Rename (already updated) opus-1.1 to opus-1.1.4

This commit is contained in:
Zack Middleton 2017-05-23 10:53:22 -05:00
parent 853110d5d4
commit ef8ad54421
279 changed files with 140 additions and 140 deletions

View file

@ -0,0 +1,112 @@
/***********************************************************************
Copyright (C) 2014 Vidyo
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions
are met:
- Redistributions of source code must retain the above copyright notice,
this list of conditions and the following disclaimer.
- Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in the
documentation and/or other materials provided with the distribution.
- Neither the name of Internet Society, IETF or IETF Trust, nor the
names of specific contributors, may be used to endorse or promote
products derived from this software without specific prior written
permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
POSSIBILITY OF SUCH DAMAGE.
***********************************************************************/
#ifdef HAVE_CONFIG_H
#include "config.h"
#endif
#include <arm_neon.h>
#include "main.h"
#include "stack_alloc.h"
#include "NSQ.h"
#include "celt/cpu_support.h"
#include "celt/arm/armcpu.h"
opus_int32 silk_noise_shape_quantizer_short_prediction_neon(const opus_int32 *buf32, const opus_int32 *coef32, opus_int order)
{
int32x4_t coef0 = vld1q_s32(coef32);
int32x4_t coef1 = vld1q_s32(coef32 + 4);
int32x4_t coef2 = vld1q_s32(coef32 + 8);
int32x4_t coef3 = vld1q_s32(coef32 + 12);
int32x4_t a0 = vld1q_s32(buf32 - 15);
int32x4_t a1 = vld1q_s32(buf32 - 11);
int32x4_t a2 = vld1q_s32(buf32 - 7);
int32x4_t a3 = vld1q_s32(buf32 - 3);
int32x4_t b0 = vqdmulhq_s32(coef0, a0);
int32x4_t b1 = vqdmulhq_s32(coef1, a1);
int32x4_t b2 = vqdmulhq_s32(coef2, a2);
int32x4_t b3 = vqdmulhq_s32(coef3, a3);
int32x4_t c0 = vaddq_s32(b0, b1);
int32x4_t c1 = vaddq_s32(b2, b3);
int32x4_t d = vaddq_s32(c0, c1);
int64x2_t e = vpaddlq_s32(d);
int64x1_t f = vadd_s64(vget_low_s64(e), vget_high_s64(e));
opus_int32 out = vget_lane_s32(vreinterpret_s32_s64(f), 0);
out += silk_RSHIFT( order, 1 );
return out;
}
opus_int32 silk_NSQ_noise_shape_feedback_loop_neon(const opus_int32 *data0, opus_int32 *data1, const opus_int16 *coef, opus_int order)
{
opus_int32 out;
if (order == 8)
{
int32x4_t a00 = vdupq_n_s32(data0[0]);
int32x4_t a01 = vld1q_s32(data1); /* data1[0] ... [3] */
int32x4_t a0 = vextq_s32 (a00, a01, 3); /* data0[0] data1[0] ...[2] */
int32x4_t a1 = vld1q_s32(data1 + 3); /* data1[3] ... [6] */
/*TODO: Convert these once in advance instead of once per sample, like
silk_noise_shape_quantizer_short_prediction_neon() does.*/
int16x8_t coef16 = vld1q_s16(coef);
int32x4_t coef0 = vmovl_s16(vget_low_s16(coef16));
int32x4_t coef1 = vmovl_s16(vget_high_s16(coef16));
/*This is not bit-exact with the C version, since we do not drop the
lower 16 bits of each multiply, but wait until the end to truncate
precision. This is an encoder-specific calculation (and unlike
silk_noise_shape_quantizer_short_prediction_neon(), is not meant to
simulate what the decoder will do). We still could use vqdmulhq_s32()
like silk_noise_shape_quantizer_short_prediction_neon() and save
half the multiplies, but the speed difference is not large, since we
then need two extra adds.*/
int64x2_t b0 = vmull_s32(vget_low_s32(a0), vget_low_s32(coef0));
int64x2_t b1 = vmlal_s32(b0, vget_high_s32(a0), vget_high_s32(coef0));
int64x2_t b2 = vmlal_s32(b1, vget_low_s32(a1), vget_low_s32(coef1));
int64x2_t b3 = vmlal_s32(b2, vget_high_s32(a1), vget_high_s32(coef1));
int64x1_t c = vadd_s64(vget_low_s64(b3), vget_high_s64(b3));
int64x1_t cS = vrshr_n_s64(c, 15);
int32x2_t d = vreinterpret_s32_s64(cS);
out = vget_lane_s32(d, 0);
vst1q_s32(data1, a0);
vst1q_s32(data1 + 4, a1);
return out;
}
return silk_NSQ_noise_shape_feedback_loop_c(data0, data1, coef, order);
}

