[Opus] Update to v1.2.1

third_party/opus/src/silk/SigProc_FIX.h

 /***********************************************************************
 Copyright (c) 2006-2011, Skype Limited. All rights reserved.
 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.
@@ skipping 17 matching lines @@
 #ifndef SILK_SIGPROC_FIX_H
 #define SILK_SIGPROC_FIX_H
 
 #ifdef  __cplusplus
 extern "C"
 {
 #endif
 
 /*#define silk_MACRO_COUNT */          /* Used to enable WMOPS counting */
 
-#define SILK_MAX_ORDER_LPC            16            /* max order of the LPC analysis in schur() and k2a() */
+#define SILK_MAX_ORDER_LPC            24            /* max order of the LPC analysis in schur() and k2a() */
 
 #include <string.h>                                 /* for memset(), memcpy(), memmove() */
 #include "typedef.h"
 #include "resampler_structs.h"
 #include "macros.h"
 #include "cpu_support.h"
 
 #if defined(OPUS_X86_MAY_HAVE_SSE4_1)
 #include "x86/SigProc_FIX_sse.h"
 #endif
 
+#if (defined(OPUS_ARM_ASM) || defined(OPUS_ARM_MAY_HAVE_NEON_INTR))
+#include "arm/biquad_alt_arm.h"
+#include "arm/LPC_inv_pred_gain_arm.h"
+#endif
+
 /********************************************************************/
 /*                    SIGNAL PROCESSING FUNCTIONS                   */
 /********************************************************************/
 
 /*!
  * Initialize/reset the resampler state for a given pair of input/output sampling rates
 */
 opus_int silk_resampler_init(
     silk_resampler_state_struct *S,                 /* I/O  Resampler state                                             */
     opus_int32                  Fs_Hz_in,           /* I    Input sampling rate (Hz)                                    */
@@ skipping 29 matching lines @@
     opus_int16                  *out,               /* O    Output signal [ floor(2*inLen/3) ]                          */
     const opus_int16            *in,                /* I    Input signal [ inLen ]                                      */
     opus_int32                  inLen               /* I    Number of input samples                                     */
 );
 
 /*!
  * second order ARMA filter;
  * slower than biquad() but uses more precise coefficients
  * can handle (slowly) varying coefficients
  */
-void silk_biquad_alt(
+void silk_biquad_alt_stride1(
     const opus_int16            *in,                /* I     input signal                                               */
     const opus_int32            *B_Q28,             /* I     MA coefficients [3]                                        */
     const opus_int32            *A_Q28,             /* I     AR coefficients [2]                                        */
     opus_int32                  *S,                 /* I/O   State vector [2]                                           */
     opus_int16                  *out,               /* O     output signal                                              */
-    const opus_int32            len,                /* I     signal length (must be even)                               */
-    opus_int                    stride              /* I     Operate on interleaved signal if > 1                       */
+    const opus_int32            len                 /* I     signal length (must be even)                               */
+);
+
+void silk_biquad_alt_stride2_c(
+    const opus_int16            *in,                /* I     input signal                                               */
+    const opus_int32            *B_Q28,             /* I     MA coefficients [3]                                        */
+    const opus_int32            *A_Q28,             /* I     AR coefficients [2]                                        */
+    opus_int32                  *S,                 /* I/O   State vector [4]                                           */
+    opus_int16                  *out,               /* O     output signal                                              */
+    const opus_int32            len                 /* I     signal length (must be even)                               */
 );
 
 /* Variable order MA prediction error filter. */
 void silk_LPC_analysis_filter(
     opus_int16                  *out,               /* O    Output signal                                               */
     const opus_int16            *in,                /* I    Input signal                                                */
     const opus_int16            *B,                 /* I    MA prediction coefficients, Q12 [order]                     */
     const opus_int32            len,                /* I    Signal length                                               */
     const opus_int32            d,                  /* I    Filter order                                                */
     int                         arch                /* I    Run-time architecture                                       */
 );
 
 /* Chirp (bandwidth expand) LP AR filter */
 void silk_bwexpander(
     opus_int16                  *ar,                /* I/O  AR filter to be expanded (without leading 1)                */
     const opus_int              d,                  /* I    Length of ar                                                */
     opus_int32                  chirp_Q16           /* I    Chirp factor (typically in the range 0 to 1)                */
 );
 
 /* Chirp (bandwidth expand) LP AR filter */
 void silk_bwexpander_32(
     opus_int32                  *ar,                /* I/O  AR filter to be expanded (without leading 1)                */
     const opus_int              d,                  /* I    Length of ar                                                */
     opus_int32                  chirp_Q16           /* I    Chirp factor in Q16                                         */
 );
 
