mirror of https://github.com/mstorsjo/fdk-aac.git
Merge "Use dynamic scaling depending on autocorr length to avoid signed integer overflow." into sc-dev am: a37553e327
Original change: https://googleplex-android-review.googlesource.com/c/platform/external/aac/+/14355879 Change-Id: I3c35d6347cafc9ace7bbcefd9d15213f257c5cae
This commit is contained in:
commit
23308594fb
|
@ -1,7 +1,7 @@
|
||||||
/* -----------------------------------------------------------------------------
|
/* -----------------------------------------------------------------------------
|
||||||
Software License for The Fraunhofer FDK AAC Codec Library for Android
|
Software License for The Fraunhofer FDK AAC Codec Library for Android
|
||||||
|
|
||||||
© Copyright 1995 - 2018 Fraunhofer-Gesellschaft zur Förderung der angewandten
|
© Copyright 1995 - 2020 Fraunhofer-Gesellschaft zur Förderung der angewandten
|
||||||
Forschung e.V. All rights reserved.
|
Forschung e.V. All rights reserved.
|
||||||
|
|
||||||
1. INTRODUCTION
|
1. INTRODUCTION
|
||||||
|
@ -102,11 +102,6 @@ amm-info@iis.fraunhofer.de
|
||||||
|
|
||||||
#include "autocorr2nd.h"
|
#include "autocorr2nd.h"
|
||||||
|
|
||||||
/* If the accumulator does not provide enough overflow bits,
|
|
||||||
products have to be shifted down in the autocorrelation below. */
|
|
||||||
#define SHIFT_FACTOR (5)
|
|
||||||
#define SHIFT >> (SHIFT_FACTOR)
|
|
||||||
|
|
||||||
/*!
|
/*!
|
||||||
*
|
*
|
||||||
* \brief Calculate second order autocorrelation using 2 accumulators
|
* \brief Calculate second order autocorrelation using 2 accumulators
|
||||||
|
@ -126,45 +121,49 @@ INT autoCorr2nd_real(
|
||||||
|
|
||||||
const FIXP_DBL *realBuf = reBuffer;
|
const FIXP_DBL *realBuf = reBuffer;
|
||||||
|
|
||||||
|
const int len_scale = fMax(DFRACT_BITS - fNormz((FIXP_DBL)(len / 2)), 1);
|
||||||
/*
|
/*
|
||||||
r11r,r22r
|
r11r,r22r
|
||||||
r01r,r12r
|
r01r,r12r
|
||||||
r02r
|
r02r
|
||||||
*/
|
*/
|
||||||
pReBuf = realBuf - 2;
|
pReBuf = realBuf - 2;
|
||||||
accu5 = ((fMultDiv2(pReBuf[0], pReBuf[2]) + fMultDiv2(pReBuf[1], pReBuf[3]))
|
accu5 =
|
||||||
SHIFT);
|
((fMultDiv2(pReBuf[0], pReBuf[2]) + fMultDiv2(pReBuf[1], pReBuf[3])) >>
|
||||||
|
len_scale);
|
||||||
pReBuf++;
|
pReBuf++;
|
||||||
|
|
||||||
/* len must be even */
|
/* len must be even */
|
||||||
accu1 = fPow2Div2(pReBuf[0]) SHIFT;
|
accu1 = fPow2Div2(pReBuf[0]) >> len_scale;
|
||||||
accu3 = fMultDiv2(pReBuf[0], pReBuf[1]) SHIFT;
|
accu3 = fMultDiv2(pReBuf[0], pReBuf[1]) >> len_scale;
|
||||||
pReBuf++;
|
pReBuf++;
|
||||||
|
|
||||||
for (j = (len - 2) >> 1; j != 0; j--, pReBuf += 2) {
|
for (j = (len - 2) >> 1; j != 0; j--, pReBuf += 2) {
|
||||||
accu1 += ((fPow2Div2(pReBuf[0]) + fPow2Div2(pReBuf[1])) SHIFT);
|
