Merge fdk-aac 'v2.0.3' into dabplus2dabplus2

8 files changed, 76 insertions, 74 deletions

diff --git a/libFDK/include/FDK_archdef.h b/libFDK/include/FDK_archdef.h
index b4fef8a..f3254f2 100644
--- a/libFDK/include/FDK_archdef.h
+++ b/libFDK/include/FDK_archdef.h
@@ -222,8 +222,18 @@ amm-info@iis.fraunhofer.de
 #define POW2COEFF_16BIT
 #define LDCOEFF_16BIT
 
+#elif defined(__riscv)
+#define ARCH_PREFER_MULT_32x16
+#define SINETABLE_16BIT
+#define POW2COEFF_16BIT
+#define LDCOEFF_16BIT
+#define WINDOWTABLE_16BIT
+
 #elif defined(__powerpc__)
 #define ARCH_PREFER_MULT_32x32
+
+#elif defined(__s390x__)
+#define ARCH_PREFER_MULT_32x32
 #define ARCH_PREFER_MULT_32x16
 #define SINETABLE_16BIT
 #define POW2COEFF_16BIT
diff --git a/libFDK/include/fft.h b/libFDK/include/fft.h
index d394046..4ef62b4 100644
--- a/libFDK/include/fft.h
+++ b/libFDK/include/fft.h
@@ -139,7 +139,7 @@ void ifft(int length, FIXP_DBL *pInput, INT *scalefactor);
  * bit scale headroom. The values are interleaved, real/imag pairs.
  */
 LNK_SECTION_CODE_L1
-static FDK_FORCEINLINE void fft_4(FIXP_DBL *x) {
+static inline void fft_4(FIXP_DBL *x) {
   FIXP_DBL a00, a10, a20, a30, tmp0, tmp1;
 
   a00 = (x[0] + x[4]) >> 1; /* Re A + Re B */
@@ -168,7 +168,7 @@ static FDK_FORCEINLINE void fft_4(FIXP_DBL *x) {
 
 #ifndef FUNCTION_fft_8
 LNK_SECTION_CODE_L1
-static FDK_FORCEINLINE void fft_8(FIXP_DBL *x) {
+static inline void fft_8(FIXP_DBL *x) {
   FIXP_SPK w_PiFOURTH = {{FIXP_SGL(0x5A82), FIXP_SGL(0x5A82)}};
 
   FIXP_DBL a00, a10, a20, a30;
diff --git a/libFDK/include/nlc_dec.h b/libFDK/include/nlc_dec.h
index cca97f1..aded569 100644
--- a/libFDK/include/nlc_dec.h
+++ b/libFDK/include/nlc_dec.h
@@ -1,7 +1,7 @@
 /* -----------------------------------------------------------------------------
 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.
 
  1.    INTRODUCTION
@@ -159,9 +159,6 @@ typedef enum {
 #ifndef HUFFDEC_PARAMS
 #define HUFFDEC_PARMS
 
-#define PAIR_SHIFT 4
-#define PAIR_MASK 0xf
-
 #define MAX_ENTRIES 168
 #define HANDLE_HUFF_NODE const SHORT(*)[MAX_ENTRIES][2]
 
diff --git a/libFDK/src/FDK_hybrid.cpp b/libFDK/src/FDK_hybrid.cpp
index 08d32a8..d208abd 100644
--- a/libFDK/src/FDK_hybrid.cpp
+++ b/libFDK/src/FDK_hybrid.cpp
@@ -1,7 +1,7 @@
 /* -----------------------------------------------------------------------------
 Software License for The Fraunhofer FDK AAC Codec Library for Android
 
-© Copyright  1995 - 2019 Fraunhofer-Gesellschaft zur Förderung der angewandten
+© Copyright  1995 - 2020 Fraunhofer-Gesellschaft zur Förderung der angewandten
 Forschung e.V. All rights reserved.
 
