Merge pull request #7028 from unknownbrackets/xxhash-update

Update xxHash to r37, including XXH64

Core/Debugger/DisassemblyManager.cpp

@@ -42,7 +42,11 @@ bool isInInterval(u32 start, u32 size, u32 value)
 
 static u32 computeHash(u32 address, u32 size)
 {
-	return XXH32(Memory::GetPointer(address),size,0xBACD7814);
+#ifdef _M_X64
+	return XXH64(Memory::GetPointer(address), size, 0xBACD7814BACD7814LL);
+#else
+	return XXH32(Memory::GetPointer(address), size, 0xBACD7814);
+#endif
 }
 
 
@@ -356,7 +360,7 @@ void DisassemblyFunction::recheck()
 	if (!PSP_IsInited())
 		return;
 
-	u32 newHash = computeHash(address,size);
+	HashType newHash = computeHash(address,size);
 	if (hash != newHash)
 	{
 		hash = newHash;
@@ -819,7 +823,7 @@ void DisassemblyData::recheck()
 	if (!PSP_IsInited())
 		return;
 
-	u32 newHash = computeHash(address,size);
+	HashType newHash = computeHash(address,size);
 	if (newHash != hash)
 	{
 		hash = newHash;

Core/Debugger/DisassemblyManager.h

@@ -21,6 +21,12 @@
 #include "Core/Debugger/SymbolMap.h"
 #include "Core/MIPS/MIPSAnalyst.h"
 
+#ifdef _M_X64
+typedef u64 HashType;
+#else
+typedef u32 HashType;
+#endif
+
 enum DisassemblyLineType { DISTYPE_OPCODE, DISTYPE_MACRO, DISTYPE_DATA, DISTYPE_OTHER };
 
 struct DisassemblyLineInfo
@@ -79,7 +85,7 @@ private:
 
 	u32 address;
 	u32 size;
-	u32 hash;
+	HashType hash;
 	std::vector<BranchLine> lines;
 	std::map<u32,DisassemblyEntry*> entries;
 	std::vector<u32> lineAddresses;
@@ -155,7 +161,7 @@ private:
 
 	u32 address;
 	u32 size;
-	u32 hash;
+	HashType hash;
 	DataType type;
 	std::map<u32,DataEntry> lines;
 	std::vector<u32> lineAddresses;

GPU/Common/TextureDecoder.cpp

@@ -157,7 +157,8 @@ void DoUnswizzleTex16Basic(const u8 *texptr, u32 *ydestp, int bxc, int byc, u32
 #ifndef _M_SSE
 QuickTexHashFunc DoQuickTexHash = &QuickTexHashBasic;
 UnswizzleTex16Func DoUnswizzleTex16 = &DoUnswizzleTex16Basic;
-ReliableHashFunc DoReliableHash = &XXH32;
+ReliableHash32Func DoReliableHash32 = &XXH32;
+ReliableHash64Func DoReliableHash64 = &XXH64;
 #endif
 
 // This has to be done after CPUDetect has done its magic.
@@ -168,7 +169,7 @@ void SetupTextureDecoder() {
 		DoUnswizzleTex16 = &DoUnswizzleTex16NEON;
 #ifndef IOS
 		// Not sure if this is safe on iOS, it's had issues with xxhash.
-		DoReliableHash = &ReliableHashNEON;
+		DoReliableHash32 = &ReliableHash32NEON;
 #endif
 	}
 #endif

GPU/Common/TextureDecoder.h

@@ -32,7 +32,16 @@ void DoUnswizzleTex16Basic(const u8 *texptr, u32 *ydestp, int bxc, int byc, u32
 #define DoUnswizzleTex16 DoUnswizzleTex16Basic
 
 #include "ext/xxhash.h"
+#define DoReliableHash32 XXH32
+#define DoReliableHash64 XXH64
+
+#ifdef _M_X64
+#define DoReliableHash XXH64
+typedef u64 ReliableHashType;
+#else
 #define DoReliableHash XXH32
+typedef u32 ReliableHashType;
+#endif
 #else
 typedef u32 (*QuickTexHashFunc)(const void *checkp, u32 size);
 extern QuickTexHashFunc DoQuickTexHash;
@@ -40,8 +49,14 @@ extern QuickTexHashFunc DoQuickTexHash;
 typedef void (*UnswizzleTex16Func)(const u8 *texptr, u32 *ydestp, int bxc, int byc, u32 pitch, u32 rowWidth);
 extern UnswizzleTex16Func DoUnswizzleTex16;
 
-typedef u32 (*ReliableHashFunc)(const void *input, int len, u32 seed);
-extern ReliableHashFunc DoReliableHash;
+typedef u32 (*ReliableHash32Func)(const void *input, size_t len, u32 seed);
+extern ReliableHash32Func DoReliableHash32;
+
+typedef u64 (*ReliableHash64Func)(const void *input, size_t len, u64 seed);
+extern ReliableHash64Func DoReliableHash64;
+
+#define DoReliableHash DoReliableHash32
+typedef u32 ReliableHashType;
 #endif
 
 // All these DXT structs are in the reverse order, as compared to PC.

