|
yading@10
|
1 /*
|
|
yading@10
|
2 * ARM NEON optimised FFT
|
|
yading@10
|
3 *
|
|
yading@10
|
4 * Copyright (c) 2009 Mans Rullgard <mans@mansr.com>
|
|
yading@10
|
5 * Copyright (c) 2009 Naotoshi Nojiri
|
|
yading@10
|
6 *
|
|
yading@10
|
7 * This algorithm (though not any of the implementation details) is
|
|
yading@10
|
8 * based on libdjbfft by D. J. Bernstein.
|
|
yading@10
|
9 *
|
|
yading@10
|
10 * This file is part of FFmpeg.
|
|
yading@10
|
11 *
|
|
yading@10
|
12 * FFmpeg is free software; you can redistribute it and/or
|
|
yading@10
|
13 * modify it under the terms of the GNU Lesser General Public
|
|
yading@10
|
14 * License as published by the Free Software Foundation; either
|
|
yading@10
|
15 * version 2.1 of the License, or (at your option) any later version.
|
|
yading@10
|
16 *
|
|
yading@10
|
17 * FFmpeg is distributed in the hope that it will be useful,
|
|
yading@10
|
18 * but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
yading@10
|
19 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
|
yading@10
|
20 * Lesser General Public License for more details.
|
|
yading@10
|
21 *
|
|
yading@10
|
22 * You should have received a copy of the GNU Lesser General Public
|
|
yading@10
|
23 * License along with FFmpeg; if not, write to the Free Software
|
|
yading@10
|
24 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
|
|
yading@10
|
25 */
|
|
yading@10
|
26
|
|
yading@10
|
27 #include "libavutil/arm/asm.S"
|
|
yading@10
|
28
|
|
yading@10
|
29 #define M_SQRT1_2 0.70710678118654752440
|
|
yading@10
|
30
|
|
yading@10
|
31
|
|
yading@10
|
32 function fft4_neon
|
|
yading@10
|
33 vld1.32 {d0-d3}, [r0,:128]
|
|
yading@10
|
34
|
|
yading@10
|
35 vext.32 q8, q1, q1, #1 @ i2,r3 d3=i3,r2
|
|
yading@10
|
36 vsub.f32 d6, d0, d1 @ r0-r1,i0-i1
|
|
yading@10
|
37 vsub.f32 d7, d16, d17 @ r3-r2,i2-i3
|
|
yading@10
|
38 vadd.f32 d4, d0, d1 @ r0+r1,i0+i1
|
|
yading@10
|
39 vadd.f32 d5, d2, d3 @ i2+i3,r2+r3
|
|
yading@10
|
40 vadd.f32 d1, d6, d7
|
|
yading@10
|
41 vsub.f32 d3, d6, d7
|
|
yading@10
|
42 vadd.f32 d0, d4, d5
|
|
yading@10
|
43 vsub.f32 d2, d4, d5
|
|
yading@10
|
44
|
|
yading@10
|
45 vst1.32 {d0-d3}, [r0,:128]
|
|
yading@10
|
46
|
|
yading@10
|
47 bx lr
|
|
yading@10
|
48 endfunc
|
|
yading@10
|
49
|
|
yading@10
|
50 function fft8_neon
|
|
yading@10
|
51 mov r1, r0
|
|
yading@10
|
52 vld1.32 {d0-d3}, [r1,:128]!
