|
1
|
1 / synth_1to1_3dnow works the same way as the c version of
|
|
|
2 / synth_1to1. this assembler code based 'decode-i586.s'
|
|
|
3 / (by Stefan Bieschewski <stb@acm.org>), two types of changes
|
|
|
4 / have been made:
|
|
|
5 / - use {MMX,3DNow!} instruction for reduce cpu
|
|
|
6 / - remove unused(?) local symbols
|
|
|
7 /
|
|
|
8 / useful sources of information on optimizing 3DNow! code include:
|
|
|
9 / AMD 3DNow! Technology Manual (Publication #21928)
|
|
|
10 / English: http://www.amd.com/K6/k6docs/pdf/21928d.pdf
|
|
|
11 / (Japanese: http://www.amd.com/japan/K6/k6docs/j21928c.pdf)
|
|
|
12 / AMD-K6-2 Processor Code Optimization Application Note (Publication #21924)
|
|
|
13 / English: http://www.amd.com/K6/k6docs/pdf/21924b.pdf
|
|
|
14 /
|
|
|
15 / This code was tested only AMD-K6-2 processor Linux systems,
|
|
|
16 / please tell me:
|
|
|
17 / - whether this code works on other 3DNow! capable processors
|
|
|
18 / (ex.IDT-C6-2) or not
|
|
|
19 / - whether this code works on other OSes or not
|
|
|
20 /
|
|
|
21 / by KIMURA Takuhiro <kim@hannah.ipc.miyakyo-u.ac.jp> - until 31.Mar.1998
|
|
|
22 / <kim@comtec.co.jp> - after 1.Apr.1998
|
|
|
23
|
|
|
24 / Enhancments for q-word operation by Michael Hipp
|
|
|
25
|
|
|
26 .bss
|
|
|
27 .comm buffs,4352,4
|
|
|
28 .data
|
|
|
29 .align 4
|
|
|
30 bo:
|
|
|
31 .long 1
|
|
|
32 .text
|
|
|
33 .globl synth_1to1_3dnow
|
|
|
34 synth_1to1_3dnow:
|
|
|
35 subl $12,%esp
|
|
|
36 pushl %ebp
|
|
|
37 pushl %edi
|
|
|
38 pushl %esi
|
|
|
39 pushl %ebx
|
|
|
40 movl 32(%esp),%eax
|
|
|
41 movl 40(%esp),%esi
|
|
|
42 movl $0,%edi
|
|
|
43 movl bo,%ebp
|
|
|
44 cmpl %edi,36(%esp)
|
|
|
45 jne .L48
|
|
|
46 decl %ebp
|
|
|
47 andl $15,%ebp
|
|
|
48 movl %ebp,bo
|
|
|
49 movl $buffs,%ecx
|
|
|
50 jmp .L49
|
|
|
51 .L48:
|
|
|
52 addl $2,%esi
|
|
|
53 movl $buffs+2176,%ecx
|
|
|
54 .L49:
|
|
|
55 testl $1,%ebp
|
|
|
56 je .L50
|
|
|
57 movl %ecx,%ebx
|
|
|
58 movl %ebp,16(%esp)
|
|
|
59 pushl %eax
|
|
|
60 movl 20(%esp),%edx
|
|
|
61 leal (%ebx,%edx,4),%eax
|
|
|
62 pushl %eax
|
|
|
63 movl 24(%esp),%eax
|
|
|
64 incl %eax
|
|
|
65 andl $15,%eax
|
|
|
66 leal 1088(,%eax,4),%eax
|
|
|
67 addl %ebx,%eax
|
|
|
68 jmp .L74
|
|
|