accu1 += ((fPow2Div2(pReBuf[0]) + fPow2Div2(pReBuf[1])) >> len_scale);
|
||||||
|
|
||||||
accu3 += ((fMultDiv2(pReBuf[0], pReBuf[1]) +
|
accu3 +=
|
||||||
fMultDiv2(pReBuf[1], pReBuf[2])) SHIFT);
|
((fMultDiv2(pReBuf[0], pReBuf[1]) + fMultDiv2(pReBuf[1], pReBuf[2])) >>
|
||||||
|
len_scale);
|
||||||
|
|
||||||
accu5 += ((fMultDiv2(pReBuf[0], pReBuf[2]) +
|
accu5 +=
|
||||||
fMultDiv2(pReBuf[1], pReBuf[3])) SHIFT);
|
((fMultDiv2(pReBuf[0], pReBuf[2]) + fMultDiv2(pReBuf[1], pReBuf[3])) >>
|
||||||
|
len_scale);
|
||||||
}
|
}
|
||||||
|
|
||||||
accu2 = (fPow2Div2(realBuf[-2]) SHIFT);
|
accu2 = (fPow2Div2(realBuf[-2]) >> len_scale);
|
||||||
accu2 += accu1;
|
accu2 += accu1;
|
||||||
|
|
||||||
accu1 += (fPow2Div2(realBuf[len - 2]) SHIFT);
|
accu1 += (fPow2Div2(realBuf[len - 2]) >> len_scale);
|
||||||
|
|
||||||
accu4 = (fMultDiv2(realBuf[-1], realBuf[-2]) SHIFT);
|
accu4 = (fMultDiv2(realBuf[-1], realBuf[-2]) >> len_scale);
|
||||||
accu4 += accu3;
|
accu4 += accu3;
|
||||||
|
|
||||||
accu3 += (fMultDiv2(realBuf[len - 1], realBuf[len - 2]) SHIFT);
|
accu3 += (fMultDiv2(realBuf[len - 1], realBuf[len - 2]) >> len_scale);
|
||||||
|
|
||||||
mScale = CntLeadingZeros(
|
mScale = CntLeadingZeros(
|
||||||
(accu1 | accu2 | fAbs(accu3) | fAbs(accu4) | fAbs(accu5))) -
|
(accu1 | accu2 | fAbs(accu3) | fAbs(accu4) | fAbs(accu5))) -
|
||||||
1;
|
1;
|
||||||
autoCorrScaling = mScale - 1 - SHIFT_FACTOR; /* -1 because of fMultDiv2*/
|
autoCorrScaling = mScale - 1 - len_scale; /* -1 because of fMultDiv2*/
|
||||||
|
|
||||||
/* Scale to common scale factor */
|
/* Scale to common scale factor */
|
||||||
ac->r11r = accu1 << mScale;
|
ac->r11r = accu1 << mScale;
|
||||||
|
@ -190,7 +189,7 @@ INT autoCorr2nd_cplx(
|
||||||
const FIXP_DBL *imBuffer, /*!< Pointer to imag part of input samples */
|
const FIXP_DBL *imBuffer, /*!< Pointer to imag part of input samples */
|
||||||
const int len /*!< Number of input samples (should be smaller than 128) */
|
const int len /*!< Number of input samples (should be smaller than 128) */
|
||||||
) {
|
) {
|
||||||
int j, autoCorrScaling, mScale, len_scale;
|
int j, autoCorrScaling, mScale;
|
||||||
|
|
||||||
FIXP_DBL accu0, accu1, accu2, accu3, accu4, accu5, accu6, accu7, accu8;
|
FIXP_DBL accu0, accu1, accu2, accu3, accu4, accu5, accu6, accu7, accu8;
|
||||||
|
|
||||||
|
@ -199,7 +198,7 @@ INT autoCorr2nd_cplx(
|
||||||
const FIXP_DBL *realBuf = reBuffer;
|
const FIXP_DBL *realBuf = reBuffer;
|
||||||
const FIXP_DBL *imagBuf = imBuffer;
|
const FIXP_DBL *imagBuf = imBuffer;
|
||||||
|
|
||||||
(len > 64) ? (len_scale = 6) : (len_scale = 5);
|
const int len_scale = fMax(DFRACT_BITS - fNormz((FIXP_DBL)len), 1);
|
||||||
/*
|
/*
|
||||||
r00r,
|
r00r,
|
||||||
r11r,r22r
|
r11r,r22r
|
||||||
|
|
Loading…
Reference in New Issue