Debian Codegraph @

src/libmpg123/synth_x86_64_float.S

raw
1 /*
2 synth_x86_64_float: SSE optimized synth for x86-64 (float output version)
3
4 copyright 1995-2026 by the mpg123 project - free software under the terms of the LGPL 2.1
5 see COPYING and AUTHORS files in distribution or http://mpg123.org
6 initially written by Taihei Monma
7 */
8
9 #include "x86_64_defs.h"
10 #include "mangle.h"
11
12 #ifdef IS_MSABI
13 /* short *window; */
14 #define ARG0 %r10
15 /* short *b0; */
16 #define ARG1 %rdx
17 /* short *samples; */
18 #define ARG2 %r8
19 /* int bo1; */
20 #define ARG3 %r9
21 #else
22 /* real *window; */
23 #define ARG0 %rdi
24 /* real *b0; */
25 #define ARG1 %rsi
26 /* real *samples; */
27 #define ARG2 %rdx
28 /* int bo1; */
29 #define ARG3 %rcx
30 #endif
31
32 #define XMMREG_SCALE %xmm15 /* {1/32768.0, 1/32768.0, 1/32768.0, 1/32768.0} */
33
34 /*
35 int synth_1to1_real_x86_64_asm(real *window, real *b0, real *samples, int bo1);
36 return value: number of clipped samples (0)
37 */
38
39 RODATA
40 ALIGN32
41 ASM_NAME(INT123_scale_x86_64):
42 .long 939524096
43 .long 939524096
44 .long 939524096
45 .long 939524096
46 .text
47 ALIGN16
48 .globl ASM_NAME(INT123_synth_1to1_real_x86_64_asm)
49 ASM_NAME(INT123_synth_1to1_real_x86_64_asm):
50 #ifdef IS_MSABI /* should save xmm6-15 */
51 movq %rcx, ARG0
52 subq $120, %rsp /* stack alignment + 7 xmm registers */
53 movaps %xmm6, (%rsp)
54 movaps %xmm7, 16(%rsp)
55 movaps %xmm8, 32(%rsp)
56 movaps %xmm9, 48(%rsp)
57 movaps %xmm10, 64(%rsp)
58 movaps %xmm11, 80(%rsp)
59 movaps %xmm15, 96(%rsp)
60 #endif
61
62 leaq ASM_NAME(INT123_scale_x86_64)(%rip), %rax
63 movaps (%rax), XMMREG_SCALE
64
65 andq $0xf, ARG3
66 shlq $2, ARG3
67 leaq 64(ARG0), ARG0
68 subq ARG3, ARG0
69
70 movl $4, %ecx
71
72 ALIGN16
73 1:
74 movups (ARG0), %xmm8
75 movups 16(ARG0), %xmm1
76 movups 32(ARG0), %xmm2
77 movups 48(ARG0), %xmm3
78 movups 128(ARG0), %xmm9
79 movups 144(ARG0), %xmm5
80 movups 160(ARG0), %xmm6
81 movups 176(ARG0), %xmm7
82 mulps (ARG1), %xmm8
83 mulps 16(ARG1), %xmm1
84 mulps 32(ARG1), %xmm2
85 mulps 48(ARG1), %xmm3
86 mulps 64(ARG1), %xmm9
87 mulps 80(ARG1), %xmm5
88 mulps 96(ARG1), %xmm6
89 mulps 112(ARG1), %xmm7
90
91 addps %xmm1, %xmm8
92 addps %xmm2, %xmm3
93 addps %xmm5, %xmm9
94 addps %xmm7, %xmm6
95 addps %xmm3, %xmm8
96 addps %xmm6, %xmm9
97 leaq 256(ARG0), ARG0
98 leaq 128(ARG1), ARG1
99
100 movups (ARG0), %xmm10
101 movups 16(ARG0), %xmm1
102 movups 32(ARG0), %xmm2
103 movups 48(ARG0), %xmm3
104 movups 128(ARG0), %xmm11
105 movups 144(ARG0), %xmm5
106 movups 160(ARG0), %xmm6
107 movups 176(ARG0), %xmm7
108 mulps (ARG1), %xmm10
109 mulps 16(ARG1), %xmm1
110 mulps 32(ARG1), %xmm2
111 mulps 48(ARG1), %xmm3
112 mulps 64(ARG1), %xmm11
113 mulps 80(ARG1), %xmm5
114 mulps 96(ARG1), %xmm6
115 mulps 112(ARG1), %xmm7
116
117 addps %xmm1, %xmm10
118 addps %xmm2, %xmm3
119 addps %xmm5, %xmm11
120 addps %xmm7, %xmm6
121 addps %xmm3, %xmm10
122 addps %xmm6, %xmm11
123 leaq 256(ARG0), ARG0