View file

@ -0,0 +1,113 @@
/***********************************************************************
Copyright (C) 2014 Vidyo
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions
are met:
- Redistributions of source code must retain the above copyright notice,
this list of conditions and the following disclaimer.
- Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in the
documentation and/or other materials provided with the distribution.
- Neither the name of Internet Society, IETF or IETF Trust, nor the
names of specific contributors, may be used to endorse or promote
products derived from this software without specific prior written
permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
POSSIBILITY OF SUCH DAMAGE.
***********************************************************************/
#ifndef SILK_NSQ_NEON_H
#define SILK_NSQ_NEON_H
#include "cpu_support.h"
#undef silk_short_prediction_create_arch_coef
/* For vectorized calc, reverse a_Q12 coefs, convert to 32-bit, and shift for vqdmulhq_s32. */
static OPUS_INLINE void silk_short_prediction_create_arch_coef_neon(opus_int32 *out, const opus_int16 *in, opus_int order)
{
out[15] = in[0] << 15;
out[14] = in[1] << 15;
out[13] = in[2] << 15;
out[12] = in[3] << 15;
out[11] = in[4] << 15;
out[10] = in[5] << 15;
out[9] = in[6] << 15;
out[8] = in[7] << 15;
out[7] = in[8] << 15;
out[6] = in[9] << 15;
if (order == 16)
{
out[5] = in[10] << 15;
out[4] = in[11] << 15;
out[3] = in[12] << 15;
out[2] = in[13] << 15;
out[1] = in[14] << 15;
out[0] = in[15] << 15;
}
else
{
out[5] = 0;
out[4] = 0;
out[3] = 0;
out[2] = 0;
out[1] = 0;
out[0] = 0;
}
}
#if defined(OPUS_ARM_PRESUME_NEON_INTR)
#define silk_short_prediction_create_arch_coef(out, in, order) \
(silk_short_prediction_create_arch_coef_neon(out, in, order))
#elif defined(OPUS_HAVE_RTCD) && defined(OPUS_ARM_MAY_HAVE_NEON_INTR)
#define silk_short_prediction_create_arch_coef(out, in, order) \
do { if (arch == OPUS_ARCH_ARM_NEON) { silk_short_prediction_create_arch_coef_neon(out, in, order); } } while (0)
#endif
opus_int32 silk_noise_shape_quantizer_short_prediction_neon(const opus_int32 *buf32, const opus_int32 *coef32, opus_int order);
opus_int32 silk_NSQ_noise_shape_feedback_loop_neon(const opus_int32 *data0, opus_int32 *data1, const opus_int16 *coef, opus_int order);
#if defined(OPUS_ARM_PRESUME_NEON_INTR)
#undef silk_noise_shape_quantizer_short_prediction
#define silk_noise_shape_quantizer_short_prediction(in, coef, coefRev, order, arch) \
((void)arch,silk_noise_shape_quantizer_short_prediction_neon(in, coefRev, order))
#undef silk_NSQ_noise_shape_feedback_loop
#define silk_NSQ_noise_shape_feedback_loop(data0, data1, coef, order, arch) ((void)arch,silk_NSQ_noise_shape_feedback_loop_neon(data0, data1, coef, order))
#elif defined(OPUS_HAVE_RTCD) && defined(OPUS_ARM_MAY_HAVE_NEON_INTR)
/* silk_noise_shape_quantizer_short_prediction implementations take different parameters based on arch
(coef vs. coefRev) so can't use the usual IMPL table implementation */
#undef silk_noise_shape_quantizer_short_prediction
#define silk_noise_shape_quantizer_short_prediction(in, coef, coefRev, order, arch) \
(arch == OPUS_ARCH_ARM_NEON ? \
silk_noise_shape_quantizer_short_prediction_neon(in, coefRev, order) : \
silk_noise_shape_quantizer_short_prediction_c(in, coef, order))
extern opus_int32
(*const SILK_NSQ_NOISE_SHAPE_FEEDBACK_LOOP_IMPL[OPUS_ARCHMASK+1])(
const opus_int32 *data0, opus_int32 *data1, const opus_int16 *coef,
opus_int order);
#undef silk_NSQ_noise_shape_feedback_loop
#define silk_NSQ_noise_shape_feedback_loop(data0, data1, coef, order, arch) \
(SILK_NSQ_NOISE_SHAPE_FEEDBACK_LOOP_IMPL[(arch)&OPUS_ARCHMASK](data0, data1, \
coef, order))
#endif
#endif /* SILK_NSQ_NEON_H */

