246 lines
11 KiB
C#
246 lines
11 KiB
C#
/* Copyright (c) 2006-2011 Skype Limited. All Rights Reserved
|
|
Ported to C# by Logan Stromberg
|
|
|
|
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.
|
|
*/
|
|
|
|
namespace Concentus.Silk
|
|
{
|
|
using Concentus.Common;
|
|
using Concentus.Common.CPlusPlus;
|
|
using Concentus.Silk.Enums;
|
|
using Concentus.Silk.Structs;
|
|
using System.Diagnostics;
|
|
|
|
internal static class LinearAlgebra
|
|
{
|
|
/* Solves Ax = b, assuming A is symmetric */
|
|
internal static void silk_solve_LDL(
|
|
int[] A, /* I Pointer to symetric square matrix A */
|
|
int A_ptr,
|
|
int M, /* I Size of matrix */
|
|
int[] b, /* I Pointer to b vector */
|
|
int[] x_Q16 /* O Pointer to x solution vector */
|
|
)
|
|
{
|
|
Inlines.OpusAssert(M <= SilkConstants.MAX_MATRIX_SIZE);
|
|
int[] L_Q16 = new int[M * M];
|
|
int[] Y = new int[SilkConstants.MAX_MATRIX_SIZE];
|
|
|
|
// [Porting note] This is an interleaved array. Formerly it was an array of data structures laid out thus:
|
|
//private struct inv_D_t
|
|
//{
|
|
// int Q36_part;
|
|
// int Q48_part;
|
|
//}
|
|
int[] inv_D = new int[SilkConstants.MAX_MATRIX_SIZE * 2];
|
|
|
|
/***************************************************
|
|
Factorize A by LDL such that A = L*D*L',
|
|
where L is lower triangular with ones on diagonal
|
|
****************************************************/
|
|
silk_LDL_factorize(A, A_ptr, M, L_Q16, inv_D);
|
|
|
|
/****************************************************
|
|
* substitute D*L'*x = Y. ie:
|
|
L*D*L'*x = b => L*Y = b <=> Y = inv(L)*b
|
|
******************************************************/
|
|
silk_LS_SolveFirst(L_Q16, M, b, Y);
|
|
|
|
/****************************************************
|
|
D*L'*x = Y <=> L'*x = inv(D)*Y, because D is
|
|
diagonal just multiply with 1/d_i
|
|
****************************************************/
|
|
silk_LS_divide_Q16(Y, inv_D, M);
|
|
|
|
/****************************************************
|
|
x = inv(L') * inv(D) * Y
|
|
*****************************************************/
|
|
silk_LS_SolveLast(L_Q16, M, Y, x_Q16);
|
|
|
|
}
|
|
|
|
/* Factorize square matrix A into LDL form */
|
|
private static void silk_LDL_factorize(
|
|
int[] A, /* I/O Pointer to Symetric Square Matrix */
|
|
int A_ptr,
|
|
int M, /* I Size of Matrix */
|
|
int[] L_Q16, /* I/O Pointer to Square Upper triangular Matrix */
|
|
int[] inv_D /* I/O Pointer to vector holding inverted diagonal elements of D */
|
|
)
|
|
{
|
|
int i, j, k, status, loop_count;
|
|
int[] scratch1;
|
|
int scratch1_ptr;
|
|
int[] scratch2;
|
|
int scratch2_ptr;
|
|
int diag_min_value, tmp_32, err;
|
|
int[] v_Q0 = new int[M]; /*SilkConstants.MAX_MATRIX_SIZE*/
|
|
int[] D_Q0 = new int[M]; /*SilkConstants.MAX_MATRIX_SIZE*/
|
|
int one_div_diag_Q36, one_div_diag_Q40, one_div_diag_Q48;
|
|
|
|
Inlines.OpusAssert(M <= SilkConstants.MAX_MATRIX_SIZE);
|
|
|
|
status = 1;
|
|
diag_min_value = Inlines.silk_max_32(Inlines.silk_SMMUL(Inlines.silk_ADD_SAT32(A[A_ptr], A[A_ptr + Inlines.silk_SMULBB(M, M) - 1]), ((int)((TuningParameters.FIND_LTP_COND_FAC) * ((long)1 << (31)) + 0.5))/*Inlines.SILK_CONST(TuningParameters.FIND_LTP_COND_FAC, 31)*/), 1 << 9);
|
|
for (loop_count = 0; loop_count < M && status == 1; loop_count++)
|
|
{
|
|
status = 0;
|
|
for (j = 0; j < M; j++)
|
|
{
|
|
scratch1 = L_Q16;
|
|
scratch1_ptr = Inlines.MatrixGetPointer(j, 0, M);
|
|
tmp_32 = 0;
|
|
for (i = 0; i < j; i++)
|
|
{
|
|
v_Q0[i] = Inlines.silk_SMULWW(D_Q0[i], scratch1[scratch1_ptr + i]); /* Q0 */
|
|
tmp_32 = Inlines.silk_SMLAWW(tmp_32, v_Q0[i], scratch1[scratch1_ptr + i]); /* Q0 */
|
|
}
|
|
