1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
|
#!/bin/sh
#
# If your shell doesn't support functions (true for some BSD users),
# you might try using GNU's bash.
#
#ident "@(#) m88k-move.sh 1-Sep-92"
#
# This file provided by Data General, February 1990.
#
# This script generates the necessary movstr library functions
# for the m88100. These functions are called from the expansion
# of movstrsi. There are eight modules created by this script,
# each with multiple entry points. One module, moveSI64n
# implements a word aligned loop; the other modules, moveXINx
# implement a straight line copy of N bytes in mode XI.
#
# By analysis of the best memcpy function, it can be determined
# what appear to be certain magic numbers. For example, a
# memcpy of 13 bytes, where the pointers are determined at run
# time to be word aligned takes 28 cycles. A call to
# __movstrQI13x13 also takes 28 cycles. The break even point
# for a HImode copy is 38 bytes. Just to be on the safe side,
# these are bumped to 16 and 48 respectively.
#
# The smaller, odd-remainder modules are provided to help
# mitigate the overhead of copying the last bytes.
#
# Changes to these functions should not be taken lightly if you
# want to be able to link programs built with older movstr
# parameters.
#
#.Revision History
#
# 1-Sep-92 Stan Cox Added moveDI96x, moveDI41x through moveDI47x.
# 2-Jan-92 Tom Wood Renamed files to comply with SVR3 14 char limit.
# 26-Oct-90 Tom Wood Delete movstr.h; moved to out-m88k.c.
# 17-Oct-90 Tom Wood Files are named *.asm rather than *.s.
# 11-Sep-90 Jeffrey Friedl
# On my BSD 4.3 awk and my GNU-awk, only the
# first character of an argument to -F is passed
# through, so I can't get this to work.
# 5-Sep-90 Ray Essick/Tom Wood
# Added a -no-tdesc option.
# 27-Aug-90 Vince Guarna/Tom Wood
# Version 3 assembler syntax (-abi).
# 16-Aug-90 Ron Guilmette
# Avoid problems on a Sparc. The common
# denominator among shells seems to be '...\'
# rather than '...\\'.
# 15-Aug-90 Ron Guilmette
# Avoid awk syntax errors on a Sun by not using
# the `!' operator.
# 22-Feb-90 Tom Wood Created.
# 20-Jun-90 Tom Wood Emit file directives.
#
#.End]=--------------------------------------------------------------*/
usage() {
echo "usage: $0 [ -abi ] [ -no-tdesc ]" 1>&2
exit 1
}
awk_flag="-F:";
awk_begin="BEGIN { "
ps=""; us="_"; tf="x"; la="@L"; fb="8"; nt="";
do_file() {
echo " file $1";
}
while [ $# -gt 0 ] ; do
case $1 in
-no-tdesc) awk_begin="$awk_begin no_tdesc=1;"
nt=";";;
-abi) awk_begin="$awk_begin abi=1;"
ps="#"; us=""; tf="a"; la=".L"; fb="16";
do_file() {
echo ' version "03.00"';
echo " file $1";
};;
*) usage;;
