Fix float to int32_t sample conversion

1 files changed, 5 insertions, 2 deletions

diff --git a/src/DspChunk.cpp b/src/DspChunk.cpp
index 962338e..90c10ab 100644
--- a/src/DspChunk.cpp
+++ b/src/DspChunk.cpp
@@ -4,6 +4,9 @@
 namespace SaneAudioRenderer

 {

     static_assert((int32_t{-1} >> 31) == -1 && (int64_t{-1} >> 63) == -1, "Code relies on right signed shift UB");

+    static_assert((int64_t)(float)((int64_t)INT32_MAX + 1) == ((int64_t)INT32_MAX + 1), "Rounding error");

+    static_assert((int32_t)(float)(INT32_MAX - 127) == (INT32_MAX - 127), "Rounding error");

+    static_assert((int32_t)(double)INT32_MAX == INT32_MAX, "Rounding error");

 

     namespace

     {

@@ -161,13 +164,13 @@ namespace SaneAudioRenderer
         template <>

         inline void ConvertSample<DspFormat::Float, DspFormat::Pcm24>(const float& input,  int24_t& output)

         {

-            PackPcm24((int32_t)(input * INT32_MAX), (char*)output.d);

+            PackPcm24((int32_t)(input * (INT32_MAX - 127)), (char*)output.d);

         }

 

         template <>

         inline void ConvertSample<DspFormat::Float, DspFormat::Pcm32>(const float& input, int32_t& output)

         {

-            output = (int32_t)(input * INT32_MAX);

+            output = (int32_t)(input * (INT32_MAX - 127));

         }

 

         template <>