GPU/Common/TextureDecoderNEON.cpp

@@ -177,7 +177,7 @@ void DoUnswizzleTex16NEON(const u8 *texptr, u32 *ydestp, int bxc, int byc, u32 p
 #  define XXH_rotl32(x,r) ((x << r) | (x >> (32 - r)))
 #endif
 
-u32 ReliableHashNEON(const void *input, int len, u32 seed) {
+u32 ReliableHash32NEON(const void *input, size_t len, u32 seed) {
 	const u8 *p = (const u8 *)input;
 	const u8 *const bEnd = p + len;
 	U32 h32;

GPU/Common/TextureDecoderNEON.h

@@ -19,4 +19,4 @@
 
 u32 QuickTexHashNEON(const void *checkp, u32 size);
 void DoUnswizzleTex16NEON(const u8 *texptr, u32 *ydestp, int bxc, int byc, u32 pitch, u32 rowWidth);
-u32 ReliableHashNEON(const void *input, int len, u32 seed);
+u32 ReliableHash32NEON(const void *input, size_t len, u32 seed);

GPU/Directx9/TextureCacheDX9.cpp

@@ -798,7 +798,7 @@ void TextureCacheDX9::UpdateCurrentClut() {
 	// If not, we're going to hash random data, which hopefully doesn't cause a performance issue.
 	const u32 clutExtendedBytes = clutTotalBytes_ + clutBaseBytes;
 
-	clutHash_ = DoReliableHash((const char *)clutBufRaw_, clutExtendedBytes, 0xC0108888);
+	clutHash_ = DoReliableHash32((const char *)clutBufRaw_, clutExtendedBytes, 0xC0108888);
 	clutBuf_ = clutBufRaw_;
 
 	// Special optimization: fonts typically draw clut4 with just alpha values in a single color.

GPU/Directx9/TransformPipelineDX9.cpp

@@ -462,7 +462,7 @@ inline u32 ComputeMiniHashRange(const void *ptr, size_t sz) {
 		size_t step = sz / 4;
 		u32 hash = 0;
 		for (size_t i = 0; i < sz; i += step) {
-			hash += DoReliableHash(p + i, 100, 0x3A44B9C4);
+			hash += DoReliableHash32(p + i, 100, 0x3A44B9C4);
 		}
 		return hash;
 	} else {
@@ -509,8 +509,8 @@ void TransformDrawEngineDX9::MarkUnreliable(VertexArrayInfoDX9 *vai) {
 	}
 }
 
-u32 TransformDrawEngineDX9::ComputeHash() {
-	u32 fullhash = 0;
+ReliableHashType TransformDrawEngineDX9::ComputeHash() {
+	ReliableHashType fullhash = 0;
 	const int vertexSize = dec_->GetDecVtxFmt().stride;
 	const int indexSize = (dec_->VertexType() & GE_VTYPE_IDX_MASK) == GE_VTYPE_IDX_16BIT ? 2 : 1;
 
@@ -645,7 +645,7 @@ void TransformDrawEngineDX9::DoFlush() {
 			case VertexArrayInfoDX9::VAI_NEW:
 				{
 					// Haven't seen this one before.
-					u32 dataHash = ComputeHash();
+					ReliableHashType dataHash = ComputeHash();
 					vai->hash = dataHash;
 					vai->minihash = ComputeMiniHash();
 					vai->status = VertexArrayInfoDX9::VAI_HASHING;
@@ -670,7 +670,7 @@ void TransformDrawEngineDX9::DoFlush() {
 					if (vai->drawsUntilNextFullHash == 0) {
 						// Let's try to skip a full hash if mini would fail.
 						const u32 newMiniHash = ComputeMiniHash();
-						u32 newHash = vai->hash;
+						ReliableHashType newHash = vai->hash;
 						if (newMiniHash == vai->minihash) {
 							newHash = ComputeHash();
 						}

GPU/Directx9/TransformPipelineDX9.h

@@ -51,11 +51,12 @@ enum {
 	VAI_FLAG_VERTEXFULLALPHA = 1,
 };
 
-
-// Don't bother storing information about draws smaller than this.
-enum {
-	VERTEX_CACHE_THRESHOLD = 20,
-};
+// Avoiding the full include of TextureDecoder.h.
+#ifdef _M_X64
+typedef u64 ReliableHashType;
+#else
+typedef u32 ReliableHashType;
+#endif
 
 // Try to keep this POD.
 class VertexArrayInfoDX9 {
@@ -81,7 +82,7 @@ public:
 		VAI_UNRELIABLE,  // never cache
 	};
 
-	u32 hash;
+	ReliableHashType hash;
 	u32 minihash;
 
 	Status status;
@@ -191,7 +192,7 @@ private:
 	IDirect3DVertexDeclaration9 *SetupDecFmtForDraw(VSShader *vshader, const DecVtxFormat &decFmt, u32 pspFmt);
 
 	u32 ComputeMiniHash();
-	u32 ComputeHash();  // Reads deferred vertex data.
+	ReliableHashType ComputeHash();  // Reads deferred vertex data.
 	void MarkUnreliable(VertexArrayInfoDX9 *vai);
 
 	VertexDecoder *GetVertexDecoder(u32 vtype);

GPU/GLES/TextureCache.cpp

@@ -921,7 +921,7 @@ void TextureCache::UpdateCurrentClut() {
 	// If not, we're going to hash random data, which hopefully doesn't cause a performance issue.
 	const u32 clutExtendedBytes = clutTotalBytes_ + clutBaseBytes;
 
