Update common_audio
Corresponds to upstream commit 524e9b043e7e86fd72353b987c9d5f6a1ebf83e1 Update notes: * Moved src/ to webrtc/ to easily diff against the third_party/webrtc in the chromium tree * ARM/NEON/MIPS support is not yet hooked up * Tests have not been copied
This commit is contained in:
Arun Raghavan
97
webrtc/common_audio/signal_processing/include/real_fft.h
Normal file
97
webrtc/common_audio/signal_processing/include/real_fft.h
Normal file
@ -0,0 +1,97 @@
|
||||
/*
|
||||
* Copyright (c) 2012 The WebRTC project authors. All Rights Reserved.
|
||||
*
|
||||
* Use of this source code is governed by a BSD-style license
|
||||
* that can be found in the LICENSE file in the root of the source
|
||||
* tree. An additional intellectual property rights grant can be found
|
||||
* in the file PATENTS. All contributing project authors may
|
||||
* be found in the AUTHORS file in the root of the source tree.
|
||||
*/
|
||||
|
||||
#ifndef WEBRTC_COMMON_AUDIO_SIGNAL_PROCESSING_INCLUDE_REAL_FFT_H_
|
||||
#define WEBRTC_COMMON_AUDIO_SIGNAL_PROCESSING_INCLUDE_REAL_FFT_H_
|
||||
|
||||
#include "webrtc/typedefs.h"
|
||||
|
||||
// For ComplexFFT(), the maximum fft order is 10;
|
||||
// for OpenMax FFT in ARM, it is 12;
|
||||
// WebRTC APM uses orders of only 7 and 8.
|
||||
enum {kMaxFFTOrder = 10};
|
||||
|
||||
struct RealFFT;
|
||||
|
||||
#ifdef __cplusplus
|
||||
extern "C" {
|
||||
#endif
|
||||
|
||||
struct RealFFT* WebRtcSpl_CreateRealFFT(int order);
|
||||
void WebRtcSpl_FreeRealFFT(struct RealFFT* self);
|
||||
|
||||
// Compute an FFT for a real-valued signal of length of 2^order,
|
||||
// where 1 < order <= MAX_FFT_ORDER. Transform length is determined by the
|
||||
// specification structure, which must be initialized prior to calling the FFT
|
||||
// function with WebRtcSpl_CreateRealFFT().
|
||||
// The relationship between the input and output sequences can
|
||||
// be expressed in terms of the DFT, i.e.:
|
||||
// x[n] = (2^(-scalefactor)/N) . SUM[k=0,...,N-1] X[k].e^(jnk.2.pi/N)
|
||||
// n=0,1,2,...N-1
|
||||
// N=2^order.
|
||||
// The conjugate-symmetric output sequence is represented using a CCS vector,
|
||||
// which is of length N+2, and is organized as follows:
|
||||
// Index: 0 1 2 3 4 5 . . . N-2 N-1 N N+1
|
||||
// Component: R0 0 R1 I1 R2 I2 . . . R[N/2-1] I[N/2-1] R[N/2] 0
|
||||
// where R[n] and I[n], respectively, denote the real and imaginary components
|
||||
// for FFT bin 'n'. Bins are numbered from 0 to N/2, where N is the FFT length.
|
||||
// Bin index 0 corresponds to the DC component, and bin index N/2 corresponds to
|
||||
// the foldover frequency.
|
||||
//
|
||||
// Input Arguments:
|
||||
// self - pointer to preallocated and initialized FFT specification structure.
|
||||
// real_data_in - the input signal. For an ARM Neon platform, it must be
|
||||
// aligned on a 32-byte boundary.
|
||||
//
|
||||
// Output Arguments:
|
||||
// complex_data_out - the output complex signal with (2^order + 2) 16-bit
|
||||
// elements. For an ARM Neon platform, it must be different
|
||||
// from real_data_in, and aligned on a 32-byte boundary.
|
||||
//
|
||||
// Return Value:
|
||||
// 0 - FFT calculation is successful.
|
||||
// -1 - Error with bad arguments (NULL pointers).
|
||||
int WebRtcSpl_RealForwardFFT(struct RealFFT* self,
|
||||
const int16_t* real_data_in,
|
||||
int16_t* complex_data_out);
|
||||
|
||||
// Compute the inverse FFT for a conjugate-symmetric input sequence of length of
|
||||
// 2^order, where 1 < order <= MAX_FFT_ORDER. Transform length is determined by
|
||||
// the specification structure, which must be initialized prior to calling the
|
||||
// FFT function with WebRtcSpl_CreateRealFFT().
