Merge remote-tracking branch 'mstorjo/master' into dabplus2

9 files changed, 54 insertions, 724 deletions

diff --git a/libFDK/include/FDK_bitbuffer.h b/libFDK/include/FDK_bitbuffer.h
index ed0b2f6..19a24b3 100644
--- a/libFDK/include/FDK_bitbuffer.h
+++ b/libFDK/include/FDK_bitbuffer.h
@@ -113,7 +113,6 @@ typedef struct {
   UINT ValidBits;
   UINT ReadOffset;
   UINT WriteOffset;
-  UINT BitCnt;
   UINT BitNdx;
 
   UCHAR *Buffer;
@@ -159,15 +158,10 @@ void FDK_pushBack(HANDLE_FDK_BITBUF hBitBuffer, const UINT numberOfBits,
 void FDK_pushForward(HANDLE_FDK_BITBUF hBitBuffer, const UINT numberOfBits,
                      UCHAR config);
 
-void FDK_byteAlign(HANDLE_FDK_BITBUF hBitBuffer, UCHAR config);
-
 UINT FDK_getValidBits(HANDLE_FDK_BITBUF hBitBuffer);
 
 INT FDK_getFreeBits(HANDLE_FDK_BITBUF hBitBuffer);
 
-void FDK_setBitCnt(HANDLE_FDK_BITBUF hBitBuffer, const UINT value);
-INT FDK_getBitCnt(HANDLE_FDK_BITBUF hBitBuffer);
-
 void FDK_Feed(HANDLE_FDK_BITBUF hBitBuffer, const UCHAR inputBuffer[],
               const UINT bufferSize, UINT *bytesValid);
 
diff --git a/libFDK/include/FDK_bitstream.h b/libFDK/include/FDK_bitstream.h
index 49eeeaf..f799026 100644
--- a/libFDK/include/FDK_bitstream.h
+++ b/libFDK/include/FDK_bitstream.h
@@ -481,21 +481,6 @@ FDK_INLINE void FDKsyncCacheBwd(HANDLE_FDK_BITSTREAM hBitStream) {
 }
 
 /**
- * \brief Byte Alignment Function.
- *        This function performs the byte_alignment() syntactic function on the
- * input stream, i.e. some bits will be discarded/padded so that the next bits
- * to be read/written will be aligned on a byte boundary with respect to
- * the bit position 0.
- *
- * \param  hBitStream HANDLE_FDK_BITSTREAM handle
- * \return void
- */
-FDK_INLINE void FDKbyteAlign(HANDLE_FDK_BITSTREAM hBitStream) {
-  FDKsyncCache(hBitStream);
-  FDK_byteAlign(&hBitStream->hBitBuf, (UCHAR)hBitStream->ConfigCache);
-}
-
-/**
  * \brief Byte Alignment Function with anchor
  *        This function performs the byte_alignment() syntactic function on the
  * input stream, i.e. some bits will be discarded so that the next bits to be
@@ -604,37 +589,6 @@ FDK_INLINE INT FDKgetFreeBits(HANDLE_FDK_BITSTREAM hBitStream) {
 }
 
 /**
- * \brief reset bitcounter in bitBuffer to zero.
- * \param hBitStream HANDLE_FDK_BITSTREAM handle
- * \return void
- */
-FDK_INLINE void FDKresetBitCnt(HANDLE_FDK_BITSTREAM hBitStream) {
-  FDKsyncCache(hBitStream);
-  FDK_setBitCnt(&hBitStream->hBitBuf, 0);
-}
-
-/**
- * \brief set bitcoutner in bitBuffer to given value.
- * \param hBitStream HANDLE_FDK_BITSTREAM handle
- * \param value new value to be assigned to the bit counter
- * \return void
- */
-FDK_INLINE void FDKsetBitCnt(HANDLE_FDK_BITSTREAM hBitStream, UINT value) {
-  FDKsyncCache(hBitStream);
-  FDK_setBitCnt(&hBitStream->hBitBuf, value);
-}
-
-/**
- * \brief get bitcounter state from bitBuffer.
- * \param hBitStream HANDLE_FDK_BITSTREAM handle
- * \return current bit counter value
- */
-FDK_INLINE INT FDKgetBitCnt(HANDLE_FDK_BITSTREAM hBitStream) {
-  FDKsyncCache(hBitStream);
-  return FDK_getBitCnt(&hBitStream->hBitBuf);
-}
-
-/**
  * \brief Fill the BitBuffer with a number of input bytes from  external source.
  *        The bytesValid variable returns the number of ramaining valid bytes in
  * extern inputBuffer.
diff --git a/libFDK/include/scale.h b/libFDK/include/scale.h
index 07bd3af..30fa089 100644
--- a/libFDK/include/scale.h
+++ b/libFDK/include/scale.h
@@ -268,11 +268,11 @@ inline void scaleValueInPlace(FIXP_DBL *value, /*!< Value */
  * to avoid problems when inverting the sign of the result.
  */
 #ifndef SATURATE_LEFT_SHIFT_ALT
-#define SATURATE_LEFT_SHIFT_ALT(src, scale, dBits)                       \
-  (((LONG)(src) > ((LONG)(((1U) << ((dBits)-1)) - 1) >> (scale)))        \
-       ? (LONG)(((1U) << ((dBits)-1)) - 1)                               \
-       : ((LONG)(src) < ~((LONG)(((1U) << ((dBits)-1)) - 2) >> (scale))) \
-             ? ~((LONG)(((1U) << ((dBits)-1)) - 2))                      \
+#define SATURATE_LEFT_SHIFT_ALT(src, scale, dBits)                        \
+  (((LONG)(src) > ((LONG)(((1U) << ((dBits)-1)) - 1) >> (scale)))         \
+       ? (LONG)(((1U) << ((dBits)-1)) - 1)                                \
+       : ((LONG)(src) <= ~((LONG)(((1U) << ((dBits)-1)) - 1) >> (scale))) \
+             ? ~((LONG)(((1U) << ((dBits)-1)) - 2))                       \
              : ((LONG)(src) << (scale)))
 #endif
 
diff --git a/libFDK/src/FDK_bitbuffer.cpp b/libFDK/src/FDK_bitbuffer.cpp
index a990c58..98905ea 100644
--- a/libFDK/src/FDK_bitbuffer.cpp
+++ b/libFDK/src/FDK_bitbuffer.cpp
@@ -128,7 +128,6 @@ void FDK_InitBitBuffer(HANDLE_FDK_BITBUF hBitBuf, UCHAR *pBuffer, UINT bufSize,
   hBitBuf->ValidBits = validBits;
   hBitBuf->ReadOffset = 0;
   hBitBuf->WriteOffset = 0;
-  hBitBuf->BitCnt = 0;
   hBitBuf->BitNdx = 0;
 
   hBitBuf->Buffer = pBuffer;
@@ -151,7 +150,6 @@ void FDK_ResetBitBuffer(HANDLE_FDK_BITBUF hBitBuf) {
   hBitBuf->ValidBits = 0;
   hBitBuf->ReadOffset = 0;
   hBitBuf->WriteOffset = 0;
-  hBitBuf->BitCnt = 0;
   hBitBuf->BitNdx = 0;
 }
 