-	clutHash_ = DoReliableHash((const char *)clutBufRaw_, clutExtendedBytes, 0xC0108888);
+	clutHash_ = DoReliableHash32((const char *)clutBufRaw_, clutExtendedBytes, 0xC0108888);
 
 	// Avoid a copy when we don't need to convert colors.
 	if (UseBGRA8888() || clutFormat != GE_CMODE_32BIT_ABGR8888) {

GPU/GLES/TransformPipeline.cpp

@@ -444,7 +444,7 @@ inline u32 ComputeMiniHashRange(const void *ptr, size_t sz) {
 		size_t step = sz / 4;
 		u32 hash = 0;
 		for (size_t i = 0; i < sz; i += step) {
-			hash += DoReliableHash(p + i, 100, 0x3A44B9C4);
+			hash += DoReliableHash32(p + i, 100, 0x3A44B9C4);
 		}
 		return hash;
 	} else {
@@ -491,8 +491,8 @@ void TransformDrawEngine::MarkUnreliable(VertexArrayInfo *vai) {
 	}
 }
 
-u32 TransformDrawEngine::ComputeHash() {
-	u32 fullhash = 0;
+ReliableHashType TransformDrawEngine::ComputeHash() {
+	ReliableHashType fullhash = 0;
 	const int vertexSize = dec_->GetDecVtxFmt().stride;
 	const int indexSize = (dec_->VertexType() & GE_VTYPE_IDX_MASK) == GE_VTYPE_IDX_16BIT ? 2 : 1;
 
@@ -633,7 +633,7 @@ void TransformDrawEngine::DoFlush() {
 			case VertexArrayInfo::VAI_NEW:
 				{
 					// Haven't seen this one before.
-					u32 dataHash = ComputeHash();
+					ReliableHashType dataHash = ComputeHash();
 					vai->hash = dataHash;
 					vai->minihash = ComputeMiniHash();
 					vai->status = VertexArrayInfo::VAI_HASHING;
@@ -658,7 +658,7 @@ void TransformDrawEngine::DoFlush() {
 					if (vai->drawsUntilNextFullHash == 0) {
 						// Let's try to skip a full hash if mini would fail.
 						const u32 newMiniHash = ComputeMiniHash();
-						u32 newHash = vai->hash;
+						ReliableHashType newHash = vai->hash;
 						if (newMiniHash == vai->minihash) {
 							newHash = ComputeHash();
 						}

GPU/GLES/TransformPipeline.h

@@ -53,6 +53,13 @@ enum {
 	VAI_FLAG_VERTEXFULLALPHA = 1,
 };
 
+// Avoiding the full include of TextureDecoder.h.
+#ifdef _M_X64
+typedef u64 ReliableHashType;
+#else
+typedef u32 ReliableHashType;
+#endif
+
 // Try to keep this POD.
 class VertexArrayInfo {
 public:
@@ -77,7 +84,7 @@ public:
 		VAI_UNRELIABLE,  // never cache
 	};
 
-	u32 hash;
+	ReliableHashType hash;
 	u32 minihash;
 
 	Status status;
@@ -188,7 +195,7 @@ private:
 	void FreeBuffer(GLuint buf);
 
 	u32 ComputeMiniHash();
-	u32 ComputeHash();  // Reads deferred vertex data.
+	ReliableHashType ComputeHash();  // Reads deferred vertex data.
 	void MarkUnreliable(VertexArrayInfo *vai);
 
 	VertexDecoder *GetVertexDecoder(u32 vtype);

ext/xxhash.c

@@ -28,6 +28,7 @@ OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 
 You can contact the author at :
 - xxHash source repository : http://code.google.com/p/xxhash/
+- public discussion board : https://groups.google.com/forum/#!forum/lz4c
 */
 
 
@@ -69,7 +70,7 @@ You can contact the author at :
 
 #ifdef _MSC_VER    // Visual Studio
 #  define FORCE_INLINE static __forceinline
-#else 
+#else
 #  ifdef __GNUC__
 #    define FORCE_INLINE static inline __attribute__((always_inline))
 #  else
@@ -77,19 +78,27 @@ You can contact the author at :
 #  endif
 #endif
 
-
 //**************************************
 // Includes & Memory related functions
 //**************************************
 #include "xxhash.h"
-// Modify the local functions below should you wish to use some other memory related routines
+// Modify the local functions below should you wish to use some other memory routines
 // for malloc(), free()
 #include <stdlib.h>
-FORCE_INLINE void* XXH_malloc(size_t s) { return malloc(s); }
-FORCE_INLINE void  XXH_free  (void* p)  { free(p); }
+FORCE_INLINE void* XXH_malloc(size_t s)
+{
+    return malloc(s);
+}
+FORCE_INLINE void  XXH_free  (void* p)
+{
+    free(p);
+}
 // for memcpy()
 #include <string.h>
-FORCE_INLINE void* XXH_memcpy(void* dest, const void* src, size_t size) { return memcpy(dest,src,size); }
+FORCE_INLINE void* XXH_memcpy(void* dest, const void* src, size_t size)
+{
+    return memcpy(dest,src,size);
+}
 
