AuroraMiddleware/VulkanMemoryAllocator
1
0
Fork 0
mirror of https://github.com/GPUOpen-LibrariesAndSDKs/VulkanMemoryAllocator synced 2024-11-05 12:20:07 +00:00
Code Issues Projects Releases Wiki Activity

Optimized bit functions, added intrinsics for GCC and Clang

Browse Source 
Code by @medranSolus
This commit is contained in:
Adam Sawicki 2022-01-24 14:58:42 +01:00
parent 12d128d8f7
commit 518907bbc5
1 changed files with 32 additions and 6 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

38
include/vk_mem_alloc.h
View File

@ -3064,12 +3064,18 @@ class VmaAllocationObjectAllocator;
// Returns number of bits set to 1 in (v). // Returns number of bits set to 1 in (v).
static inline uint32_t VmaCountBitsSet(uint32_t v) static inline uint32_t VmaCountBitsSet(uint32_t v)
{ {
#ifdef _MSC_VER
return __popcnt(v);
#elif defined __GNUC__ || defined __clang__
return static_cast<uint32_t>(__builtin_popcount(v));
#else
uint32_t c = v - ((v >> 1) & 0x55555555); uint32_t c = v - ((v >> 1) & 0x55555555);
c = ((c >> 2) & 0x33333333) + (c & 0x33333333); c = ((c >> 2) & 0x33333333) + (c & 0x33333333);
c = ((c >> 4) + c) & 0x0F0F0F0F; c = ((c >> 4) + c) & 0x0F0F0F0F;
c = ((c >> 8) + c) & 0x00FF00FF; c = ((c >> 8) + c) & 0x00FF00FF;
c = ((c >> 16) + c) & 0x0000FFFF; c = ((c >> 16) + c) & 0x0000FFFF;
return c; return c;
#endif
} }
static inline uint8_t VmaBitScanLSB(uint64_t mask) static inline uint8_t VmaBitScanLSB(uint64_t mask)
@ -3078,15 +3084,20 @@ static inline uint8_t VmaBitScanLSB(uint64_t mask)
unsigned long pos; unsigned long pos;
if (_BitScanForward64(&pos, mask)) if (_BitScanForward64(&pos, mask))
return static_cast<uint8_t>(pos); return static_cast<uint8_t>(pos);
return UINT8_MAX;
#elif defined __GNUC__ || defined __clang__
return static_cast<uint8_t>__builtin_ffsll(mask)) - 1U;
#else #else
uint8_t pos = 0; uint8_t pos = 0;
uint64_t bit = 1;
do do
{ {
if (mask & (1ULL << pos)) if (mask & bit)
return pos; return pos;
bit <<= 1;
} while (pos++ < 63); } while (pos++ < 63);
#endif
return UINT8_MAX; return UINT8_MAX;
#endif
} }
static inline uint8_t VmaBitScanLSB(uint32_t mask) static inline uint8_t VmaBitScanLSB(uint32_t mask)
@ -3095,15 +3106,20 @@ static inline uint8_t VmaBitScanLSB(uint32_t mask)
unsigned long pos; unsigned long pos;
if (_BitScanForward(&pos, mask)) if (_BitScanForward(&pos, mask))
return static_cast<uint8_t>(pos); return static_cast<uint8_t>(pos);
return UINT8_MAX;
#elif defined __GNUC__ || defined __clang__
return static_cast<uint8_t>__builtin_ffsl(mask)) - 1U;
#else #else
uint8_t pos = 0; uint8_t pos = 0;
uint32_t bit = 1;
do do
{ {
if (mask & (1UL << pos)) if (mask & bit)
return pos; return pos;
bit <<= 1;
} while (pos++ < 31); } while (pos++ < 31);
#endif
return UINT8_MAX; return UINT8_MAX;
#endif
} }
static inline uint8_t VmaBitScanMSB(uint64_t mask) static inline uint8_t VmaBitScanMSB(uint64_t mask)
@ -3112,12 +3128,17 @@ static inline uint8_t VmaBitScanMSB(uint64_t mask)
unsigned long pos; unsigned long pos;
if (_BitScanReverse64(&pos, mask)) if (_BitScanReverse64(&pos, mask))
return static_cast<uint8_t>(pos); return static_cast<uint8_t>(pos);
#elif defined __GNUC__ || defined __clang__
if (mask)
return static_cast<uint8_t>(__builtin_clzll(mask));
#else #else
uint8_t pos = 63; uint8_t pos = 63;
uint64_t bit = 1 << 63;
do do
{ {
if (mask & (1ULL << pos)) if (mask & bit)
return pos; return pos;
bit >>= 1;
} while (pos-- > 0); } while (pos-- > 0);
#endif #endif
return UINT8_MAX; return UINT8_MAX;
@ -3129,12 +3150,17 @@ static inline uint8_t VmaBitScanMSB(uint32_t mask)
unsigned long pos; unsigned long pos;
if (_BitScanReverse(&pos, mask)) if (_BitScanReverse(&pos, mask))
return static_cast<uint8_t>(pos); return static_cast<uint8_t>(pos);
#elif defined __GNUC__ || defined __clang__
if (mask)
return static_cast<uint8_t>(__builtin_clzl(mask));
#else #else
uint8_t pos = 31; uint8_t pos = 31;
uint32_t bit = 1 << 31;
do do
{ {
if (mask & (1UL << pos)) if (mask & bit)
return pos; return pos;
bit >>= 1;
} while (pos-- > 0); } while (pos-- > 0);
#endif #endif
return UINT8_MAX; return UINT8_MAX;