124 leaq 128(ARG1), ARG1
125
126 movaps %xmm8, %xmm0
127 movaps %xmm10, %xmm1
128 unpcklps %xmm9, %xmm8
129 unpcklps %xmm11, %xmm10
130 unpckhps %xmm9, %xmm0
131 unpckhps %xmm11, %xmm1
132 movaps %xmm8, %xmm2
133 movaps %xmm0, %xmm3
134 movlhps %xmm10, %xmm8
135 movhlps %xmm2, %xmm10
136 movlhps %xmm1, %xmm0
137 movhlps %xmm3, %xmm1
138 subps %xmm10, %xmm8
139 subps %xmm1, %xmm0
140 addps %xmm8, %xmm0
141
142 movups (ARG2), %xmm1
143 movups 16(ARG2), %xmm2
144 mulps XMMREG_SCALE, %xmm0
145 shufps $0xdd, %xmm2, %xmm1
146 movaps %xmm0, %xmm2
147 unpcklps %xmm1, %xmm0
148 unpckhps %xmm1, %xmm2
149 movups %xmm0, (ARG2)
150 movups %xmm2, 16(ARG2)
151
152 leaq 32(ARG2), ARG2
153 decl %ecx
154 jnz 1b
155
156 movl $4, %ecx
157
158 ALIGN16
159 1:
160 movups (ARG0), %xmm8
161 movups 16(ARG0), %xmm1
162 movups 32(ARG0), %xmm2
163 movups 48(ARG0), %xmm3
164 movups 128(ARG0), %xmm9
165 movups 144(ARG0), %xmm5
166 movups 160(ARG0), %xmm6
167 movups 176(ARG0), %xmm7
168 mulps (ARG1), %xmm8
169 mulps 16(ARG1), %xmm1
170 mulps 32(ARG1), %xmm2
171 mulps 48(ARG1), %xmm3
172 mulps -64(ARG1), %xmm9
173 mulps -48(ARG1), %xmm5
174 mulps -32(ARG1), %xmm6
175 mulps -16(ARG1), %xmm7
176
177 addps %xmm1, %xmm8
178 addps %xmm2, %xmm3
179 addps %xmm5, %xmm9
180 addps %xmm7, %xmm6
181 addps %xmm3, %xmm8
182 addps %xmm6, %xmm9
183 leaq 256(ARG0), ARG0
184 leaq -128(ARG1), ARG1
185
186 movups (ARG0), %xmm10
187 movups 16(ARG0), %xmm1
188 movups 32(ARG0), %xmm2
189 movups 48(ARG0), %xmm3
190 movups 128(ARG0), %xmm11
191 movups 144(ARG0), %xmm5
192 movups 160(ARG0), %xmm6
193 movups 176(ARG0), %xmm7
194 mulps (ARG1), %xmm10
195 mulps 16(ARG1), %xmm1
196 mulps 32(ARG1), %xmm2
197 mulps 48(ARG1), %xmm3
198 mulps -64(ARG1), %xmm11
199 mulps -48(ARG1), %xmm5
200 mulps -32(ARG1), %xmm6
201 mulps -16(ARG1), %xmm7
202
203 addps %xmm1, %xmm10
204 addps %xmm2, %xmm3
205 addps %xmm5, %xmm11
206 addps %xmm7, %xmm6
207 addps %xmm3, %xmm10
208 addps %xmm6, %xmm11
209 leaq 256(ARG0), ARG0
210 leaq -128(ARG1), ARG1
211
212 movaps %xmm8, %xmm0
213 movaps %xmm10, %xmm1
214 unpcklps %xmm9, %xmm8
215 unpcklps %xmm11, %xmm10
216 unpckhps %xmm9, %xmm0
217 unpckhps %xmm11, %xmm1
218 movaps %xmm8, %xmm2
219 movaps %xmm0, %xmm3
220 movlhps %xmm10, %xmm8
221 movhlps %xmm2, %xmm10
222 movlhps %xmm1, %xmm0
223 movhlps %xmm3, %xmm1
224 addps %xmm10, %xmm8
225 addps %xmm1, %xmm0
226 addps %xmm8, %xmm0
227
228 movups (ARG2), %xmm1
229 movups 16(ARG2), %xmm2
230 mulps XMMREG_SCALE, %xmm0
231 shufps $0xdd, %xmm2, %xmm1
232 movaps %xmm0, %xmm2
233 unpcklps %xmm1, %xmm0
234 unpckhps %xmm1, %xmm2
235 movups %xmm0, (ARG2)
236 movups %xmm2, 16(ARG2)
237
238 leaq 32(ARG2), ARG2
239 decl %ecx
240 jnz 1b
241
242 xorl %eax, %eax
243
244 #ifdef IS_MSABI
245 movaps (%rsp), %xmm6
246 movaps 16(%rsp), %xmm7
247 movaps 32(%rsp), %xmm8
248 movaps 48(%rsp), %xmm9
249 movaps 64(%rsp), %xmm10
250 movaps 80(%rsp), %xmm11
251 movaps 96(%rsp), %xmm15
252 addq $120, %rsp
253 #endif
254 ret
255
256 NONEXEC_STACK