 
 //**************************************
@@ -97,17 +106,17 @@ FORCE_INLINE void* XXH_memcpy(void* dest, const void* src, size_t size) { return
 //**************************************
 #if defined (__STDC_VERSION__) && __STDC_VERSION__ >= 199901L   // C99
 # include <stdint.h>
-  typedef uint8_t  BYTE;
-  typedef uint16_t U16;
-  typedef uint32_t U32;
-  typedef  int32_t S32;
-  typedef uint64_t U64;
+typedef uint8_t  BYTE;
+typedef uint16_t U16;
+typedef uint32_t U32;
+typedef  int32_t S32;
+typedef uint64_t U64;
 #else
-  typedef unsigned char      BYTE;
-  typedef unsigned short     U16;
-  typedef unsigned int       U32;
-  typedef   signed int       S32;
-  typedef unsigned long long U64;
+typedef unsigned char      BYTE;
+typedef unsigned short     U16;
+typedef unsigned int       U32;
+typedef   signed int       S32;
+typedef unsigned long long U64;
 #endif
 
 #if defined(__GNUC__)  && !defined(XXH_USE_UNALIGNED_ACCESS)
@@ -124,13 +133,21 @@ FORCE_INLINE void* XXH_memcpy(void* dest, const void* src, size_t size) { return
 #  endif
 #endif
 
-typedef struct _U32_S { U32 v; } _PACKED U32_S;
+typedef struct _U32_S
+{
+    U32 v;
+} _PACKED U32_S;
+typedef struct _U64_S
+{
+    U64 v;
+} _PACKED U64_S;
 
 #if !defined(XXH_USE_UNALIGNED_ACCESS) && !defined(__GNUC__)
 #  pragma pack(pop)
 #endif
 
 #define A32(x) (((U32_S *)(x))->v)
+#define A64(x) (((U64_S *)(x))->v)
 
 
 //***************************************
@@ -141,20 +158,37 @@ typedef struct _U32_S { U32 v; } _PACKED U32_S;
 // Note : although _rotl exists for minGW (GCC under windows), performance seems poor
 #if defined(_MSC_VER)
 #  define XXH_rotl32(x,r) _rotl(x,r)
+#  define XXH_rotl64(x,r) _rotl64(x,r)
 #else
 #  define XXH_rotl32(x,r) ((x << r) | (x >> (32 - r)))
+#  define XXH_rotl64(x,r) ((x << r) | (x >> (64 - r)))
 #endif
 
 #if defined(_MSC_VER)     // Visual Studio
 #  define XXH_swap32 _byteswap_ulong
+#  define XXH_swap64 _byteswap_uint64
 #elif GCC_VERSION >= 403
 #  define XXH_swap32 __builtin_bswap32
+#  define XXH_swap64 __builtin_bswap64
 #else
-static inline U32 XXH_swap32 (U32 x) {
+static inline U32 XXH_swap32 (U32 x)
+{
     return  ((x << 24) & 0xff000000 ) |
-        ((x <<  8) & 0x00ff0000 ) |
-        ((x >>  8) & 0x0000ff00 ) |
-        ((x >> 24) & 0x000000ff );}
+            ((x <<  8) & 0x00ff0000 ) |
+            ((x >>  8) & 0x0000ff00 ) |
+            ((x >> 24) & 0x000000ff );
+}
+static inline U64 XXH_swap64 (U64 x)
+{
+    return  ((x << 56) & 0xff00000000000000ULL) |
+            ((x << 40) & 0x00ff000000000000ULL) |
+            ((x << 24) & 0x0000ff0000000000ULL) |
+            ((x << 8)  & 0x000000ff00000000ULL) |
+            ((x >> 8)  & 0x00000000ff000000ULL) |
+            ((x >> 24) & 0x0000000000ff0000ULL) |
+            ((x >> 40) & 0x000000000000ff00ULL) |
+            ((x >> 56) & 0x00000000000000ffULL);
+}
 #endif
 
 
@@ -167,13 +201,18 @@ static inline U32 XXH_swap32 (U32 x) {
 #define PRIME32_4    668265263U
 #define PRIME32_5    374761393U
 
+#define PRIME64_1 11400714785074694791ULL
+#define PRIME64_2 14029467366897019727ULL
+#define PRIME64_3  1609587929392839161ULL
+#define PRIME64_4  9650029242287828579ULL
+#define PRIME64_5  2870177450012600261ULL
 
 //**************************************
 // Architecture Macros
 //**************************************
 typedef enum { XXH_bigEndian=0, XXH_littleEndian=1 } XXH_endianess;
 #ifndef XXH_CPU_LITTLE_ENDIAN   // It is possible to define XXH_CPU_LITTLE_ENDIAN externally, for example using a compiler switch
-    static const int one = 1;
+static const int one = 1;
 #   define XXH_CPU_LITTLE_ENDIAN   (*(char*)(&one))
 #endif
 
@@ -190,27 +229,48 @@ typedef enum { XXH_bigEndian=0, XXH_littleEndian=1 } XXH_endianess;
 typedef enum { XXH_aligned, XXH_unaligned } XXH_alignment;
 
