Browse code
ARM: NEON fixed-point forward MDCT
Signed-off-by: Mans Rullgard <mans@mansr.com>
Mans Rullgard authored on 2011/03/29 03:39:44Showing 3 changed files
| ... | ... |
@@ -22,11 +22,21 @@ |
| 22 | 22 |
#include "libavcodec/fft.h" |
| 23 | 23 |
|
| 24 | 24 |
void ff_fft_fixed_calc_neon(FFTContext *s, FFTComplex *z); |
| 25 |
+void ff_mdct_fixed_calc_neon(FFTContext *s, FFTSample *o, const FFTSample *i); |
|
| 26 |
+void ff_mdct_fixed_calcw_neon(FFTContext *s, FFTDouble *o, const FFTSample *i); |
|
| 25 | 27 |
|
| 26 | 28 |
av_cold void ff_fft_fixed_init_arm(FFTContext *s) |
| 27 | 29 |
{
|
| 28 | 30 |
if (HAVE_NEON) {
|
| 29 | 31 |
s->fft_permutation = FF_FFT_PERM_SWAP_LSBS; |
| 30 | 32 |
s->fft_calc = ff_fft_fixed_calc_neon; |
| 33 |
+ |
|
| 34 |
+#if CONFIG_MDCT |
|
| 35 |
+ if (!s->inverse && s->mdct_bits >= 5) {
|
|
| 36 |
+ s->mdct_permutation = FF_MDCT_PERM_INTERLEAVE; |
|
| 37 |
+ s->mdct_calc = ff_mdct_fixed_calc_neon; |
|
| 38 |
+ s->mdct_calcw = ff_mdct_fixed_calcw_neon; |
|
| 39 |
+ } |
|
| 40 |
+#endif |
|
| 31 | 41 |
} |
| 32 | 42 |
} |
| 33 | 43 |
new file mode 100644 |
| ... | ... |
@@ -0,0 +1,195 @@ |
| 0 |
+/* |
|
| 1 |
+ * Copyright (c) 2011 Mans Rullgard <mans@mansr.com> |
|
| 2 |
+ * |
|
| 3 |
+ * This file is part of Libav. |
|
| 4 |
+ * |
|
| 5 |
+ * Libav is free software; you can redistribute it and/or |
|
| 6 |
+ * modify it under the terms of the GNU Lesser General Public |
|
| 7 |
+ * License as published by the Free Software Foundation; either |
|
| 8 |
+ * version 2.1 of the License, or (at your option) any later version. |
|
| 9 |
+ * |
|
| 10 |
+ * Libav is distributed in the hope that it will be useful, |
|
| 11 |
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of |
|
| 12 |
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
|
| 13 |
+ * Lesser General Public License for more details. |
|
| 14 |
+ * |
|
| 15 |
+ * You should have received a copy of the GNU Lesser General Public |
|
| 16 |
+ * License along with Libav; if not, write to the Free Software |
|
| 17 |
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA |
|
| 18 |
+ */ |
|
| 19 |
+ |
|
| 20 |
+#include "asm.S" |
|
| 21 |
+ |
|
| 22 |
+ preserve8 |
|
| 23 |
+ |
|
| 24 |
+.macro prerot dst, rt |
|
| 25 |
+ lsr r3, r6, #2 @ n4 |
|
| 26 |
+ add \rt, r4, r6, lsr #1 @ revtab + n4 |
|
| 27 |
+ add r9, r3, r3, lsl #1 @ n3 |
|
| 28 |
+ add r8, r7, r6 @ tcos + n4 |
|
| 29 |
+ add r3, r2, r6, lsr #1 @ in + n4 |
|
| 30 |
+ add r9, r2, r9, lsl #1 @ in + n3 |
|
| 31 |
+ sub r8, r8, #16 |
|
| 32 |
+ sub r10, r3, #16 |
|
| 33 |
+ sub r11, r9, #16 |
|
| 34 |
+ mov r12, #-16 |
|
| 35 |
+1: |
|
| 36 |
+ vld2.16 {d0,d1}, [r9, :128]!
|
|
| 37 |
+ vld2.16 {d2,d3}, [r11,:128], r12
|
|
| 38 |
+ vld2.16 {d4,d5}, [r3, :128]!
