mirror of
https://github.com/libretro/RetroArch
synced 2025-02-20 15:40:44 +00:00
Stylistic cleanups in CC resampler.
This commit is contained in:
Themaister
parent
0c57a1726b
commit
4d9ff7d147
@ -1,10 +1,21 @@
|
||||
/*
|
||||
* Convoluted Cosine Resampler
|
||||
* Copyright (C) 2014 - Ali Bouhlel ( aliaspider@gmail.com )
|
||||
/* RetroArch - A frontend for libretro.
|
||||
* Copyright (C) 2010-2014 - Hans-Kristian Arntzen
|
||||
* Copyright (C) 2014 - Ali Bouhlel ( aliaspider@gmail.com )
|
||||
*
|
||||
* RetroArch is free software: you can redistribute it and/or modify it under the terms
|
||||
* of the GNU General Public License as published by the Free Software Found-
|
||||
* ation, either version 3 of the License, or (at your option) any later version.
|
||||
*
|
||||
* licence: GPLv3
|
||||
* RetroArch 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 General Public License for more details.
|
||||
*
|
||||
* You should have received a copy of the GNU General Public License along with RetroArch.
|
||||
* If not, see <http://www.gnu.org/licenses/>.
|
||||
*/
|
||||
|
||||
// Convoluted Cosine Resampler
|
||||
|
||||
#include "resampler.h"
|
||||
#include "../libretro.h"
|
||||
#include "../performance.h"
|
||||
@ -19,206 +30,162 @@
|
||||
#define RARCH_LOG(...) fprintf(stderr, __VA_ARGS__)
|
||||
#endif
|
||||
|
||||
typedef struct audio_frame_float
|
||||
{
|
||||
float l;
|
||||
float r;
|
||||
} audio_frame_float_t;
|
||||
|
||||
typedef struct audio_frame_int16
|
||||
{
|
||||
int16_t l;
|
||||
int16_t r;
|
||||
} audio_frame_int16_t;
|
||||
|
||||
#ifdef _MIPS_ARCH_ALLEGREX1
|
||||
typedef struct rarch_CC_resampler
|
||||
{
|
||||
int dummy;
|
||||
}rarch_CC_resampler_t;
|
||||
|
||||
static void resampler_CC_process(void *re_, struct resampler_data *data)
|
||||
{
|
||||
(void)re_;
|
||||
// rarch_CC_resampler_t *re = (rarch_CC_resampler_t*)re_;
|
||||
float ratio,fraction;
|
||||
|
||||
typedef struct audio_frame_float
|
||||
{
|
||||
float l;
|
||||
float r;
|
||||
}audio_frame_float_t;
|
||||
|
||||
typedef struct audio_frame_int16
|
||||
{
|
||||
int16_t l;
|
||||
int16_t r;
|
||||
}audio_frame_int16_t;
|
||||
|
||||
float ratio, fraction;
|
||||
|
||||
audio_frame_float_t *inp = (audio_frame_float_t*)data->data_in;
|
||||
audio_frame_float_t *inp_max = inp + data->input_frames;
|
||||
audio_frame_float_t *outp = (audio_frame_float_t*)data->data_out;
|
||||
|
||||
__asm__ (
|
||||
".set push\n"
|
||||
".set noreorder\n"
|
||||
".set push\n"
|
||||
".set noreorder\n"
|
||||
|
||||
"mtv %2, s700 \n" // 700 = data->ratio = b
|
||||
// "vsat0.s s700, s700 \n"
