FFmpeg
 All Data Structures Namespaces Files Functions Variables Typedefs Enumerations Enumerator Macros Groups Pages
acelp_filters_mips.c
Go to the documentation of this file.
1  /*
2  * Copyright (c) 2012
3  * MIPS Technologies, Inc., California.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  * notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  * notice, this list of conditions and the following disclaimer in the
12  * documentation and/or other materials provided with the distribution.
13  * 3. Neither the name of the MIPS Technologies, Inc., nor the names of its
14  * contributors may be used to endorse or promote products derived from
15  * this software without specific prior written permission.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE MIPS TECHNOLOGIES, INC. ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED. IN NO EVENT SHALL THE MIPS TECHNOLOGIES, INC. BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  *
29  * Author: Nedeljko Babic (nbabic@mips.com)
30  *
31  * various filters for ACELP-based codecs optimized for MIPS
32  *
33  * This file is part of FFmpeg.
34  *
35  * FFmpeg is free software; you can redistribute it and/or
36  * modify it under the terms of the GNU Lesser General Public
37  * License as published by the Free Software Foundation; either
38  * version 2.1 of the License, or (at your option) any later version.
39  *
40  * FFmpeg is distributed in the hope that it will be useful,
41  * but WITHOUT ANY WARRANTY; without even the implied warranty of
42  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
43  * Lesser General Public License for more details.
44  *
45  * You should have received a copy of the GNU Lesser General Public
46  * License along with FFmpeg; if not, write to the Free Software
47  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
48  */
49 
50 /**
51  * @file
52  * Reference: libavcodec/acelp_filters.c
53  */
54 #include "config.h"
55 #include "libavutil/attributes.h"
57 #include "libavutil/mips/asmdefs.h"
58 
59 #if HAVE_INLINE_ASM
60 static void ff_acelp_interpolatef_mips(float *out, const float *in,
61  const float *filter_coeffs, int precision,
62  int frac_pos, int filter_length, int length)
63 {
64  int n, i;
65  int prec = precision * 4;
66  int fc_offset = precision - frac_pos;
67  float in_val_p, in_val_m, fc_val_p, fc_val_m;
68 
69  for (n = 0; n < length; n++) {
70  /**
71  * four pointers are defined in order to minimize number of
72  * computations done in inner loop
73  */
74  const float *p_in_p = &in[n];
75  const float *p_in_m = &in[n-1];
76  const float *p_filter_coeffs_p = &filter_coeffs[frac_pos];
77  const float *p_filter_coeffs_m = filter_coeffs + fc_offset;
78  float v = 0;
79 
80  for (i = 0; i < filter_length;i++) {
81  __asm__ volatile (
82  "lwc1 %[in_val_p], 0(%[p_in_p]) \n\t"
83  "lwc1 %[fc_val_p], 0(%[p_filter_coeffs_p]) \n\t"
84  "lwc1 %[in_val_m], 0(%[p_in_m]) \n\t"
85  "lwc1 %[fc_val_m], 0(%[p_filter_coeffs_m]) \n\t"
86  PTR_ADDIU "%[p_in_p], %[p_in_p], 4 \n\t"
87  "madd.s %[v],%[v], %[in_val_p],%[fc_val_p] \n\t"