  1.    INTRODUCTION
@@ -539,11 +539,11 @@ static void dualChannelFiltering(const FIXP_DBL *const pQmfReal,
   i6 = pQmfImag[pReadIdx[6]] >> 2;
 
   FDK_ASSERT((invert == 0) || (invert == 1));
-  mHybridReal[0 + invert] = (r6 + r1) << 1;
-  mHybridImag[0 + invert] = (i6 + i1) << 1;
+  mHybridReal[0 + invert] = SATURATE_LEFT_SHIFT((r6 + r1), 1, DFRACT_BITS);
+  mHybridImag[0 + invert] = SATURATE_LEFT_SHIFT((i6 + i1), 1, DFRACT_BITS);
 
-  mHybridReal[1 - invert] = (r6 - r1) << 1;
-  mHybridImag[1 - invert] = (i6 - i1) << 1;
+  mHybridReal[1 - invert] = SATURATE_LEFT_SHIFT((r6 - r1), 1, DFRACT_BITS);
+  mHybridImag[1 - invert] = SATURATE_LEFT_SHIFT((i6 - i1), 1, DFRACT_BITS);
 }
 
 static void fourChannelFiltering(const FIXP_DBL *const pQmfReal,
@@ -766,15 +766,15 @@ static void eightChannelFiltering(const FIXP_DBL *const pQmfReal,
     mHybridReal[3] = pfft[FFT_IDX_R(1)] << sc;
     mHybridImag[3] = pfft[FFT_IDX_I(1)] << sc;
 
-    mHybridReal[4] = pfft[FFT_IDX_R(2)] << sc;
-    mHybridReal[4] += pfft[FFT_IDX_R(5)] << sc;
-    mHybridImag[4] = pfft[FFT_IDX_I(2)] << sc;
-    mHybridImag[4] += pfft[FFT_IDX_I(5)] << sc;
+    mHybridReal[4] = SATURATE_LEFT_SHIFT(
+        (pfft[FFT_IDX_R(2)] + pfft[FFT_IDX_R(5)]), sc, DFRACT_BITS);
+    mHybridImag[4] = SATURATE_LEFT_SHIFT(
+        (pfft[FFT_IDX_I(2)] + pfft[FFT_IDX_I(5)]), sc, DFRACT_BITS);
 
-    mHybridReal[5] = pfft[FFT_IDX_R(3)] << sc;
-    mHybridReal[5] += pfft[FFT_IDX_R(4)] << sc;
-    mHybridImag[5] = pfft[FFT_IDX_I(3)] << sc;
-    mHybridImag[5] += pfft[FFT_IDX_I(4)] << sc;
+    mHybridReal[5] = SATURATE_LEFT_SHIFT(
+        (pfft[FFT_IDX_R(3)] + pfft[FFT_IDX_R(4)]), sc, DFRACT_BITS);
+    mHybridImag[5] = SATURATE_LEFT_SHIFT(
+        (pfft[FFT_IDX_I(3)] + pfft[FFT_IDX_I(4)]), sc, DFRACT_BITS);
   } else {
     for (k = 0; k < 8; k++) {
       mHybridReal[k] = pfft[FFT_IDX_R(k)] << sc;
diff --git a/libFDK/src/autocorr2nd.cpp b/libFDK/src/autocorr2nd.cpp
index 718a555..8c5673c 100644
--- a/libFDK/src/autocorr2nd.cpp
+++ b/libFDK/src/autocorr2nd.cpp
@@ -1,7 +1,7 @@
 /* -----------------------------------------------------------------------------
 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.
 
  1.    INTRODUCTION
@@ -102,11 +102,6 @@ amm-info@iis.fraunhofer.de
 
 #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
@@ -126,45 +121,49 @@ INT autoCorr2nd_real(
 
   const FIXP_DBL *realBuf = reBuffer;
 
+  const int len_scale = fMax(DFRACT_BITS - fNormz((FIXP_DBL)(len / 2)), 1);
   /*
     r11r,r22r
     r01r,r12r
     r02r
   */
   pReBuf = realBuf - 2;
-  accu5 = ((fMultDiv2(pReBuf[0], pReBuf[2]) + fMultDiv2(pReBuf[1], pReBuf[3]))
-               SHIFT);
+  accu5 =
+      ((fMultDiv2(pReBuf[0], pReBuf[2]) + fMultDiv2(pReBuf[1], pReBuf[3])) >>
+       len_scale);
   pReBuf++;
 