View file

@ -0,0 +1,47 @@
/***********************************************************************
Copyright (C) 2013 Xiph.Org Foundation and contributors
Copyright (c) 2013 Parrot
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions
are met:
- Redistributions of source code must retain the above copyright notice,
this list of conditions and the following disclaimer.
- Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in the
documentation and/or other materials provided with the distribution.
- Neither the name of Internet Society, IETF or IETF Trust, nor the
names of specific contributors, may be used to endorse or promote
products derived from this software without specific prior written
permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
POSSIBILITY OF SUCH DAMAGE.
***********************************************************************/
#ifndef SILK_SIGPROC_FIX_ARMv4_H
#define SILK_SIGPROC_FIX_ARMv4_H
#undef silk_MLA
static OPUS_INLINE opus_int32 silk_MLA_armv4(opus_int32 a, opus_int32 b,
opus_int32 c)
{
opus_int32 res;
__asm__(
"#silk_MLA\n\t"
"mla %0, %1, %2, %3\n\t"
: "=&r"(res)
: "r"(b), "r"(c), "r"(a)
);
return res;
}
#define silk_MLA(a, b, c) (silk_MLA_armv4(a, b, c))
#endif

View file

@ -0,0 +1,61 @@
/***********************************************************************
Copyright (c) 2006-2011, Skype Limited. All rights reserved.
Copyright (c) 2013 Parrot
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions
are met:
- Redistributions of source code must retain the above copyright notice,
this list of conditions and the following disclaimer.
- Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in the
documentation and/or other materials provided with the distribution.
- Neither the name of Internet Society, IETF or IETF Trust, nor the
names of specific contributors, may be used to endorse or promote
products derived from this software without specific prior written
permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
POSSIBILITY OF SUCH DAMAGE.
***********************************************************************/
#ifndef SILK_SIGPROC_FIX_ARMv5E_H
#define SILK_SIGPROC_FIX_ARMv5E_H
#undef silk_SMULTT
static OPUS_INLINE opus_int32 silk_SMULTT_armv5e(opus_int32 a, opus_int32 b)
{
opus_int32 res;
__asm__(
"#silk_SMULTT\n\t"
"smultt %0, %1, %2\n\t"
: "=r"(res)
: "%r"(a), "r"(b)
);
return res;
}
#define silk_SMULTT(a, b) (silk_SMULTT_armv5e(a, b))
#undef silk_SMLATT
static OPUS_INLINE opus_int32 silk_SMLATT_armv5e(opus_int32 a, opus_int32 b,
opus_int32 c)
{
opus_int32 res;
__asm__(
"#silk_SMLATT\n\t"
"smlatt %0, %1, %2, %3\n\t"
: "=r"(res)
: "%r"(b), "r"(c), "r"(a)
);
return res;
}
#define silk_SMLATT(a, b, c) (silk_SMLATT_armv5e(a, b, c))
#endif