|
||||
// For a transform of length M, the input sequence is represented using a packed
|
||||
// CCS vector of length M+2, which is explained in the comments for
|
||||
// WebRtcSpl_RealForwardFFTC above.
|
||||
//
|
||||
// Input Arguments:
|
||||
// self - pointer to preallocated and initialized FFT specification structure.
|
||||
// complex_data_in - the input complex signal with (2^order + 2) 16-bit
|
||||
// elements. For an ARM Neon platform, it must be aligned on
|
||||
// a 32-byte boundary.
|
||||
//
|
||||
// Output Arguments:
|
||||
// real_data_out - the output real signal. For an ARM Neon platform, it must
|
||||
// be different to complex_data_in, and aligned on a 32-byte
|
||||
// boundary.
|
||||
//
|
||||
// Return Value:
|
||||
// 0 or a positive number - a value that the elements in the |real_data_out|
|
||||
// should be shifted left with in order to get
|
||||
// correct physical values.
|
||||
// -1 - Error with bad arguments (NULL pointers).
|
||||
int WebRtcSpl_RealInverseFFT(struct RealFFT* self,
|
||||
const int16_t* complex_data_in,
|
||||
int16_t* real_data_out);
|
||||
|
||||
#ifdef __cplusplus
|
||||
}
|
||||
#endif
|
||||
|
||||
#endif // WEBRTC_COMMON_AUDIO_SIGNAL_PROCESSING_INCLUDE_REAL_FFT_H_
|
||||
File diff suppressed because it is too large
Load Diff
173
webrtc/common_audio/signal_processing/include/spl_inl.h
Normal file
173
webrtc/common_audio/signal_processing/include/spl_inl.h
Normal file
@ -0,0 +1,173 @@
|
||||
/*
|
||||
* Copyright (c) 2011 The WebRTC project authors. All Rights Reserved.
|
||||
*
|
||||
* Use of this source code is governed by a BSD-style license
|
||||
* that can be found in the LICENSE file in the root of the source
|
||||
* tree. An additional intellectual property rights grant can be found
|
||||
* in the file PATENTS. All contributing project authors may
|
||||
* be found in the AUTHORS file in the root of the source tree.
|
||||
*/
|
||||
|
||||
|
||||
// This header file includes the inline functions in
|
||||
// the fix point signal processing library.
|
||||
|
||||
#ifndef WEBRTC_SPL_SPL_INL_H_
|
||||
#define WEBRTC_SPL_SPL_INL_H_
|
||||
|
||||
#ifdef WEBRTC_ARCH_ARM_V7
|
||||
#include "webrtc/common_audio/signal_processing/include/spl_inl_armv7.h"
|
||||
#else
|
||||
|
||||
#if defined(MIPS32_LE)
|
||||
#include "webrtc/common_audio/signal_processing/include/spl_inl_mips.h"
|
||||
#endif
|
||||
|
||||
#if !defined(MIPS_DSP_R1_LE)
|
||||
static __inline int16_t WebRtcSpl_SatW32ToW16(int32_t value32) {
|
||||
int16_t out16 = (int16_t) value32;
|
||||
|
||||
if (value32 > 32767)
|
||||
out16 = 32767;
|
||||
else if (value32 < -32768)
|
||||
out16 = -32768;
|
||||
|
||||
return out16;
|
||||
}
|
||||
|
||||
static __inline int32_t WebRtcSpl_AddSatW32(int32_t l_var1, int32_t l_var2) {
|
||||
int32_t l_sum;
|
||||
|
||||
// Perform long addition
|
||||
l_sum = l_var1 + l_var2;
|
||||
|
||||
if (l_var1 < 0) { // Check for underflow.
|
||||
if ((l_var2 < 0) && (l_sum >= 0)) {
|
||||
l_sum = (int32_t)0x80000000;
|
||||
}
|
||||
} else { // Check for overflow.
|
||||
if ((l_var2 > 0) && (l_sum < 0)) {
|
||||
l_sum = (int32_t)0x7FFFFFFF;
|
||||
}
|
||||
}
|
||||
|
||||
return l_sum;
|
||||
}
|
||||
|
||||
static __inline int32_t WebRtcSpl_SubSatW32(int32_t l_var1, int32_t l_var2) {
|
||||
int32_t l_diff;
|
||||
|
||||
// Perform subtraction.
