author | Mark J. Nelson <Mark.J.Nelson@Sun.COM> |
Wed, 06 Aug 2008 16:29:39 -0600 | |
changeset 7298 | b69e27387f74 |
parent 6812 | febeba71273d |
permissions | -rw-r--r-- |
0 | 1 |
/* |
2 |
* CDDL HEADER START |
|
3 |
* |
|
4 |
* The contents of this file are subject to the terms of the |
|
6812 | 5 |
* Common Development and Distribution License (the "License"). |
6 |
* You may not use this file except in compliance with the License. |
|
0 | 7 |
* |
8 |
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE |
|
9 |
* or http://www.opensolaris.org/os/licensing. |
|
10 |
* See the License for the specific language governing permissions |
|
11 |
* and limitations under the License. |
|
12 |
* |
|
13 |
* When distributing Covered Code, include this CDDL HEADER in each |
|
14 |
* file and include the License file at usr/src/OPENSOLARIS.LICENSE. |
|
15 |
* If applicable, add the following below this CDDL HEADER, with the |
|
16 |
* fields enclosed by brackets "[]" replaced with your own identifying |
|
17 |
* information: Portions Copyright [yyyy] [name of copyright owner] |
|
18 |
* |
|
19 |
* CDDL HEADER END |
|
20 |
*/ |
|
6812 | 21 |
|
0 | 22 |
/* |
6812 | 23 |
* Copyright 2008 Sun Microsystems, Inc. All rights reserved. |
0 | 24 |
* Use is subject to license terms. |
25 |
*/ |
|
26 |
||
7298
b69e27387f74
6733918 Teamware has retired, please welcome your new manager, Mercurial
Mark J. Nelson <Mark.J.Nelson@Sun.COM>
parents:
6812
diff
changeset
|
27 |
.file "memset.s" |
6812 | 28 |
|
0 | 29 |
/* |
30 |
* char *memset(sp, c, n) |
|
31 |
* |
|
32 |
* Set an array of n chars starting at sp to the character c. |
|
33 |
* Return sp. |
|
34 |
* |
|
35 |
* Fast assembler language version of the following C-program for memset |
|
36 |
* which represents the `standard' for the C-library. |
|
37 |
* |
|
38 |
* void * |
|
39 |
* memset(void *sp1, int c, size_t n) |
|
40 |
* { |
|
41 |
* if (n != 0) { |
|
42 |
* char *sp = sp1; |
|
43 |
* do { |
|
44 |
* *sp++ = (char)c; |
|
45 |
* } while (--n != 0); |
|
46 |
* } |
|
47 |
* return (sp1); |
|
48 |
* } |
|
49 |
*/ |
|
50 |
||
51 |
#include <sys/asm_linkage.h> |
|
52 |
#include <sys/sun4asi.h> |
|
53 |
||
54 |
ANSI_PRAGMA_WEAK(memset,function) |
|
55 |
||
56 |
#define ALIGN8(X) (((X) + 7) & ~7) |
|
57 |
||
58 |
.section ".text" |
|
59 |
.align 32 |
|
60 |
||
61 |
ENTRY(memset) |
|
62 |
cmp %o2, 12 ! if small counts, just write bytes |
|
63 |
bgeu,pn %ncc, .wrbig |
|
64 |
mov %o0, %o5 ! copy sp1 before using it |
|
65 |
||
66 |
.wrchar: |
|
67 |