88  PTR_ADDIU "%[p_in_m], %[p_in_m], -4 \n\t"
89  PTR_ADDU "%[p_filter_coeffs_p],%[p_filter_coeffs_p], %[prec] \n\t"
90  PTR_ADDU "%[p_filter_coeffs_m],%[p_filter_coeffs_m], %[prec] \n\t"
91  "madd.s %[v],%[v],%[in_val_m], %[fc_val_m] \n\t"
92 
93  : [v] "+&f" (v),[p_in_p] "+r" (p_in_p), [p_in_m] "+r" (p_in_m),
94  [p_filter_coeffs_p] "+r" (p_filter_coeffs_p),
95  [in_val_p] "=&f" (in_val_p), [in_val_m] "=&f" (in_val_m),
96  [fc_val_p] "=&f" (fc_val_p), [fc_val_m] "=&f" (fc_val_m),
97  [p_filter_coeffs_m] "+r" (p_filter_coeffs_m)
98  : [prec] "r" (prec)
99  : "memory"
100  );
101  }
102  out[n] = v;
103  }
104 }
105 
106 static void ff_acelp_apply_order_2_transfer_function_mips(float *out, const float *in,
107  const float zero_coeffs[2],
108  const float pole_coeffs[2],
109  float gain, float mem[2], int n)
110 {
111  /**
112  * loop is unrolled eight times
113  */
114 
115  __asm__ volatile (
116  "lwc1 $f0, 0(%[mem]) \n\t"
117  "blez %[n], ff_acelp_apply_order_2_transfer_function_end%= \n\t"
118  "lwc1 $f1, 4(%[mem]) \n\t"
119  "lwc1 $f2, 0(%[pole_coeffs]) \n\t"
120  "lwc1 $f3, 4(%[pole_coeffs]) \n\t"
121  "lwc1 $f4, 0(%[zero_coeffs]) \n\t"
122  "lwc1 $f5, 4(%[zero_coeffs]) \n\t"
123 
124  "ff_acelp_apply_order_2_transfer_function_madd%=: \n\t"
125 
126  "lwc1 $f6, 0(%[in]) \n\t"
127  "mul.s $f9, $f3, $f1 \n\t"
128  "mul.s $f7, $f2, $f0 \n\t"
129  "msub.s $f7, $f7, %[gain], $f6 \n\t"
130  "sub.s $f7, $f7, $f9 \n\t"
131  "madd.s $f8, $f7, $f4, $f0 \n\t"
132  "madd.s $f8, $f8, $f5, $f1 \n\t"
133  "lwc1 $f11, 4(%[in]) \n\t"
134  "mul.s $f12, $f3, $f0 \n\t"
135  "mul.s $f13, $f2, $f7 \n\t"
136  "msub.s $f13, $f13, %[gain], $f11 \n\t"
137  "sub.s $f13, $f13, $f12 \n\t"
138  "madd.s $f14, $f13, $f4, $f7 \n\t"
139  "madd.s $f14, $f14, $f5, $f0 \n\t"
140  "swc1 $f8, 0(%[out]) \n\t"
141  "lwc1 $f6, 8(%[in]) \n\t"
142  "mul.s $f9, $f3, $f7 \n\t"
143  "mul.s $f15, $f2, $f13 \n\t"
144  "msub.s $f15, $f15, %[gain], $f6 \n\t"
145  "sub.s $f15, $f15, $f9 \n\t"
146  "madd.s $f8, $f15, $f4, $f13 \n\t"
147  "madd.s $f8, $f8, $f5, $f7 \n\t"
148  "swc1 $f14, 4(%[out]) \n\t"
149  "lwc1 $f11, 12(%[in]) \n\t"
150  "mul.s $f12, $f3, $f13 \n\t"
151  "mul.s $f16, $f2, $f15 \n\t"
152  "msub.s $f16, $f16, %[gain], $f11 \n\t"
153  "sub.s $f16, $f16, $f12 \n\t"
154  "madd.s $f14, $f16, $f4, $f15 \n\t"
155  "madd.s $f14, $f14, $f5, $f13 \n\t"
156  "swc1 $f8, 8(%[out]) \n\t"
157  "lwc1 $f6, 16(%[in]) \n\t"
158  "mul.s $f9, $f3, $f15 \n\t"
159  "mul.s $f7, $f2, $f16 \n\t"
160  "msub.s $f7, $f7, %[gain], $f6 \n\t"
161  "sub.s $f7, $f7, $f9 \n\t"
162  "madd.s $f8, $f7, $f4, $f16 \n\t"
163  "madd.s $f8, $f8, $f5, $f15 \n\t"
164  "swc1 $f14, 12(%[out]) \n\t"
165  "lwc1 $f11, 20(%[in]) \n\t"
166  "mul.s $f12, $f3, $f16 \n\t"
167  "mul.s $f13, $f2, $f7 \n\t"
168  "msub.s $f13, $f13, %[gain], $f11 \n\t"
169  "sub.s $f13, $f13, $f12 \n\t"
170  "madd.s $f14, $f13, $f4, $f7 \n\t"
171  "madd.s $f14, $f14, $f5, $f16 \n\t"
172  "swc1 $f8, 16(%[out]) \n\t"
173  "lwc1 $f6, 24(%[in]) \n\t"
174  "mul.s $f9, $f3, $f7 \n\t"