 /* Compute inverse of LPC prediction gain, and                           */
 /* test if LPC coefficients are stable (all poles within unit circle)    */
-opus_int32 silk_LPC_inverse_pred_gain(              /* O   Returns inverse prediction gain in energy domain, Q30        */
+opus_int32 silk_LPC_inverse_pred_gain_c(            /* O   Returns inverse prediction gain in energy domain, Q30        */
     const opus_int16            *A_Q12,             /* I   Prediction coefficients, Q12 [order]                         */
     const opus_int              order               /* I   Prediction order                                             */
 );
 
-/* For input in Q24 domain */
-opus_int32 silk_LPC_inverse_pred_gain_Q24(          /* O    Returns inverse prediction gain in energy domain, Q30       */
-    const opus_int32            *A_Q24,             /* I    Prediction coefficients [order]                             */
-    const opus_int              order               /* I    Prediction order                                            */
-);
-
 /* Split signal in two decimated bands using first-order allpass filters */
 void silk_ana_filt_bank_1(
     const opus_int16            *in,                /* I    Input signal [N]                                            */
     opus_int32                  *S,                 /* I/O  State vector [2]                                            */
     opus_int16                  *outL,              /* O    Low band [N/2]                                              */
     opus_int16                  *outH,              /* O    High band [N/2]                                             */
     const opus_int32            N                   /* I    Number of input samples                                     */
 );
 
+#if !defined(OVERRIDE_silk_biquad_alt_stride2)
+#define silk_biquad_alt_stride2(in, B_Q28, A_Q28, S, out, len, arch) ((void)(arch), silk_biquad_alt_stride2_c(in, B_Q28, A_Q28, S, out, len))
+#endif
+
+#if !defined(OVERRIDE_silk_LPC_inverse_pred_gain)
+#define silk_LPC_inverse_pred_gain(A_Q12, order, arch)     ((void)(arch), silk_LPC_inverse_pred_gain_c(A_Q12, order))
+#endif
+
 /********************************************************************/
 /*                        SCALAR FUNCTIONS                          */
 /********************************************************************/
 
 /* Approximation of 128 * log2() (exact inverse of approx 2^() below) */
 /* Convert input to a log scale    */
 opus_int32 silk_lin2log(
     const opus_int32            inLin               /* I  input in linear scale                                         */
 );
 
@@ skipping 99 matching lines @@
 void silk_A2NLSF(
     opus_int16                  *NLSF,              /* O    Normalized Line Spectral Frequencies in Q15 (0..2^15-1) [d] */
     opus_int32                  *a_Q16,             /* I/O  Monic whitening filter coefficients in Q16 [d]              */
     const opus_int              d                   /* I    Filter order (must be even)                                 */
 );
 
 /* compute whitening filter coefficients from normalized line spectral frequencies */
 void silk_NLSF2A(
     opus_int16                  *a_Q12,             /* O    monic whitening filter coefficients in Q12,  [ d ]          */
     const opus_int16            *NLSF,              /* I    normalized line spectral frequencies in Q15, [ d ]          */
-    const opus_int              d                   /* I    filter order (should be even)                               */
+    const opus_int              d,                  /* I    filter order (should be even)                               */
+    int                         arch                /* I    Run-time architecture                                       */
+);
+
+/* Convert int32 coefficients to int16 coefs and make sure there's no wrap-around */
+void silk_LPC_fit(
+    opus_int16                  *a_QOUT,            /* O    Output signal                                               */
+    opus_int32                  *a_QIN,             /* I/O  Input signal                                                */
+    const opus_int              QOUT,               /* I    Input Q domain                                              */
+    const opus_int              QIN,                /* I    Input Q domain                                              */
+    const opus_int              d                   /* I    Filter order                                                */
 );
 
 void silk_insertion_sort_increasing(
     opus_int32                  *a,                 /* I/O   Unsorted / Sorted vector                                   */
     opus_int                    *idx,               /* O     Index vector for the sorted elements                       */
     const opus_int              L,                  /* I     Vector length                                              */
     const opus_int              K                   /* I     Number of correctly sorted positions                       */
 );
 
 void silk_insertion_sort_decreasing_int16(
@@ skipping 179 matching lines @@
 #define silk_SUB_SAT64(a, b)                ((((a)-(b)) & 0x8000000000000000LL) == 0 ?                                               \
                                             (( (a) & ((b)^0x8000000000000000LL) & 0x8000000000000000LL) ? silk_int64_MIN : (a)-(b)) : \
                                             ((((a)^0x8000000000000000LL) & (b)  & 0x8000000000000000LL) ? silk_int64_MAX : (a)-(b)) )
 