|
||||
"vrcp.s s701, s700 \n" // 701 = 1.0 / b
|
||||
"vadd.s s702, s700, s700 \n" // 702 = 2 * b
|
||||
"vmul.s s703, s700, s710 \n" // 703 = b * pi
|
||||
"mtv %2, s700 \n" // 700 = data->ratio = b
|
||||
// "vsat0.s s700, s700 \n"
|
||||
"vrcp.s s701, s700 \n" // 701 = 1.0 / b
|
||||
"vadd.s s702, s700, s700 \n" // 702 = 2 * b
|
||||
"vmul.s s703, s700, s710 \n" // 703 = b * pi
|
||||
|
||||
"mfv %0, s701 \n"
|
||||
"mfv %1, s730 \n"
|
||||
"mfv %0, s701 \n"
|
||||
"mfv %1, s730 \n"
|
||||
|
||||
".set pop\n"
|
||||
:"=r"(ratio),"=r"(fraction): "r"((float)data->ratio)
|
||||
".set pop\n"
|
||||
: "=r"(ratio), "=r"(fraction)
|
||||
: "r"((float)data->ratio)
|
||||
);
|
||||
|
||||
while(true)
|
||||
for (;;)
|
||||
{
|
||||
while ((fraction < ratio))
|
||||
while (fraction < ratio)
|
||||
{
|
||||
__asm__ (
|
||||
".set push \n"
|
||||
".set noreorder \n"
|
||||
".set push \n"
|
||||
".set noreorder \n"
|
||||
|
||||
"lv.s s620, 0(%1) \n"
|
||||
"lv.s s621, 4(%1) \n"
|
||||
"lv.s s620, 0(%1) \n"
|
||||
"lv.s s621, 4(%1) \n"
|
||||
|
||||
"vsub.s s731, s701, s730 \n"
|
||||
"vsub.s s731, s701, s730 \n"
|
||||
|
||||
"vadd.q c600, c730[-X,Y,-X,Y], c730[1/2,1/2,-1/2,-1/2]\n"
|
||||
"vadd.q c600, c730[-X,Y,-X,Y], c730[1/2,1/2,-1/2,-1/2]\n"
|
||||
|
||||
"vmul.q c610, c600, c700[Z,Z,Z,Z] \n" //*2*b
|
||||
"vmul.q c600, c600, c700[W,W,W,W] \n" //*b*pi
|
||||
"vsin.q c610, c610 \n"
|
||||
"vadd.q c600, c600, c610 \n"
|
||||
"vmul.q c610, c600, c700[Z,Z,Z,Z] \n" //*2*b
|
||||
"vmul.q c600, c600, c700[W,W,W,W] \n" //*b*pi
|
||||
"vsin.q c610, c610 \n"
|
||||
"vadd.q c600, c600, c610 \n"
|
||||
|
||||
"vmul.q c600[-1:1,-1:1,-1:1,-1:1], c600, c710[Y,Y,Y,Y] \n"
|
||||
"vmul.q c600[-1:1,-1:1,-1:1,-1:1], c600, c710[Y,Y,Y,Y] \n"
|
||||
|
||||
"vsub.p c600, c600, c602 \n"
|
||||
"vsub.p c600, c600, c602 \n"
|
||||
|
||||
"vmul.q c620, c620[X,Y,X,Y], c600[X,X,Y,Y] \n"
|
||||
"vmul.q c620, c620[X,Y,X,Y], c600[X,X,Y,Y] \n"
|
||||
|
||||
"vadd.q c720, c720, c620 \n"
|
||||
"vadd.q c720, c720, c620 \n"
|
||||
|
||||
|
||||
"vadd.s s730, s730, s730[1] \n"
|
||||
"mfv %0, s730 \n"
|
||||
"vadd.s s730, s730, s730[1] \n"
|
||||
"mfv %0, s730 \n"
|
||||
|
||||
".set pop \n"
|
||||
: "=r"(fraction)
|
||||
: "r"(inp));
|
||||
|
||||
".set pop \n"
|
||||
:"=r"(fraction): "r"(inp)
|
||||
);
|
||||
inp++;
|
||||
if (inp == inp_max)
|
||||
goto done;
|
||||
}
|
||||
__asm__ (
|
||||
".set push \n"
|
||||
".set noreorder \n"
|
||||
".set push \n"
|
||||
".set noreorder \n"
|
||||
|
||||
"vmul.p c720, c720, c720[1/2,1/2] \n"
|
||||
"sv.s s720, 0(%1) \n"
|
||||
"sv.s s721, 4(%1) \n"
|
||||
"vmov.q c720, c720[Z,W,0,0] \n"