@@ -161,7 +159,6 @@ INT FDK_get(HANDLE_FDK_BITBUF hBitBuf, const UINT numberOfBits) {
   UINT bitOffset = hBitBuf->BitNdx & 0x07;
 
   hBitBuf->BitNdx = (hBitBuf->BitNdx + numberOfBits) & (hBitBuf->bufBits - 1);
-  hBitBuf->BitCnt += numberOfBits;
   hBitBuf->ValidBits -= numberOfBits;
 
   UINT byteMask = hBitBuf->bufSize - 1;
@@ -184,7 +181,6 @@ INT FDK_get(HANDLE_FDK_BITBUF hBitBuf, const UINT numberOfBits) {
 INT FDK_get32(HANDLE_FDK_BITBUF hBitBuf) {
   UINT BitNdx = hBitBuf->BitNdx + 32;
   hBitBuf->BitNdx = BitNdx & (hBitBuf->bufBits - 1);
-  hBitBuf->BitCnt += 32;
   hBitBuf->ValidBits = (UINT)((INT)hBitBuf->ValidBits - (INT)32);
 
   UINT byteOffset = (BitNdx - 1) >> 3;
@@ -223,7 +219,6 @@ INT FDK_getBwd(HANDLE_FDK_BITBUF hBitBuf, const UINT numberOfBits) {
   int i;
 
   hBitBuf->BitNdx = (hBitBuf->BitNdx - numberOfBits) & (hBitBuf->bufBits - 1);
-  hBitBuf->BitCnt -= numberOfBits;
   hBitBuf->ValidBits += numberOfBits;
 
   UINT tx = hBitBuf->Buffer[(byteOffset - 3) & byteMask] << 24 |
@@ -256,7 +251,6 @@ void FDK_put(HANDLE_FDK_BITBUF hBitBuf, UINT value, const UINT numberOfBits) {
     UINT bitOffset = hBitBuf->BitNdx & 0x7;
 
     hBitBuf->BitNdx = (hBitBuf->BitNdx + numberOfBits) & (hBitBuf->bufBits - 1);
-    hBitBuf->BitCnt += numberOfBits;
     hBitBuf->ValidBits += numberOfBits;
 
     UINT byteMask = hBitBuf->bufSize - 1;
@@ -307,7 +301,6 @@ void FDK_putBwd(HANDLE_FDK_BITBUF hBitBuf, UINT value,
   int i;
 
   hBitBuf->BitNdx = (hBitBuf->BitNdx - numberOfBits) & (hBitBuf->bufBits - 1);
-  hBitBuf->BitCnt -= numberOfBits;
   hBitBuf->ValidBits -= numberOfBits;
 
   /* in place turn around */
@@ -344,7 +337,6 @@ void FDK_putBwd(HANDLE_FDK_BITBUF hBitBuf, UINT value,
 #ifndef FUNCTION_FDK_pushBack
 void FDK_pushBack(HANDLE_FDK_BITBUF hBitBuf, const UINT numberOfBits,
                   UCHAR config) {
-  hBitBuf->BitCnt = (UINT)((INT)hBitBuf->BitCnt - (INT)numberOfBits);
   hBitBuf->ValidBits =
       (config == 0) ? (UINT)((INT)hBitBuf->ValidBits + (INT)numberOfBits)
                     : ((UINT)((INT)hBitBuf->ValidBits - (INT)numberOfBits));
@@ -355,7 +347,6 @@ void FDK_pushBack(HANDLE_FDK_BITBUF hBitBuf, const UINT numberOfBits,
 
 void FDK_pushForward(HANDLE_FDK_BITBUF hBitBuf, const UINT numberOfBits,
                      UCHAR config) {
-  hBitBuf->BitCnt = (UINT)((INT)hBitBuf->BitCnt + (INT)numberOfBits);
   hBitBuf->ValidBits =
       (config == 0) ? ((UINT)((INT)hBitBuf->ValidBits - (INT)numberOfBits))
                     : (UINT)((INT)hBitBuf->ValidBits + (INT)numberOfBits);
@@ -363,19 +354,6 @@ void FDK_pushForward(HANDLE_FDK_BITBUF hBitBuf, const UINT numberOfBits,
       (UINT)((INT)hBitBuf->BitNdx + (INT)numberOfBits) & (hBitBuf->bufBits - 1);
 }
 
-void FDK_byteAlign(HANDLE_FDK_BITBUF hBitBuf, UCHAR config) {
-  INT alignment = hBitBuf->BitCnt & 0x07;
-
-  if (alignment) {
-    if (config == 0)
-      FDK_pushForward(hBitBuf, 8 - alignment, config); /* BS_READER */
-    else
-      FDK_put(hBitBuf, 0, 8 - alignment); /* BS_WRITER */
-  }
-
-  hBitBuf->BitCnt = 0;
-}
-
 #ifndef FUNCTION_FDK_getValidBits
 UINT FDK_getValidBits(HANDLE_FDK_BITBUF hBitBuf) { return hBitBuf->ValidBits; }
 #endif /* #ifndef FUNCTION_FDK_getValidBits */
@@ -384,12 +362,6 @@ INT FDK_getFreeBits(HANDLE_FDK_BITBUF hBitBuf) {
   return (hBitBuf->bufBits - hBitBuf->ValidBits);
 }
 
-void FDK_setBitCnt(HANDLE_FDK_BITBUF hBitBuf, const UINT value) {
-  hBitBuf->BitCnt = value;
-}
-
-INT FDK_getBitCnt(HANDLE_FDK_BITBUF hBitBuf) { return hBitBuf->BitCnt; }
-
 void FDK_Feed(HANDLE_FDK_BITBUF hBitBuf, const UCHAR *RESTRICT inputBuffer,
               const UINT bufferSize, UINT *bytesValid) {
   inputBuffer = &inputBuffer[bufferSize - *bytesValid];
@@ -438,7 +410,6 @@ void CopyAlignedBlock(HANDLE_FDK_BITBUF h_BitBufSrc, UCHAR *RESTRICT dstBuffer,
 
   h_BitBufSrc->BitNdx =
       (h_BitBufSrc->BitNdx + bToRead) & (h_BitBufSrc->bufBits - 1);
-  h_BitBufSrc->BitCnt += bToRead;
   h_BitBufSrc->ValidBits -= bToRead;
 }
 
diff --git a/libFDK/src/FDK_decorrelate.cpp b/libFDK/src/FDK_decorrelate.cpp
index 8e665b7..c5de79a 100644
--- a/libFDK/src/FDK_decorrelate.cpp
+++ b/libFDK/src/FDK_decorrelate.cpp
@@ -1086,8 +1086,9 @@ static INT DuckerCalcEnergy(DUCKER_INSTANCE *const self,
     pb = SpatialDecGetProcessingBand(maxHybBand, self->mapHybBands2ProcBands);
     for (qs = startHybBand; qs <= maxHybBand; qs++) {
       pb = SpatialDecGetProcessingBand(qs, self->mapHybBands2ProcBands);
-      energy[pb] +=
-          (fPow2Div2(inputReal[qs] << clz) + fPow2Div2(inputImag[qs] << clz));
+      energy[pb] =
+          fAddSaturate(energy[pb], fPow2Div2(inputReal[qs] << clz) +
+                                       fPow2Div2(inputImag[qs] << clz));
     }
     pb++;
 