|
|
| 39 |
+ vld2.16 {d6,d7}, [r10,:128], r12
|
|
| 40 |
+ vld2.16 {d16,d17},[r7, :128]! @ cos, sin
|
|
| 41 |
+ vld2.16 {d18,d19},[r8, :128], r12
|
|
| 42 |
+ vrev64.16 q1, q1 |
|
| 43 |
+ vrev64.16 q3, q3 |
|
| 44 |
+ vrev64.16 q9, q9 |
|
| 45 |
+ vneg.s16 d0, d0 |
|
| 46 |
+ vneg.s16 d2, d2 |
|
| 47 |
+ vneg.s16 d16, d16 |
|
| 48 |
+ vneg.s16 d18, d18 |
|
| 49 |
+ vhsub.s16 d0, d0, d3 @ re |
|
| 50 |
+ vhsub.s16 d4, d7, d4 @ im |
|
| 51 |
+ vhsub.s16 d6, d6, d5 |
|
| 52 |
+ vhsub.s16 d2, d2, d1 |
|
| 53 |
+ vmull.s16 q10, d0, d16 |
|
| 54 |
+ vmlsl.s16 q10, d4, d17 |
|
| 55 |
+ vmull.s16 q11, d0, d17 |
|
| 56 |
+ vmlal.s16 q11, d4, d16 |
|
| 57 |
+ vmull.s16 q12, d6, d18 |
|
| 58 |
+ vmlsl.s16 q12, d2, d19 |
|
| 59 |
+ vmull.s16 q13, d6, d19 |
|
| 60 |
+ vmlal.s16 q13, d2, d18 |
|
| 61 |
+ vshrn.s32 d0, q10, #15 |
|
| 62 |
+ vshrn.s32 d1, q11, #15 |
|
| 63 |
+ vshrn.s32 d2, q12, #15 |
|
| 64 |
+ vshrn.s32 d3, q13, #15 |
|
| 65 |
+ vzip.16 d0, d1 |
|
| 66 |
+ vzip.16 d2, d3 |
|
| 67 |
+ ldrh lr, [r4], #2 |
|
| 68 |
+ ldrh r2, [\rt, #-2]! |
|
| 69 |
+ add lr, \dst, lr, lsl #2 |
|
| 70 |
+ add r2, \dst, r2, lsl #2 |
|
| 71 |
+ vst1.32 {d0[0]}, [lr,:32]
|
|
| 72 |
+ vst1.32 {d2[0]}, [r2,:32]
|
|
| 73 |
+ ldrh lr, [r4], #2 |
|
| 74 |
+ ldrh r2, [\rt, #-2]! |
|
| 75 |
+ add lr, \dst, lr, lsl #2 |
|
| 76 |
+ add r2, \dst, r2, lsl #2 |
|
| 77 |
+ vst1.32 {d0[1]}, [lr,:32]
|
|
| 78 |
+ vst1.32 {d2[1]}, [r2,:32]
|
|
| 79 |
+ ldrh lr, [r4], #2 |
|
| 80 |
+ ldrh r2, [\rt, #-2]! |
|
| 81 |
+ add lr, \dst, lr, lsl #2 |
|
| 82 |
+ add r2, \dst, r2, lsl #2 |
|
| 83 |
+ vst1.32 {d1[0]}, [lr,:32]
|
|
| 84 |
+ vst1.32 {d3[0]}, [r2,:32]
|
|
| 85 |
+ ldrh lr, [r4], #2 |
|
| 86 |
+ ldrh r2, [\rt, #-2]! |
|
| 87 |
+ add lr, \dst, lr, lsl #2 |
|
| 88 |
+ add r2, \dst, r2, lsl #2 |
|
| 89 |
+ vst1.32 {d1[1]}, [lr,:32]
|
|
| 90 |
+ vst1.32 {d3[1]}, [r2,:32]
|
|
| 91 |
+ subs r6, r6, #32 |
|
| 92 |
+ bgt 1b |
|
| 93 |
+.endm |
|
| 94 |
+ |
|
| 95 |
+function ff_mdct_fixed_calc_neon, export=1 |
|
| 96 |
+ push {r1,r4-r11,lr}
|
|
| 97 |
+ |
|
| 98 |
+ ldr r4, [r0, #8] @ revtab |
|
| 99 |
+ ldr r6, [r0, #16] @ mdct_size; n |
|
| 100 |
+ ldr r7, [r0, #24] @ tcos |
|
| 101 |
+ |
|
| 102 |
+ prerot r1, r5 |
|
| 103 |
+ |
|
| 104 |
+ mov r4, r0 |
|
| 105 |
+ bl X(ff_fft_fixed_calc_neon) |
|
| 106 |
+ |
|
| 107 |
+ pop {r5}
|
|
| 108 |
+ mov r12, #-16 |
|
| 109 |
+ ldr r6, [r4, #16] @ mdct_size; n |
|
| 110 |
+ ldr r7, [r4, #24] @ tcos |
|
| 111 |
+ add r5, r5, r6, lsr #1 |
|
| 112 |
+ add r7, r7, r6, lsr #1 |
|
| 113 |
+ sub r1, r5, #16 |
|
| 114 |
+ sub r2, r7, #16 |
|
| 115 |
+1: |
|
| 116 |
+ vld2.16 {d4,d5}, [r7,:128]!