|
||||
"vsub.s s730, s730, s701 \n"
|
||||
"mfv %0, s730 \n"
|
||||
"vmul.p c720, c720, c720[1/2,1/2] \n"
|
||||
"sv.s s720, 0(%1) \n"
|
||||
"sv.s s721, 4(%1) \n"
|
||||
"vmov.q c720, c720[Z,W,0,0] \n"
|
||||
"vsub.s s730, s730, s701 \n"
|
||||
"mfv %0, s730 \n"
|
||||
|
||||
".set pop \n"
|
||||
: "=r"(fraction)
|
||||
: "r"(outp));
|
||||
|
||||
".set pop \n"
|
||||
:"=r"(fraction): "r"(outp)
|
||||
);
|
||||
outp++;
|
||||
}
|
||||
|
||||
// The VFPU state is assumed to remain intact in-between calls to resampler_CC_process.
|
||||
|
||||
done:
|
||||
data->output_frames = (outp - (audio_frame_float_t*)data->data_out);
|
||||
data->output_frames = outp - (audio_frame_float_t*)data->data_out;
|
||||
}
|
||||
|
||||
|
||||
static void resampler_CC_free(void *re_)
|
||||
{
|
||||
rarch_CC_resampler_t *re = (rarch_CC_resampler_t*)re_;
|
||||
if (re)
|
||||
free(re);
|
||||
(void)re_;
|
||||
}
|
||||
|
||||
static void *resampler_CC_init(double bandwidth_mod)
|
||||
{
|
||||
rarch_CC_resampler_t *re = (rarch_CC_resampler_t*)calloc(1, sizeof(rarch_CC_resampler_t));
|
||||
if (!re)
|
||||
return NULL;
|
||||
|
||||
__asm__ (
|
||||
".set push\n"
|
||||
".set noreorder\n"
|
||||
".set push\n"
|
||||
".set noreorder\n"
|
||||
|
||||
"vcst.s s710, VFPU_PI \n" // 710 = pi
|
||||
"vcst.s s711, VFPU_1_PI \n" // 711 = 1.0 / (pi)
|
||||
"vcst.s s710, VFPU_PI \n" // 710 = pi
|
||||
"vcst.s s711, VFPU_1_PI \n" // 711 = 1.0 / (pi)
|
||||
|
||||
"vzero.q c720 \n"
|
||||
"vzero.q c730 \n"
|
||||
"vzero.q c720 \n"
|
||||
"vzero.q c730 \n"
|
||||
|
||||
".set pop\n"
|
||||
);
|
||||
".set pop\n");
|
||||
|
||||
RARCH_LOG("\nConvoluted Cosine resampler (VFPU): \n");
|
||||
return re;
|
||||
return (void*)-1;
|
||||
}
|
||||
#else
|
||||
|
||||
|
||||
//#define HAVE_SSE_MATHFUN_H
|
||||
#if defined(__SSE2__) && defined(HAVE_SSE_MATHFUN_H)
|
||||
#define USE_SSE2
|
||||
#include "sse_mathfun.h"
|
||||
|
||||
static inline float _mm_sin(float x)
|
||||
{
|
||||
static float temp;
|
||||
__m128 vector = _mm_set1_ps(x);
|
||||
vector = sin_ps(vector);
|
||||
_mm_store1_ps(&temp,vector);
|
||||
return temp;
|
||||
}
|
||||
static inline float _mm_cos(float x)
|
||||
{
|
||||
static float temp;
|
||||
__m128 vector = _mm_set1_ps(x);
|
||||
vector = cos_ps(vector);
|
||||
_mm_store1_ps(&temp,vector);
|
||||
return temp;
|
||||
}
|
||||
|
||||
#define sin(x) _mm_sin(x)
|
||||
#define cos(x) _mm_cos(x)
|
||||
#endif
|
||||
|
||||
|
||||
typedef struct audio_frame_float
|
||||
{
|
||||
float l;
|
||||
float r;
|
||||
}audio_frame_float_t;
|
||||
|
||||
// C reference version. Not optimized.