175  "mul.s $f15, $f2, $f13 \n\t"
176  "msub.s $f15, $f15, %[gain], $f6 \n\t"
177  "sub.s $f1, $f15, $f9 \n\t"
178  "madd.s $f8, $f1, $f4, $f13 \n\t"
179  "madd.s $f8, $f8, $f5, $f7 \n\t"
180  "swc1 $f14, 20(%[out]) \n\t"
181  "lwc1 $f11, 28(%[in]) \n\t"
182  "mul.s $f12, $f3, $f13 \n\t"
183  "mul.s $f16, $f2, $f1 \n\t"
184  "msub.s $f16, $f16, %[gain], $f11 \n\t"
185  "sub.s $f0, $f16, $f12 \n\t"
186  "madd.s $f14, $f0, $f4, $f1 \n\t"
187  "madd.s $f14, $f14, $f5, $f13 \n\t"
188  "swc1 $f8, 24(%[out]) \n\t"
189  PTR_ADDIU "%[out], 32 \n\t"
190  PTR_ADDIU "%[in], 32 \n\t"
191  "addiu %[n], -8 \n\t"
192  "swc1 $f14, -4(%[out]) \n\t"
193  "bnez %[n], ff_acelp_apply_order_2_transfer_function_madd%= \n\t"
194  "swc1 $f1, 4(%[mem]) \n\t"
195  "swc1 $f0, 0(%[mem]) \n\t"
196 
197  "ff_acelp_apply_order_2_transfer_function_end%=: \n\t"
198 
199  : [out] "+r" (out),
200  [in] "+r" (in), [gain] "+f" (gain),
201  [n] "+r" (n), [mem] "+r" (mem)
202  : [zero_coeffs] "r" (zero_coeffs),
203  [pole_coeffs] "r" (pole_coeffs)
204  : "$f0", "$f1", "$f2", "$f3", "$f4", "$f5",
205  "$f6", "$f7", "$f8", "$f9", "$f10", "$f11",
206  "$f12", "$f13", "$f14", "$f15", "$f16", "memory"
207  );
208 }
209 #endif /* HAVE_INLINE_ASM */
210 
212 {
213 #if HAVE_INLINE_ASM
214  c->acelp_interpolatef = ff_acelp_interpolatef_mips;
215  c->acelp_apply_order_2_transfer_function = ff_acelp_apply_order_2_transfer_function_mips;
216 #endif
217 }
float v
MIPS assembly defines from sys/asm.h but rewritten for use with C inline assembly (rather than from w...
void(* acelp_apply_order_2_transfer_function)(float *out, const float *in, const float zero_coeffs[2], const float pole_coeffs[2], float gain, float mem[2], int n)
Apply an order 2 rational transfer function in-place.
Definition: acelp_filters.h:47
Macro definitions for various function/variable attributes.
int mem
Definition: avisynth_c.h:684
GLsizei GLsizei * length
Definition: opengl_enc.c:115
int n
Definition: avisynth_c.h:547
uint8_t pi<< 24) CONV_FUNC_GROUP(AV_SAMPLE_FMT_FLT, float, AV_SAMPLE_FMT_U8, uint8_t,(*(constuint8_t *) pi-0x80)*(1.0f/(1<< 7))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_DBL, double, AV_SAMPLE_FMT_U8, uint8_t,(*(constuint8_t *) pi-0x80)*(1.0/(1<< 7))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_U8, uint8_t, AV_SAMPLE_FMT_S16, int16_t,(*(constint16_t *) pi >>8)+0x80) CONV_FUNC_GROUP(AV_SAMPLE_FMT_FLT, float, AV_SAMPLE_FMT_S16, int16_t,*(constint16_t *) pi *(1.0f/(1<< 15))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_DBL, double, AV_SAMPLE_FMT_S16, int16_t,*(constint16_t *) pi *(1.0/(1<< 15))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_U8, uint8_t, AV_SAMPLE_FMT_S32, int32_t,(*(constint32_t *) pi >>24)+0x80) CONV_FUNC_GROUP(AV_SAMPLE_FMT_FLT, float, AV_SAMPLE_FMT_S32, int32_t,*(constint32_t *) pi *(1.0f/(1U<< 31))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_DBL, double, AV_SAMPLE_FMT_S32, int32_t,*(constint32_t *) pi *(1.0/(1U<< 31))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_U8, uint8_t, AV_SAMPLE_FMT_FLT, float, av_clip_uint8(lrintf(*(constfloat *) pi *(1<< 7))+0x80)) CONV_FUNC_GROUP(AV_SAMPLE_FMT_S16, int16_t, AV_SAMPLE_FMT_FLT, float, av_clip_int16(lrintf(*(constfloat *) pi *(1<< 15)))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_S32, int32_t, AV_SAMPLE_FMT_FLT, float, av_clipl_int32(llrintf(*(constfloat *) pi *(1U<< 31)))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_U8, uint8_t, AV_SAMPLE_FMT_DBL, double, av_clip_uint8(lrint(*(constdouble *) pi *(1<< 7))+0x80)) CONV_FUNC_GROUP(AV_SAMPLE_FMT_S16, int16_t, AV_SAMPLE_FMT_DBL, double, av_clip_int16(lrint(*(constdouble *) pi *(1<< 15)))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_S32, int32_t, AV_SAMPLE_FMT_DBL, double, av_clipl_int32(llrint(*(constdouble *) pi *(1U<< 31))))#defineSET_CONV_FUNC_GROUP(ofmt, ifmt) staticvoidset_generic_function(AudioConvert *ac){}voidff_audio_convert_free(AudioConvert **ac){if(!*ac) return;ff_dither_free(&(*ac) ->dc);av_freep(ac);}AudioConvert *ff_audio_convert_alloc(AVAudioResampleContext *avr, enumAVSampleFormatout_fmt, enumAVSampleFormatin_fmt, intchannels, intsample_rate, intapply_map){AudioConvert *ac;intin_planar, out_planar;ac=av_mallocz(sizeof(*ac));if(!ac) returnNULL;ac->avr=avr;ac->out_fmt=out_fmt;ac->in_fmt=in_fmt;ac->channels=channels;ac->apply_map=apply_map;if(avr->dither_method!=AV_RESAMPLE_DITHER_NONE &&av_get_packed_sample_fmt(out_fmt)==AV_SAMPLE_FMT_S16 &&av_get_bytes_per_sample(in_fmt)>2){ac->dc=ff_dither_alloc(avr, out_fmt, in_fmt, channels, sample_rate, apply_map);if(!ac->dc){av_free(ac);returnNULL;}returnac;}in_planar=ff_sample_fmt_is_planar(in_fmt, channels);out_planar=ff_sample_fmt_is_planar(out_fmt, channels);if(in_planar==out_planar){ac->func_type=CONV_FUNC_TYPE_FLAT;ac->planes=in_planar?ac->channels:1;}elseif(in_planar) ac->func_type=CONV_FUNC_TYPE_INTERLEAVE;elseac->func_type=CONV_FUNC_TYPE_DEINTERLEAVE;set_generic_function(ac);if(ARCH_AARCH64) ff_audio_convert_init_aarch64(ac);if(ARCH_ARM) ff_audio_convert_init_arm(ac);if(ARCH_X86) ff_audio_convert_init_x86(ac);returnac;}intff_audio_convert(AudioConvert *ac, AudioData *out, AudioData *in){intuse_generic=1;intlen=in->nb_samples;intp;if(ac->dc){av_log(ac->avr, AV_LOG_TRACE,"%dsamples-audio_convert:%sto%s(dithered)\n", len, av_get_sample_fmt_name(ac->in_fmt), av_get_sample_fmt_name(ac->out_fmt));returnff_convert_dither(ac-> in
#define PTR_ADDIU
Definition: asmdefs.h:41
static double c[64]
uint8_t pi<< 24) CONV_FUNC_GROUP(AV_SAMPLE_FMT_FLT, float, AV_SAMPLE_FMT_U8, uint8_t,(*(constuint8_t *) pi-0x80)*(1.0f/(1<< 7))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_DBL, double, AV_SAMPLE_FMT_U8, uint8_t,(*(constuint8_t *) pi-0x80)*(1.0/(1<< 7))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_U8, uint8_t, AV_SAMPLE_FMT_S16, int16_t,(*(constint16_t *) pi >>8)+0x80) CONV_FUNC_GROUP(AV_SAMPLE_FMT_FLT, float, AV_SAMPLE_FMT_S16, int16_t,*(constint16_t *) pi *(1.0f/(1<< 15))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_DBL, double, AV_SAMPLE_FMT_S16, int16_t,*(constint16_t *) pi *(1.0/(1<< 