@@ -1100,7 +1101,7 @@ static INT DuckerCalcEnergy(DUCKER_INSTANCE *const self,
       FIXP_DBL nrg = 0;
       qs_next = (int)self->qs_next[pb];
       for (; qs < qs_next; qs++) {
-        nrg += fPow2Div2(inputReal[qs] << clz);
+        nrg = fAddSaturate(nrg, fPow2Div2(inputReal[qs] << clz));
       }
       energy[pb] = nrg;
     }
@@ -1138,13 +1139,14 @@ static INT DuckerCalcEnergy(DUCKER_INSTANCE *const self,
 #else
     for (qs = startHybBand; qs <= maxHybBand; qs++) {
       int pb = SpatialDecGetProcessingBand(qs, self->mapHybBands2ProcBands);
-      energy[pb] +=
-          (fPow2Div2(inputReal[qs] << clz) + fPow2Div2(inputImag[qs] << clz));
+      energy[pb] =
+          fAddSaturate(energy[pb], fPow2Div2(inputReal[qs] << clz) +
+                                       fPow2Div2(inputImag[qs] << clz));
     }
 
     for (; qs <= maxHybridBand; qs++) {
       int pb = SpatialDecGetProcessingBand(qs, self->mapHybBands2ProcBands);
-      energy[pb] += fPow2Div2(inputReal[qs] << clz);
+      energy[pb] = fAddSaturate(energy[pb], fPow2Div2(inputReal[qs] << clz));
     }
 #endif /* FUNCTION_DuckerCalcEnergy_func4 */
   }
@@ -1237,9 +1239,9 @@ static INT DuckerApply(DUCKER_INSTANCE *const self,
 
       tmp1 = scaleValue(tmp1, scaleSmoothDirRevNrg_asExponent);
       tmp2 = scaleValue(tmp2, scaleSmoothDirRevNrg_asExponent);
-      tmp1 = fMultAddDiv2(tmp1, directNrg[pb] >> scaleDirectNrg,
+      tmp1 = fMultAddDiv2(tmp1, scaleValue(directNrg[pb], -scaleDirectNrg),
                           DUCK_ONE_MINUS_ALPHA_X4_FDK);
-      tmp2 = fMultAddDiv2(tmp2, reverbNrg[pb] >> scaleReverbNrg,
+      tmp2 = fMultAddDiv2(tmp2, scaleValue(reverbNrg[pb], -scaleReverbNrg),
                           DUCK_ONE_MINUS_ALPHA_X4_FDK);
     } else {
       tmp1 = fMultAddDiv2(tmp1, directNrg[pb], DUCK_ONE_MINUS_ALPHA_X4_FDK);
diff --git a/libFDK/src/FDK_qmf_domain.cpp b/libFDK/src/FDK_qmf_domain.cpp
index 043a372..3245deb 100644
--- a/libFDK/src/FDK_qmf_domain.cpp
+++ b/libFDK/src/FDK_qmf_domain.cpp
@@ -786,6 +786,7 @@ void FDK_QmfDomain_ClearRequested(HANDLE_FDK_QMF_DOMAIN_GC hgc) {
   hgc->flags_requested = 0;
   hgc->nInputChannels_requested = 0;
   hgc->nOutputChannels_requested = 0;
+  hgc->parkChannel_requested = 0;
   hgc->nBandsAnalysis_requested = 0;
   hgc->nBandsSynthesis_requested = 0;
   hgc->nQmfTimeSlots_requested = 0;
@@ -798,6 +799,7 @@ static void FDK_QmfDomain_ClearConfigured(HANDLE_FDK_QMF_DOMAIN_GC hgc) {
   hgc->flags = 0;
   hgc->nInputChannels = 0;
   hgc->nOutputChannels = 0;
+  hgc->parkChannel = 0;
   hgc->nBandsAnalysis = 0;
   hgc->nBandsSynthesis = 0;
   hgc->nQmfTimeSlots = 0;
@@ -887,10 +889,6 @@ QMF_DOMAIN_ERROR FDK_QmfDomain_Configure(HANDLE_FDK_QMF_DOMAIN hqd) {
         !(hgc->flags & (QMF_FLAG_CLDFB | QMF_FLAG_MPSLDFB))) {
       hgc->flags_requested |= QMF_FLAG_DOWNSAMPLED;
     }
-    if ((hgc->flags_requested & QMF_FLAG_MPSLDFB) &&
-        (hgc->flags_requested & QMF_FLAG_CLDFB)) {
-      hgc->flags_requested &= ~QMF_FLAG_CLDFB;
-    }
 
     hasChanged = 1;
   }
@@ -899,6 +897,10 @@ QMF_DOMAIN_ERROR FDK_QmfDomain_Configure(HANDLE_FDK_QMF_DOMAIN hqd) {
 
   /* 5. set requested flags */
   if (hgc->flags != hgc->flags_requested) {
+    if ((hgc->flags_requested & QMF_FLAG_MPSLDFB) &&
+        (hgc->flags_requested & QMF_FLAG_CLDFB)) {
+      hgc->flags_requested &= ~QMF_FLAG_CLDFB;
+    }
     hgc->flags = hgc->flags_requested;
     hasChanged = 1;
   }
@@ -981,9 +983,8 @@ QMF_DOMAIN_ERROR FDK_QmfDomain_Configure(HANDLE_FDK_QMF_DOMAIN hqd) {
   }
 