|
||||
typedef struct rarch_CC_resampler
|
||||
{
|
||||
audio_frame_float_t buffer[4];
|
||||
float distance;
|
||||
void (*process)(void *re, struct resampler_data *data);
|
||||
} rarch_CC_resampler_t;
|
||||
|
||||
}rarch_CC_resampler_t;
|
||||
|
||||
|
||||
|
||||
static inline float cc_int(float x, float b){
|
||||
float val = x * b * M_PI + sin(x * b * M_PI);
|
||||
return (val > M_PI)? M_PI : (val < -M_PI)? -M_PI : val;
|
||||
static inline float cc_int(float x, float b)
|
||||
{
|
||||
float val = x * b * M_PI + sinf(x * b * M_PI);
|
||||
return (val > M_PI) ? M_PI : (val < -M_PI) ? -M_PI : val;
|
||||
}
|
||||
|
||||
static inline float cc_kernel(float x, float b){
|
||||
static inline float cc_kernel(float x, float b)
|
||||
{
|
||||
return (cc_int(x + 0.5, b) - cc_int(x - 0.5, b)) / (2.0 * M_PI);
|
||||
}
|
||||
|
||||
static inline void add_to(const audio_frame_float_t* source,audio_frame_float_t* target, float ratio){
|
||||
static inline void add_to(const audio_frame_float_t *source, audio_frame_float_t *target, float ratio)
|
||||
{
|
||||
target->l += source->l * ratio;
|
||||
target->r += source->r * ratio;
|
||||
}
|
||||
|
||||
static void resampler_CC_downsample(void *re_, struct resampler_data *data)
|
||||
{
|
||||
|
||||
rarch_CC_resampler_t *re = (rarch_CC_resampler_t*)re_;
|
||||
|
||||
audio_frame_float_t *inp = (audio_frame_float_t*)data->data_in;
|
||||
@ -227,9 +194,9 @@ static void resampler_CC_downsample(void *re_, struct resampler_data *data)
|
||||
|
||||
float ratio = 1.0 / data->ratio;
|
||||
|
||||
float b = data->ratio; // cutoff frequency
|
||||
float b = data->ratio; // cutoff frequency
|
||||
|
||||
while(inp != inp_max)
|
||||
while (inp != inp_max)
|
||||
{
|
||||
add_to(inp, re->buffer + 0, cc_kernel(re->distance, b));
|
||||
add_to(inp, re->buffer + 1, cc_kernel(re->distance - ratio, b));
|
||||
@ -240,7 +207,7 @@ static void resampler_CC_downsample(void *re_, struct resampler_data *data)
|
||||
|
||||
if (re->distance > (ratio + 0.5))
|
||||
{
|
||||
*outp=re->buffer[0];
|
||||
*outp = re->buffer[0];
|
||||
|
||||
re->buffer[0] = re->buffer[1];
|
||||
re->buffer[1] = re->buffer[2];
|
||||
@ -248,12 +215,12 @@ static void resampler_CC_downsample(void *re_, struct resampler_data *data)
|
||||
re->buffer[2].l = 0.0;
|
||||
re->buffer[2].r = 0.0;
|
||||
|
||||
re->distance-=ratio;
|
||||
re->distance -= ratio;
|
||||
outp++;
|
||||
}
|
||||
}
|
||||
|
||||
data->output_frames = (outp - (audio_frame_float_t*)data->data_out);
|
||||
data->output_frames = outp - (audio_frame_float_t*)data->data_out;
|
||||
}
|