|
|
yading@10
|
53 vld1.32 {d16-d19}, [r1,:128]
|
|
yading@10
|
54
|
|
yading@10
|
55 movw r2, #0x04f3 @ sqrt(1/2)
|
|
yading@10
|
56 movt r2, #0x3f35
|
|
yading@10
|
57 eor r3, r2, #1<<31
|
|
yading@10
|
58 vdup.32 d31, r2
|
|
yading@10
|
59
|
|
yading@10
|
60 vext.32 q11, q1, q1, #1 @ i2,r3,i3,r2
|
|
yading@10
|
61 vadd.f32 d4, d16, d17 @ r4+r5,i4+i5
|
|
yading@10
|
62 vmov d28, r3, r2
|
|
yading@10
|
63 vadd.f32 d5, d18, d19 @ r6+r7,i6+i7
|
|
yading@10
|
64 vsub.f32 d17, d16, d17 @ r4-r5,i4-i5
|
|
yading@10
|
65 vsub.f32 d19, d18, d19 @ r6-r7,i6-i7
|
|
yading@10
|
66 vrev64.32 d29, d28
|
|
yading@10
|
67 vadd.f32 d20, d0, d1 @ r0+r1,i0+i1
|
|
yading@10
|
68 vadd.f32 d21, d2, d3 @ r2+r3,i2+i3
|
|
yading@10
|
69 vmul.f32 d26, d17, d28 @ -a2r*w,a2i*w
|
|
yading@10
|
70 vext.32 q3, q2, q2, #1
|
|
yading@10
|
71 vmul.f32 d27, d19, d29 @ a3r*w,-a3i*w
|
|
yading@10
|
72 vsub.f32 d23, d22, d23 @ i2-i3,r3-r2
|
|
yading@10
|
73 vsub.f32 d22, d0, d1 @ r0-r1,i0-i1
|
|
yading@10
|
74 vmul.f32 d24, d17, d31 @ a2r*w,a2i*w
|
|
yading@10
|
75 vmul.f32 d25, d19, d31 @ a3r*w,a3i*w
|
|
yading@10
|
76 vadd.f32 d0, d20, d21
|
|
yading@10
|
77 vsub.f32 d2, d20, d21
|
|
yading@10
|
78 vadd.f32 d1, d22, d23
|
|
yading@10
|
79 vrev64.32 q13, q13
|
|
yading@10
|
80 vsub.f32 d3, d22, d23
|
|
yading@10
|
81 vsub.f32 d6, d6, d7
|
|
yading@10
|
82 vadd.f32 d24, d24, d26 @ a2r+a2i,a2i-a2r t1,t2
|
|
yading@10
|
83 vadd.f32 d25, d25, d27 @ a3r-a3i,a3i+a3r t5,t6
|
|
yading@10
|
84 vadd.f32 d7, d4, d5
|
|
yading@10
|
85 vsub.f32 d18, d2, d6
|
|
yading@10
|
86 vext.32 q13, q12, q12, #1
|
|
yading@10
|
87 vadd.f32 d2, d2, d6
|
|
yading@10
|
88 vsub.f32 d16, d0, d7
|
|
yading@10
|
89 vadd.f32 d5, d25, d24
|
|
yading@10
|
90 vsub.f32 d4, d26, d27
|
|
yading@10
|
91 vadd.f32 d0, d0, d7
|
|
yading@10
|
92 vsub.f32 d17, d1, d5
|
|
yading@10
|
93 vsub.f32 d19, d3, d4
|
|
yading@10
|
94 vadd.f32 d3, d3, d4
|
|
yading@10
|
95 vadd.f32 d1, d1, d5
|
|
yading@10
|
96
|
|
yading@10
|
97 vst1.32 {d16-d19}, [r1,:128]
|
|
yading@10
|
98 vst1.32 {d0-d3}, [r0,:128]
|
|
yading@10
|
99
|
|
yading@10
|
100 bx lr
|
|
yading@10
|
101 endfunc
|
|
yading@10
|
102
|
|
yading@10
|
103 function fft16_neon
|
|
yading@10
|
104 movrel r1, mppm
|
|
yading@10
|
105 vld1.32 {d16-d19}, [r0,:128]! @ q8{r0,i0,r1,i1} q9{r2,i2,r3,i3}
|
|
yading@10
|
106 pld [r0, #32]
|
|
yading@10
|
107 vld1.32 {d2-d3}, [r1,:128]
|
|
yading@10
