swresample/arm: add ff_resample_common_apply_filter_{x4,x8}_{float,s16}_neon

This commit is contained in:
Matthieu Bouron 2016-05-08 18:56:03 +00:00
parent 17ea1cd559
commit f6265a5cbc
5 changed files with 199 additions and 2 deletions

View File

@ -1,5 +1,8 @@
OBJS += arm/audio_convert_init.o
OBJS += arm/audio_convert_init.o \
arm/resample_init.o
OBJS-$(CONFIG_NEON_CLOBBER_TEST) += arm/neontest.o
NEON-OBJS += arm/audio_convert_neon.o
NEON-OBJS += arm/audio_convert_neon.o \
arm/resample.o

View File

@ -0,0 +1,77 @@
/*
* Copyright (c) 2016 Matthieu Bouron <matthieu.bouron stupeflix.com>
*
* This file is part of FFmpeg.
*
* FFmpeg is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
* FFmpeg is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with FFmpeg; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
#include "libavutil/arm/asm.S"
function ff_resample_common_apply_filter_x4_float_neon, export=1
vmov.f32 q0, #0.0 @ accumulator
1: vld1.32 {q1}, [r1]! @ src
vld1.32 {q2}, [r2]! @ filter
vmla.f32 q0, q1, q2 @ accumulator += src + {0..3} * filter + {0..3}
subs r3, #4 @ filter_length -= 4
bgt 1b @ loop until filter_length
vpadd.f32 d0, d0, d1 @ pair adding of the 4x32-bit accumulated values
vpadd.f32 d0, d0, d0 @ pair adding of the 4x32-bit accumulator values
vst1.32 {d0[0]}, [r0] @ write accumulator
mov pc, lr
endfunc
function ff_resample_common_apply_filter_x8_float_neon, export=1
vmov.f32 q0, #0.0 @ accumulator
1: vld1.32 {q1}, [r1]! @ src
vld1.32 {q2}, [r2]! @ filter
vld1.32 {q8}, [r1]! @ src
vld1.32 {q9}, [r2]! @ filter
vmla.f32 q0, q1, q2 @ accumulator += src + {0..3} * filter + {0..3}
vmla.f32 q0, q8, q9 @ accumulator += src + {4..7} * filter + {4..7}
subs r3, #8 @ filter_length -= 8
bgt 1b @ loop until filter_length
vpadd.f32 d0, d0, d1 @ pair adding of the 4x32-bit accumulated values
vpadd.f32 d0, d0, d0 @ pair adding of the 4x32-bit accumulator values
vst1.32 {d0[0]}, [r0] @ write accumulator
mov pc, lr
endfunc
function ff_resample_common_apply_filter_x4_s16_neon, export=1
vmov.s32 q0, #0 @ accumulator
1: vld1.16 {d2}, [r1]! @ src
vld1.16 {d4}, [r2]! @ filter
vmlal.s16 q0, d2, d4 @ accumulator += src + {0..3} * filter + {0..3}
subs r3, #4 @ filter_length -= 4
bgt 1b @ loop until filter_length
vpadd.s32 d0, d0, d1 @ pair adding of the 4x32-bit accumulated values
vpadd.s32 d0, d0, d0 @ pair adding of the 4x32-bit accumulator values
vst1.32 {d0[0]}, [r0] @ write accumulator
mov pc, lr
endfunc
function ff_resample_common_apply_filter_x8_s16_neon, export=1
vmov.s32 q0, #0 @ accumulator
1: vld1.16 {q1}, [r1]! @ src
vld1.16 {q2}, [r2]! @ filter
vmlal.s16 q0, d2, d4 @ accumulator += src + {0..3} * filter + {0..3}
vmlal.s16 q0, d3, d5 @ accumulator += src + {4..7} * filter + {4..7}
subs r3, #8 @ filter_length -= 8
bgt 1b @ loop until filter_length
vpadd.s32 d0, d0, d1 @ pair adding of the 4x32-bit accumulated values
vpadd.s32 d0, d0, d0 @ pair adding of the 4x32-bit accumulator values
vst1.32 {d0[0]}, [r0] @ write accumulator
mov pc, lr
endfunc