|
||||
l_diff = l_var1 - l_var2;
|
||||
|
||||
if (l_var1 < 0) { // Check for underflow.
|
||||
if ((l_var2 > 0) && (l_diff > 0)) {
|
||||
l_diff = (int32_t)0x80000000;
|
||||
}
|
||||
} else { // Check for overflow.
|
||||
if ((l_var2 < 0) && (l_diff < 0)) {
|
||||
l_diff = (int32_t)0x7FFFFFFF;
|
||||
}
|
||||
}
|
||||
|
||||
return l_diff;
|
||||
}
|
||||
|
||||
static __inline int16_t WebRtcSpl_AddSatW16(int16_t a, int16_t b) {
|
||||
return WebRtcSpl_SatW32ToW16((int32_t) a + (int32_t) b);
|
||||
}
|
||||
|
||||
static __inline int16_t WebRtcSpl_SubSatW16(int16_t var1, int16_t var2) {
|
||||
return WebRtcSpl_SatW32ToW16((int32_t) var1 - (int32_t) var2);
|
||||
}
|
||||
#endif // #if !defined(MIPS_DSP_R1_LE)
|
||||
|
||||
#if !defined(MIPS32_LE)
|
||||
static __inline int16_t WebRtcSpl_GetSizeInBits(uint32_t n) {
|
||||
int16_t bits;
|
||||
|
||||
if (0xFFFF0000 & n) {
|
||||
bits = 16;
|
||||
} else {
|
||||
bits = 0;
|
||||
}
|
||||
if (0x0000FF00 & (n >> bits)) bits += 8;
|
||||
if (0x000000F0 & (n >> bits)) bits += 4;
|
||||
if (0x0000000C & (n >> bits)) bits += 2;
|
||||
if (0x00000002 & (n >> bits)) bits += 1;
|
||||
if (0x00000001 & (n >> bits)) bits += 1;
|
||||
|
||||
return bits;
|
||||
}
|
||||
|
||||
static __inline int16_t WebRtcSpl_NormW32(int32_t a) {
|
||||
int16_t zeros;
|
||||
|
||||
if (a == 0) {
|
||||
return 0;
|
||||
}
|
||||
else if (a < 0) {
|
||||
a = ~a;
|
||||
}
|
||||
|
||||
if (!(0xFFFF8000 & a)) {
|
||||
zeros = 16;
|
||||
} else {
|
||||
zeros = 0;
|
||||
}
|
||||
if (!(0xFF800000 & (a << zeros))) zeros += 8;
|
||||
if (!(0xF8000000 & (a << zeros))) zeros += 4;
|
||||
if (!(0xE0000000 & (a << zeros))) zeros += 2;
|
||||
if (!(0xC0000000 & (a << zeros))) zeros += 1;
|
||||
|
||||
return zeros;
|
||||
}
|
||||
|
||||
static __inline int16_t WebRtcSpl_NormU32(uint32_t a) {
|
||||
int16_t zeros;
|
||||
|
||||
if (a == 0) return 0;
|
||||
|
||||
if (!(0xFFFF0000 & a)) {
|
||||
zeros = 16;
|
||||
} else {
|
||||
zeros = 0;
|
||||
}
|
||||
if (!(0xFF000000 & (a << zeros))) zeros += 8;
|
||||
if (!(0xF0000000 & (a << zeros))) zeros += 4;
|
||||
if (!(0xC0000000 & (a << zeros))) zeros += 2;
|
||||
if (!(0x80000000 & (a << zeros))) zeros += 1;
|
||||
|
||||
return zeros;
|
||||
}
|
||||
|
||||
static __inline int16_t WebRtcSpl_NormW16(int16_t a) {
|
||||
int16_t zeros;
|
||||
|
||||
if (a == 0) {
|
||||
return 0;
|
||||
}
|
||||
else if (a < 0) {
|
||||
a = ~a;
|
||||
}
|
||||
|
||||
if (!(0xFF80 & a)) {
|
||||
zeros = 8;
|
||||
} else {
|
||||
zeros = 0;
|
||||
}
|
||||
if (!(0xF800 & (a << zeros))) zeros += 4;
|
||||
if (!(0xE000 & (a << zeros))) zeros += 2;
|
||||
if (!(0xC000 & (a << zeros))) zeros += 1;
|
||||
|
||||
return zeros;
|
||||
}
|
||||
|
||||
static __inline int32_t WebRtc_MulAccumW16(int16_t a, int16_t b, int32_t c) {
|
||||
return (a * b + c);
|
||||
}
|
||||
#endif // #if !defined(MIPS32_LE)
|
||||
|
||||
#endif // WEBRTC_ARCH_ARM_V7
|
||||
|
||||
#endif // WEBRTC_SPL_SPL_INL_H_
|
||||
136
webrtc/common_audio/signal_processing/include/spl_inl_armv7.h
Normal file
136
webrtc/common_audio/signal_processing/include/spl_inl_armv7.h
Normal file
@ -0,0 +1,136 @@
|
||||
/*
|
||||
* Copyright (c) 2012 The WebRTC project authors. All Rights Reserved.