   /* len must be even */
-  accu1 = fPow2Div2(pReBuf[0]) SHIFT;
-  accu3 = fMultDiv2(pReBuf[0], pReBuf[1]) SHIFT;
+  accu1 = fPow2Div2(pReBuf[0]) >> len_scale;
+  accu3 = fMultDiv2(pReBuf[0], pReBuf[1]) >> len_scale;
   pReBuf++;
 
   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]) +
-               fMultDiv2(pReBuf[1], pReBuf[2])) SHIFT);
+    accu3 +=
+        ((fMultDiv2(pReBuf[0], pReBuf[1]) + fMultDiv2(pReBuf[1], pReBuf[2])) >>
+         len_scale);
 
-    accu5 += ((fMultDiv2(pReBuf[0], pReBuf[2]) +
-               fMultDiv2(pReBuf[1], pReBuf[3])) SHIFT);
+    accu5 +=
+        ((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;
 
-  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;
 
-  accu3 += (fMultDiv2(realBuf[len - 1], realBuf[len - 2]) SHIFT);
+  accu3 += (fMultDiv2(realBuf[len - 1], realBuf[len - 2]) >> len_scale);
 
   mScale = CntLeadingZeros(
                (accu1 | accu2 | fAbs(accu3) | fAbs(accu4) | fAbs(accu5))) -
            1;
-  autoCorrScaling = mScale - 1 - SHIFT_FACTOR; /* -1 because of fMultDiv2*/
+  autoCorrScaling = mScale - 1 - len_scale; /* -1 because of fMultDiv2*/
 
   /* Scale to common scale factor */
   ac->r11r = accu1 << mScale;
@@ -190,7 +189,7 @@ INT autoCorr2nd_cplx(
     const FIXP_DBL *imBuffer, /*!< Pointer to imag part of input samples */
     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;
 
@@ -199,7 +198,7 @@ INT autoCorr2nd_cplx(
   const FIXP_DBL *realBuf = reBuffer;
   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,
     r11r,r22r
diff --git a/libFDK/src/dct.cpp b/libFDK/src/dct.cpp
index bd26736..35507b5 100644
--- a/libFDK/src/dct.cpp
+++ b/libFDK/src/dct.cpp
@@ -1,7 +1,7 @@
 /* -----------------------------------------------------------------------------
 Software License for The Fraunhofer FDK AAC Codec Library for Android
 
-© Copyright  1995 - 2019 Fraunhofer-Gesellschaft zur Förderung der angewandten
+© Copyright  1995 - 2020 Fraunhofer-Gesellschaft zur Förderung der angewandten
 Forschung e.V. All rights reserved.
 
  1.    INTRODUCTION
@@ -305,9 +305,8 @@ void dct_II(
 
   {
     for (i = 0; i < M; i++) {
-      tmp[i] = pDat[2 * i] >> 1; /* dit_fft expects 1 bit scaled input values */
-      tmp[L - 1 - i] =
-          pDat[2 * i + 1] >> 1; /* dit_fft expects 1 bit scaled input values */
+      tmp[i] = pDat[2 * i] >> 2;
+      tmp[L - 1 - i] = pDat[2 * i + 1] >> 2;
     }
   }
 
@@ -337,15 +336,14 @@ void dct_II(
     a1 = ((pTmp_0[0] >> 1) + (pTmp_1[0] >> 1));
     a2 = ((pTmp_0[1] >> 1) - (pTmp_1[1] >> 1));
 
-    cplxMultDiv2(&accu3, &accu4, (a1 + accu2), -(accu1 + a2),
-                 sin_twiddle[i * inc]);
-    pDat[L - i] = accu4;
-    pDat[i] = accu3;
+    cplxMult(&accu3, &accu4, (accu1 + a2), (a1 + accu2), sin_twiddle[i * inc]);
+    pDat[L - i] = -accu3;
+    pDat[i] = accu4;
 