View File

@ -0,0 +1,115 @@
/*
* Audio resampling
*
* Copyright (c) 2004-2012 Michael Niedermayer <michaelni@gmx.at>
*
* This file is part of FFmpeg.
*
* FFmpeg is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
* FFmpeg is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with FFmpeg; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
#include "config.h"
#include "libavutil/cpu.h"
#include "libavutil/avassert.h"
#include "libavutil/arm/cpu.h"
#include "libswresample/resample.h"
#define DECLARE_RESAMPLE_COMMON_TEMPLATE(TYPE, DELEM, FELEM, FELEM2, OUT) \
\
void ff_resample_common_apply_filter_x4_##TYPE##_neon(FELEM2 *acc, const DELEM *src, \
const FELEM *filter, int length); \
\
void ff_resample_common_apply_filter_x8_##TYPE##_neon(FELEM2 *acc, const DELEM *src, \
const FELEM *filter, int length); \
\
static int ff_resample_common_##TYPE##_neon(ResampleContext *c, void *dest, const void *source, \
int n, int update_ctx) \
{ \
DELEM *dst = dest; \
const DELEM *src = source; \
int dst_index; \
int index= c->index; \
int frac= c->frac; \
int sample_index = index >> c->phase_shift; \
int x4_aligned_filter_length = c->filter_length & ~3; \
int x8_aligned_filter_length = c->filter_length & ~7; \
\
index &= c->phase_mask; \
for (dst_index = 0; dst_index < n; dst_index++) { \
FELEM *filter = ((FELEM *) c->filter_bank) + c->filter_alloc * index; \
\
FELEM2 val=0; \
int i = 0; \
if (x8_aligned_filter_length >= 8) { \
ff_resample_common_apply_filter_x8_##TYPE##_neon(&val, &src[sample_index], \
filter, x8_aligned_filter_length); \
i += x8_aligned_filter_length; \
\
} else if (x4_aligned_filter_length >= 4) { \
ff_resample_common_apply_filter_x4_##TYPE##_neon(&val, &src[sample_index], \
filter, x4_aligned_filter_length); \
i += x4_aligned_filter_length; \
} \
for (; i < c->filter_length; i++) { \
val += src[sample_index + i] * (FELEM2)filter[i]; \
} \
OUT(dst[dst_index], val); \
\
frac += c->dst_incr_mod; \
index += c->dst_incr_div; \
if (frac >= c->src_incr) { \
frac -= c->src_incr; \
index++; \
} \
sample_index += index >> c->phase_shift; \
index &= c->phase_mask; \
} \
\
if(update_ctx){ \
c->frac= frac; \
c->index= index; \
} \
\
return sample_index; \
} \
#define OUT(d, v) d = v
DECLARE_RESAMPLE_COMMON_TEMPLATE(float, float, float, float, OUT)
#undef OUT
#define OUT(d, v) (v) = ((v) + (1<<(14)))>>15; (d) = av_clip_int16(v)
DECLARE_RESAMPLE_COMMON_TEMPLATE(s16, int16_t, int16_t, int32_t, OUT)
#undef OUT
av_cold void swri_resample_dsp_arm_init(ResampleContext *c)
{
int cpu_flags = av_get_cpu_flags();
if (!have_neon(cpu_flags))
return;
switch(c->format) {
case AV_SAMPLE_FMT_FLTP:
if (!c->linear)
c->dsp.resample = ff_resample_common_float_neon;
break;
case AV_SAMPLE_FMT_S16P:
if (!c->linear)
c->dsp.resample = ff_resample_common_s16_neon;
break;
}
}

View File

@ -60,5 +60,6 @@ typedef struct ResampleContext {
void swri_resample_dsp_init(ResampleContext *c);
void swri_resample_dsp_x86_init(ResampleContext *c);
void swri_resample_dsp_arm_init(ResampleContext *c);
#endif /* SWRESAMPLE_RESAMPLE_H */

View File

@ -65,4 +65,5 @@ void swri_resample_dsp_init(ResampleContext *c)
}
if (ARCH_X86) swri_resample_dsp_x86_init(c);
else if (ARCH_ARM) swri_resample_dsp_arm_init(c);
}