deccc %o2 ! byte clearing loop |
|
68 |
inc %o5 |
|
69 |
bgeu,a,pt %ncc, .wrchar |
|
70 |
stb %o1, [%o5 + -1] ! we've already incremented the address |
|
71 |
||
72 |
retl |
|
73 |
.empty ! next instruction is safe, %o0 still good |
|
74 |
||
75 |
.wrbig: |
|
76 |
andcc %o5, 7, %o3 ! is sp1 aligned on a 8 byte bound |
|
77 |
bz,pt %ncc, .blkchk ! already double aligned |
|
78 |
and %o1, 0xff, %o1 ! o1 is (char)c |
|
79 |
sub %o3, 8, %o3 ! -(bytes till double aligned) |
|
80 |
add %o2, %o3, %o2 ! update o2 with new count |
|
81 |
||
82 |
! Set -(%o3) bytes till sp1 double aligned |
|
83 |
1: stb %o1, [%o5] ! there is at least 1 byte to set |
|
84 |
inccc %o3 ! byte clearing loop |
|
85 |
bl,pt %ncc, 1b |
|
86 |
inc %o5 |
|
87 |
||
88 |
||
89 |
! Now sp1 is double aligned (sp1 is found in %o5) |
|
90 |
.blkchk: |
|
91 |
sll %o1, 8, %o3 |
|
92 |
or %o1, %o3, %o1 ! now o1 has 2 bytes of c |
|
93 |
||
94 |
sll %o1, 16, %o3 |
|
95 |
or %o1, %o3, %o1 ! now o1 has 4 bytes of c |
|
96 |
||
97 |
cmp %o2, 4095 ! if large count use Block ld/st |
|
98 |
||
99 |
sllx %o1, 32, %o3 |
|
100 |
or %o1, %o3, %o1 ! now o1 has 8 bytes of c |
|
101 |
||
102 |
bgu,a,pn %ncc, .blkwr ! Do block write for large count |
|
103 |
andcc %o5, 63, %o3 ! is sp1 block aligned? |
|
104 |
||
105 |
and %o2, 24, %o3 ! o3 is {0, 8, 16, 24} |
|
106 |
||
107 |
1: subcc %o3, 8, %o3 ! double-word loop |
|
108 |
add %o5, 8, %o5 |
|
109 |
bgeu,a,pt %ncc, 1b |
|
110 |
stx %o1, [%o5 - 8] ! already incremented the address |
|
111 |
||
112 |
andncc %o2, 31, %o4 ! o4 has 32 byte aligned count |
|
113 |
bz,pn %ncc, 3f ! First instruction of icache line |
|
114 |
2: |
|
115 |
subcc %o4, 32, %o4 ! main loop, 32 bytes per iteration |
|
116 |
stx %o1, [%o5 - 8] |
|
117 |
stx %o1, [%o5] |
|
118 |
stx %o1, [%o5 + 8] |
|
119 |
stx %o1, [%o5 + 16] |
|
120 |
bnz,pt %ncc, 2b |
|
121 |
add %o5, 32, %o5 |
|
122 |
||
123 |
3: |
|
124 |
and %o2, 7, %o2 ! o2 has the remaining bytes (<8) |
|
125 |
||
126 |
4: |
|
127 |
deccc %o2 ! byte clearing loop |
|
128 |
inc %o5 |
|
129 |
bgeu,a,pt %ncc, 4b |
|
130 |
stb %o1, [%o5 - 9] ! already incremented the address |
|
131 |
||
132 |
retl |
|
133 |
nop ! %o0 still preserved |
|
134 |
||
135 |
.blkwr: |
|
136 |
bz,pn %ncc, .blalign ! now block aligned |
|
137 |
sub %o3, 64, %o3 ! o3 is -(bytes till block aligned) |
|
138 |
add %o2, %o3, %o2 ! o2 is the remainder |
|
139 |
||
140 |
! Store -(%o3) bytes till dst is block (64 byte) aligned. |
|
141 |
! Use double word stores. |
|
142 |
! Recall that dst is already double word aligned |
