AnotherFoxGuy/polyvox
1
0
Fork 0
Code Issues Pull Requests Activity

Going back to building key by shifting instead of using bitfield.

Browse Source
This commit is contained in:
David Williams
2015-03-21 08:05:58 +01:00
parent 0e995b5140
commit 0d638f9837
3 changed files with 30 additions and 38 deletions
Download Patch File Download Diff File Expand all files Collapse all files

46
include/PolyVox/PagedVolume.h
View File

@ -281,7 +281,7 @@ namespace PolyVox
PagedVolume& operator=(const PagedVolume& rhs);
private:
struct ChunkKey
/*struct ChunkKey
{
ChunkKey(int32_t x, int32_t y, int32_t z) : iZPos(z), iYPos(y), iXPos(x), iValid(~0) {}
int64_t iZPos : 10;
@ -292,39 +292,31 @@ namespace PolyVox
// so we don't need to have a seperate check for that before using it.
int64_t iValid : 2;
};
static_assert(sizeof(ChunkKey) == sizeof(int64_t), "");
static_assert(sizeof(ChunkKey) == sizeof(int64_t), "");*/
int64_t posToChunkKey(int32_t iXPos, int32_t iYPos, int32_t iZPos) const
uint64_t posToChunkKey(int32_t iXPos, int32_t iYPos, int32_t iZPos) const
{
iXPos = iXPos >> m_uChunkSideLengthPower;
iYPos = iYPos >> m_uChunkSideLengthPower;
iZPos = iZPos >> m_uChunkSideLengthPower;
ChunkKey chunkKey(iXPos, iYPos, iZPos);
//chunkKey.iValid = ~0;
//chunkKey.iValid = 2;
// In general, bit shifting signed integers is dubious because of potential undefined or implementation defied behaviour.
// However, it seems that in practice most compilers and architectures work in the same way (http://stackoverflow.com/a/6488645).
static_assert((int64_t(-3) / 4) == 0, "fdfs");
static_assert((int64_t(-3) >> 2) == -1, "fdfs");
/*chunkKey.iXPos = iXPos;
chunkKey.iYPos = iYPos;
chunkKey.iZPos = iZPos;*/
// If this kind of casting ever causes problems there are
// other solutions here: http://stackoverflow.com/a/2468738
int64_t iKeyAsInt64 = force_cast<int64_t>(chunkKey);
return iKeyAsInt64;
// Bit-shifting of signed integer values has various issues with undefined or implementation-defined behaviour.
// Therefore we cast to unsigned to avoid these (we only care about the bit pattern anyway, not the actual value).
const uint64_t uXPos = static_cast<uint64_t>(iXPos);
const uint64_t uYPos = static_cast<uint64_t>(iYPos);
const uint64_t uZPos = static_cast<uint64_t>(iZPos);
const uint64_t xKey = ((uXPos >> m_uChunkSideLengthPower) & 0x3FF) << 20;
const uint64_t yKey = ((uYPos >> m_uChunkSideLengthPower) & 0x3FF) << 10;
const uint64_t zKey = ((uZPos >> m_uChunkSideLengthPower) & 0x3FF);
const uint64_t key = 0x80000000 | xKey | yKey | zKey;
return key;
}
Chunk* getChunk(int64_t iKeyAsInt32) const;
Chunk* getChunk(uint64_t iKeyAsInt32) const;
mutable int64_t m_v3dLastAccessedChunkKey = 0;
mutable uint64_t m_v3dLastAccessedChunkKey = 0;
mutable Chunk* m_pLastAccessedChunk = nullptr;
mutable std::unordered_map<uint32_t, std::unique_ptr< Chunk > > m_mapChunks;
mutable std::unordered_map<uint64_t, std::unique_ptr< Chunk > > m_mapChunks;
mutable uint32_t m_uTimestamper = 0;