1 | /* Optimized memset for Huawei Kunpeng processor. |
2 | Copyright (C) 2012-2024 Free Software Foundation, Inc. |
3 | |
4 | This file is part of the GNU C Library. |
5 | |
6 | The GNU C Library is free software; you can redistribute it and/or |
7 | modify it under the terms of the GNU Lesser General Public |
8 | License as published by the Free Software Foundation; either |
9 | version 2.1 of the License, or (at your option) any later version. |
10 | |
11 | The GNU C Library is distributed in the hope that it will be useful, |
12 | but WITHOUT ANY WARRANTY; without even the implied warranty of |
13 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
14 | Lesser General Public License for more details. |
15 | |
16 | You should have received a copy of the GNU Lesser General Public |
17 | License along with the GNU C Library. If not, see |
18 | <https://www.gnu.org/licenses/>. */ |
19 | |
20 | #include <sysdep.h> |
21 | #include <sysdeps/aarch64/memset-reg.h> |
22 | |
23 | /* Assumptions: |
24 | * |
25 | * ARMv8-a, AArch64, unaligned accesses |
26 | * |
27 | */ |
28 | |
29 | ENTRY (__memset_kunpeng) |
30 | |
31 | PTR_ARG (0) |
32 | SIZE_ARG (2) |
33 | |
34 | dup v0.16B, valw |
35 | add dstend, dstin, count |
36 | |
37 | cmp count, 128 |
38 | b.hs L(set_long) |
39 | |
40 | cmp count, 16 |
41 | b.lo L(less16) |
42 | |
43 | /* Set 16..127 bytes. */ |
44 | str q0, [dstin] |
45 | tbnz count, 6, L(set127) |
46 | str q0, [dstend, -16] |
47 | tbz count, 5, 1f |
48 | str q0, [dstin, 16] |
49 | str q0, [dstend, -32] |
50 | 1: ret |
51 | |
52 | .p2align 4 |
53 | /* Set 64..127 bytes. Write 64 bytes from the start and |
54 | 64 bytes from the end. */ |
55 | L(set127): |
56 | stp q0, q0, [dstin, 16] |
57 | str q0, [dstin, 48] |
58 | stp q0, q0, [dstend, -64] |
59 | stp q0, q0, [dstend, -32] |
60 | ret |
61 | |
62 | .p2align 4 |
63 | /* Set 0..15 bytes. */ |
64 | L(less16): |
65 | tbz count, 3, L(less8) |
66 | str d0, [dstin] |
67 | str d0, [dstend, -8] |
68 | ret |
69 | L(less8): |
70 | tbz count, 2, 2f |
71 | str s0, [dstin] |
72 | str s0, [dstend, -4] |
73 | ret |
74 | 2: cbz count, 3f |
75 | str b0, [dstin] |
76 | tbz count, 1, 3f |
77 | str h0, [dstend, -2] |
78 | 3: ret |
79 | |
80 | .p2align 4 |
81 | L(set_long): |
82 | bic dst, dstin, 15 |
83 | str q0, [dstin] |
84 | sub count, dstend, dst /* Count is 16 too large. */ |
85 | sub dst, dst, 16 /* Dst is biased by -32. */ |
86 | sub count, count, 64 + 16 + 1 /* Adjust count and bias for loop. */ |
87 | 1: stp q0, q0, [dst, 32] |
88 | stp q0, q0, [dst, 64]! |
89 | subs count, count, 64 |
90 | b.lo 1f |
91 | stp q0, q0, [dst, 32] |
92 | stp q0, q0, [dst, 64]! |
93 | subs count, count, 64 |
94 | b.lo 1f |
95 | stp q0, q0, [dst, 32] |
96 | stp q0, q0, [dst, 64]! |
97 | subs count, count, 64 |
98 | b.lo 1f |
99 | stp q0, q0, [dst, 32] |
100 | stp q0, q0, [dst, 64]! |
101 | subs count, count, 64 |
102 | b.hs 1b |
103 | |
104 | 1: stp q0, q0, [dstend, -64] |
105 | stp q0, q0, [dstend, -32] |
106 | ret |
107 | |
108 | END (__memset_kunpeng) |
109 | |