|
108 vext.32 q13, q9, q9, #1
|
|
yading@10
|
109 vld1.32 {d22-d25}, [r0,:128]! @ q11{r4,i4,r5,i5} q12{r6,i5,r7,i7}
|
|
yading@10
|
110 vadd.f32 d4, d16, d17
|
|
yading@10
|
111 vsub.f32 d5, d16, d17
|
|
yading@10
|
112 vadd.f32 d18, d18, d19
|
|
yading@10
|
113 vsub.f32 d19, d26, d27
|
|
yading@10
|
114
|
|
yading@10
|
115 vadd.f32 d20, d22, d23
|
|
yading@10
|
116 vsub.f32 d22, d22, d23
|
|
yading@10
|
117 vsub.f32 d23, d24, d25
|
|
yading@10
|
118 vadd.f32 q8, q2, q9 @ {r0,i0,r1,i1}
|
|
yading@10
|
119 vadd.f32 d21, d24, d25
|
|
yading@10
|
120 vmul.f32 d24, d22, d2
|
|
yading@10
|
121 vsub.f32 q9, q2, q9 @ {r2,i2,r3,i3}
|
|
yading@10
|
122 vmul.f32 d25, d23, d3
|
|
yading@10
|
123 vuzp.32 d16, d17 @ {r0,r1,i0,i1}
|
|
yading@10
|
124 vmul.f32 q1, q11, d2[1]
|
|
yading@10
|
125 vuzp.32 d18, d19 @ {r2,r3,i2,i3}
|
|
yading@10
|
126 vrev64.32 q12, q12
|
|
yading@10
|
127 vadd.f32 q11, q12, q1 @ {t1a,t2a,t5,t6}
|
|
yading@10
|
128 vld1.32 {d24-d27}, [r0,:128]! @ q12{r8,i8,r9,i9} q13{r10,i10,r11,i11}
|
|
yading@10
|
129 vzip.32 q10, q11
|
|
yading@10
|
130 vld1.32 {d28-d31}, [r0,:128] @ q14{r12,i12,r13,i13} q15{r14,i14,r15,i15}
|
|
yading@10
|
131 vadd.f32 d0, d22, d20
|
|
yading@10
|
132 vadd.f32 d1, d21, d23
|
|
yading@10
|
133 vsub.f32 d2, d21, d23
|
|
yading@10
|
134 vsub.f32 d3, d22, d20
|
|
yading@10
|
135 sub r0, r0, #96
|
|
yading@10
|
136 vext.32 q13, q13, q13, #1
|
|
yading@10
|
137 vsub.f32 q10, q8, q0 @ {r4,r5,i4,i5}
|
|
yading@10
|
138 vadd.f32 q8, q8, q0 @ {r0,r1,i0,i1}
|
|
yading@10
|
139 vext.32 q15, q15, q15, #1
|
|
yading@10
|
140 vsub.f32 q11, q9, q1 @ {r6,r7,i6,i7}
|
|
yading@10
|
141 vswp d25, d26 @ q12{r8,i8,i10,r11} q13{r9,i9,i11,r10}
|
|
yading@10
|
142 vadd.f32 q9, q9, q1 @ {r2,r3,i2,i3}
|
|
yading@10
|
143 vswp d29, d30 @ q14{r12,i12,i14,r15} q15{r13,i13,i15,r14}
|
|
yading@10
|
144 vadd.f32 q0, q12, q13 @ {t1,t2,t5,t6}
|
|
yading@10
|
145 vadd.f32 q1, q14, q15 @ {t1a,t2a,t5a,t6a}
|
|
yading@10
|
146 movrelx r2, X(ff_cos_16)
|
|
yading@10
|
147 vsub.f32 q13, q12, q13 @ {t3,t4,t7,t8}
|
|
yading@10
|
148 vrev64.32 d1, d1
|
|
yading@10
|
149 vsub.f32 q15, q14, q15 @ {t3a,t4a,t7a,t8a}
|
|
yading@10
|
150 vrev64.32 d3, d3
|
|
yading@10
|
151 movrel r3, pmmp
|
|
yading@10
|
152 vswp d1, d26 @ q0{t1,t2,t3,t4} q13{t6,t5,t7,t8}
|
|
yading@10
|
153 vswp d3, d30 @ q1{t1a,t2a,t3a,t4a} q15{t6a,t5a,t7a,t8a}
|
|
yading@10
|
154 vadd.f32 q12, q0, q13 @ {r8,i8,r9,i9}