||||
|
||||
#ifndef min
|
||||
@ -262,17 +229,16 @@ static void resampler_CC_downsample(void *re_, struct resampler_data *data)
|
||||
|
||||
static void resampler_CC_upsample(void *re_, struct resampler_data *data)
|
||||
{
|
||||
|
||||
rarch_CC_resampler_t *re = (rarch_CC_resampler_t*)re_;
|
||||
|
||||
audio_frame_float_t *inp = (audio_frame_float_t*)data->data_in;
|
||||
audio_frame_float_t *inp_max = inp + data->input_frames;
|
||||
audio_frame_float_t *outp = (audio_frame_float_t*)data->data_out;
|
||||
|
||||
float b = min(data->ratio, 1.00); // cutoff frequency
|
||||
float b = min(data->ratio, 1.00); // cutoff frequency
|
||||
float ratio = 1.0 / data->ratio;
|
||||
|
||||
while(inp != inp_max)
|
||||
while (inp != inp_max)
|
||||
{
|
||||
re->buffer[0] = re->buffer[1];
|
||||
re->buffer[1] = re->buffer[2];
|
||||
@ -286,7 +252,7 @@ static void resampler_CC_upsample(void *re_, struct resampler_data *data)
|
||||
outp->l = 0.0;
|
||||
outp->r = 0.0;
|
||||
|
||||
for (i=0; i!=4; i++)
|
||||
for (i = 0; i < 4; i++)
|
||||
{
|
||||
temp = cc_kernel(re->distance + 1.0 - i, b);
|
||||
outp->l += re->buffer[i].l * temp;
|
||||
@ -297,12 +263,11 @@ static void resampler_CC_upsample(void *re_, struct resampler_data *data)
|
||||
outp++;
|
||||
}
|
||||
|
||||
re->distance-= 1.0;
|
||||
re->distance -= 1.0;
|
||||
inp++;
|
||||
}
|
||||
|
||||
data->output_frames = (outp - (audio_frame_float_t*)data->data_out);
|
||||
|
||||
data->output_frames = outp - (audio_frame_float_t*)data->data_out;
|
||||
}
|
||||
|
||||
static void resampler_CC_process(void *re_, struct resampler_data *data)
|
||||
@ -325,15 +290,15 @@ static void *resampler_CC_init(double bandwidth_mod)
|
||||
if (!re)
|
||||
return NULL;
|
||||
|
||||
for (i=0; i!=4 ; i++)
|
||||
for (i = 0; i < 4; i++)
|
||||
{
|
||||
re->buffer[i].l=0.0;
|
||||
re->buffer[i].r=0.0;
|
||||
re->buffer[i].l = 0.0;
|
||||
re->buffer[i].r = 0.0;
|
||||
}
|
||||
|
||||
RARCH_LOG("Convoluted Cosine resampler (C) : ");
|
||||
|
||||
if (bandwidth_mod < 0.75) // variations of data->ratio around 0.75 are safer than around 1.0 for both up/downsampler.
|
||||
if (bandwidth_mod < 0.75) // variations of data->ratio around 0.75 are safer than around 1.0 for both up/downsampler.
|
||||
{
|
||||
RARCH_LOG("CC_downsample @%f \n", bandwidth_mod);
|
||||
re->process = resampler_CC_downsample;
|
||||
@ -356,3 +321,4 @@ const rarch_resampler_t CC_resampler = {
|
||||
resampler_CC_free,
|
||||
"CC",
|
||||
};
|
||||
|
||||
|
||||
109
audio/utils.c
109
audio/utils.c
@ -1,5 +1,6 @@
|
||||
/* RetroArch - A frontend for libretro.