View file

@ -0,0 +1,55 @@
/***********************************************************************
Copyright (C) 2014 Vidyo
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions
are met:
- Redistributions of source code must retain the above copyright notice,
this list of conditions and the following disclaimer.
- Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in the
documentation and/or other materials provided with the distribution.
- Neither the name of Internet Society, IETF or IETF Trust, nor the
names of specific contributors, may be used to endorse or promote
products derived from this software without specific prior written
permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
POSSIBILITY OF SUCH DAMAGE.
***********************************************************************/
#ifdef HAVE_CONFIG_H
# include "config.h"
#endif
#include "NSQ.h"
#if defined(OPUS_HAVE_RTCD)
# if (defined(OPUS_ARM_MAY_HAVE_NEON_INTR) && \
!defined(OPUS_ARM_PRESUME_NEON_INTR))
/*There is no table for silk_noise_shape_quantizer_short_prediction because the
NEON version takes different parameters than the C version.
Instead RTCD is done via if statements at the call sites.
See NSQ_neon.h for details.*/
opus_int32
(*const SILK_NSQ_NOISE_SHAPE_FEEDBACK_LOOP_IMPL[OPUS_ARCHMASK+1])(
const opus_int32 *data0, opus_int32 *data1, const opus_int16 *coef,
opus_int order) = {
silk_NSQ_noise_shape_feedback_loop_c, /* ARMv4 */
silk_NSQ_noise_shape_feedback_loop_c, /* EDSP */
silk_NSQ_noise_shape_feedback_loop_c, /* Media */
silk_NSQ_noise_shape_feedback_loop_neon, /* NEON */
};
# endif
#endif /* OPUS_HAVE_RTCD */

View file

@ -0,0 +1,39 @@
/***********************************************************************
Copyright (C) 2015 Vidyo
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions
are met:
- Redistributions of source code must retain the above copyright notice,
this list of conditions and the following disclaimer.
- Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in the
documentation and/or other materials provided with the distribution.
- Neither the name of Internet Society, IETF or IETF Trust, nor the
names of specific contributors, may be used to endorse or promote
products derived from this software without specific prior written
permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
POSSIBILITY OF SUCH DAMAGE.
***********************************************************************/
#ifndef SILK_MACROS_ARM64_H
#define SILK_MACROS_ARM64_H
#include <arm_neon.h>
#undef silk_ADD_SAT32
#define silk_ADD_SAT32(a, b) (vqadds_s32((a), (b)))
#undef silk_SUB_SAT32
#define silk_SUB_SAT32(a, b) (vqsubs_s32((a), (b)))
#endif /* SILK_MACROS_ARM64_H */

View file

@ -0,0 +1,103 @@
/***********************************************************************
Copyright (C) 2013 Xiph.Org Foundation and contributors.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions
are met:
- Redistributions of source code must retain the above copyright notice,
this list of conditions and the following disclaimer.
- Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in the
documentation and/or other materials provided with the distribution.
- Neither the name of Internet Society, IETF or IETF Trust, nor the
names of specific contributors, may be used to endorse or promote
products derived from this software without specific prior written
permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
POSSIBILITY OF SUCH DAMAGE.
***********************************************************************/
#ifndef SILK_MACROS_ARMv4_H
#define SILK_MACROS_ARMv4_H
/* (a32 * (opus_int32)((opus_int16)(b32))) >> 16 output have to be 32bit int */
#undef silk_SMULWB
static OPUS_INLINE opus_int32 silk_SMULWB_armv4(opus_int32 a, opus_int16 b)
{
unsigned rd_lo;
int rd_hi;
__asm__(
"#silk_SMULWB\n\t"
"smull %0, %1, %2, %3\n\t"
: "=&r"(rd_lo), "=&r"(rd_hi)
: "%r"(a), "r"(b<<16)
);
return rd_hi;
}
#define silk_SMULWB(a, b) (silk_SMULWB_armv4(a, b))
/* a32 + (b32 * (opus_int32)((opus_int16)(c32))) >> 16 output have to be 32bit int */
#undef silk_SMLAWB
#define silk_SMLAWB(a, b, c) ((a) + silk_SMULWB(b, c))
/* (a32 * (b32 >> 16)) >> 16 */
#undef silk_SMULWT
static OPUS_INLINE opus_int32 silk_SMULWT_armv4(opus_int32 a, opus_int32 b)
{
unsigned rd_lo;
int rd_hi;
__asm__(
"#silk_SMULWT\n\t"
"smull %0, %1, %2, %3\n\t"
: "=&r"(rd_lo), "=&r"(rd_hi)
: "%r"(a), "r"(b&~0xFFFF)
);
return rd_hi;
}
#define silk_SMULWT(a, b) (silk_SMULWT_armv4(a, b))
/* a32 + (b32 * (c32 >> 16)) >> 16 */
#undef silk_SMLAWT
#define silk_SMLAWT(a, b, c) ((a) + silk_SMULWT(b, c))
/* (a32 * b32) >> 16 */
#undef silk_SMULWW
static OPUS_INLINE opus_int32 silk_SMULWW_armv4(opus_int32 a, opus_int32 b)
{
unsigned rd_lo;
int rd_hi;
__asm__(
"#silk_SMULWW\n\t"
"smull %0, %1, %2, %3\n\t"
: "=&r"(rd_lo), "=&r"(rd_hi)
: "%r"(a), "r"(b)
);
return (rd_hi<<16)+(rd_lo>>16);
}
#define silk_SMULWW(a, b) (silk_SMULWW_armv4(a, b))
#undef silk_SMLAWW
static OPUS_INLINE opus_int32 silk_SMLAWW_armv4(opus_int32 a, opus_int32 b,
opus_int32 c)
{
unsigned rd_lo;
int rd_hi;
__asm__(
"#silk_SMLAWW\n\t"
"smull %0, %1, %2, %3\n\t"
: "=&r"(rd_lo), "=&r"(rd_hi)
: "%r"(b), "r"(c)
);
return a+(rd_hi<<16)+(rd_lo>>16);
}
#define silk_SMLAWW(a, b, c) (silk_SMLAWW_armv4(a, b, c))
#endif /* SILK_MACROS_ARMv4_H */