|
|
yading@10
|
155 vadd.f32 q14, q1, q15 @ {r12,i12,r13,i13}
|
|
yading@10
|
156 vld1.32 {d4-d5}, [r2,:64]
|
|
yading@10
|
157 vsub.f32 q13, q0, q13 @ {r10,i10,r11,i11}
|
|
yading@10
|
158 vsub.f32 q15, q1, q15 @ {r14,i14,r15,i15}
|
|
yading@10
|
159 vswp d25, d28 @ q12{r8,i8,r12,i12} q14{r9,i9,r13,i13}
|
|
yading@10
|
160 vld1.32 {d6-d7}, [r3,:128]
|
|
yading@10
|
161 vrev64.32 q1, q14
|
|
yading@10
|
162 vmul.f32 q14, q14, d4[1]
|
|
yading@10
|
163 vmul.f32 q1, q1, q3
|
|
yading@10
|
164 vmla.f32 q14, q1, d5[1] @ {t1a,t2a,t5a,t6a}
|
|
yading@10
|
165 vswp d27, d30 @ q13{r10,i10,r14,i14} q15{r11,i11,r15,i15}
|
|
yading@10
|
166 vzip.32 q12, q14
|
|
yading@10
|
167 vadd.f32 d0, d28, d24
|
|
yading@10
|
168 vadd.f32 d1, d25, d29
|
|
yading@10
|
169 vsub.f32 d2, d25, d29
|
|
yading@10
|
170 vsub.f32 d3, d28, d24
|
|
yading@10
|
171 vsub.f32 q12, q8, q0 @ {r8,r9,i8,i9}
|
|
yading@10
|
172 vadd.f32 q8, q8, q0 @ {r0,r1,i0,i1}
|
|
yading@10
|
173 vsub.f32 q14, q10, q1 @ {r12,r13,i12,i13}
|
|
yading@10
|
174 mov r1, #32
|
|
yading@10
|
175 vadd.f32 q10, q10, q1 @ {r4,r5,i4,i5}
|
|
yading@10
|
176 vrev64.32 q0, q13
|
|
yading@10
|
177 vmul.f32 q13, q13, d5[0]
|
|
yading@10
|
178 vrev64.32 q1, q15
|
|
yading@10
|
179 vmul.f32 q15, q15, d5[1]
|
|
yading@10
|
180 vst2.32 {d16-d17},[r0,:128], r1
|
|
yading@10
|
181 vmul.f32 q0, q0, q3
|
|
yading@10
|
182 vst2.32 {d20-d21},[r0,:128], r1
|
|
yading@10
|
183 vmul.f32 q1, q1, q3
|
|
yading@10
|
184 vmla.f32 q13, q0, d5[0] @ {t1,t2,t5,t6}
|
|
yading@10
|
185 vmla.f32 q15, q1, d4[1] @ {t1a,t2a,t5a,t6a}
|
|
yading@10
|
186 vst2.32 {d24-d25},[r0,:128], r1
|
|
yading@10
|
187 vst2.32 {d28-d29},[r0,:128]
|
|
yading@10
|
188 vzip.32 q13, q15
|
|
yading@10
|
189 sub r0, r0, #80
|
|
yading@10
|
190 vadd.f32 d0, d30, d26
|
|
yading@10
|
191 vadd.f32 d1, d27, d31
|
|
yading@10
|
192 vsub.f32 d2, d27, d31
|
|
yading@10
|
193 vsub.f32 d3, d30, d26
|
|
yading@10
|
194 vsub.f32 q13, q9, q0 @ {r10,r11,i10,i11}
|
|
yading@10
|
195 vadd.f32 q9, q9, q0 @ {r2,r3,i2,i3}
|
|
yading@10
|
196 vsub.f32 q15, q11, q1 @ {r14,r15,i14,i15}
|
|
yading@10
|
197 vadd.f32 q11, q11, q1 @ {r6,r7,i6,i7}
|
|
yading@10
|
198 vst2.32 {d18-d19},[r0,:128], r1
|
|
yading@10
|
199 vst2.32 {d22-d23},[r0,:128], r1
|
|
yading@10
|
200 vst2.32 {d26-d27},[r0,:128], r1
|
|
yading@10
|
201 vst2.32 {d30-d31},[r0,:128]
|
|
yading@10
|
202 bx lr
|
|