 bail:
-  if (err == QMF_DOMAIN_OUT_OF_MEMORY) {
-    FDK_QmfDomain_FreePersistentMemory(hqd);
-    FDK_QmfDomain_ClearConfigured(&hqd->globalConf);
+  if (err) {
+    FDK_QmfDomain_FreeMem(hqd);
   }
   return err;
 }
diff --git a/libFDK/src/arm/dct_arm.cpp b/libFDK/src/arm/dct_arm.cpp
deleted file mode 100644
index dd66109..0000000
--- a/libFDK/src/arm/dct_arm.cpp
+++ /dev/null
@@ -1,572 +0,0 @@
-/* -----------------------------------------------------------------------------
-Software License for The Fraunhofer FDK AAC Codec Library for Android
-
-© Copyright  1995 - 2018 Fraunhofer-Gesellschaft zur Förderung der angewandten
-Forschung e.V. All rights reserved.
-
- 1.    INTRODUCTION
-The Fraunhofer FDK AAC Codec Library for Android ("FDK AAC Codec") is software
-that implements the MPEG Advanced Audio Coding ("AAC") encoding and decoding
-scheme for digital audio. This FDK AAC Codec software is intended to be used on
-a wide variety of Android devices.
-
-AAC's HE-AAC and HE-AAC v2 versions are regarded as today's most efficient
-general perceptual audio codecs. AAC-ELD is considered the best-performing
-full-bandwidth communications codec by independent studies and is widely
-deployed. AAC has been standardized by ISO and IEC as part of the MPEG
-specifications.
-
-Patent licenses for necessary patent claims for the FDK AAC Codec (including
-those of Fraunhofer) may be obtained through Via Licensing
-(www.vialicensing.com) or through the respective patent owners individually for
-the purpose of encoding or decoding bit streams in products that are compliant
-with the ISO/IEC MPEG audio standards. Please note that most manufacturers of
-Android devices already license these patent claims through Via Licensing or
-directly from the patent owners, and therefore FDK AAC Codec software may
-already be covered under those patent licenses when it is used for those
-licensed purposes only.
-
-Commercially-licensed AAC software libraries, including floating-point versions
-with enhanced sound quality, are also available from Fraunhofer. Users are
-encouraged to check the Fraunhofer website for additional applications
-information and documentation.
-
-2.    COPYRIGHT LICENSE
-
-Redistribution and use in source and binary forms, with or without modification,
-are permitted without payment of copyright license fees provided that you
-satisfy the following conditions:
-
-You must retain the complete text of this software license in redistributions of
-the FDK AAC Codec or your modifications thereto in source code form.
-
-You must retain the complete text of this software license in the documentation
-and/or other materials provided with redistributions of the FDK AAC Codec or
-your modifications thereto in binary form. You must make available free of
-charge copies of the complete source code of the FDK AAC Codec and your
-modifications thereto to recipients of copies in binary form.
-
-The name of Fraunhofer may not be used to endorse or promote products derived
-from this library without prior written permission.
-
-You may not charge copyright license fees for anyone to use, copy or distribute
-the FDK AAC Codec software or your modifications thereto.
-
-Your modified versions of the FDK AAC Codec must carry prominent notices stating
-that you changed the software and the date of any change. For modified versions
-of the FDK AAC Codec, the term "Fraunhofer FDK AAC Codec Library for Android"
-must be replaced by the term "Third-Party Modified Version of the Fraunhofer FDK
-AAC Codec Library for Android."
-
-3.    NO PATENT LICENSE
-
-NO EXPRESS OR IMPLIED LICENSES TO ANY PATENT CLAIMS, including without
-limitation the patents of Fraunhofer, ARE GRANTED BY THIS SOFTWARE LICENSE.
-Fraunhofer provides no warranty of patent non-infringement with respect to this
-software.
-
-You may use this FDK AAC Codec software or modifications thereto only for
-purposes that are authorized by appropriate patent licenses.
-
-4.    DISCLAIMER
-
-This FDK AAC Codec software is provided by Fraunhofer on behalf of the copyright
-holders and contributors "AS IS" and WITHOUT ANY EXPRESS OR IMPLIED WARRANTIES,
-including but not limited to the implied warranties of merchantability and
-fitness for a particular purpose. IN NO EVENT SHALL THE COPYRIGHT HOLDER 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), arising in any way out of the use of
-this software, even if advised of the possibility of such damage.
-
-5.    CONTACT INFORMATION
-
-Fraunhofer Institute for Integrated Circuits IIS
-Attention: Audio and Multimedia Departments - FDK AAC LL
-Am Wolfsmantel 33
-91058 Erlangen, Germany
-
-www.iis.fraunhofer.de/amm
-amm-info@iis.fraunhofer.de
------------------------------------------------------------------------------ */
-
-/******************* Library for basic calculation routines ********************
-
-   Author(s):
-
-   Description:
-
-*******************************************************************************/
-
-#ifdef FUNCTION_dct_IV_func1
-
-/*
-   Note: This assembler routine is here, because the ARM926 compiler does
-         not encode the inline assembler with optimal speed.
-         With this version, we save 2 cycles per loop iteration.
-*/
-
-__asm void dct_IV_func1(int i, const FIXP_SPK *twiddle,
-                        FIXP_DBL *RESTRICT pDat_0, FIXP_DBL *RESTRICT pDat_1) {
-  /* Register map:
-     r0   i
-     r1   twiddle
-     r2   pDat_0
-     r3   pDat_1
-     r4   accu1
-     r5   accu2
-     r6   accu3
-     r7   accu4
-     r8   val_tw
-     r9   accuX
-  */
-  PUSH{r4 - r9}
-
-  /* 44 cycles for 2 iterations = 22 cycles/iteration */
-  dct_IV_loop1_start
-      /*  First iteration */
-      LDR r8,
-      [r1],
-# 4 // val_tw = *twiddle++;
-      LDR r5,
-      [ r2, #0 ]  // accu2 = pDat_0[0]
-      LDR r4,
-      [ r3, #0 ]  // accu1 = pDat_1[0]
-
-      SMULWT r9,
-      r5,
-      r8  // accuX = accu2*val_tw.l
-          SMULWB r5,
-      r5,
-      r8  // accu2 = accu2*val_tw.h
-          RSB r9,
-      r9,
-# 0 // accuX =-accu2*val_tw.l