View file

@ -0,0 +1,213 @@
/***********************************************************************
Copyright (c) 2006-2011, Skype Limited. All rights reserved.
Copyright (c) 2013 Parrot
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions
are met:
- Redistributions of source code must retain the above copyright notice,
this list of conditions and the following disclaimer.
- Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in the
documentation and/or other materials provided with the distribution.
- Neither the name of Internet Society, IETF or IETF Trust, nor the
names of specific contributors, may be used to endorse or promote
products derived from this software without specific prior written
permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
POSSIBILITY OF SUCH DAMAGE.
***********************************************************************/
#ifndef SILK_MACROS_ARMv5E_H
#define SILK_MACROS_ARMv5E_H
/* (a32 * (opus_int32)((opus_int16)(b32))) >> 16 output have to be 32bit int */
#undef silk_SMULWB
static OPUS_INLINE opus_int32 silk_SMULWB_armv5e(opus_int32 a, opus_int16 b)
{
int res;
__asm__(
"#silk_SMULWB\n\t"
"smulwb %0, %1, %2\n\t"
: "=r"(res)
: "r"(a), "r"(b)
);
return res;
}
#define silk_SMULWB(a, b) (silk_SMULWB_armv5e(a, b))
/* a32 + (b32 * (opus_int32)((opus_int16)(c32))) >> 16 output have to be 32bit int */
#undef silk_SMLAWB
static OPUS_INLINE opus_int32 silk_SMLAWB_armv5e(opus_int32 a, opus_int32 b,
opus_int16 c)
{
int res;
__asm__(
"#silk_SMLAWB\n\t"
"smlawb %0, %1, %2, %3\n\t"
: "=r"(res)
: "r"(b), "r"(c), "r"(a)
);
return res;
}
#define silk_SMLAWB(a, b, c) (silk_SMLAWB_armv5e(a, b, c))
/* (a32 * (b32 >> 16)) >> 16 */
#undef silk_SMULWT
static OPUS_INLINE opus_int32 silk_SMULWT_armv5e(opus_int32 a, opus_int32 b)
{
int res;
__asm__(
"#silk_SMULWT\n\t"
"smulwt %0, %1, %2\n\t"
: "=r"(res)
: "r"(a), "r"(b)
);
return res;
}
#define silk_SMULWT(a, b) (silk_SMULWT_armv5e(a, b))
/* a32 + (b32 * (c32 >> 16)) >> 16 */
#undef silk_SMLAWT
static OPUS_INLINE opus_int32 silk_SMLAWT_armv5e(opus_int32 a, opus_int32 b,
opus_int32 c)
{
int res;
__asm__(
"#silk_SMLAWT\n\t"
"smlawt %0, %1, %2, %3\n\t"
: "=r"(res)
: "r"(b), "r"(c), "r"(a)
);
return res;
}
#define silk_SMLAWT(a, b, c) (silk_SMLAWT_armv5e(a, b, c))
/* (opus_int32)((opus_int16)(a3))) * (opus_int32)((opus_int16)(b32)) output have to be 32bit int */
#undef silk_SMULBB