yading@10
|
203 endfunc
|
|
yading@10
|
204
|
|
yading@10
|
205 function fft_pass_neon
|
|
yading@10
|
206 push {r4-r6,lr}
|
|
yading@10
|
207 mov r6, r2 @ n
|
|
yading@10
|
208 lsl r5, r2, #3 @ 2 * n * sizeof FFTSample
|
|
yading@10
|
209 lsl r4, r2, #4 @ 2 * n * sizeof FFTComplex
|
|
yading@10
|
210 lsl r2, r2, #5 @ 4 * n * sizeof FFTComplex
|
|
yading@10
|
211 add r3, r2, r4
|
|
yading@10
|
212 add r4, r4, r0 @ &z[o1]
|
|
yading@10
|
213 add r2, r2, r0 @ &z[o2]
|
|
yading@10
|
214 add r3, r3, r0 @ &z[o3]
|
|
yading@10
|
215 vld1.32 {d20-d21},[r2,:128] @ {z[o2],z[o2+1]}
|
|
yading@10
|
216 movrel r12, pmmp
|
|
yading@10
|
217 vld1.32 {d22-d23},[r3,:128] @ {z[o3],z[o3+1]}
|
|
yading@10
|
218 add r5, r5, r1 @ wim
|
|
yading@10
|
219 vld1.32 {d6-d7}, [r12,:128] @ pmmp
|
|
yading@10
|
220 vswp d21, d22
|
|
yading@10
|
221 vld1.32 {d4}, [r1,:64]! @ {wre[0],wre[1]}
|
|
yading@10
|
222 sub r5, r5, #4 @ wim--
|
|
yading@10
|
223 vrev64.32 q1, q11
|
|
yading@10
|
224 vmul.f32 q11, q11, d4[1]
|
|
yading@10
|
225 vmul.f32 q1, q1, q3
|
|
yading@10
|
226 vld1.32 {d5[0]}, [r5,:32] @ d5[0] = wim[-1]
|
|
yading@10
|
227 vmla.f32 q11, q1, d5[0] @ {t1a,t2a,t5a,t6a}
|
|
yading@10
|
228 vld2.32 {d16-d17},[r0,:128] @ {z[0],z[1]}
|
|
yading@10
|
229 sub r6, r6, #1 @ n--
|
|
yading@10
|
230 vld2.32 {d18-d19},[r4,:128] @ {z[o1],z[o1+1]}
|
|
yading@10
|
231 vzip.32 q10, q11
|
|
yading@10
|
232 vadd.f32 d0, d22, d20
|
|
yading@10
|
233 vadd.f32 d1, d21, d23
|
|
yading@10
|
234 vsub.f32 d2, d21, d23
|
|
yading@10
|
235 vsub.f32 d3, d22, d20
|
|
yading@10
|
236 vsub.f32 q10, q8, q0
|
|
yading@10
|
237 vadd.f32 q8, q8, q0
|
|
yading@10
|
238 vsub.f32 q11, q9, q1
|
|
yading@10
|
239 vadd.f32 q9, q9, q1
|
|
yading@10
|
240 vst2.32 {d20-d21},[r2,:128]! @ {z[o2],z[o2+1]}
|
|
yading@10
|
241 vst2.32 {d16-d17},[r0,:128]! @ {z[0],z[1]}
|
|
yading@10
|
242 vst2.32 {d22-d23},[r3,:128]! @ {z[o3],z[o3+1]}
|
|
yading@10
|
243 vst2.32 {d18-d19},[r4,:128]! @ {z[o1],z[o1+1]}
|
|
yading@10
|
244 sub r5, r5, #8 @ wim -= 2
|
|
yading@10
|
245 1:
|
|
yading@10
|
246 vld1.32 {d20-d21},[r2,:128] @ {z[o2],z[o2+1]}
|
|
yading@10
|
247 vld1.32 {d22-d23},[r3,:128] @ {z[o3],z[o3+1]}
|
|
yading@10
|
248 vswp d21, d22
|
|
yading@10
|
249 vld1.32 {d4}, [r1]! @ {wre[0],wre[1]}
|
|
yading@10
|
250 vrev64.32 q0, q10
|
|
yading@10
|
251 vmul.f32 q10, q10, d4[0]
|
|
yading@10
|
252 vrev64.32 q1, q11
|
|
yading@10
|