 FORCE_INLINE U32 XXH_readLE32_align(const U32* ptr, XXH_endianess endian, XXH_alignment align)
-{ 
+{
     if (align==XXH_unaligned)
-        return endian==XXH_littleEndian ? A32(ptr) : XXH_swap32(A32(ptr)); 
+        return endian==XXH_littleEndian ? A32(ptr) : XXH_swap32(A32(ptr));
     else
-        return endian==XXH_littleEndian ? *ptr : XXH_swap32(*ptr); 
+        return endian==XXH_littleEndian ? *ptr : XXH_swap32(*ptr);
 }
 
-FORCE_INLINE U32 XXH_readLE32(const U32* ptr, XXH_endianess endian) { return XXH_readLE32_align(ptr, endian, XXH_unaligned); }
+FORCE_INLINE U32 XXH_readLE32(const U32* ptr, XXH_endianess endian)
+{
+    return XXH_readLE32_align(ptr, endian, XXH_unaligned);
+}
+
+FORCE_INLINE U64 XXH_readLE64_align(const U64* ptr, XXH_endianess endian, XXH_alignment align)
+{
+    if (align==XXH_unaligned)
+        return endian==XXH_littleEndian ? A64(ptr) : XXH_swap64(A64(ptr));
+    else
+        return endian==XXH_littleEndian ? *ptr : XXH_swap64(*ptr);
+}
+
+FORCE_INLINE U64 XXH_readLE64(const U64* ptr, XXH_endianess endian)
+{
+    return XXH_readLE64_align(ptr, endian, XXH_unaligned);
+}
 
 
 //****************************
 // Simple Hash Functions
 //****************************
-FORCE_INLINE U32 XXH32_endian_align(const void* input, int len, U32 seed, XXH_endianess endian, XXH_alignment align)
+FORCE_INLINE U32 XXH32_endian_align(const void* input, size_t len, U32 seed, XXH_endianess endian, XXH_alignment align)
 {
     const BYTE* p = (const BYTE*)input;
-    const BYTE* const bEnd = p + len;
+    const BYTE* bEnd = p + len;
     U32 h32;
+#define XXH_get32bits(p) XXH_readLE32_align((const U32*)p, endian, align)
 
 #ifdef XXH_ACCEPT_NULL_INPUT_POINTER
-    if (p==NULL) { len=0; p=(const BYTE*)(size_t)16; }
+    if (p==NULL)
+    {
+        len=0;
+        bEnd=p=(const BYTE*)(size_t)16;
+    }
 #endif
 
     if (len>=16)
@@ -226,11 +286,24 @@ FORCE_INLINE U32 XXH32_endian_align(const void* input, int len, U32 seed, XXH_en
 #if defined(ARM) && defined(__GNUC__)
             __builtin_prefetch(p + 0xc0, 0, 0);
 #endif
-            v1 += XXH_readLE32_align((const U32*)p, endian, align) * PRIME32_2; v1 = XXH_rotl32(v1, 13); v1 *= PRIME32_1; p+=4;
-            v2 += XXH_readLE32_align((const U32*)p, endian, align) * PRIME32_2; v2 = XXH_rotl32(v2, 13); v2 *= PRIME32_1; p+=4;
-            v3 += XXH_readLE32_align((const U32*)p, endian, align) * PRIME32_2; v3 = XXH_rotl32(v3, 13); v3 *= PRIME32_1; p+=4;
-            v4 += XXH_readLE32_align((const U32*)p, endian, align) * PRIME32_2; v4 = XXH_rotl32(v4, 13); v4 *= PRIME32_1; p+=4;
-        } while (p<=limit);
+            v1 += XXH_get32bits(p) * PRIME32_2;
+            v1 = XXH_rotl32(v1, 13);
+            v1 *= PRIME32_1;
+            p+=4;
+            v2 += XXH_get32bits(p) * PRIME32_2;
+            v2 = XXH_rotl32(v2, 13);
+            v2 *= PRIME32_1;
+            p+=4;
+            v3 += XXH_get32bits(p) * PRIME32_2;
+            v3 = XXH_rotl32(v3, 13);
+            v3 *= PRIME32_1;
+            p+=4;
+            v4 += XXH_get32bits(p) * PRIME32_2;
+            v4 = XXH_rotl32(v4, 13);
+            v4 *= PRIME32_1;
+            p+=4;
+        }
+        while (p<=limit);
 
         h32 = XXH_rotl32(v1, 1) + XXH_rotl32(v2, 7) + XXH_rotl32(v3, 12) + XXH_rotl32(v4, 18);
     }
@@ -241,9 +314,9 @@ FORCE_INLINE U32 XXH32_endian_align(const void* input, int len, U32 seed, XXH_en
 
     h32 += (U32) len;
 
-    while (p<=bEnd-4)
+    while (p+4<=bEnd)
     {
-        h32 += XXH_readLE32_align((const U32*)p, endian, align) * PRIME32_3;
+        h32 += XXH_get32bits(p) * PRIME32_3;
         h32  = XXH_rotl32(h32, 17) * PRIME32_4 ;
         p+=4;
     }
@@ -265,18 +338,19 @@ FORCE_INLINE U32 XXH32_endian_align(const void* input, int len, U32 seed, XXH_en
 }
 