|
||||
*
|
||||
* Use of this source code is governed by a BSD-style license
|
||||
* that can be found in the LICENSE file in the root of the source
|
||||
* tree. An additional intellectual property rights grant can be found
|
||||
* in the file PATENTS. All contributing project authors may
|
||||
* be found in the AUTHORS file in the root of the source tree.
|
||||
*/
|
||||
|
||||
|
||||
/* This header file includes the inline functions for ARM processors in
|
||||
* the fix point signal processing library.
|
||||
*/
|
||||
|
||||
#ifndef WEBRTC_SPL_SPL_INL_ARMV7_H_
|
||||
#define WEBRTC_SPL_SPL_INL_ARMV7_H_
|
||||
|
||||
/* TODO(kma): Replace some assembly code with GCC intrinsics
|
||||
* (e.g. __builtin_clz).
|
||||
*/
|
||||
|
||||
/* This function produces result that is not bit exact with that by the generic
|
||||
* C version in some cases, although the former is at least as accurate as the
|
||||
* later.
|
||||
*/
|
||||
static __inline int32_t WEBRTC_SPL_MUL_16_32_RSFT16(int16_t a, int32_t b) {
|
||||
int32_t tmp = 0;
|
||||
__asm __volatile ("smulwb %0, %1, %2":"=r"(tmp):"r"(b), "r"(a));
|
||||
return tmp;
|
||||
}
|
||||
|
||||
static __inline int32_t WEBRTC_SPL_MUL_16_16(int16_t a, int16_t b) {
|
||||
int32_t tmp = 0;
|
||||
__asm __volatile ("smulbb %0, %1, %2":"=r"(tmp):"r"(a), "r"(b));
|
||||
return tmp;
|
||||
}
|
||||
|
||||
// TODO(kma): add unit test.
|
||||
static __inline int32_t WebRtc_MulAccumW16(int16_t a, int16_t b, int32_t c) {
|
||||
int32_t tmp = 0;
|
||||
__asm __volatile ("smlabb %0, %1, %2, %3":"=r"(tmp):"r"(a), "r"(b), "r"(c));
|
||||
return tmp;
|
||||
}
|
||||
|
||||
static __inline int16_t WebRtcSpl_AddSatW16(int16_t a, int16_t b) {
|
||||
int32_t s_sum = 0;
|
||||
|
||||
__asm __volatile ("qadd16 %0, %1, %2":"=r"(s_sum):"r"(a), "r"(b));
|
||||
|
||||
return (int16_t) s_sum;
|
||||
}
|
||||
|
||||
static __inline int32_t WebRtcSpl_AddSatW32(int32_t l_var1, int32_t l_var2) {
|
||||
int32_t l_sum = 0;
|
||||
|
||||
__asm __volatile ("qadd %0, %1, %2":"=r"(l_sum):"r"(l_var1), "r"(l_var2));
|
||||
|
||||
return l_sum;
|
||||
}
|
||||
|
||||
static __inline int32_t WebRtcSpl_SubSatW32(int32_t l_var1, int32_t l_var2) {
|
||||
int32_t l_sub = 0;
|
||||
|
||||
__asm __volatile ("qsub %0, %1, %2":"=r"(l_sub):"r"(l_var1), "r"(l_var2));
|
||||
|
||||
return l_sub;
|
||||
}
|
||||
|
||||
static __inline int16_t WebRtcSpl_SubSatW16(int16_t var1, int16_t var2) {