-    cplxMultDiv2(&accu3, &accu4, (a1 - accu2), -(accu1 - a2),
-                 sin_twiddle[(M - i) * inc]);
-    pDat[M + i] = accu4;
-    pDat[M - i] = accu3;
+    cplxMult(&accu3, &accu4, (accu1 - a2), (a1 - accu2),
+             sin_twiddle[(M - i) * inc]);
+    pDat[M + i] = -accu3;
+    pDat[M - i] = accu4;
 
     /* Create index helper variables for (4*i)*inc indexed equivalent values of
      * short tables. */
@@ -356,12 +354,12 @@ void dct_II(
     }
   }
 
-  cplxMultDiv2(&accu1, &accu2, tmp[M], tmp[M + 1], sin_twiddle[(M / 2) * inc]);
+  cplxMult(&accu1, &accu2, tmp[M], tmp[M + 1], sin_twiddle[(M / 2) * inc]);
   pDat[L - (M / 2)] = accu2;
   pDat[M / 2] = accu1;
 
-  pDat[0] = (tmp[0] >> 1) + (tmp[1] >> 1);
-  pDat[M] = fMult(((tmp[0] >> 1) - (tmp[1] >> 1)),
+  pDat[0] = tmp[0] + tmp[1];
+  pDat[M] = fMult(tmp[0] - tmp[1],
                   sin_twiddle[M * inc].v.re); /* cos((PI/(2*L))*M); */
 
   *pDat_e += 2;
diff --git a/libFDK/src/fft.cpp b/libFDK/src/fft.cpp
index 4e6fdd2..c9ee784 100644
--- a/libFDK/src/fft.cpp
+++ b/libFDK/src/fft.cpp
@@ -170,7 +170,7 @@ amm-info@iis.fraunhofer.de
 
 /* Performs the FFT of length 2. Input vector unscaled, output vector scaled
  * with factor 0.5 */
-static FDK_FORCEINLINE void fft2(FIXP_DBL *RESTRICT pDat) {
+static inline void fft2(FIXP_DBL *RESTRICT pDat) {
   FIXP_DBL r1, i1;
   FIXP_DBL r2, i2;
 
@@ -196,7 +196,7 @@ static FDK_FORCEINLINE void fft2(FIXP_DBL *RESTRICT pDat) {
 
 #ifndef FUNCTION_fft3
 /* Performs the FFT of length 3 according to the algorithm after winograd. */
-static FDK_FORCEINLINE void fft3(FIXP_DBL *RESTRICT pDat) {
+static inline void fft3(FIXP_DBL *RESTRICT pDat) {
   FIXP_DBL r1, r2;
   FIXP_DBL s1, s2;
   FIXP_DBL pD;
@@ -233,7 +233,7 @@ static FDK_FORCEINLINE void fft3(FIXP_DBL *RESTRICT pDat) {
 
 /* performs the FFT of length 5 according to the algorithm after winograd */
 /* This version works with a prescale of 2 instead of 3 */
-static FDK_FORCEINLINE void fft5(FIXP_DBL *RESTRICT pDat) {
+static inline void fft5(FIXP_DBL *RESTRICT pDat) {
   FIXP_DBL r1, r2, r3, r4;
   FIXP_DBL s1, s2, s3, s4;
   FIXP_DBL t;
diff --git a/libFDK/src/nlc_dec.cpp b/libFDK/src/nlc_dec.cpp
index 8b2f97d..3733d98 100644
--- a/libFDK/src/nlc_dec.cpp
+++ b/libFDK/src/nlc_dec.cpp
@@ -1,7 +1,7 @@
 /* -----------------------------------------------------------------------------
 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.
 