-      SMLAWT r5, r4, r8,
-      r5  // accu2 = accu2*val_tw.h + accu1*val_tw.l
-          SMLAWB r4,
-      r4, r8,
-      r9  // accu1 = accu1*val_tw.h - accu2*val_tw.l
-
-          LDR r8,
-      [r1],
-# 4 // val_tw = *twiddle++;
-      LDR r7,
-      [ r3, # - 4 ]  // accu4 = pDat_1[-1]
-      LDR r6,
-      [ r2, #4 ]  // accu3 = pDat_0[1]
-
-      SMULWB r9,
-      r7,
-      r8  // accuX = accu4*val_tw.h
-          SMULWT r7,
-      r7,
-      r8  // accu4 = accu4*val_tw.l
-          RSB r9,
-      r9,
-# 0 // accuX =-accu4*val_tw.h
-      SMLAWB r7, r6, r8,
-      r7  // accu4 = accu4*val_tw.l+accu3*val_tw.h
-          SMLAWT r6,
-      r6, r8,
-      r9  // accu3 = accu3*val_tw.l-accu4*val_tw.h
-
-          STR r5,
-      [r2],
-# 4 // *pDat_0++ = accu2
-      STR r4, [r2],
-# 4 // *pDat_0++ = accu1
-      STR r6, [r3],
-#- 4 // *pDat_1-- = accu3
-      STR r7, [r3],
-#- 4 // *pDat_1-- = accu4
-
-      /*  Second iteration */
-      LDR r8, [r1],
-# 4 // val_tw = *twiddle++;
-      LDR r5,
-      [ r2, #0 ]  // accu2 = pDat_0[0]
-      LDR r4,
-      [ r3, #0 ]  // accu1 = pDat_1[0]
-
-      SMULWT r9,
-      r5,
-      r8  // accuX = accu2*val_tw.l
-          SMULWB r5,
-      r5,
-      r8  // accu2 = accu2*val_tw.h
-          RSB r9,
-      r9,
-# 0 // accuX =-accu2*val_tw.l
-      SMLAWT r5, r4, r8,
-      r5  // accu2 = accu2*val_tw.h + accu1*val_tw.l
-          SMLAWB r4,
-      r4, r8,
-      r9  // accu1 = accu1*val_tw.h - accu2*val_tw.l
-
-          LDR r8,
-      [r1],
-# 4 // val_tw = *twiddle++;
-      LDR r7,
-      [ r3, # - 4 ]  // accu4 = pDat_1[-1]
-      LDR r6,
-      [ r2, #4 ]  // accu3 = pDat_0[1]
-
-      SMULWB r9,
-      r7,
-      r8  // accuX = accu4*val_tw.h
-          SMULWT r7,
-      r7,
-      r8  // accu4 = accu4*val_tw.l
-          RSB r9,
-      r9,
-# 0 // accuX =-accu4*val_tw.h
-      SMLAWB r7, r6, r8,
-      r7  // accu4 = accu4*val_tw.l+accu3*val_tw.h
-          SMLAWT r6,
-      r6, r8,
-      r9  // accu3 = accu3*val_tw.l-accu4*val_tw.h
-
-          STR r5,
-      [r2],
-# 4 // *pDat_0++ = accu2
-      STR r4, [r2],
-# 4 // *pDat_0++ = accu1
-      STR r6, [r3],
-#- 4 // *pDat_1-- = accu3
-      STR r7, [r3],
-#- 4 // *pDat_1-- = accu4
-
-      SUBS r0, r0,
-# 1 BNE dct_IV_loop1_start
-
-      POP { r4 - r9 }
-
-  BX lr
-}
-
-#endif /* FUNCTION_dct_IV_func1 */
-
-#ifdef FUNCTION_dct_IV_func2
-
-/* __attribute__((noinline)) */
-static inline void dct_IV_func2(int i, const FIXP_SPK *twiddle,
-                                FIXP_DBL *pDat_0, FIXP_DBL *pDat_1, int inc) {
-  FIXP_DBL accu1, accu2, accu3, accu4, accuX;
-  LONG val_tw;
-
-  accu1 = pDat_1[-2];
-  accu2 = pDat_1[-1];
-
-  *--pDat_1 = -(pDat_0[1] >> 1);
-  *pDat_0++ = (pDat_0[0] >> 1);
-
-  twiddle += inc;
-
-  __asm {
-    LDR     val_tw, [twiddle], inc, LSL #2  // val_tw = *twiddle; twiddle += inc
-    B       dct_IV_loop2_2nd_part
-
-        /* 42 cycles for 2 iterations = 21 cycles/iteration */
-dct_IV_loop2:
-    SMULWT  accuX, accu2, val_tw
-    SMULWB  accu2, accu2, val_tw
-    RSB     accuX, accuX, #0
-    SMLAWB  accuX, accu1, val_tw, accuX
-    SMLAWT  accu2, accu1, val_tw, accu2
-    STR     accuX, [pDat_0], #4
-    STR     accu2, [pDat_1, #-4] !
-
-    LDR     accu4, [pDat_0, #4]
-    LDR     accu3, [pDat_0]
-    SMULWB  accuX, accu4, val_tw
-    SMULWT  accu4, accu4, val_tw
-    RSB     accuX, accuX, #0
-    SMLAWT  accuX, accu3, val_tw, accuX
-    SMLAWB  accu4, accu3, val_tw, accu4
-
-    LDR     accu1, [pDat_1, #-8]
-    LDR     accu2, [pDat_1, #-4]
-
-    LDR     val_tw, [twiddle], inc, LSL #2  // val_tw = *twiddle; twiddle += inc
-
-    STR     accuX, [pDat_1, #-4] !
-    STR     accu4, [pDat_0], #4
-
-dct_IV_loop2_2nd_part:
-    SMULWT  accuX, accu2, val_tw
-    SMULWB  accu2, accu2, val_tw
-    RSB     accuX, accuX, #0
-    SMLAWB  accuX, accu1, val_tw, accuX
-    SMLAWT  accu2, accu1, val_tw, accu2
-    STR     accuX, [pDat_0], #4
-    STR     accu2, [pDat_1, #-4] !
-
-    LDR     accu4, [pDat_0, #4]
-    LDR     accu3, [pDat_0]
-    SMULWB  accuX, accu4, val_tw
-    SMULWT  accu4, accu4, val_tw
-    RSB     accuX, accuX, #0
-    SMLAWT  accuX, accu3, val_tw, accuX
-    SMLAWB  accu4, accu3, val_tw, accu4
-
-    LDR     accu1, [pDat_1, #-8]
-    LDR     accu2, [pDat_1, #-4]
-
-    STR     accuX, [pDat_1, #-4] !
-    STR     accu4, [pDat_0], #4
-
-    LDR     val_tw, [twiddle], inc, LSL #2  // val_tw = *twiddle; twiddle += inc
-
-    SUBS    i, i, #1
-    BNE     dct_IV_loop2
-  }
-
-  /* Last Sin and Cos value pair are the same */
-  accu1 = fMultDiv2(accu1, WTC(0x5a82799a));
-  accu2 = fMultDiv2(accu2, WTC(0x5a82799a));
-
-  *--pDat_1 = accu1 + accu2;
-  *pDat_0++ = accu1 - accu2;
-}
-#endif /* FUNCTION_dct_IV_func2 */
-
-#ifdef FUNCTION_dst_IV_func1
-
-__asm void dst_IV_func1(int i, const FIXP_SPK *twiddle, FIXP_DBL *pDat_0,
-                        FIXP_DBL *pDat_1) {
-  /* Register map:
-     r0   i
-     r1   twiddle
-     r2   pDat_0
-     r3   pDat_1
-     r4   accu1
-     r5   accu2
-     r6   accu3
-     r7   accu4
-     r8   val_tw
-     r9   accuX
-  */
-  PUSH{r4 - r9}
-
-  dst_IV_loop1 LDR r8,
-      [r1],
-# 4 // val_tw = *twiddle++
-      LDR r5,
-      [r2]  // accu2 = pDat_0[0]
-      LDR r6,
-      [ r2, #4 ]  // accu3 = pDat_0[1]
-      RSB r5,
-      r5,
-# 0 // accu2 = -accu2
-      SMULWT r9, r5,
-      r8  // accuX = (-accu2)*val_tw.l
-          LDR r4,
-      [ r3, # - 4 ]  // accu1 = pDat_1[-1]
-      RSB r9,
-      r9,
-# 0 // accuX = -(-accu2)*val_tw.l
-      SMLAWB r9, r4, r8,
-      r9  // accuX = accu1*val_tw.h-(-accu2)*val_tw.l
-          SMULWT r4,
-      r4,
-      r8  // accu1 = accu1*val_tw.l
-          LDR r7,
-      [ r3, # - 8 ]  // accu4 = pDat_1[-2]
-      SMLAWB r5,
-      r5, r8,
-      r4  // accu2 = (-accu2)*val_tw.t+accu1*val_tw.l
-          LDR r8,
-      [r1],
-# 4 // val_tw = *twiddle++
-      STR r5, [r2],
-# 4 // *pDat_0++ = accu2
-      STR r9, [r2],
-# 4 // *pDat_0++ = accu1 (accuX)
-      RSB r7, r7,
-# 0 // accu4 = -accu4
-      SMULWB r5, r7,
-      r8  // accu2 = (-accu4)*val_tw.h