69 .L50:
|
|
|
70 leal 1088(%ecx),%ebx
|
|
|
71 leal 1(%ebp),%edx
|
|
|
72 movl %edx,16(%esp)
|
|
|
73 pushl %eax
|
|
|
74 leal 1092(%ecx,%ebp,4),%eax
|
|
|
75 pushl %eax
|
|
|
76 leal (%ecx,%ebp,4),%eax
|
|
|
77 .L74:
|
|
|
78 pushl %eax
|
|
|
79 call dct64_3dnow
|
|
|
80 addl $12,%esp
|
|
|
81 movl 16(%esp),%edx
|
|
|
82 leal 0(,%edx,4),%edx
|
|
|
83 movl $decwin+64,%eax
|
|
|
84 movl %eax,%ecx
|
|
|
85 subl %edx,%ecx
|
|
|
86 movl $16,%ebp
|
|
|
87
|
|
|
88 .L55:
|
|
|
89 movq (%ecx),%mm4
|
|
|
90 movq (%ebx),%mm3
|
|
|
91 movq 8(%ecx),%mm0
|
|
|
92 movq 8(%ebx),%mm1
|
|
|
93 pfmul %mm3,%mm4
|
|
|
94
|
|
|
95 movq 16(%ecx),%mm2
|
|
|
96 pfmul %mm1,%mm0
|
|
|
97 movq 16(%ebx),%mm3
|
|
|
98 pfadd %mm0,%mm4
|
|
|
99
|
|
|
100 movq 24(%ecx),%mm0
|
|
|
101 pfmul %mm2,%mm3
|
|
|
102 movq 24(%ebx),%mm1
|
|
|
103 pfadd %mm3,%mm4
|
|
|
104
|
|
|
105 movq 32(%ecx),%mm2
|
|
|
106 pfmul %mm1,%mm0
|
|
|
107 movq 32(%ebx),%mm3
|
|
|
108 pfadd %mm0,%mm4
|
|
|
109
|
|
|
110 movq 40(%ecx),%mm0
|
|
|
111 pfmul %mm2,%mm3
|
|
|
112 movq 40(%ebx),%mm1
|
|
|
113 pfadd %mm3,%mm4
|
|
|
114
|
|
|
115 movq 48(%ecx),%mm2
|
|
|
116 pfmul %mm1,%mm0
|
|
|
117 movq 48(%ebx),%mm3
|
|
|
118 pfadd %mm0,%mm4
|
|
|
119
|
|
|
120 movq 56(%ecx),%mm0
|
|
|
121 pfmul %mm2,%mm3
|
|
|
122 movq 56(%ebx),%mm1
|
|
|
123 pfadd %mm3,%mm4
|
|
|
124
|
|
|
125 pfmul %mm1,%mm0
|
|
|
126 pfadd %mm0,%mm4
|
|
|
127
|
|
|
128 movq %mm4,%mm0
|
|
|
129 psrlq $32,%mm0
|
|
|
130 pfsub %mm0,%mm4
|
|
|
131
|
|
|
132 pf2id %mm4,%mm4
|
|
|
133 movd %mm4,%eax
|
|
|
134
|
|
|
135 sar $16,%eax
|
|
|
136 movw %ax,(%esi)
|
|
|
137
|
|
|
138 addl $64,%ebx
|
|
|
139 subl $-128,%ecx
|
|
|
140 addl $4,%esi
|
|
|
141 decl %ebp
|
|
|
142 jnz .L55
|
|
|
143
|
|
|
144 / --- end of loop 1 ---
|
|
|
145
|
|
|
146 movd (%ecx),%mm2
|
|
|
147 movd (%ebx),%mm1
|
|
|
148 pfmul %mm1,%mm2
|
|
|
149
|
|
|
150 movd 8(%ecx),%mm0
|
|
|
151 movd 8(%ebx),%mm1
|
|
|
152 pfmul %mm0,%mm1
|
|
|
153 pfadd %mm1,%mm2
|
|
|
154
|
|
|
155 movd 16(%ecx),%mm0
|
|
|
156 movd 16(%ebx),%mm1
|
|
|
157 pfmul %mm0,%mm1
|
|
|
158 pfadd %mm1,%mm2
|
|
|
159
|
|
|
160 movd 24(%ecx),%mm0
|
|
|
161 movd 24(%ebx),%mm1
|
|
|
162 pfmul %mm0,%mm1
|
|
|
163 pfadd %mm1,%mm2
|
|
|
164
|
|
|