|
||||
* Copyright (C) 2010-2014 - Hans-Kristian Arntzen
|
||||
* Copyright (C) 2014 - Ali Bouhlel ( aliaspider@gmail.com )
|
||||
*
|
||||
* RetroArch is free software: you can redistribute it and/or modify it under the terms
|
||||
* of the GNU General Public License as published by the Free Software Found-
|
||||
@ -167,59 +168,58 @@ static void audio_convert_float_to_s16_neon(int16_t *out, const float *in, size_
|
||||
void audio_convert_s16_to_float_ALLEGREX(float *out,
|
||||
const int16_t *in, size_t samples, float gain)
|
||||
{
|
||||
|
||||
#ifdef DEBUG
|
||||
// make sure the buffer is 16 byte aligned, this should be the default behaviour of malloc in the PSPSDK
|
||||
rarch_assert(((uint32_t)out & 0xF) == 0);
|
||||
// Make sure the buffer is 16 byte aligned, this should be the default behaviour of malloc in the PSPSDK.
|
||||
// Only the output buffer can be assumed to be 16-byte aligned.
|
||||
rarch_assert(((uintptr_t)out & 0xf) == 0);
|
||||
#endif
|
||||
|
||||
size_t i;
|
||||
gain = gain / 0x8000;
|
||||
__asm__ (
|
||||
".set push \n"
|
||||
".set noreorder \n"
|
||||
"mtv %0, s200 \n"
|
||||
".set pop \n"
|
||||
::"r"(gain)
|
||||
);
|
||||
".set push \n"
|
||||
".set noreorder \n"
|
||||
"mtv %0, s200 \n"
|
||||
".set pop \n"
|
||||
::"r"(gain));
|
||||
|
||||
for (i = 0; (i+16) <= samples; i+=16)
|
||||
for (i = 0; i + 16 <= samples; i += 16)
|
||||
{
|
||||
__asm__ (
|
||||
".set push \n"
|
||||
".set noreorder \n"
|
||||
".set push \n"
|
||||
".set noreorder \n"
|
||||
|
||||
"lv.s s100, 0(%0) \n"
|
||||
"lv.s s101, 4(%0) \n"
|
||||
"lv.s s110, 8(%0) \n"
|
||||
"lv.s s111, 12(%0) \n"
|
||||
"lv.s s120, 16(%0) \n"
|
||||
"lv.s s121, 20(%0) \n"
|
||||
"lv.s s130, 24(%0) \n"
|
||||
"lv.s s131, 28(%0) \n"
|
||||
"lv.s s100, 0(%0) \n"
|
||||
"lv.s s101, 4(%0) \n"
|
||||
"lv.s s110, 8(%0) \n"
|
||||
"lv.s s111, 12(%0) \n"
|
||||
"lv.s s120, 16(%0) \n"
|
||||
"lv.s s121, 20(%0) \n"
|
||||
"lv.s s130, 24(%0) \n"
|
||||
"lv.s s131, 28(%0) \n"
|
||||
|
||||
"vs2i.p c100, c100 \n"
|
||||
"vs2i.p c110, c110 \n"
|
||||
"vs2i.p c120, c120 \n"
|
||||
"vs2i.p c130, c130 \n"
|
||||
"vs2i.p c100, c100 \n"
|
||||
"vs2i.p c110, c110 \n"
|
||||
"vs2i.p c120, c120 \n"
|
||||
"vs2i.p c130, c130 \n"
|
||||
|
||||
"vi2f.q c100, c100, 16 \n"
|
||||
"vi2f.q c110, c110, 16 \n"
|
||||
"vi2f.q c120, c120, 16 \n"
|
||||
"vi2f.q c130, c130, 16 \n"
|
||||
"vi2f.q c100, c100, 16 \n"
|
||||
"vi2f.q c110, c110, 16 \n"
|
||||