  1.    INTRODUCTION
@@ -568,12 +568,12 @@ bail:
 static ERROR_t huff_decode(HANDLE_FDK_BITSTREAM strm, SCHAR* out_data_1,
                            SCHAR* out_data_2, DATA_TYPE data_type,
                            DIFF_TYPE diff_type_1, DIFF_TYPE diff_type_2,
-                           int num_val, int* cdg_scheme, int ldMode) {
+                           int num_val, PAIRING* pairing_scheme, int ldMode) {
   ERROR_t err = HUFFDEC_OK;
+  CODING_SCHEME coding_scheme = HUFF_1D;
   DIFF_TYPE diff_type;
 
   int i = 0;
-  ULONG data = 0;
 
   SCHAR pair_vec[28][2];
 
@@ -596,15 +596,13 @@ static ERROR_t huff_decode(HANDLE_FDK_BITSTREAM strm, SCHAR* out_data_1,
   int hufYY;
 
   /* Coding scheme */
-  data = FDKreadBits(strm, 1);
-  *cdg_scheme = (data << PAIR_SHIFT);
+  coding_scheme = (CODING_SCHEME)FDKreadBits(strm, 1);
 
-  if (*cdg_scheme >> PAIR_SHIFT == HUFF_2D) {
+  if (coding_scheme == HUFF_2D) {
     if ((out_data_1 != NULL) && (out_data_2 != NULL) && (ldMode == 0)) {
-      data = FDKreadBits(strm, 1);
-      *cdg_scheme = (*cdg_scheme | data);
+      *pairing_scheme = (PAIRING)FDKreadBits(strm, 1);
     } else {
-      *cdg_scheme = (*cdg_scheme | FREQ_PAIR);
+      *pairing_scheme = FREQ_PAIR;
     }
   }
 
@@ -613,7 +611,7 @@ static ERROR_t huff_decode(HANDLE_FDK_BITSTREAM strm, SCHAR* out_data_1,
     hufYY2 = diff_type_2;
   }
 
-  switch (*cdg_scheme >> PAIR_SHIFT) {
+  switch (coding_scheme) {
     case HUFF_1D:
       p0_flag[0] = (diff_type_1 == DIFF_FREQ);
       p0_flag[1] = (diff_type_2 == DIFF_FREQ);
@@ -634,7 +632,7 @@ static ERROR_t huff_decode(HANDLE_FDK_BITSTREAM strm, SCHAR* out_data_1,
 
     case HUFF_2D:
 
-      switch (*cdg_scheme & PAIR_MASK) {
+      switch (*pairing_scheme) {
         case FREQ_PAIR:
 
           if (out_data_1 != NULL) {
@@ -843,7 +841,7 @@ ERROR_t EcDataPairDec(DECODER_TYPE DECODER, HANDLE_FDK_BITSTREAM strm,
   SCHAR* pDataVec[2] = {NULL, NULL};
 
   DIFF_TYPE diff_type[2] = {DIFF_FREQ, DIFF_FREQ};
-  int cdg_scheme = HUFF_1D;
+  PAIRING pairing = FREQ_PAIR;
   DIRECTION direction = BACKWARDS;
 
   switch (data_type) {
@@ -959,7 +957,7 @@ ERROR_t EcDataPairDec(DECODER_TYPE DECODER, HANDLE_FDK_BITSTREAM strm,
     }
     /* Huffman decoding */
     err = huff_decode(strm, pDataVec[0], pDataVec[1], data_type, diff_type[0],
-                      diff_type[1], dataBands, &cdg_scheme,
+                      diff_type[1], dataBands, &pairing,
                       (DECODER == SAOC_DECODER));
     if (err != HUFFDEC_OK) {
       return HUFFDEC_NOTOK;
@@ -986,8 +984,8 @@ ERROR_t EcDataPairDec(DECODER_TYPE DECODER, HANDLE_FDK_BITSTREAM strm,
         }
       }
 
-      mixed_time_pair = (diff_type[0] != diff_type[1]) &&
-                        ((cdg_scheme & PAIR_MASK) == TIME_PAIR);
+      mixed_time_pair =
+          (diff_type[0] != diff_type[1]) && (pairing == TIME_PAIR);
 
       if (direction == BACKWARDS) {
         if (diff_type[0] == DIFF_FREQ) {