 
-U32 XXH32(const void* input, int len, U32 seed)
+unsigned int XXH32 (const void* input, size_t len, unsigned seed)
 {
 #if 0
     // Simple version, good for code maintenance, but unfortunately slow for small inputs
-    void* state = XXH32_init(seed);
-    XXH32_update(state, input, len);
-    return XXH32_digest(state);
+    XXH32_state_t state;
+    XXH32_reset(&state, seed);
+    XXH32_update(&state, input, len);
+    return XXH32_digest(&state);
 #else
     XXH_endianess endian_detected = (XXH_endianess)XXH_CPU_LITTLE_ENDIAN;
 
 #  if !defined(XXH_USE_UNALIGNED_ACCESS)
-    if (!(((size_t)input) & 3))   // Input is aligned, let's leverage the speed advantage
+    if ((((size_t)input) & 3) == 0)   // Input is aligned, let's leverage the speed advantage
     {
         if ((endian_detected==XXH_littleEndian) || XXH_FORCE_NATIVE_FORMAT)
             return XXH32_endian_align(input, len, seed, XXH_littleEndian, XXH_aligned);
@@ -292,12 +366,156 @@ U32 XXH32(const void* input, int len, U32 seed)
 #endif
 }
 
+FORCE_INLINE U64 XXH64_endian_align(const void* input, size_t len, U64 seed, XXH_endianess endian, XXH_alignment align)
+{
+    const BYTE* p = (const BYTE*)input;
+    const BYTE* bEnd = p + len;
+    U64 h64;
+#define XXH_get64bits(p) XXH_readLE64_align((const U64*)p, endian, align)
 
-//****************************
-// Advanced Hash Functions
-//****************************
+#ifdef XXH_ACCEPT_NULL_INPUT_POINTER
+    if (p==NULL)
+    {
+        len=0;
+        bEnd=p=(const BYTE*)(size_t)32;
+    }
+#endif
 
-struct XXH_state32_t
+    if (len>=32)
+    {
+        const BYTE* const limit = bEnd - 32;
+        U64 v1 = seed + PRIME64_1 + PRIME64_2;
+        U64 v2 = seed + PRIME64_2;
+        U64 v3 = seed + 0;
+        U64 v4 = seed - PRIME64_1;
+
+        do
+        {
+#if defined(ARM) && defined(__GNUC__)
+			// TODO: Validate that this helps as it does with XXH32.
+            __builtin_prefetch(p + 0xc0, 0, 0);
+#endif
+            v1 += XXH_get64bits(p) * PRIME64_2;
+            p+=8;
+            v1 = XXH_rotl64(v1, 31);
+            v1 *= PRIME64_1;
+            v2 += XXH_get64bits(p) * PRIME64_2;
+            p+=8;
+            v2 = XXH_rotl64(v2, 31);
+            v2 *= PRIME64_1;
+            v3 += XXH_get64bits(p) * PRIME64_2;
+            p+=8;
+            v3 = XXH_rotl64(v3, 31);
+            v3 *= PRIME64_1;
+            v4 += XXH_get64bits(p) * PRIME64_2;
+            p+=8;
+            v4 = XXH_rotl64(v4, 31);
+            v4 *= PRIME64_1;
+        }
+        while (p<=limit);
+
+        h64 = XXH_rotl64(v1, 1) + XXH_rotl64(v2, 7) + XXH_rotl64(v3, 12) + XXH_rotl64(v4, 18);
+
+        v1 *= PRIME64_2;
+        v1 = XXH_rotl64(v1, 31);
+        v1 *= PRIME64_1;
+        h64 ^= v1;
+        h64 = h64 * PRIME64_1 + PRIME64_4;
+
+        v2 *= PRIME64_2;
+        v2 = XXH_rotl64(v2, 31);
+        v2 *= PRIME64_1;
+        h64 ^= v2;
+        h64 = h64 * PRIME64_1 + PRIME64_4;
+
+        v3 *= PRIME64_2;
+        v3 = XXH_rotl64(v3, 31);
+        v3 *= PRIME64_1;
+        h64 ^= v3;
+        h64 = h64 * PRIME64_1 + PRIME64_4;
+
+        v4 *= PRIME64_2;
+        v4 = XXH_rotl64(v4, 31);
+        v4 *= PRIME64_1;
+        h64 ^= v4;
+        h64 = h64 * PRIME64_1 + PRIME64_4;
+    }
+    else
+    {
+        h64  = seed + PRIME64_5;
+    }
+
+    h64 += (U64) len;
+
+    while (p+8<=bEnd)
+    {
+        U64 k1 = XXH_get64bits(p);
+        k1 *= PRIME64_2;
+        k1 = XXH_rotl64(k1,31);
+        k1 *= PRIME64_1;
+        h64 ^= k1;
+        h64 = XXH_rotl64(h64,27) * PRIME64_1 + PRIME64_4;
+        p+=8;
+    }
+
+    if (p+4<=bEnd)
+    {
+        h64 ^= (U64)(XXH_get32bits(p)) * PRIME64_1;
+        h64 = XXH_rotl64(h64, 23) * PRIME64_2 + PRIME64_3;
+        p+=4;
+    }
+
+    while (p<bEnd)
+    {
+        h64 ^= (*p) * PRIME64_5;
+        h64 = XXH_rotl64(h64, 11) * PRIME64_1;
+        p++;
+    }
+
+    h64 ^= h64 >> 33;
+    h64 *= PRIME64_2;
+    h64 ^= h64 >> 29;
+    h64 *= PRIME64_3;
+    h64 ^= h64 >> 32;
+
+    return h64;
+}
+
+
+unsigned long long XXH64 (const void* input, size_t len, unsigned long long seed)
+{
+#if 0
+    // Simple version, good for code maintenance, but unfortunately slow for small inputs
+    XXH64_state_t state;
+    XXH64_reset(&state, seed);
+    XXH64_update(&state, input, len);
+    return XXH64_digest(&state);
+#else
+    XXH_endianess endian_detected = (XXH_endianess)XXH_CPU_LITTLE_ENDIAN;
+
+#  if !defined(XXH_USE_UNALIGNED_ACCESS)
+    if ((((size_t)input) & 7)==0)   // Input is aligned, let's leverage the speed advantage
+    {
+        if ((endian_detected==XXH_littleEndian) || XXH_FORCE_NATIVE_FORMAT)
+            return XXH64_endian_align(input, len, seed, XXH_littleEndian, XXH_aligned);
+        else
+            return XXH64_endian_align(input, len, seed, XXH_bigEndian, XXH_aligned);
+    }
+#  endif
+
+    if ((endian_detected==XXH_littleEndian) || XXH_FORCE_NATIVE_FORMAT)
+        return XXH64_endian_align(input, len, seed, XXH_littleEndian, XXH_unaligned);
+    else
+        return XXH64_endian_align(input, len, seed, XXH_bigEndian, XXH_unaligned);
+#endif
+}
+
+/****************************************************
+ *  Advanced Hash Functions
+****************************************************/
+
+/*** Allocation ***/
+typedef struct
 {
     U64 total_len;
     U32 seed;
@@ -305,21 +523,51 @@ struct XXH_state32_t
     U32 v2;
     U32 v3;
     U32 v4;
-    int memsize;
+    U32 memsize;
     char memory[16];
+} XXH_istate32_t;
+
+typedef struct
+{
+    U64 total_len;
+    U64 seed;
+    U64 v1;
+    U64 v2;
+    U64 v3;
+    U64 v4;
+    U32 memsize;
+    char memory[32];
+} XXH_istate64_t;
+
+
+XXH32_state_t* XXH32_createState(void)
+{
+    XXH_STATIC_ASSERT(sizeof(XXH32_state_t) >= sizeof(XXH_istate32_t));   // A compilation error here means XXH32_state_t is not large enough
+    return (XXH32_state_t*)malloc(sizeof(XXH32_state_t));
+}
+XXH_errorcode XXH32_freeState(XXH32_state_t* statePtr)
+{
+    free(statePtr);
+    return XXH_OK;
+};
+
+XXH64_state_t* XXH64_createState(void)
+{
+    XXH_STATIC_ASSERT(sizeof(XXH64_state_t) >= sizeof(XXH_istate64_t));   // A compilation error here means XXH64_state_t is not large enough
+    return (XXH64_state_t*)malloc(sizeof(XXH64_state_t));
+}
+XXH_errorcode XXH64_freeState(XXH64_state_t* statePtr)
+{
+    free(statePtr);
+    return XXH_OK;
 };
 