"vi2f.q c120, c120, 16 \n"
|
||||
"vi2f.q c130, c130, 16 \n"
|
||||
|
||||
"vmscl.q e100, e100, s200 \n"
|
||||
"vmscl.q e100, e100, s200 \n"
|
||||
|
||||
"sv.q c100, 0(%1) \n"
|
||||
"sv.q c110, 16(%1) \n"
|
||||
"sv.q c120, 32(%1) \n"
|
||||
"sv.q c130, 48(%1) \n"
|
||||
"sv.q c100, 0(%1) \n"
|
||||
"sv.q c110, 16(%1) \n"
|
||||
"sv.q c120, 32(%1) \n"
|
||||
"sv.q c130, 48(%1) \n"
|
||||
|
||||
".set pop \n"
|
||||
::"r"(in+i),"r"(out+i)
|
||||
);
|
||||
".set pop \n"
|
||||
:: "r"(in + i), "r"(out + i));
|
||||
}
|
||||
|
||||
for (;i != samples; i++)
|
||||
for (; i < samples; i++)
|
||||
out[i] = (float)in[i] * gain;
|
||||
}
|
||||
|
||||
@ -227,39 +227,38 @@ void audio_convert_float_to_s16_ALLEGREX(int16_t *out,
|
||||
const float *in, size_t samples)
|
||||
{
|
||||
#ifdef DEBUG
|
||||
// make sure the buffers are 16 byte aligned, this should be the default behaviour of malloc in the PSPSDK
|
||||
rarch_assert(((uint32_t)in & 0xF) == 0);
|
||||
rarch_assert(((uint32_t)out & 0xF) == 0);
|
||||
// Make sure the buffers are 16 byte aligned, this should be the default behaviour of malloc in the PSPSDK.
|
||||
// Both buffers are allocated by RetroArch, so can assume alignment.
|
||||
rarch_assert(((uintptr_t)in & 0xf) == 0);
|
||||
rarch_assert(((uintptr_t)out & 0xf) == 0);
|
||||
#endif
|
||||
|
||||
size_t i;
|
||||
for (i = 0; (i+8) <= samples; i+=8)
|
||||
for (i = 0; i + 8 <= samples; i += 8)
|
||||
{
|
||||
__asm__ (
|
||||
".set push \n"
|
||||
".set noreorder \n"
|
||||
".set push \n"
|
||||
".set noreorder \n"
|
||||
|
||||
"lv.q c100, 0(%0) \n"
|
||||
"lv.q c110, 16(%0) \n"
|
||||
"lv.q c100, 0(%0) \n"
|
||||
"lv.q c110, 16(%0) \n"
|
||||
|
||||
"vf2in.q c100, c100, 31 \n"
|
||||
"vf2in.q c110, c110, 31 \n"
|
||||
"vi2s.q c100, c100 \n"
|
||||
"vi2s.q c102, c110 \n"
|
||||
"vf2in.q c100, c100, 31 \n"
|
||||
"vf2in.q c110, c110, 31 \n"
|
||||
"vi2s.q c100, c100 \n"
|
||||
"vi2s.q c102, c110 \n"
|
||||
|
||||
"sv.q c100, 0(%1) \n"
|
||||
"sv.q c100, 0(%1) \n"
|
||||
|
||||
".set pop \n"
|
||||
::"r"(in+i),"r"(out+i)
|
||||
);
|
||||
".set pop \n"
|
||||
:: "r"(in + i), "r"(out + i));
|
||||
}
|
||||
|
||||
for (;i != samples; i++)
|
||||
for (; i < samples; i++)
|
||||
{
|
||||
int32_t val = (int32_t)(in[i] * 0x8000);
|
||||
out[i] = (val > 0x7FFF) ? 0x7FFF : (val < -0x8000 ? -0x8000 : (int16_t)val);
|
||||
}
|
||||
|
||||
}
|
||||
#endif
|
||||
|
||||
|
||||
Loading…
x
Reference in New Issue
Block a user