tmp_32 = Inlines.silk_SUB32(Inlines.MatrixGet(A, A_ptr, j, j, M), tmp_32);
|
|
|
|
if (tmp_32 < diag_min_value)
|
|
{
|
|
tmp_32 = Inlines.silk_SUB32(Inlines.silk_SMULBB(loop_count + 1, diag_min_value), tmp_32);
|
|
/* Matrix not positive semi-definite, or ill conditioned */
|
|
for (i = 0; i < M; i++)
|
|
{
|
|
Inlines.MatrixSet(A, A_ptr, i, i, M, Inlines.silk_ADD32(Inlines.MatrixGet(A, A_ptr, i, i, M), tmp_32));
|
|
}
|
|
status = 1;
|
|
break;
|
|
}
|
|
D_Q0[j] = tmp_32; /* always < max(Correlation) */
|
|
|
|
/* two-step division */
|
|
one_div_diag_Q36 = Inlines.silk_INVERSE32_varQ(tmp_32, 36); /* Q36 */
|
|
one_div_diag_Q40 = Inlines.silk_LSHIFT(one_div_diag_Q36, 4); /* Q40 */
|
|
err = Inlines.silk_SUB32((int)1 << 24, Inlines.silk_SMULWW(tmp_32, one_div_diag_Q40)); /* Q24 */
|
|
one_div_diag_Q48 = Inlines.silk_SMULWW(err, one_div_diag_Q40); /* Q48 */
|
|
|
|
/* Save 1/Ds */
|
|
inv_D[(j * 2) + 0] = one_div_diag_Q36;
|
|
inv_D[(j * 2) + 1] = one_div_diag_Q48;
|
|
|
|
Inlines.MatrixSet(L_Q16, j, j, M, 65536); /* 1.0 in Q16 */
|
|
scratch1 = A;
|
|
scratch1_ptr = Inlines.MatrixGetPointer(j, 0, M) + A_ptr;
|
|
scratch2 = L_Q16;
|
|
scratch2_ptr = Inlines.MatrixGetPointer(j + 1, 0, M);
|
|
for (i = j + 1; i < M; i++)
|
|
{
|
|
tmp_32 = 0;
|
|
for (k = 0; k < j; k++)
|
|
{
|
|
tmp_32 = Inlines.silk_SMLAWW(tmp_32, v_Q0[k], scratch2[scratch2_ptr + k]); /* Q0 */
|
|
}
|
|
tmp_32 = Inlines.silk_SUB32(scratch1[scratch1_ptr + i], tmp_32); /* always < max(Correlation) */
|
|
|
|
/* tmp_32 / D_Q0[j] : Divide to Q16 */
|
|
Inlines.MatrixSet(L_Q16, i, j, M, Inlines.silk_ADD32(Inlines.silk_SMMUL(tmp_32, one_div_diag_Q48),
|
|
Inlines.silk_RSHIFT(Inlines.silk_SMULWW(tmp_32, one_div_diag_Q36), 4)));
|
|
|
|
/* go to next column */
|
|
scratch2_ptr += M;
|
|
}
|
|
}
|
|
}
|
|
|
|
Inlines.OpusAssert(status == 0);
|
|
}
|
|
|
|
private static void silk_LS_divide_Q16(
|
|
int[] T, /* I/O Numenator vector */
|
|
int[] inv_D, /* I 1 / D vector */
|
|
int M /* I dimension */
|
|
)
|
|
{
|
|
int i;
|
|
int tmp_32;
|
|
int one_div_diag_Q36, one_div_diag_Q48;
|
|
|
|
for (i = 0; i < M; i++)
|
|
{
|
|
one_div_diag_Q36 = inv_D[(i * 2) + 0];
|
|
one_div_diag_Q48 = inv_D[(i * 2) + 1];
|
|
|
|
tmp_32 = T[i];
|
|
T[i] = Inlines.silk_ADD32(Inlines.silk_SMMUL(tmp_32, one_div_diag_Q48), Inlines.silk_RSHIFT(Inlines.silk_SMULWW(tmp_32, one_div_diag_Q36), 4));
|
|
}
|
|
}
|
|
|
|
/* Solve Lx = b, when L is lower triangular and has ones on the diagonal */
|
|
private static void silk_LS_SolveFirst(
|
|
int[] L_Q16, /* I Pointer to Lower Triangular Matrix */
|
|
int M, /* I Dim of Matrix equation */
|
|
int[] b, /* I b Vector */
|
|
int[] x_Q16 /* O x Vector */
|
|
)
|
|
{
|
|
int i, j;
|
|
int ptr32;
|
|
int tmp_32;
|
|
|
|
for (i = 0; i < M; i++)
|
|
{
|
|
ptr32 = Inlines.MatrixGetPointer(i, 0, M);
|
|
tmp_32 = 0;
|
|
for (j = 0; j < i; j++)
|
|
{
|
|
tmp_32 = Inlines.silk_SMLAWW(tmp_32, L_Q16[ptr32 + j], x_Q16[j]);
|
|
}
|
|
x_Q16[i] = Inlines.silk_SUB32(b[i], tmp_32);
|
|
}
|
|
}
|
|
|
|
/* Solve L^t*x = b, where L is lower triangular with ones on the diagonal */
|
|
private static void silk_LS_SolveLast(
|
|
int[] L_Q16, /* I Pointer to Lower Triangular Matrix */
|
|
int M, /* I Dim of Matrix equation */
|
|
int[] b, /* I b Vector */
|
|
int[] x_Q16 /* O x Vector */
|
|
)
|
|
{
|
|
int i, j;
|
|
int ptr32;
|
|
int tmp_32;
|
|
|
|
for (i = M - 1; i >= 0; i--)
|
|
{
|
|
ptr32 = Inlines.MatrixGetPointer(0, i, M);
|
|
tmp_32 = 0;
|
|
for (j = M - 1; j > i; j--)
|
|
{
|
|
tmp_32 = Inlines.silk_SMLAWW(tmp_32, L_Q16[ptr32 + Inlines.silk_SMULBB(j, M)], x_Q16[j]);
|
|
}
|
|
x_Q16[i] = Inlines.silk_SUB32(b[i], tmp_32);
|
|
}
|
|
}
|
|
}
|
|
}
|