165 movd 32(%ecx),%mm0
|
|
|
166 movd 32(%ebx),%mm1
|
|
|
167 pfmul %mm0,%mm1
|
|
|
168 pfadd %mm1,%mm2
|
|
|
169
|
|
|
170 movd 40(%ecx),%mm0
|
|
|
171 movd 40(%ebx),%mm1
|
|
|
172 pfmul %mm0,%mm1
|
|
|
173 pfadd %mm1,%mm2
|
|
|
174
|
|
|
175 movd 48(%ecx),%mm0
|
|
|
176 movd 48(%ebx),%mm1
|
|
|
177 pfmul %mm0,%mm1
|
|
|
178 pfadd %mm1,%mm2
|
|
|
179
|
|
|
180 movd 56(%ecx),%mm0
|
|
|
181 movd 56(%ebx),%mm1
|
|
|
182 pfmul %mm0,%mm1
|
|
|
183 pfadd %mm1,%mm2
|
|
|
184
|
|
|
185 pf2id %mm2,%mm2
|
|
|
186 movd %mm2,%eax
|
|
|
187
|
|
|
188 sar $16,%eax
|
|
|
189
|
|
|
190 movw %ax,(%esi)
|
|
|
191
|
|
|
192 addl $-64,%ebx
|
|
|
193 addl $4,%esi
|
|
|
194 addl $256,%ecx
|
|
|
195 movl $15,%ebp
|
|
|
196
|
|
|
197 .L68:
|
|
|
198 psubd %mm0,%mm0
|
|
|
199
|
|
|
200 movq (%ebx),%mm1
|
|
|
201 movq (%ecx),%mm2
|
|
|
202 pfmul %mm1,%mm2
|
|
|
203 pfsub %mm2,%mm0
|
|
|
204
|
|
|
205 movq 8(%ebx),%mm3
|
|
|
206 movq 8(%ecx),%mm4
|
|
|
207 pfmul %mm3,%mm4
|
|
|
208 pfsub %mm4,%mm0
|
|
|
209
|
|
|
210 movq 16(%ebx),%mm1
|
|
|
211 movq 16(%ecx),%mm2
|
|
|
212 pfmul %mm1,%mm2
|
|
|
213 pfsub %mm2,%mm0
|
|
|
214
|
|
|
215 movq 24(%ebx),%mm3
|
|
|
216 movq 24(%ecx),%mm4
|
|
|
217 pfmul %mm3,%mm4
|
|
|
218 pfsub %mm4,%mm0
|
|
|
219
|
|
|
220 movq 32(%ebx),%mm1
|
|
|
221 movq 32(%ecx),%mm2
|
|
|
222 pfmul %mm1,%mm2
|
|
|
223 pfsub %mm2,%mm0
|
|
|
224
|
|
|
225 movq 40(%ebx),%mm3
|
|
|
226 movq 40(%ecx),%mm4
|
|
|
227 pfmul %mm3,%mm4
|
|
|
228 pfsub %mm4,%mm0
|
|
|
229
|
|
|
230 movq 48(%ebx),%mm1
|
|
|
231 movq 48(%ecx),%mm2
|
|
|
232 pfmul %mm1,%mm2
|
|
|
233 pfsub %mm2,%mm0
|
|
|
234
|
|
|
235 movq 56(%ebx),%mm3
|
|
|
236 movq 56(%ecx),%mm4
|
|
|
237 pfmul %mm3,%mm4
|
|
|
238 pfsub %mm4,%mm0
|
|
|
239
|
|
|
240 pfacc %mm0,%mm0
|
|
|
241
|
|
|
242 pf2id %mm0,%mm0
|
|
|
243 movd %mm0,%eax
|
|
|
244
|
|
|
245 sar $16,%eax
|
|
|
246
|
|
|
247 movw %ax,(%esi)
|
|
|
248
|
|
|
249 addl $-64,%ebx
|
|
|
250 subl $-128,%ecx
|
|
|
251 addl $4,%esi
|
|
|
252 decl %ebp
|
|
|
253 jnz .L68
|
|
|
254
|
|
|
255 / --- end of loop 2
|
|
|
256
|
|
|
257 femms
|
|
|
258
|
|
|
259 movl %edi,%eax
|
|
|
260 popl %ebx
|
|
|
261 popl %esi
|
|
|
262 popl %edi
|
|
|
263 popl %ebp
|
|
|
264 addl $12,%esp
|
|
|
265 ret
|