|
143 |
1: |
|
144 |
stx %o1, [%o5] |
|
145 |
addcc %o3, 8, %o3 |
|
146 |
bl,pt %ncc, 1b |
|
147 |
add %o5, 8, %o5 |
|
148 |
||
149 |
! sp1 is block aligned |
|
150 |
.blalign: |
|
151 |
rd %fprs, %g1 ! g1 = fprs |
|
152 |
||
153 |
and %o2, 63, %o3 ! calc bytes left after blk store. |
|
154 |
||
155 |
andcc %g1, 0x4, %g1 ! fprs.du = fprs.dl = 0 |
|
156 |
bz,a %ncc, 2f ! Is fprs.fef == 0 |
|
157 |
wr %g0, 0x4, %fprs ! fprs.fef = 1 |
|
158 |
2: |
|
159 |
brnz,pn %o1, 3f ! %o1 is safe to check all 64-bits |
|
160 |
andn %o2, 63, %o4 ! calc size of blocks in bytes |
|
161 |
fzero %d0 |
|
162 |
fzero %d2 |
|
163 |
fzero %d4 |
|
164 |
fzero %d6 |
|
165 |
fmuld %d0, %d0, %d8 |
|
166 |
fzero %d10 |
|
167 |
ba 4f |
|
168 |
fmuld %d0, %d0, %d12 |
|
169 |
||
170 |
3: |
|
171 |
! allocate 8 bytes of scratch space on the stack |
|
172 |
add %sp, -SA(16), %sp |
|
173 |
stx %o1, [%sp + STACK_BIAS + ALIGN8(MINFRAME)] ! move %o1 to %d0 |
|
174 |
ldd [%sp + STACK_BIAS + ALIGN8(MINFRAME)], %d0 |
|
175 |
||
176 |
fmovd %d0, %d2 |
|
177 |
add %sp, SA(16), %sp ! deallocate the scratch space |
|
178 |
fmovd %d0, %d4 |
|
179 |
fmovd %d0, %d6 |
|
180 |
fmovd %d0, %d8 |
|
181 |
fmovd %d0, %d10 |
|
182 |
fmovd %d0, %d12 |
|
183 |
4: |
|
184 |
fmovd %d0, %d14 |
|
185 |
||
186 |
! 1st quadrant has 64 bytes of c |
|
187 |
! instructions 32-byte aligned here |
|
188 |
||
189 |
stda %d0, [%o5]ASI_BLK_P |
|
190 |
subcc %o4, 64, %o4 |
|
191 |
bgu,pt %ncc, 4b |
|
192 |
add %o5, 64, %o5 |
|
193 |
||
194 |
! Set the remaining doubles |
|
195 |
subcc %o3, 8, %o3 ! Can we store any doubles? |
|
196 |
blu,pn %ncc, 6f |
|
197 |
and %o2, 7, %o2 ! calc bytes left after doubles |
|
198 |
||
199 |
5: |
|
200 |
std %d0, [%o5] ! store the doubles |
|
201 |
subcc %o3, 8, %o3 |
|
202 |
bgeu,pt %ncc, 5b |
|
203 |
add %o5, 8, %o5 |
|
204 |
6: |
|
205 |
! Set the remaining bytes |
|
206 |
brz %o2, .exit ! safe to check all 64-bits |
|
207 |
||
208 |
#if 0 |
|
209 |
! Terminate the copy with a partial store. (bug 1200071 does not apply) |
|
210 |
! The data should be at d0 |
|
211 |
dec %o2 ! needed to get the mask right |
|
212 |
edge8n %g0, %o2, %o4 |
|
213 |
stda %d0, [%o5]%o4, ASI_PST8_P |
|
214 |
#else |
|
215 |
7: |
|
216 |
deccc %o2 |
|
217 |
stb %o1, [%o5] |
|
218 |
bgu,pt %ncc, 7b |
|
219 |
inc %o5 |
|
220 |
#endif |
|
221 |
||
222 |
.exit: |
|
223 |
membar #StoreLoad|#StoreStore |
|
224 |
retl ! %o0 was preserved |
|
225 |
wr %g1, %g0, %fprs ! fprs = g1 restore fprs |
|
226 |
||
227 |
SET_SIZE(memset) |