diff options
| author | Jean-Michel Trivi <jmtrivi@google.com> | 2013-08-08 17:38:33 -0700 |
|---|---|---|
| committer | Jean-Michel Trivi <jmtrivi@google.com> | 2013-08-20 15:01:21 -0700 |
| commit | ed247dfa54234c1115c35a05327955b29b161e8a (patch) | |
| tree | 9e4488bf844e217bf16d8cd3f001738d14ed56fd | |
| parent | 577fcbb570d023be4cea9564292dd2bd95f40c3b (diff) | |
| download | fdk-aac-ed247dfa54234c1115c35a05327955b29b161e8a.tar.gz fdk-aac-ed247dfa54234c1115c35a05327955b29b161e8a.tar.bz2 fdk-aac-ed247dfa54234c1115c35a05327955b29b161e8a.zip | |
Encoder quantizer fix
AAC Encoder: Saturate quantizer shift value to prevent
undefined behaviour.
In very rare cases the shift value 'totalShift = (16-4)-(3*(totalShift>>2))'
can be greater than accu data width. If you apply a shift with more then
31 bit the result depends on the architecture and is not defined in C.
For certain platforms zeros are shifted in. That would be our desired behaviour.
On other platforms the shift will be applied as modulo. For example >>34
would be applied as >>2. To prevent this discrepancy the shift value
is limited/saturated to DFRACT_BITS-1. 'accu >>= fixMin(totalShift,DFRACT_BITS-1)'.
Bug 9428126
Change-Id: I27177654c4dc22cf899bc35dad9cdd040dccb02d
| -rw-r--r-- | libAACenc/src/aacenc_lib.cpp | 2 | ||||
| -rw-r--r-- | libAACenc/src/quantize.cpp | 4 |
2 files changed, 3 insertions, 3 deletions
diff --git a/libAACenc/src/aacenc_lib.cpp b/libAACenc/src/aacenc_lib.cpp index 2f95ab1..e229ed6 100644 --- a/libAACenc/src/aacenc_lib.cpp +++ b/libAACenc/src/aacenc_lib.cpp @@ -98,7 +98,7 @@ amm-info@iis.fraunhofer.de /* Encoder library info */ #define AACENCODER_LIB_VL0 3 #define AACENCODER_LIB_VL1 4 -#define AACENCODER_LIB_VL2 1 +#define AACENCODER_LIB_VL2 2 #define AACENCODER_LIB_TITLE "AAC Encoder" #define AACENCODER_LIB_BUILD_DATE __DATE__ #define AACENCODER_LIB_BUILD_TIME __TIME__ diff --git a/libAACenc/src/quantize.cpp b/libAACenc/src/quantize.cpp index da6f85e..5380e35 100644 --- a/libAACenc/src/quantize.cpp +++ b/libAACenc/src/quantize.cpp @@ -127,7 +127,7 @@ static void FDKaacEnc_quantizeLines(INT gain, accu = fMultDiv2(FDKaacEnc_mTab_3_4[tabIndex],FDKaacEnc_quantTableE[totalShift&3]); totalShift = (16-4)-(3*(totalShift>>2)); FDK_ASSERT(totalShift >=0); /* MAX_QUANT_VIOLATION */ - accu>>=totalShift; + accu >>= fixMin(totalShift,DFRACT_BITS-1); quaSpectrum[line] = (SHORT)(-((LONG)(k + accu) >> (DFRACT_BITS-1-16))); } else if(accu > FL2FXCONST_DBL(0.0f)) @@ -140,7 +140,7 @@ static void FDKaacEnc_quantizeLines(INT gain, accu = fMultDiv2(FDKaacEnc_mTab_3_4[tabIndex],FDKaacEnc_quantTableE[totalShift&3]); totalShift = (16-4)-(3*(totalShift>>2)); FDK_ASSERT(totalShift >=0); /* MAX_QUANT_VIOLATION */ - accu>>=totalShift; + accu >>= fixMin(totalShift,DFRACT_BITS-1); quaSpectrum[line] = (SHORT)((LONG)(k + accu) >> (DFRACT_BITS-1-16)); } else |