esac
shift
done
rm -f move?I*[xn].s move?I*[xn].asm
#.Implementation_continued[=-----------------------------------------------
#
# This generates the word aligned loop. The loop is entered
# from the callable entry points ___movstrSI64nN, where at
# least N bytes will be copied. r2 is the destination pointer
# offset by 4, r3 is the source pointer offset by 4, r6 is the
# loop count. Thus, the total bytes moved is 64 * r6 + N. The
# first value is is preloaded into r4 or r5 (r4 if N/4 is odd;
# r5 if N/4 is even). Upon returning, r2 and r3 have been
# updated and may be used for the remainder bytes to move.
#
# The code for this loop is generated by the awk program
# following. Edit *it*, not what it generates!
#
#.End]=------------------------------------------------------------------*/
gen_movstrN() {
awk $awk_flag "$awk_begin"'
if (abi) {
ps="#"; us=""; tf="a"; la=".L"; fb=16;
} else {
ps=""; us="_"; tf="x"; la="@L"; fb=8;
}
}
NR == 1 && NF == 4 {
mode = $1; suffix = $2; align = $3; count = $4;
ld = align; st = 0;
printf "; The following was calculated using awk.\n";
printf "\ttext\n";
printf "\talign\t%d\n", fb;
printf "%sloop%s%d:\n", la, mode, count * align;
printf "\taddu\t%sr3,%sr3,%d\n", ps, ps, count * align;
printf "\taddu\t%sr2,%sr2,%d\n", ps, ps, count * align;
printf "\tsubu\t%sr6,%sr6,1\n", ps, ps;
for (r = count + 1; r >= 1; r--) {
evenp = r % 2;
name = sprintf("__%smovstr%s%dn%d", us, mode, count * align, r * align);
if (r > 1) {
printf "\tglobal\t%s\n", name;
printf "%s:\n", name;
}
if (r > 2) {
printf "\tld%s\t%sr%d,%sr3,%d\n", suffix, ps, 4 + evenp, ps, ld;
printf "\tst%s\t%sr%d,%sr2,%d\n", suffix, ps, 5 - evenp, ps, st;
} else if (r == 2) {
printf "\tld%s\t%sr%d,%sr3,%d\n", suffix, ps, 4 + evenp, ps, ld;
printf "\tbcnd.n\t%sgt0,%sr6,%sloop%s%d\n", ps, ps, la, mode, count * align;
printf "\tst%s\t%sr%d,%sr2,%d\n", suffix, ps, 5 - evenp, ps, st;
printf "\tjmp.n\t%sr1\n", ps;
} else {
printf "\tst%s\t%sr%d,%sr2,%d\n", suffix, ps, 5 - evenp, ps, st;
}
ld += align; st += align;
}
if (!no_tdesc) {
printf "%send%s%d:\n", la, mode, count * align;
printf "\tsection\t.tdesc,\"%s\"\n", tf;
printf "\tword\t0x42,1,%sloop%s%d", la, mode, count * align;
printf ",%send%s%d\n", la, mode, count * align;
printf "\tword\t0x0100001f,0,1,0\n";
printf "\ttext\n";
}
printf "; End of awk generated code.\n";
exit;
}'
}
(do_file '"movstrSI64n.s"';
echo 'SI::4:16' | gen_movstrN) > moveSI64n.asm
#.Implementation_continued[=-----------------------------------------------
#
# This generates the even-remainder, straight-line modules.
# The code is entered from the callable entry points
# ___movstrXINxM, where exactly M bytes will be copied in XI
# mode. r2 is the destination pointer, r3 is the source
# pointer, neither being offset. The first value is preloaded
# into r4 or r5 (r4 if M-N/B is even; r5 if M-N/B is odd, where
# B is the mode size of XI). Upon returning, r2 and r3 have not
# been changed.
#
# The code for these cases is generated by the awk program
# following. Edit *it*, not what it generates!
#
#.End]=------------------------------------------------------------------*/
gen_movstrX0() {
awk $awk_flag "$awk_begin"'
if (abi) {
ps="#"; us=""; tf="a"; la=".L"; fb=16;
} else {
ps=""; us="_"; tf="x"; la="@L"; fb=8;
}
}
NR == 1 && NF == 4 {
mode = $1; suffix = $2; align = $3; bytes = $4;
ld = align; st = 0; count = bytes / align;
reg[0] = 4; if (align == 8) reg[1] = 6; else reg[1] = 5;
printf "; The following was calculated using awk.\n";
printf "\ttext\n";
printf "\talign\t%d\n", fb;
for (r = count; r >= 1; r--) {
evenp = r % 2;
name = sprintf("__%smovstr%s%dx%d", us, mode, count * align, r * align);
if (r > 1) {
printf "\tglobal\t%s\n", name;
printf "%s:\n", name;
}
if (r == 1)
printf "\tjmp.n\t%sr1\n", ps;
else
printf "\tld%s\t%sr%d,%sr3,%d\n", suffix, ps, reg[evenp], ps, ld;
printf "\tst%s\t%sr%d,%sr2,%d\n", suffix, ps, reg[1-evenp], ps, st;
ld += align; st += align;
}
if (!no_tdesc) {
printf "%send%s%dx:\n", la, mode, count * align;
printf "\tsection\t.tdesc,\"%s\"\n", tf;
printf "\tword\t0x42,1,__%smovstr%s%dx%d", us, mode, count * align, count * align;
printf ",%send%s%dx\n", la, mode, count * align;
printf "\tword\t0x0100001f,0,1,0\n";
printf "\ttext\n";
}
printf "; End of awk generated code.\n"
exit;
}'
}
(do_file '"movstrQI16x.s"';
echo 'QI:.b:1:16' | gen_movstrX0) > moveQI16x.asm
(do_file '"movstrHI48x.s"';
echo 'HI:.h:2:48' | gen_movstrX0) > moveHI48x.asm
(do_file '"movstrSI96x.s"';
echo 'SI::4:96' | gen_movstrX0) > moveSI96x.asm
(do_file '"movstrDI96x.s"';