-          SMULWB r4,
-      r6,
-      r8  // accu1 = (-accu4)*val_tw.l
-          RSB r5,
-      r5,
-# 0 // accu2 = -(-accu4)*val_tw.h
-      SMLAWT r6, r6, r8,
-      r5  // accu3 = (-accu4)*val_tw.l-(-accu3)*val_tw.h
-          SMLAWT r7,
-      r7, r8,
-      r4  // accu4 = (-accu3)*val_tw.l+(-accu4)*val_tw.h
-          STR r6,
-      [ r3, # - 4 ] !  // *--pDat_1 = accu3
-      STR r7,
-      [ r3, # - 4 ] !  // *--pDat_1 = accu4
-
-      LDR r8,
-      [r1],
-# 4 // val_tw = *twiddle++
-      LDR r5,
-      [r2]  // accu2 = pDat_0[0]
-      LDR r6,
-      [ r2, #4 ]  // accu3 = pDat_0[1]
-      RSB r5,
-      r5,
-# 0 // accu2 = -accu2
-      SMULWT r9, r5,
-      r8  // accuX = (-accu2)*val_tw.l
-          LDR r4,
-      [ r3, # - 4 ]  // accu1 = pDat_1[-1]
-      RSB r9,
-      r9,
-# 0 // accuX = -(-accu2)*val_tw.l
-      SMLAWB r9, r4, r8,
-      r9  // accuX = accu1*val_tw.h-(-accu2)*val_tw.l
-          SMULWT r4,
-      r4,
-      r8  // accu1 = accu1*val_tw.l
-          LDR r7,
-      [ r3, # - 8 ]  // accu4 = pDat_1[-2]
-      SMLAWB r5,
-      r5, r8,
-      r4  // accu2 = (-accu2)*val_tw.t+accu1*val_tw.l
-          LDR r8,
-      [r1],
-# 4 // val_tw = *twiddle++
-      STR r5, [r2],
-# 4 // *pDat_0++ = accu2
-      STR r9, [r2],
-# 4 // *pDat_0++ = accu1 (accuX)
-      RSB r7, r7,
-# 0 // accu4 = -accu4
-      SMULWB r5, r7,
-      r8  // accu2 = (-accu4)*val_tw.h
-          SMULWB r4,
-      r6,
-      r8  // accu1 = (-accu4)*val_tw.l
-          RSB r5,
-      r5,
-# 0 // accu2 = -(-accu4)*val_tw.h
-      SMLAWT r6, r6, r8,
-      r5  // accu3 = (-accu4)*val_tw.l-(-accu3)*val_tw.h
-          SMLAWT r7,
-      r7, r8,
-      r4  // accu4 = (-accu3)*val_tw.l+(-accu4)*val_tw.h
-          STR r6,
-      [ r3, # - 4 ] !  // *--pDat_1 = accu3
-      STR r7,
-      [ r3, # - 4 ] !  // *--pDat_1 = accu4
-
-      SUBS r0,
-      r0,
-# 4 // i-= 4
-      BNE dst_IV_loop1
-
-          POP{r4 - r9} BX lr
-}
-#endif /* FUNCTION_dst_IV_func1 */
-
-#ifdef FUNCTION_dst_IV_func2
-
-/* __attribute__((noinline)) */
-static inline void dst_IV_func2(int i, const FIXP_SPK *twiddle,
-                                FIXP_DBL *RESTRICT pDat_0,
-                                FIXP_DBL *RESTRICT pDat_1, int inc) {
-  FIXP_DBL accu1, accu2, accu3, accu4;
-  LONG val_tw;
-
-  accu4 = pDat_0[0];
-  accu3 = pDat_0[1];
-  accu4 >>= 1;
-  accu3 >>= 1;
-  accu4 = -accu4;
-
-  accu1 = pDat_1[-1];
-  accu2 = pDat_1[0];
-
-  *pDat_0++ = accu3;
-  *pDat_1-- = accu4;
-
-  __asm {
-    B       dst_IV_loop2_2nd_part
-
-        /* 50 cycles for 2 iterations = 25 cycles/iteration */
-
-dst_IV_loop2:
-
-    LDR     val_tw, [twiddle], inc, LSL #2  // val_tw = *twiddle; twiddle += inc
-
-    RSB     accu2, accu2, #0  // accu2 = -accu2
-    RSB     accu1, accu1, #0  // accu1 = -accu1
-    SMULWT  accu3, accu2, val_tw  // accu3 = (-accu2)*val_tw.l
-    SMULWT  accu4, accu1, val_tw  // accu4 = (-accu1)*val_tw.l
-    RSB     accu3, accu3, #0  // accu3 = -accu2*val_tw.l
-    SMLAWB  accu1, accu1, val_tw, accu3  // accu1 = -accu1*val_tw.h-(-accu2)*val_tw.l
-    SMLAWB  accu2, accu2, val_tw, accu4  // accu2 = (-accu1)*val_tw.l+(-accu2)*val_tw.h
-    STR     accu1, [pDat_1], #-4  // *pDat_1-- = accu1
-    STR     accu2, [pDat_0], #4  // *pDat_0++ = accu2
-
-    LDR     accu4, [pDat_0]  // accu4 = pDat_0[0]
-    LDR     accu3, [pDat_0, #4]  // accu3 = pDat_0[1]
-
-    RSB     accu4, accu4, #0  // accu4 = -accu4
-    RSB     accu3, accu3, #0  // accu3 = -accu3
-
-    SMULWB  accu1, accu3, val_tw  // accu1 = (-accu3)*val_tw.h
-    SMULWT  accu2, accu3, val_tw  // accu2 = (-accu3)*val_tw.l
-    RSB     accu1, accu1, #0  // accu1 = -(-accu3)*val_tw.h
-    SMLAWT  accu3, accu4, val_tw, accu1  // accu3 = (-accu4)*val_tw.l-(-accu3)*val_tw.h
-    SMLAWB  accu4, accu4, val_tw, accu2  // accu4 = (-accu3)*val_tw.l+(-accu4)*val_tw.h
-
-    LDR     accu1, [pDat_1, #-4]  // accu1 = pDat_1[-1]
-    LDR     accu2, [pDat_1]  // accu2 = pDat_1[0]
-
-    STR     accu3, [pDat_0], #4  // *pDat_0++ = accu3
-    STR     accu4, [pDat_1], #-4  // *pDat_1-- = accu4
-
-dst_IV_loop2_2nd_part:
-
-    LDR     val_tw, [twiddle], inc, LSL #2  // val_tw = *twiddle; twiddle += inc
-
-    RSB     accu2, accu2, #0  // accu2 = -accu2
-    RSB     accu1, accu1, #0  // accu1 = -accu1
-    SMULWT  accu3, accu2, val_tw  // accu3 = (-accu2)*val_tw.l
-    SMULWT  accu4, accu1, val_tw  // accu4 = (-accu1)*val_tw.l
-    RSB     accu3, accu3, #0  // accu3 = -accu2*val_tw.l
-    SMLAWB  accu1, accu1, val_tw, accu3  // accu1 = -accu1*val_tw.h-(-accu2)*val_tw.l
-    SMLAWB  accu2, accu2, val_tw, accu4  // accu2 = (-accu1)*val_tw.l+(-accu2)*val_tw.h
-    STR     accu1, [pDat_1], #-4  // *pDat_1-- = accu1
-    STR     accu2, [pDat_0], #4  // *pDat_0++ = accu2
-
-    LDR     accu4, [pDat_0]  // accu4 = pDat_0[0]
-    LDR     accu3, [pDat_0, #4]  // accu3 = pDat_0[1]
-
-    RSB     accu4, accu4, #0  // accu4 = -accu4
-    RSB     accu3, accu3, #0  // accu3 = -accu3
-
-    SMULWB  accu1, accu3, val_tw  // accu1 = (-accu3)*val_tw.h
-    SMULWT  accu2, accu3, val_tw  // accu2 = (-accu3)*val_tw.l
-    RSB     accu1, accu1, #0  // accu1 = -(-accu3)*val_tw.h
-    SMLAWT  accu3, accu4, val_tw, accu1  // accu3 = (-accu4)*val_tw.l-(-accu3)*val_tw.h
-    SMLAWB  accu4, accu4, val_tw, accu2  // accu4 = (-accu3)*val_tw.l+(-accu4)*val_tw.h
-
-    LDR     accu1, [pDat_1, #-4]  // accu1 = pDat_1[-1]
-    LDR     accu2, [pDat_1]  // accu2 = pDat_1[0]
-
-    STR     accu3, [pDat_0], #4  // *pDat_0++ = accu3
-    STR     accu4, [pDat_1], #-4  // *pDat_1-- = accu4
-
-    SUBS    i, i, #1
-    BNE     dst_IV_loop2
-  }
-
-  /* Last Sin and Cos value pair are the same */
-  accu1 = fMultDiv2(-accu1, WTC(0x5a82799a));
-  accu2 = fMultDiv2(-accu2, WTC(0x5a82799a));
-
-  *pDat_0 = accu1 + accu2;
-  *pDat_1 = accu1 - accu2;
-}
-#endif /* FUNCTION_dst_IV_func2 */
diff --git a/libFDK/src/dct.cpp b/libFDK/src/dct.cpp
index a451331..776493e 100644
--- a/libFDK/src/dct.cpp
+++ b/libFDK/src/dct.cpp
@@ -124,10 +124,6 @@ amm-info@iis.fraunhofer.de
 #include "FDK_tools_rom.h"
 #include "fft.h"
 