 /* Saturation for positive input values */
 #define silk_POS_SAT32(a)                   ((a) > silk_int32_MAX ? silk_int32_MAX : (a))
 
 /* Add with saturation for positive input values */
 #define silk_ADD_POS_SAT8(a, b)             ((((a)+(b)) & 0x80)                 ? silk_int8_MAX  : ((a)+(b)))
 #define silk_ADD_POS_SAT16(a, b)            ((((a)+(b)) & 0x8000)               ? silk_int16_MAX : ((a)+(b)))
-#define silk_ADD_POS_SAT32(a, b)            ((((a)+(b)) & 0x80000000)           ? silk_int32_MAX : ((a)+(b)))
-#define silk_ADD_POS_SAT64(a, b)            ((((a)+(b)) & 0x8000000000000000LL) ? silk_int64_MAX : ((a)+(b)))
+#define silk_ADD_POS_SAT32(a, b)            ((((opus_uint32)(a)+(opus_uint32)(b)) & 0x80000000) ? silk_int32_MAX : ((a)+(b)))
 
 #define silk_LSHIFT8(a, shift)              ((opus_int8)((opus_uint8)(a)<<(shift)))         /* shift >= 0, shift < 8  */
 #define silk_LSHIFT16(a, shift)             ((opus_int16)((opus_uint16)(a)<<(shift)))       /* shift >= 0, shift < 16 */
 #define silk_LSHIFT32(a, shift)             ((opus_int32)((opus_uint32)(a)<<(shift)))       /* shift >= 0, shift < 32 */
 #define silk_LSHIFT64(a, shift)             ((opus_int64)((opus_uint64)(a)<<(shift)))       /* shift >= 0, shift < 64 */
 #define silk_LSHIFT(a, shift)               silk_LSHIFT32(a, shift)                         /* shift >= 0, shift < 32 */
 
 #define silk_RSHIFT8(a, shift)              ((a)>>(shift))                                  /* shift >= 0, shift < 8  */
 #define silk_RSHIFT16(a, shift)             ((a)>>(shift))                                  /* shift >= 0, shift < 16 */
 #define silk_RSHIFT32(a, shift)             ((a)>>(shift))                                  /* shift >= 0, shift < 32 */
@@ skipping 79 matching lines @@
 #define silk_abs_int(a)                     (((a) ^ ((a) >> (8 * sizeof(a) - 1))) - ((a) >> (8 * sizeof(a) - 1)))
 #define silk_abs_int32(a)                   (((a) ^ ((a) >> 31)) - ((a) >> 31))
 #define silk_abs_int64(a)                   (((a) >  0)  ? (a) : -(a))
 
 #define silk_sign(a)                        ((a) > 0 ? 1 : ( (a) < 0 ? -1 : 0 ))
 
 /* PSEUDO-RANDOM GENERATOR                                                          */
 /* Make sure to store the result as the seed for the next call (also in between     */
 /* frames), otherwise result won't be random at all. When only using some of the    */
 /* bits, take the most significant bits by right-shifting.                          */
-#define silk_RAND(seed)                     (silk_MLA_ovflw(907633515, (seed), 196314165))
+#define RAND_MULTIPLIER                     196314165
+#define RAND_INCREMENT                      907633515
+#define silk_RAND(seed)                     (silk_MLA_ovflw((RAND_INCREMENT), (seed), (RAND_MULTIPLIER)))
 
 /*  Add some multiplication functions that can be easily mapped to ARM. */
 
 /*    silk_SMMUL: Signed top word multiply.
           ARMv6        2 instruction cycles.
           ARMv3M+      3 instruction cycles. use SMULL and ignore LSB registers.(except xM)*/
 /*#define silk_SMMUL(a32, b32)                (opus_int32)silk_RSHIFT(silk_SMLAL(silk_SMULWB((a32), (b32)), (a32), silk_RSHIFT_ROUND((b32), 16)), 16)*/
 /* the following seems faster on x86 */
 #define silk_SMMUL(a32, b32)                (opus_int32)silk_RSHIFT64(silk_SMULL((a32), (b32)), 32)
 
@@ skipping 20 matching lines @@
 #if defined(MIPSr1_ASM)
 #include "mips/sigproc_fix_mipsr1.h"
 #endif
 
 
 #ifdef  __cplusplus
 }
 #endif
 
 #endif /* SILK_SIGPROC_FIX_H */