11 |
** mov (w8|w9|w10|w11), #?0 |
| 12 |
** bfsub za\.h\[\1, 0, vgx4\], {z0\.h - z3\.h} |
| 13 |
** ret |
| 14 |
*/ |
| 15 |
TEST_ZA_XN (sub_0_z0, svbfloat16x4_t, |
| 16 |
svsub_za16_bf16_vg1x4 (0, z0), |
| 17 |
svsub_za16_vg1x4 (0, z0)) |
| 18 |
|
| 19 |
/* |
| 20 |
** sub_w0_z0: |
| 21 |
** mov (w8|w9|w10|w11), w0 |
| 22 |
** bfsub za\.h\[\1, 0, vgx4\], {z0\.h - z3\.h} |
| 23 |
** ret |
| 24 |
*/ |
| 25 |
TEST_ZA_XN (sub_w0_z0, svbfloat16x4_t, |
| 26 |
svsub_za16_bf16_vg1x4 (w0, z0), |
| 27 |
svsub_za16_vg1x4 (w0, z0)) |
| 28 |
|
| 29 |
/* |
| 30 |
** sub_w7_z0: |
| 31 |
** mov (w8|w9|w10|w11), w7 |
| 32 |
** bfsub za\.h\[\1, 0, vgx4\], {z0\.h - z3\.h} |
| 33 |
** ret |
| 34 |
*/ |
| 35 |
TEST_ZA_XN (sub_w7_z0, svbfloat16x4_t, |
| 36 |
svsub_za16_bf16_vg1x4 (w7, z0), |
| 37 |
svsub_za16_vg1x4 (w7, z0)) |
| 38 |
|
| 39 |
/* |
| 40 |
** sub_w8_z0: |
| 41 |
** bfsub za\.h\[w8, 0, vgx4\], {z0\.h - z3\.h} |
| 42 |
** ret |
| 43 |
*/ |
| 44 |
TEST_ZA_XN (sub_w8_z0, svbfloat16x4_t, |
| 45 |
svsub_za16_bf16_vg1x4 (w8, z0), |
| 46 |
svsub_za16_vg1x4 (w8, z0)) |
| 47 |
|
| 48 |
/* |
| 49 |
** sub_w11_z0: |
| 50 |
** bfsub za\.h\[w11, 0, vgx4\], {z0\.h - z3\.h} |
| 51 |
** ret |
| 52 |
*/ |
| 53 |
TEST_ZA_XN (sub_w11_z0, svbfloat16x4_t, |
| 54 |
svsub_za16_bf16_vg1x4 (w11, z0), |
| 55 |
svsub_za16_vg1x4 (w11, z0)) |
| 56 |
|
| 57 |
|
| 58 |
/* |
| 59 |
** sub_w12_z0: |
| 60 |
** mov (w8|w9|w10|w11), w12 |
| 61 |
** bfsub za\.h\[\1, 0, vgx4\], {z0\.h - z3\.h} |
| 62 |
** ret |
| 63 |
*/ |
| 64 |
TEST_ZA_XN (sub_w12_z0, svbfloat16x4_t, |
| 65 |
svsub_za16_bf16_vg1x4 (w12, z0), |
| 66 |
svsub_za16_vg1x4 (w12, z0)) |
| 67 |
|
| 68 |
/* |
| 69 |
** sub_w8p7_z0: |
| 70 |
** bfsub za\.h\[w8, 7, vgx4\], {z0\.h - z3\.h} |
| 71 |
** ret |
| 72 |
*/ |
| 73 |
TEST_ZA_XN (sub_w8p7_z0, svbfloat16x4_t, |
| 74 |
svsub_za16_bf16_vg1x4 (w8 + 7, z0), |
| 75 |
svsub_za16_vg1x4 (w8 + 7, z0)) |
| 76 |
|
| 77 |
/* |
| 78 |
** sub_w8p8_z0: |
| 79 |
** add (w8|w9|w10|w11), w8, #?8 |
| 80 |
** bfsub za\.h\[\1, 0, vgx4\], {z0\.h - z3\.h} |
| 81 |
** ret |
| 82 |
*/ |
| 83 |
TEST_ZA_XN (sub_w8p8_z0, svbfloat16x4_t, |
| 84 |
svsub_za16_bf16_vg1x4 (w8 + 8, z0), |
| 85 |
svsub_za16_vg1x4 (w8 + 8, z0)) |
| 86 |
|
| 87 |
/* |
| 88 |
** sub_w8m1_z0: |
| 89 |
** sub (w8|w9|w10|w11), w8, #?1 |
| 90 |
** bfsub za\.h\[\1, 0, vgx4\], {z0\.h - z3\.h} |
| 91 |
** ret |
| 92 |
*/ |
| 93 |
TEST_ZA_XN (sub_w8m1_z0, svbfloat16x4_t, |
| 94 |
svsub_za16_bf16_vg1x4 (w8 - 1, z0), |
| 95 |
svsub_za16_vg1x4 (w8 - 1, z0)) |
| 96 |
|
| 97 |
/* |
| 98 |
** sub_w8_z4: |
| 99 |
** bfsub za\.h\[w8, 0, vgx4\], {z4\.h - z7\.h} |
| 100 |
** ret |
| 101 |
*/ |
| 102 |
TEST_ZA_XN (sub_w8_z4, svbfloat16x4_t, |
| 103 |
svsub_za16_bf16_vg1x4 (w8, z4), |
| 104 |
svsub_za16_vg1x4 (w8, z4)) |
| 105 |
|
| 106 |
/* Leave the assembler to check for correctness for misaligned registers. */ |
| 107 |
|
| 108 |
/* |
| 109 |
** sub_w8_z18: |
| 110 |
** mov [^\n]+ |
| 111 |
** mov [^\n]+ |
| 112 |
** mov [^\n]+ |
| 113 |
** mov [^\n]+ |
| 114 |
** bfsub za\.h\[w8, 0, vgx4\], [^\n]+ |
| 115 |
** ret |
| 116 |
*/ |
| 117 |
TEST_ZA_XN (sub_w8_z18, svbfloat16x4_t, |
| 118 |
svsub_za16_bf16_vg1x4 (w8, z18), |
| 119 |
svsub_za16_vg1x4 (w8, z18)) |
| 120 |
|
| 121 |
/* |
| 122 |
** sub_w8_z23: |
| 123 |
** mov [^\n]+ |
| 124 |
** mov [^\n]+ |
| 125 |
** mov [^\n]+ |
| 126 |
** mov [^\n]+ |
| 127 |
** bfsub za\.h\[w8, 0, vgx4\], [^\n]+ |
| 128 |
** ret |
| 129 |
*/ |
| 130 |
TEST_ZA_XN (sub_w8_z23, svbfloat16x4_t, |
| 131 |
svsub_za16_bf16_vg1x4 (w8, z23), |
| 132 |
svsub_za16_vg1x4 (w8, z23)) |
| 133 |
|
| 134 |
/* |
| 135 |
** sub_w8_z28: |
| 136 |
** bfsub za\.h\[w8, 0, vgx4\], {z28\.h - z31\.h} |
| 137 |
** ret |
| 138 |
*/ |
| 139 |
TEST_ZA_XN (sub_w8_z28, svbfloat16x4_t, |
| 140 |
svsub_za16_bf16_vg1x4 (w8, z28), |
| 141 |
svsub_za16_vg1x4 (w8, z28)) |