echo 'DI:.d:8:96' | gen_movstrX0) > moveDI96x.asm
#.Implementation_continued[=-----------------------------------------------
#
# This generates the odd-remainder, straight-line modules. The
# interface is the same as that for the even-remainder modules.
#
#.End]=------------------------------------------------------------------*/
gen_movstrXr() {
awk $awk_flag "$awk_begin"'
if (abi) {
ps="#"; us=""; tf="a"; la=".L"; fb=16;
} else {
ps=""; us="_"; tf="x"; la="@L"; fb=8;
}
}
NR == 1 && NF == 4 {
mode = $1; rem = $2; most = $3; count = $4;
suffix[1] = ".b"; suffix[2] = ".h"; suffix[4] = ""; suffix[8] = ".d";
prev = align = most;
ld = align; st = 0; total = count - rem - most;
evenp = int(total/align) % 2;
reg[0] = 4; if (align == 8) reg[1] = 6; else reg[1] = 5;
printf "; The following was calculated using awk.\n";
printf "\ttext\n";
printf "\talign\t%d\n", fb;
for (bytes = total; bytes >= 0; bytes -= align) {
if (bytes < align) {
if (bytes >= 4) align = 4;
else if (bytes >= 2) align = 2;
else align = 1;
}
name = sprintf("__%smovstr%s%dx%d", us, mode, total + most, bytes + most);
if (bytes > most) {
printf "\tglobal\t%s\n", name;
printf "%s:\n", name;
}
if (bytes == 0)
printf "\tjmp.n\t%sr1\n", ps;
else
printf "\tld%s\t%sr%d,%sr3,%d\n", suffix[align], ps, reg[evenp], ps, ld;
printf "\tst%s\t%sr%d,%sr2,%d\n", suffix[prev], ps, reg[1-evenp], ps, st;
ld += align; st += prev; prev = align;
evenp = 1 - evenp;
}
if (!no_tdesc) {
printf "%send%s%dx:\n", la, mode, total + most;
printf "\tsection\t.tdesc,\"%s\"\n", tf;
printf "\tword\t0x42,1,__%smovstr%s%dx%d", us, mode, total + most, total + most;
printf ",%send%s%dx\n", la, mode, total + most;
printf "\tword\t0x0100001f,0,1,0\n";
printf "\ttext\n";
}
printf "; End of awk generated code.\n"
exit;
}'
}
(do_file '"movstrDI47x.s"';
echo 'DI:1:8:48' | gen_movstrXr) > moveDI47x.asm
(do_file '"movstrDI46x.s"';
echo 'DI:2:8:48' | gen_movstrXr) > moveDI46x.asm
(do_file '"movstrDI45x.s"';
echo 'DI:3:8:48' | gen_movstrXr) > moveDI45x.asm
(do_file '"movstrDI44x.s"';
echo 'DI:4:8:48' | gen_movstrXr) > moveDI44x.asm
(do_file '"movstrDI43x.s"';
echo 'DI:5:8:48' | gen_movstrXr) > moveDI43x.asm
(do_file '"movstrDI42x.s"';
echo 'DI:6:8:48' | gen_movstrXr) > moveDI42x.asm
(do_file '"movstrDI41x.s"';
echo 'DI:7:8:48' | gen_movstrXr) > moveDI41x.asm
(do_file '"movstrSI47x.s"';
echo 'SI:1:4:48' | gen_movstrXr) > moveSI47x.asm
(do_file '"movstrSI46x.s"';
echo 'SI:2:4:48' | gen_movstrXr) > moveSI46x.asm
(do_file '"movstrSI45x.s"';
echo 'SI:3:4:48' | gen_movstrXr) > moveSI45x.asm
(do_file '"movstrHI15x.s"';
echo 'HI:1:2:16' | gen_movstrXr) > moveHI15x.asm
|
