hls: fix signalling of 64-bit dist2 values

This commit fixes an error in the hls: dist2 is capable of being a
64-bit value and must not be signalled as 32-bit.

tmc3/BitWriter.h

@@ -66,9 +66,17 @@ public:
 
   void writeUe(uint32_t value);
 
+  // NB: probably best to think twice before using this method
+  void writeUe64(uint64_t value);
+
   template<typename T>
   void writeUe(const T& value)
   {
+    // Avoid accidental truncation
+    constexpr bool use_writeUe64 =
+      std::is_same<uint64_t, T>::value || std::is_same<int64_t, T>::value;
+    static_assert(!use_writeUe64, "use explicit writeUe64()");
+
     writeUe(uint32_t(value));
   }
 
@@ -161,6 +169,18 @@ BitWriter<OutputIt>::writeUe(uint32_t value)
 
 //----------------------------------------------------------------------------
 
+template<class OutputIt>
+void
+BitWriter<OutputIt>::writeUe64(uint64_t value)
+{
+  value++;
+  int len = ilog2(value);
+  writeUn(len, 0);
+  writeUn(len + 1, value);
+}
+
+//----------------------------------------------------------------------------
+
 template<class OutputIt>
 void
 BitWriter<OutputIt>::writeSe(int32_t value)

tmc3/PCCMisc.h

@@ -109,6 +109,22 @@ ceilpow2(uint32_t x)
   return x + 1;
 }
 
+//---------------------------------------------------------------------------
+// Round @x up to next power of two.
+//
+inline uint64_t
+ceilpow2(uint64_t x)
+{
+  x--;
+  x = x | (x >> 1);
+  x = x | (x >> 2);
+  x = x | (x >> 4);
+  x = x | (x >> 8);
+  x = x | (x >> 16);
+  x = x | (x >> 32);
+  return x + 1;
+}
+
 //---------------------------------------------------------------------------
 // Compute \left\floor \text{log}_2(x) \right\floor.
 // NB: ilog2(0) = -1.
@@ -120,6 +136,17 @@ ilog2(uint32_t x)
   return popcnt(x) - 1;
 }
 
+//---------------------------------------------------------------------------
+// Compute \left\floor \text{log}_2(x) \right\floor.
+// NB: ilog2(0) = -1.
+
+inline int
+ilog2(uint64_t x)
+{
+  x = ceilpow2(x + 1) - 1;
+  return popcnt(uint32_t(x >> 32)) + popcnt(uint32_t(x)) - 1;
+}
+
 //---------------------------------------------------------------------------
 // Compute \left\ceil \text{log}_2(x) \right\ceil.
 // NB: ceillog2(0) = 32.

tmc3/PCCTMC3Common.h

@@ -301,7 +301,7 @@ inline void
 PCCBuildLevelOfDetail2(
   const PCCPointSet3& pointCloud,
   const int levelOfDetailCount,
-  const std::vector<int>& dist2,
+  const std::vector<int64_t>& dist2,
   std::vector<uint32_t>& numberOfPointsPerLOD,
   std::vector<uint32_t>& indexes)
 {
@@ -412,7 +412,7 @@ inline void
 PCCBuildLevelOfDetail(
   const PCCPointSet3& pointCloud,
   const int levelOfDetailCount,
-  const std::vector<int>& dist2,
+  const std::vector<int64_t>& dist2,
   std::vector<uint32_t>& numberOfPointsPerLOD,
   std::vector<uint32_t>& indexes)
 {

tmc3/hls.h

@@ -239,7 +239,7 @@ struct AttributeParameterSet {
 
   // NB: derived from num_detail_levels_minus1
   int numDetailLevels;
-  std::vector<int> dist2;
+  std::vector<int64_t> dist2;
 
   // NB: these parameters are shared by raht and lift
   std::vector<int> quant_step_size_luma;

tmc3/io_hls.cpp

@@ -253,7 +253,8 @@ write(const AttributeParameterSet& aps)
     int num_detail_levels_minus1 = aps.numDetailLevels - 1;
     bs.writeUe(num_detail_levels_minus1);
     for (int idx = 0; idx <= num_detail_levels_minus1; idx++) {
-      bs.writeUe(aps.dist2[idx]);
+      // todo(??): is this an appropriate encoding?
+      bs.writeUe64(aps.dist2[idx]);
       bs.writeUe(aps.quant_step_size_luma[idx]);
       if (chroma_quant_steps_present_flag)
         bs.writeUe(aps.quant_step_size_chroma[idx]);