|
|
| 117 |
+ vld2.16 {d6,d7}, [r2,:128], r12
|
|
| 118 |
+ vld2.16 {d0,d1}, [r5,:128]
|
|
| 119 |
+ vld2.16 {d2,d3}, [r1,:128]
|
|
| 120 |
+ vrev64.16 q3, q3 |
|
| 121 |
+ vrev64.16 q1, q1 |
|
| 122 |
+ vneg.s16 q3, q3 |
|
| 123 |
+ vneg.s16 q2, q2 |
|
| 124 |
+ vmull.s16 q11, d2, d6 |
|
| 125 |
+ vmlal.s16 q11, d3, d7 |
|
| 126 |
+ vmull.s16 q8, d0, d5 |
|
| 127 |
+ vmlsl.s16 q8, d1, d4 |
|
| 128 |
+ vmull.s16 q9, d0, d4 |
|
| 129 |
+ vmlal.s16 q9, d1, d5 |
|
| 130 |
+ vmull.s16 q10, d2, d7 |
|
| 131 |
+ vmlsl.s16 q10, d3, d6 |
|
| 132 |
+ vshrn.s32 d0, q11, #15 |
|
| 133 |
+ vshrn.s32 d1, q8, #15 |
|
| 134 |
+ vshrn.s32 d2, q9, #15 |
|
| 135 |
+ vshrn.s32 d3, q10, #15 |
|
| 136 |
+ vrev64.16 q0, q0 |
|
| 137 |
+ vst2.16 {d2,d3}, [r5,:128]!
|
|
| 138 |
+ vst2.16 {d0,d1}, [r1,:128], r12
|
|
| 139 |
+ subs r6, r6, #32 |
|
| 140 |
+ bgt 1b |
|
| 141 |
+ |
|
| 142 |
+ pop {r4-r11,pc}
|
|
| 143 |
+endfunc |
|
| 144 |
+ |
|
| 145 |
+function ff_mdct_fixed_calcw_neon, export=1 |
|
| 146 |
+ push {r1,r4-r11,lr}
|
|
| 147 |
+ |
|
| 148 |
+ ldrd r4, r5, [r0, #8] @ revtab, tmp_buf |
|
| 149 |
+ ldr r6, [r0, #16] @ mdct_size; n |
|
| 150 |
+ ldr r7, [r0, #24] @ tcos |
|
| 151 |
+ |
|
| 152 |
+ prerot r5, r1 |
|
| 153 |
+ |
|
| 154 |
+ mov r4, r0 |
|
| 155 |
+ mov r1, r5 |
|
| 156 |
+ bl X(ff_fft_fixed_calc_neon) |
|
| 157 |
+ |
|
| 158 |
+ pop {r7}
|
|
| 159 |
+ mov r12, #-16 |
|
| 160 |
+ ldr r6, [r4, #16] @ mdct_size; n |
|
| 161 |
+ ldr r9, [r4, #24] @ tcos |
|
| 162 |
+ add r5, r5, r6, lsr #1 |
|
| 163 |
+ add r7, r7, r6 |
|
| 164 |
+ add r9, r9, r6, lsr #1 |
|
| 165 |
+ sub r3, r5, #16 |
|
| 166 |
+ sub r1, r7, #16 |
|
| 167 |
+ sub r2, r9, #16 |
|
| 168 |
+1: |
|
| 169 |
+ vld2.16 {d4,d5}, [r9,:128]!
|
|
| 170 |
+ vld2.16 {d6,d7}, [r2,:128], r12
|
|
| 171 |
+ vld2.16 {d0,d1}, [r5,:128]!
|
|
| 172 |
+ vld2.16 {d2,d3}, [r3,:128], r12
|
|
| 173 |
+ vrev64.16 q3, q3 |
|
| 174 |
+ vrev64.16 q1, q1 |
|
| 175 |
+ vneg.s16 q3, q3 |
|
| 176 |
+ vneg.s16 q2, q2 |
|
| 177 |
+ vmull.s16 q8, d2, d6 |
|
| 178 |
+ vmlal.s16 q8, d3, d7 |
|
| 179 |
+ vmull.s16 q9, d0, d5 |
|
| 180 |
+ vmlsl.s16 q9, d1, d4 |
|
| 181 |
+ vmull.s16 q10, d0, d4 |
|
| 182 |
+ vmlal.s16 q10, d1, d5 |
|
| 183 |
+ vmull.s16 q11, d2, d7 |
|
| 184 |
+ vmlsl.s16 q11, d3, d6 |
|
| 185 |
+ vrev64.32 q8, q8 |
|
| 186 |
+ vrev64.32 q9, q9 |
|
| 187 |
+ vst2.32 {q10,q11},[r7,:128]!
|
|
| 188 |
+ vst2.32 {d16,d18},[r1,:128], r12
|
|
| 189 |
+ vst2.32 {d17,d19},[r1,:128], r12
|
|
| 190 |
+ subs r6, r6, #32 |
|
| 191 |
+ bgt 1b |
|
| 192 |
+ |
|
| 193 |
+ pop {r4-r11,pc}
|
|
| 194 |
+endfunc |