253 vmul.f32 q11, q11, d4[1]
|
|
yading@10
|
254 vld1.32 {d5}, [r5] @ {wim[-1],wim[0]}
|
|
yading@10
|
255 vmul.f32 q0, q0, q3
|
|
yading@10
|
256 sub r5, r5, #8 @ wim -= 2
|
|
yading@10
|
257 vmul.f32 q1, q1, q3
|
|
yading@10
|
258 vmla.f32 q10, q0, d5[1] @ {t1,t2,t5,t6}
|
|
yading@10
|
259 vmla.f32 q11, q1, d5[0] @ {t1a,t2a,t5a,t6a}
|
|
yading@10
|
260 vld2.32 {d16-d17},[r0,:128] @ {z[0],z[1]}
|
|
yading@10
|
261 subs r6, r6, #1 @ n--
|
|
yading@10
|
262 vld2.32 {d18-d19},[r4,:128] @ {z[o1],z[o1+1]}
|
|
yading@10
|
263 vzip.32 q10, q11
|
|
yading@10
|
264 vadd.f32 d0, d22, d20
|
|
yading@10
|
265 vadd.f32 d1, d21, d23
|
|
yading@10
|
266 vsub.f32 d2, d21, d23
|
|
yading@10
|
267 vsub.f32 d3, d22, d20
|
|
yading@10
|
268 vsub.f32 q10, q8, q0
|
|
yading@10
|
269 vadd.f32 q8, q8, q0
|
|
yading@10
|
270 vsub.f32 q11, q9, q1
|
|
yading@10
|
271 vadd.f32 q9, q9, q1
|
|
yading@10
|
272 vst2.32 {d20-d21}, [r2,:128]! @ {z[o2],z[o2+1]}
|
|
yading@10
|
273 vst2.32 {d16-d17}, [r0,:128]! @ {z[0],z[1]}
|
|
yading@10
|
274 vst2.32 {d22-d23}, [r3,:128]! @ {z[o3],z[o3+1]}
|
|
yading@10
|
275 vst2.32 {d18-d19}, [r4,:128]! @ {z[o1],z[o1+1]}
|
|
yading@10
|
276 bne 1b
|
|
yading@10
|
277
|
|
yading@10
|
278 pop {r4-r6,pc}
|
|
yading@10
|
279 endfunc
|
|
yading@10
|
280
|
|
yading@10
|
281 .macro def_fft n, n2, n4
|
|
yading@10
|
282 .align 6
|
|
yading@10
|
283 function fft\n\()_neon
|
|
yading@10
|
284 push {r4, lr}
|
|
yading@10
|
285 mov r4, r0
|
|
yading@10
|
286 bl fft\n2\()_neon
|
|
yading@10
|
287 add r0, r4, #\n4*2*8
|
|
yading@10
|
288 bl fft\n4\()_neon
|
|
yading@10
|
289 add r0, r4, #\n4*3*8
|
|
yading@10
|
290 bl fft\n4\()_neon
|
|
yading@10
|
291 mov r0, r4
|
|
yading@10
|
292 pop {r4, lr}
|
|
yading@10
|
293 movrelx r1, X(ff_cos_\n)
|
|
yading@10
|
294 mov r2, #\n4/2
|
|
yading@10
|
295 b fft_pass_neon
|
|
yading@10
|
296 endfunc
|
|
yading@10
|
297 .endm
|
|
yading@10
|
298
|
|
yading@10
|
299 def_fft 32, 16, 8
|
|
yading@10
|
300 def_fft 64, 32, 16
|
|
yading@10
|
301 def_fft 128, 64, 32
|
|
yading@10
|
302 def_fft 256, 128, 64
|
|
yading@10
|
303 def_fft 512, 256, 128
|
|
yading@10
|
304 def_fft 1024, 512, 256
|
|
yading@10
|
305 def_fft 2048, 1024, 512
|
|
yading@10
|
306 def_fft 4096, 2048, 1024
|
|
yading@10
|
307 def_fft 8192, 4096, 2048
|
|
yading@10
|
308 def_fft 16384, 8192, 4096
|
|
yading@10
|
309 def_fft 32768, 16384, 8192
|
|
yading@10
|