|
||||
int32_t s_sub = 0;
|
||||
|
||||
__asm __volatile ("qsub16 %0, %1, %2":"=r"(s_sub):"r"(var1), "r"(var2));
|
||||
|
||||
return (int16_t)s_sub;
|
||||
}
|
||||
|
||||
static __inline int16_t WebRtcSpl_GetSizeInBits(uint32_t n) {
|
||||
int32_t tmp = 0;
|
||||
|
||||
__asm __volatile ("clz %0, %1":"=r"(tmp):"r"(n));
|
||||
|
||||
return (int16_t)(32 - tmp);
|
||||
}
|
||||
|
||||
static __inline int16_t WebRtcSpl_NormW32(int32_t a) {
|
||||
int32_t tmp = 0;
|
||||
|
||||
if (a == 0) {
|
||||
return 0;
|
||||
}
|
||||
else if (a < 0) {
|
||||
a ^= 0xFFFFFFFF;
|
||||
}
|
||||
|
||||
__asm __volatile ("clz %0, %1":"=r"(tmp):"r"(a));
|
||||
|
||||
return (int16_t)(tmp - 1);
|
||||
}
|
||||
|
||||
static __inline int16_t WebRtcSpl_NormU32(uint32_t a) {
|
||||
int tmp = 0;
|
||||
|
||||
if (a == 0) return 0;
|
||||
|
||||
__asm __volatile ("clz %0, %1":"=r"(tmp):"r"(a));
|
||||
|
||||
return (int16_t)tmp;
|
||||
}
|
||||
|
||||
static __inline int16_t WebRtcSpl_NormW16(int16_t a) {
|
||||
int32_t tmp = 0;
|
||||
int32_t a_32 = a;
|
||||
|
||||
if (a_32 == 0) {
|
||||
return 0;
|
||||
}
|
||||
else if (a_32 < 0) {
|
||||
a_32 ^= 0xFFFFFFFF;
|
||||
}
|
||||
|
||||
__asm __volatile ("clz %0, %1":"=r"(tmp):"r"(a_32));
|
||||
|
||||
return (int16_t)(tmp - 17);
|
||||
}
|
||||
|
||||
// TODO(kma): add unit test.
|
||||
static __inline int16_t WebRtcSpl_SatW32ToW16(int32_t value32) {
|
||||
int32_t out = 0;
|
||||
|
||||
__asm __volatile ("ssat %0, #16, %1" : "=r"(out) : "r"(value32));
|
||||
|
||||
return (int16_t)out;
|
||||
}
|
||||
|
||||
#endif // WEBRTC_SPL_SPL_INL_ARMV7_H_
|
||||
225
webrtc/common_audio/signal_processing/include/spl_inl_mips.h
Normal file
225
webrtc/common_audio/signal_processing/include/spl_inl_mips.h
Normal file
@ -0,0 +1,225 @@
|
||||
/*
|
||||
* Copyright (c) 2013 The WebRTC project authors. All Rights Reserved.
|
||||
*
|
||||
* Use of this source code is governed by a BSD-style license
|
||||
* that can be found in the LICENSE file in the root of the source
|
||||
* tree. An additional intellectual property rights grant can be found
|
||||
* in the file PATENTS. All contributing project authors may
|
||||
* be found in the AUTHORS file in the root of the source tree.
|
||||
*/
|
||||
|
||||
|
||||
// This header file includes the inline functions in
|
||||
// the fix point signal processing library.