15))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_U8, uint8_t, AV_SAMPLE_FMT_S32, int32_t,(*(constint32_t *) pi >>24)+0x80) CONV_FUNC_GROUP(AV_SAMPLE_FMT_FLT, float, AV_SAMPLE_FMT_S32, int32_t,*(constint32_t *) pi *(1.0f/(1U<< 31))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_DBL, double, AV_SAMPLE_FMT_S32, int32_t,*(constint32_t *) pi *(1.0/(1U<< 31))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_U8, uint8_t, AV_SAMPLE_FMT_FLT, float, av_clip_uint8(lrintf(*(constfloat *) pi *(1<< 7))+0x80)) CONV_FUNC_GROUP(AV_SAMPLE_FMT_S16, int16_t, AV_SAMPLE_FMT_FLT, float, av_clip_int16(lrintf(*(constfloat *) pi *(1<< 15)))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_S32, int32_t, AV_SAMPLE_FMT_FLT, float, av_clipl_int32(llrintf(*(constfloat *) pi *(1U<< 31)))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_U8, uint8_t, AV_SAMPLE_FMT_DBL, double, av_clip_uint8(lrint(*(constdouble *) pi *(1<< 7))+0x80)) CONV_FUNC_GROUP(AV_SAMPLE_FMT_S16, int16_t, AV_SAMPLE_FMT_DBL, double, av_clip_int16(lrint(*(constdouble *) pi *(1<< 15)))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_S32, int32_t, AV_SAMPLE_FMT_DBL, double, av_clipl_int32(llrint(*(constdouble *) pi *(1U<< 31))))#defineSET_CONV_FUNC_GROUP(ofmt, ifmt) staticvoidset_generic_function(AudioConvert *ac){}voidff_audio_convert_free(AudioConvert **ac){if(!*ac) return;ff_dither_free(&(*ac) ->dc);av_freep(ac);}AudioConvert *ff_audio_convert_alloc(AVAudioResampleContext *avr, enumAVSampleFormatout_fmt, enumAVSampleFormatin_fmt, intchannels, intsample_rate, intapply_map){AudioConvert *ac;intin_planar, out_planar;ac=av_mallocz(sizeof(*ac));if(!ac) returnNULL;ac->avr=avr;ac->out_fmt=out_fmt;ac->in_fmt=in_fmt;ac->channels=channels;ac->apply_map=apply_map;if(avr->dither_method!=AV_RESAMPLE_DITHER_NONE &&av_get_packed_sample_fmt(out_fmt)==AV_SAMPLE_FMT_S16 &&av_get_bytes_per_sample(in_fmt)>2){ac->dc=ff_dither_alloc(avr, out_fmt, in_fmt, channels, sample_rate, apply_map);if(!ac->dc){av_free(ac);returnNULL;}returnac;}in_planar=ff_sample_fmt_is_planar(in_fmt, channels);out_planar=ff_sample_fmt_is_planar(out_fmt, channels);if(in_planar==out_planar){ac->func_type=CONV_FUNC_TYPE_FLAT;ac->planes=in_planar?ac->channels:1;}elseif(in_planar) ac->func_type=CONV_FUNC_TYPE_INTERLEAVE;elseac->func_type=CONV_FUNC_TYPE_DEINTERLEAVE;set_generic_function(ac);if(ARCH_AARCH64) ff_audio_convert_init_aarch64(ac);if(ARCH_ARM) ff_audio_convert_init_arm(ac);if(ARCH_X86) ff_audio_convert_init_x86(ac);returnac;}intff_audio_convert(AudioConvert *ac, AudioData *out, AudioData *in){intuse_generic=1;intlen=in->nb_samples;intp;if(ac->dc){av_log(ac->avr, AV_LOG_TRACE,"%dsamples-audio_convert:%sto%s(dithered)\n", len, av_get_sample_fmt_name(ac->in_fmt), av_get_sample_fmt_name(ac->out_fmt));returnff_convert_dither(ac-> out
#define PTR_ADDU
Definition: asmdefs.h:40
void ff_acelp_filter_init_mips(ACELPFContext *c)
void(* acelp_interpolatef)(float *out, const float *in, const float *filter_coeffs, int precision, int frac_pos, int filter_length, int length)
Floating point version of ff_acelp_interpolate()
Definition: acelp_filters.h:32