310 def_fft 65536, 32768, 16384
|
|
yading@10
|
311
|
|
yading@10
|
312 function ff_fft_calc_neon, export=1
|
|
yading@10
|
313 ldr r2, [r0]
|
|
yading@10
|
314 sub r2, r2, #2
|
|
yading@10
|
315 movrel r3, fft_tab_neon
|
|
yading@10
|
316 ldr r3, [r3, r2, lsl #2]
|
|
yading@10
|
317 mov r0, r1
|
|
yading@10
|
318 bx r3
|
|
yading@10
|
319 endfunc
|
|
yading@10
|
320
|
|
yading@10
|
321 function ff_fft_permute_neon, export=1
|
|
yading@10
|
322 push {r4,lr}
|
|
yading@10
|
323 mov r12, #1
|
|
yading@10
|
324 ldr r2, [r0] @ nbits
|
|
yading@10
|
325 ldr r3, [r0, #12] @ tmp_buf
|
|
yading@10
|
326 ldr r0, [r0, #8] @ revtab
|
|
yading@10
|
327 lsl r12, r12, r2
|
|
yading@10
|
328 mov r2, r12
|
|
yading@10
|
329 1:
|
|
yading@10
|
330 vld1.32 {d0-d1}, [r1,:128]!
|
|
yading@10
|
331 ldr r4, [r0], #4
|
|
yading@10
|
332 uxth lr, r4
|
|
yading@10
|
333 uxth r4, r4, ror #16
|
|
yading@10
|
334 add lr, r3, lr, lsl #3
|
|
yading@10
|
335 add r4, r3, r4, lsl #3
|
|
yading@10
|
336 vst1.32 {d0}, [lr,:64]
|
|
yading@10
|
337 vst1.32 {d1}, [r4,:64]
|
|
yading@10
|
338 subs r12, r12, #2
|
|
yading@10
|
339 bgt 1b
|
|
yading@10
|
340
|
|
yading@10
|
341 sub r1, r1, r2, lsl #3
|
|
yading@10
|
342 1:
|
|
yading@10
|
343 vld1.32 {d0-d3}, [r3,:128]!
|
|
yading@10
|
344 vst1.32 {d0-d3}, [r1,:128]!
|
|
yading@10
|
345 subs r2, r2, #4
|
|
yading@10
|
346 bgt 1b
|
|
yading@10
|
347
|
|
yading@10
|
348 pop {r4,pc}
|
|
yading@10
|
349 endfunc
|
|
yading@10
|
350
|
|
yading@10
|
351 const fft_tab_neon
|
|
yading@10
|
352 .word fft4_neon
|
|
yading@10
|
353 .word fft8_neon
|
|
yading@10
|
354 .word fft16_neon
|
|
yading@10
|
355 .word fft32_neon
|
|
yading@10
|
356 .word fft64_neon
|
|
yading@10
|
357 .word fft128_neon
|
|
yading@10
|
358 .word fft256_neon
|
|
yading@10
|
359 .word fft512_neon
|
|
yading@10
|
360 .word fft1024_neon
|
|
yading@10
|
361 .word fft2048_neon
|
|
yading@10
|
362 .word fft4096_neon
|
|
yading@10
|
363 .word fft8192_neon
|
|
yading@10
|
364 .word fft16384_neon
|
|
yading@10
|
365 .word fft32768_neon
|
|
yading@10
|
366 .word fft65536_neon
|
|
yading@10
|
367 endconst
|
|
yading@10
|
368
|
|
yading@10
|
369 const pmmp, align=4
|
|
yading@10
|
370 .float +1.0, -1.0, -1.0, +1.0
|
|
yading@10
|
371 endconst
|
|
yading@10
|
372
|
|
yading@10
|
373 const mppm, align=4
|
|
yading@10
|
374 .float -M_SQRT1_2, M_SQRT1_2, M_SQRT1_2, -M_SQRT1_2
|
|
yading@10
|
375 endconst
|