|
||||
|
||||
#ifndef WEBRTC_SPL_SPL_INL_MIPS_H_
|
||||
#define WEBRTC_SPL_SPL_INL_MIPS_H_
|
||||
|
||||
static __inline int32_t WEBRTC_SPL_MUL_16_16(int32_t a,
|
||||
int32_t b) {
|
||||
int32_t value32 = 0;
|
||||
int32_t a1 = 0, b1 = 0;
|
||||
|
||||
__asm __volatile(
|
||||
#if defined(MIPS32_R2_LE)
|
||||
"seh %[a1], %[a] \n\t"
|
||||
"seh %[b1], %[b] \n\t"
|
||||
#else
|
||||
"sll %[a1], %[a], 16 \n\t"
|
||||
"sll %[b1], %[b], 16 \n\t"
|
||||
"sra %[a1], %[a1], 16 \n\t"
|
||||
"sra %[b1], %[b1], 16 \n\t"
|
||||
#endif
|
||||
"mul %[value32], %[a1], %[b1] \n\t"
|
||||
: [value32] "=r" (value32), [a1] "=&r" (a1), [b1] "=&r" (b1)
|
||||
: [a] "r" (a), [b] "r" (b)
|
||||
: "hi", "lo"
|
||||
);
|
||||
return value32;
|
||||
}
|
||||
|
||||
static __inline int32_t WEBRTC_SPL_MUL_16_32_RSFT16(int16_t a,
|
||||
int32_t b) {
|
||||
int32_t value32 = 0, b1 = 0, b2 = 0;
|
||||
int32_t a1 = 0;
|
||||
|
||||
__asm __volatile(
|
||||
#if defined(MIPS32_R2_LE)
|
||||
"seh %[a1], %[a] \n\t"
|
||||
#else
|
||||
"sll %[a1], %[a], 16 \n\t"
|
||||
"sra %[a1], %[a1], 16 \n\t"
|
||||
#endif
|
||||
"andi %[b2], %[b], 0xFFFF \n\t"
|
||||
"sra %[b1], %[b], 16 \n\t"
|
||||
"sra %[b2], %[b2], 1 \n\t"
|
||||
"mul %[value32], %[a1], %[b1] \n\t"
|
||||
"mul %[b2], %[a1], %[b2] \n\t"
|
||||
"addiu %[b2], %[b2], 0x4000 \n\t"
|
||||
"sra %[b2], %[b2], 15 \n\t"
|
||||
"addu %[value32], %[value32], %[b2] \n\t"
|
||||
: [value32] "=&r" (value32), [b1] "=&r" (b1), [b2] "=&r" (b2),
|
||||
[a1] "=&r" (a1)
|
||||
: [a] "r" (a), [b] "r" (b)
|
||||
: "hi", "lo"
|
||||
);
|
||||
return value32;
|
||||
}
|
||||
|
||||
#if defined(MIPS_DSP_R1_LE)
|
||||
static __inline int16_t WebRtcSpl_SatW32ToW16(int32_t value32) {
|
||||
__asm __volatile(
|
||||
"shll_s.w %[value32], %[value32], 16 \n\t"
|
||||
"sra %[value32], %[value32], 16 \n\t"
|
||||
: [value32] "+r" (value32)
|
||||
:
|
||||
);
|
||||
int16_t out16 = (int16_t)value32;
|
||||
return out16;
|
||||
}
|
||||
|
||||
static __inline int16_t WebRtcSpl_AddSatW16(int16_t a, int16_t b) {
|
||||
int32_t value32 = 0;
|
||||
|
||||
__asm __volatile(
|
||||
"addq_s.ph %[value32], %[a], %[b] \n\t"
|
||||
: [value32] "=r" (value32)
|
||||
: [a] "r" (a), [b] "r" (b)
|
||||
);
|
||||
return (int16_t)value32;
|
||||
}
|
||||
|
||||
static __inline int32_t WebRtcSpl_AddSatW32(int32_t l_var1, int32_t l_var2) {
|
||||
int32_t l_sum;
|
||||
|
||||
__asm __volatile(
|
||||
"addq_s.w %[l_sum], %[l_var1], %[l_var2] \n\t"
|
||||
: [l_sum] "=r" (l_sum)
|
||||
: [l_var1] "r" (l_var1), [l_var2] "r" (l_var2)
|
||||
);
|
||||
|
||||
return l_sum;
|
||||
}
|
||||
|
||||
static __inline int16_t WebRtcSpl_SubSatW16(int16_t var1, int16_t var2) {
|
||||
int32_t value32;
|
||||
|
||||
__asm __volatile(
|
||||
"subq_s.ph %[value32], %[var1], %[var2] \n\t"
|
||||
: [value32] "=r" (value32)
|
||||
: [var1] "r" (var1), [var2] "r" (var2)
|
||||
);
|
||||
|
||||
return (int16_t)value32;