 
-int XXH32_sizeofState() 
+/*** Hash feed ***/
+
+XXH_errorcode XXH32_reset(XXH32_state_t* state_in, U32 seed)
 {
-    XXH_STATIC_ASSERT(XXH32_SIZEOFSTATE >= sizeof(struct XXH_state32_t));   // A compilation error here means XXH32_SIZEOFSTATE is not large enough
-    return sizeof(struct XXH_state32_t); 
-}
-
-
-XXH_errorcode XXH32_resetState(void* state_in, U32 seed)
-{ 
-    struct XXH_state32_t * state = (struct XXH_state32_t *) state_in;
+    XXH_istate32_t* state = (XXH_istate32_t*) state_in;
     state->seed = seed;
     state->v1 = seed + PRIME32_1 + PRIME32_2;
     state->v2 = seed + PRIME32_2;
@@ -330,18 +578,23 @@ XXH_errorcode XXH32_resetState(void* state_in, U32 seed)
     return XXH_OK;
 }
 
-
-void* XXH32_init (U32 seed)
+XXH_errorcode XXH64_reset(XXH64_state_t* state_in, unsigned long long seed)
 {
-    void* state = XXH_malloc (sizeof(struct XXH_state32_t));
-    XXH32_resetState(state, seed);
-    return state;
+    XXH_istate64_t* state = (XXH_istate64_t*) state_in;
+    state->seed = seed;
+    state->v1 = seed + PRIME64_1 + PRIME64_2;
+    state->v2 = seed + PRIME64_2;
+    state->v3 = seed + 0;
+    state->v4 = seed - PRIME64_1;
+    state->total_len = 0;
+    state->memsize = 0;
+    return XXH_OK;
 }
 
 
-FORCE_INLINE XXH_errorcode XXH32_update_endian (void* state_in, const void* input, int len, XXH_endianess endian)
+FORCE_INLINE XXH_errorcode XXH32_update_endian (XXH32_state_t* state_in, const void* input, size_t len, XXH_endianess endian)
 {
-    struct XXH_state32_t * state = (struct XXH_state32_t *) state_in;
+    XXH_istate32_t* state = (XXH_istate32_t *) state_in;
     const BYTE* p = (const BYTE*)input;
     const BYTE* const bEnd = p + len;
 