-#if defined(__arm__)
-#include "arm/dct_arm.cpp"
-#endif
-
 void dct_getTables(const FIXP_WTP **ptwiddle, const FIXP_STP **sin_twiddle,
                    int *sin_step, int length) {
   const FIXP_WTP *twiddle;
@@ -387,12 +383,6 @@ void dct_IV(FIXP_DBL *pDat, int L, int *pDat_e) {
 
   dct_getTables(&twiddle, &sin_twiddle, &sin_step, L);
 
-#ifdef FUNCTION_dct_IV_func1
-  if (M >= 4 && (M & 3) == 0) {
-    /* ARM926: 44 cycles for 2 iterations = 22 cycles/iteration */
-    dct_IV_func1(M >> 2, twiddle, &pDat[0], &pDat[L - 1]);
-  } else
-#endif /* FUNCTION_dct_IV_func1 */
   {
     FIXP_DBL *RESTRICT pDat_0 = &pDat[0];
     FIXP_DBL *RESTRICT pDat_1 = &pDat[L - 2];
@@ -410,10 +400,10 @@ void dct_IV(FIXP_DBL *pDat, int L, int *pDat_e) {
       cplxMultDiv2(&accu1, &accu2, accu1, accu2, twiddle[i]);
       cplxMultDiv2(&accu3, &accu4, accu4, accu3, twiddle[i + 1]);
 
-      pDat_0[0] = accu2;
-      pDat_0[1] = accu1;
-      pDat_1[0] = accu4;
-      pDat_1[1] = -accu3;
+      pDat_0[0] = accu2 >> 1;
+      pDat_0[1] = accu1 >> 1;
+      pDat_1[0] = accu4 >> 1;
+      pDat_1[1] = -(accu3 >> 1);
     }
     if (M & 1) {
       FIXP_DBL accu1, accu2;
@@ -423,19 +413,13 @@ void dct_IV(FIXP_DBL *pDat, int L, int *pDat_e) {
 
       cplxMultDiv2(&accu1, &accu2, accu1, accu2, twiddle[i]);
 
-      pDat_0[0] = accu2;
-      pDat_0[1] = accu1;
+      pDat_0[0] = accu2 >> 1;
+      pDat_0[1] = accu1 >> 1;
     }
   }
 
   fft(M, pDat, pDat_e);
 
-#ifdef FUNCTION_dct_IV_func2
-  if (M >= 4 && (M & 3) == 0) {
-    /* ARM926: 42 cycles for 2 iterations = 21 cycles/iteration */
-    dct_IV_func2(M >> 2, sin_twiddle, &pDat[0], &pDat[L], sin_step);
-  } else
-#endif /* FUNCTION_dct_IV_func2 */
   {
     FIXP_DBL *RESTRICT pDat_0 = &pDat[0];
     FIXP_DBL *RESTRICT pDat_1 = &pDat[L - 2];
@@ -446,20 +430,19 @@ void dct_IV(FIXP_DBL *pDat, int L, int *pDat_e) {
     accu1 = pDat_1[0];
     accu2 = pDat_1[1];
 