|
||||
}
|
||||
|
||||
static __inline int32_t WebRtcSpl_SubSatW32(int32_t l_var1, int32_t l_var2) {
|
||||
int32_t l_diff;
|
||||
|
||||
__asm __volatile(
|
||||
"subq_s.w %[l_diff], %[l_var1], %[l_var2] \n\t"
|
||||
: [l_diff] "=r" (l_diff)
|
||||
: [l_var1] "r" (l_var1), [l_var2] "r" (l_var2)
|
||||
);
|
||||
|
||||
return l_diff;
|
||||
}
|
||||
#endif
|
||||
|
||||
static __inline int16_t WebRtcSpl_GetSizeInBits(uint32_t n) {
|
||||
int bits = 0;
|
||||
int i32 = 32;
|
||||
|
||||
__asm __volatile(
|
||||
"clz %[bits], %[n] \n\t"
|
||||
"subu %[bits], %[i32], %[bits] \n\t"
|
||||
: [bits] "=&r" (bits)
|
||||
: [n] "r" (n), [i32] "r" (i32)
|
||||
);
|
||||
|
||||
return (int16_t)bits;
|
||||
}
|
||||
|
||||
static __inline int16_t WebRtcSpl_NormW32(int32_t a) {
|
||||
int zeros = 0;
|
||||
|
||||
__asm __volatile(
|
||||
".set push \n\t"
|
||||
".set noreorder \n\t"
|
||||
"bnez %[a], 1f \n\t"
|
||||
" sra %[zeros], %[a], 31 \n\t"
|
||||
"b 2f \n\t"
|
||||
" move %[zeros], $zero \n\t"
|
||||
"1: \n\t"
|
||||
"xor %[zeros], %[a], %[zeros] \n\t"
|
||||
"clz %[zeros], %[zeros] \n\t"
|
||||
"addiu %[zeros], %[zeros], -1 \n\t"
|
||||
"2: \n\t"
|
||||
".set pop \n\t"
|
||||
: [zeros]"=&r"(zeros)
|
||||
: [a] "r" (a)
|
||||
);
|
||||
|
||||
return (int16_t)zeros;
|
||||
}
|
||||
|
||||
static __inline int16_t WebRtcSpl_NormU32(uint32_t a) {
|
||||
int zeros = 0;
|
||||
|
||||
__asm __volatile(
|
||||
"clz %[zeros], %[a] \n\t"
|
||||
: [zeros] "=r" (zeros)
|
||||
: [a] "r" (a)
|
||||
);
|
||||
|
||||
return (int16_t)(zeros & 0x1f);
|
||||
}
|
||||
|
||||
static __inline int16_t WebRtcSpl_NormW16(int16_t a) {
|
||||
int zeros = 0;
|
||||
int a0 = a << 16;
|
||||
|
||||
__asm __volatile(
|
||||
".set push \n\t"
|
||||
".set noreorder \n\t"
|
||||
"bnez %[a0], 1f \n\t"
|
||||
" sra %[zeros], %[a0], 31 \n\t"
|
||||
"b 2f \n\t"
|
||||
" move %[zeros], $zero \n\t"
|
||||
"1: \n\t"
|
||||
"xor %[zeros], %[a0], %[zeros] \n\t"
|
||||
"clz %[zeros], %[zeros] \n\t"
|
||||
"addiu %[zeros], %[zeros], -1 \n\t"
|
||||
"2: \n\t"
|
||||
".set pop \n\t"
|
||||
: [zeros]"=&r"(zeros)
|
||||
: [a0] "r" (a0)
|
||||
);
|
||||
|
||||
return (int16_t)zeros;
|
||||
}
|
||||
|
||||
static __inline int32_t WebRtc_MulAccumW16(int16_t a,
|
||||
int16_t b,
|
||||
int32_t c) {
|
||||
int32_t res = 0, c1 = 0;
|
||||
__asm __volatile(
|
||||
#if defined(MIPS32_R2_LE)
|
||||
"seh %[a], %[a] \n\t"
|
||||
"seh %[b], %[b] \n\t"
|
||||
#else
|
||||
"sll %[a], %[a], 16 \n\t"
|
||||
"sll %[b], %[b], 16 \n\t"
|
||||
"sra %[a], %[a], 16 \n\t"
|
||||
"sra %[b], %[b], 16 \n\t"
|
||||
#endif
|
||||
"mul %[res], %[a], %[b] \n\t"
|
||||
"addu %[c1], %[c], %[res] \n\t"
|
||||
: [c1] "=r" (c1), [res] "=&r" (res)
|
||||
: [a] "r" (a), [b] "r" (b), [c] "r" (c)
|
||||
: "hi", "lo"
|
||||
);
|
||||
return (c1);
|
||||
}
|
||||
|
||||
#endif // WEBRTC_SPL_SPL_INL_MIPS_H_
|
||||
Reference in New Issue
Block a user