static OPUS_INLINE opus_int32 silk_SMULBB_armv5e(opus_int32 a, opus_int32 b)
{
int res;
__asm__(
"#silk_SMULBB\n\t"
"smulbb %0, %1, %2\n\t"
: "=r"(res)
: "%r"(a), "r"(b)
);
return res;
}
#define silk_SMULBB(a, b) (silk_SMULBB_armv5e(a, b))
/* a32 + (opus_int32)((opus_int16)(b32)) * (opus_int32)((opus_int16)(c32)) output have to be 32bit int */
#undef silk_SMLABB
static OPUS_INLINE opus_int32 silk_SMLABB_armv5e(opus_int32 a, opus_int32 b,
opus_int32 c)
{
int res;
__asm__(
"#silk_SMLABB\n\t"
"smlabb %0, %1, %2, %3\n\t"
: "=r"(res)
: "%r"(b), "r"(c), "r"(a)
);
return res;
}
#define silk_SMLABB(a, b, c) (silk_SMLABB_armv5e(a, b, c))
/* (opus_int32)((opus_int16)(a32)) * (b32 >> 16) */
#undef silk_SMULBT
static OPUS_INLINE opus_int32 silk_SMULBT_armv5e(opus_int32 a, opus_int32 b)
{
int res;
__asm__(
"#silk_SMULBT\n\t"
"smulbt %0, %1, %2\n\t"
: "=r"(res)
: "r"(a), "r"(b)
);
return res;
}
#define silk_SMULBT(a, b) (silk_SMULBT_armv5e(a, b))
/* a32 + (opus_int32)((opus_int16)(b32)) * (c32 >> 16) */
#undef silk_SMLABT
static OPUS_INLINE opus_int32 silk_SMLABT_armv5e(opus_int32 a, opus_int32 b,
opus_int32 c)
{
int res;
__asm__(
"#silk_SMLABT\n\t"
"smlabt %0, %1, %2, %3\n\t"
: "=r"(res)
: "r"(b), "r"(c), "r"(a)
);
return res;
}
#define silk_SMLABT(a, b, c) (silk_SMLABT_armv5e(a, b, c))
/* add/subtract with output saturated */
#undef silk_ADD_SAT32
static OPUS_INLINE opus_int32 silk_ADD_SAT32_armv5e(opus_int32 a, opus_int32 b)
{
int res;
__asm__(
"#silk_ADD_SAT32\n\t"
"qadd %0, %1, %2\n\t"
: "=r"(res)
: "%r"(a), "r"(b)
);
return res;
}
#define silk_ADD_SAT32(a, b) (silk_ADD_SAT32_armv5e(a, b))
#undef silk_SUB_SAT32
static OPUS_INLINE opus_int32 silk_SUB_SAT32_armv5e(opus_int32 a, opus_int32 b)
{
int res;
__asm__(
"#silk_SUB_SAT32\n\t"
"qsub %0, %1, %2\n\t"
: "=r"(res)
: "r"(a), "r"(b)
);
return res;
}
#define silk_SUB_SAT32(a, b) (silk_SUB_SAT32_armv5e(a, b))
#undef silk_CLZ16
static OPUS_INLINE opus_int32 silk_CLZ16_armv5(opus_int16 in16)
{
int res;
__asm__(
"#silk_CLZ16\n\t"
"clz %0, %1;\n"
: "=r"(res)
: "r"(in16<<16|0x8000)
);
return res;
}
#define silk_CLZ16(in16) (silk_CLZ16_armv5(in16))
#undef silk_CLZ32
static OPUS_INLINE opus_int32 silk_CLZ32_armv5(opus_int32 in32)
{
int res;
__asm__(
"#silk_CLZ32\n\t"
"clz %0, %1\n\t"
: "=r"(res)
: "r"(in32)
);
return res;
}
#define silk_CLZ32(in32) (silk_CLZ32_armv5(in32))
#endif /* SILK_MACROS_ARMv5E_H */