-    pDat_1[1] = -(pDat_0[1] >> 1);
-    pDat_0[0] = (pDat_0[0] >> 1);
+    pDat_1[1] = -pDat_0[1];
 
     /* 28 cycles for ARM926 */
     for (idx = sin_step, i = 1; i<(M + 1)>> 1; i++, idx += sin_step) {
       FIXP_STP twd = sin_twiddle[idx];
-      cplxMultDiv2(&accu3, &accu4, accu1, accu2, twd);
+      cplxMult(&accu3, &accu4, accu1, accu2, twd);
       pDat_0[1] = accu3;
       pDat_1[0] = accu4;
 
       pDat_0 += 2;
       pDat_1 -= 2;
 
-      cplxMultDiv2(&accu3, &accu4, pDat_0[1], pDat_0[0], twd);
+      cplxMult(&accu3, &accu4, pDat_0[1], pDat_0[0], twd);
 
       accu1 = pDat_1[0];
       accu2 = pDat_1[1];
@@ -470,8 +453,8 @@ void dct_IV(FIXP_DBL *pDat, int L, int *pDat_e) {
 
     if ((M & 1) == 0) {
       /* Last Sin and Cos value pair are the same */
-      accu1 = fMultDiv2(accu1, WTC(0x5a82799a));
-      accu2 = fMultDiv2(accu2, WTC(0x5a82799a));
+      accu1 = fMult(accu1, WTC(0x5a82799a));
+      accu2 = fMult(accu2, WTC(0x5a82799a));
 
       pDat_1[0] = accu1 + accu2;
       pDat_0[1] = accu1 - accu2;
@@ -497,11 +480,6 @@ void dst_IV(FIXP_DBL *pDat, int L, int *pDat_e) {
 
   dct_getTables(&twiddle, &sin_twiddle, &sin_step, L);
 
-#ifdef FUNCTION_dst_IV_func1
-  if ((M >= 4) && ((M & 3) == 0)) {
-    dst_IV_func1(M, twiddle, &pDat[0], &pDat[L]);
-  } else
-#endif
   {
     FIXP_DBL *RESTRICT pDat_0 = &pDat[0];
     FIXP_DBL *RESTRICT pDat_1 = &pDat[L - 2];
@@ -519,10 +497,10 @@ void dst_IV(FIXP_DBL *pDat, int L, int *pDat_e) {
       cplxMultDiv2(&accu1, &accu2, accu1, accu2, twiddle[i]);
       cplxMultDiv2(&accu3, &accu4, accu4, accu3, twiddle[i + 1]);
 
-      pDat_0[0] = accu2;
-      pDat_0[1] = accu1;
-      pDat_1[0] = accu4;
-      pDat_1[1] = -accu3;
+      pDat_0[0] = accu2 >> 1;
+      pDat_0[1] = accu1 >> 1;
+      pDat_1[0] = accu4 >> 1;
+      pDat_1[1] = -(accu3 >> 1);
     }
     if (M & 1) {
       FIXP_DBL accu1, accu2;
@@ -532,19 +510,13 @@ void dst_IV(FIXP_DBL *pDat, int L, int *pDat_e) {
 
       cplxMultDiv2(&accu1, &accu2, accu1, accu2, twiddle[i]);
 
-      pDat_0[0] = accu2;
-      pDat_0[1] = accu1;
+      pDat_0[0] = accu2 >> 1;
+      pDat_0[1] = accu1 >> 1;
     }
   }
 
   fft(M, pDat, pDat_e);
 
-#ifdef FUNCTION_dst_IV_func2
-  if ((M >= 4) && ((M & 3) == 0)) {
-    dst_IV_func2(M >> 2, sin_twiddle + sin_step, &pDat[0], &pDat[L - 1],
-                 sin_step);
-  } else
-#endif /* FUNCTION_dst_IV_func2 */
   {
     FIXP_DBL *RESTRICT pDat_0;
     FIXP_DBL *RESTRICT pDat_1;
@@ -558,20 +530,20 @@ void dst_IV(FIXP_DBL *pDat, int L, int *pDat_e) {
     accu1 = pDat_1[0];
     accu2 = pDat_1[1];
 
-    pDat_1[1] = -(pDat_0[0] >> 1);
-    pDat_0[0] = (pDat_0[1] >> 1);
+    pDat_1[1] = -pDat_0[0];
+    pDat_0[0] = pDat_0[1];
 
     for (idx = sin_step, i = 1; i<(M + 1)>> 1; i++, idx += sin_step) {
       FIXP_STP twd = sin_twiddle[idx];
 
-      cplxMultDiv2(&accu3, &accu4, accu1, accu2, twd);
+      cplxMult(&accu3, &accu4, accu1, accu2, twd);
       pDat_1[0] = -accu3;
       pDat_0[1] = -accu4;
 
       pDat_0 += 2;
       pDat_1 -= 2;
 
-      cplxMultDiv2(&accu3, &accu4, pDat_0[1], pDat_0[0], twd);
+      cplxMult(&accu3, &accu4, pDat_0[1], pDat_0[0], twd);
 
       accu1 = pDat_1[0];
       accu2 = pDat_1[1];
@@ -582,8 +554,8 @@ void dst_IV(FIXP_DBL *pDat, int L, int *pDat_e) {
 
     if ((M & 1) == 0) {
       /* Last Sin and Cos value pair are the same */
-      accu1 = fMultDiv2(accu1, WTC(0x5a82799a));
-      accu2 = fMultDiv2(accu2, WTC(0x5a82799a));
+      accu1 = fMult(accu1, WTC(0x5a82799a));
+      accu2 = fMult(accu2, WTC(0x5a82799a));
 
       pDat_0[1] = -accu1 - accu2;
       pDat_1[0] = accu2 - accu1;
diff --git a/libFDK/src/nlc_dec.cpp b/libFDK/src/nlc_dec.cpp
index 8a8ccfd..6e98ce0 100644
--- a/libFDK/src/nlc_dec.cpp
+++ b/libFDK/src/nlc_dec.cpp
@@ -647,6 +647,10 @@ static ERROR_t huff_decode(HANDLE_FDK_BITSTREAM strm, SCHAR* out_data_1,
             }
             df_rest_flag_1 = num_val_1_int % 2;
             if (df_rest_flag_1) num_val_1_int -= 1;
+            if (num_val_1_int < 0) {
+              err = HUFFDEC_NOTOK;
+              goto bail;
+            }
           }
           if (out_data_2 != NULL) {
             if (diff_type_2 == DIFF_FREQ) {
@@ -658,6 +662,10 @@ static ERROR_t huff_decode(HANDLE_FDK_BITSTREAM strm, SCHAR* out_data_1,
             }
             df_rest_flag_2 = num_val_2_int % 2;
             if (df_rest_flag_2) num_val_2_int -= 1;
+            if (num_val_2_int < 0) {
+              err = HUFFDEC_NOTOK;
+              goto bail;
+            }
           }
 
           if (out_data_1 != NULL) {