@@ -354,7 +607,7 @@ FORCE_INLINE XXH_errorcode XXH32_update_endian (void* state_in, const void* inpu
     if (state->memsize + len < 16)   // fill in tmp buffer
     {
         XXH_memcpy(state->memory + state->memsize, input, len);
-        state->memsize +=  len;
+        state->memsize += (U32)len;
         return XXH_OK;
     }
 
@@ -363,10 +616,22 @@ FORCE_INLINE XXH_errorcode XXH32_update_endian (void* state_in, const void* inpu
         XXH_memcpy(state->memory + state->memsize, input, 16-state->memsize);
         {
             const U32* p32 = (const U32*)state->memory;
-            state->v1 += XXH_readLE32(p32, endian) * PRIME32_2; state->v1 = XXH_rotl32(state->v1, 13); state->v1 *= PRIME32_1; p32++;
-            state->v2 += XXH_readLE32(p32, endian) * PRIME32_2; state->v2 = XXH_rotl32(state->v2, 13); state->v2 *= PRIME32_1; p32++; 
-            state->v3 += XXH_readLE32(p32, endian) * PRIME32_2; state->v3 = XXH_rotl32(state->v3, 13); state->v3 *= PRIME32_1; p32++;
-            state->v4 += XXH_readLE32(p32, endian) * PRIME32_2; state->v4 = XXH_rotl32(state->v4, 13); state->v4 *= PRIME32_1; p32++;
+            state->v1 += XXH_readLE32(p32, endian) * PRIME32_2;
+            state->v1 = XXH_rotl32(state->v1, 13);
+            state->v1 *= PRIME32_1;
+            p32++;
+            state->v2 += XXH_readLE32(p32, endian) * PRIME32_2;
+            state->v2 = XXH_rotl32(state->v2, 13);
+            state->v2 *= PRIME32_1;
+            p32++;
+            state->v3 += XXH_readLE32(p32, endian) * PRIME32_2;
+            state->v3 = XXH_rotl32(state->v3, 13);
+            state->v3 *= PRIME32_1;
+            p32++;
+            state->v4 += XXH_readLE32(p32, endian) * PRIME32_2;
+            state->v4 = XXH_rotl32(state->v4, 13);
+            state->v4 *= PRIME32_1;
+            p32++;
         }
         p += 16-state->memsize;
         state->memsize = 0;
@@ -382,11 +647,24 @@ FORCE_INLINE XXH_errorcode XXH32_update_endian (void* state_in, const void* inpu
 
         do
         {
-            v1 += XXH_readLE32((const U32*)p, endian) * PRIME32_2; v1 = XXH_rotl32(v1, 13); v1 *= PRIME32_1; p+=4;
-            v2 += XXH_readLE32((const U32*)p, endian) * PRIME32_2; v2 = XXH_rotl32(v2, 13); v2 *= PRIME32_1; p+=4;
-            v3 += XXH_readLE32((const U32*)p, endian) * PRIME32_2; v3 = XXH_rotl32(v3, 13); v3 *= PRIME32_1; p+=4;
-            v4 += XXH_readLE32((const U32*)p, endian) * PRIME32_2; v4 = XXH_rotl32(v4, 13); v4 *= PRIME32_1; p+=4;
-        } while (p<=limit);
+            v1 += XXH_readLE32((const U32*)p, endian) * PRIME32_2;
+            v1 = XXH_rotl32(v1, 13);
+            v1 *= PRIME32_1;
+            p+=4;
+            v2 += XXH_readLE32((const U32*)p, endian) * PRIME32_2;
+            v2 = XXH_rotl32(v2, 13);
+            v2 *= PRIME32_1;
+            p+=4;
+            v3 += XXH_readLE32((const U32*)p, endian) * PRIME32_2;
+            v3 = XXH_rotl32(v3, 13);
+            v3 *= PRIME32_1;
+            p+=4;
+            v4 += XXH_readLE32((const U32*)p, endian) * PRIME32_2;
+            v4 = XXH_rotl32(v4, 13);
+            v4 *= PRIME32_1;
+            p+=4;
+        }
+        while (p<=limit);
 
         state->v1 = v1;
         state->v2 = v2;
@@ -403,10 +681,10 @@ FORCE_INLINE XXH_errorcode XXH32_update_endian (void* state_in, const void* inpu
     return XXH_OK;
 }
 
-XXH_errorcode XXH32_update (void* state_in, const void* input, int len)
+XXH_errorcode XXH32_update (XXH32_state_t* state_in, const void* input, size_t len)
 {
     XXH_endianess endian_detected = (XXH_endianess)XXH_CPU_LITTLE_ENDIAN;
-    
+
     if ((endian_detected==XXH_littleEndian) || XXH_FORCE_NATIVE_FORMAT)
         return XXH32_update_endian(state_in, input, len, XXH_littleEndian);
     else
@@ -415,9 +693,9 @@ XXH_errorcode XXH32_update (void* state_in, const void* input, int len)
 
 
 
-FORCE_INLINE U32 XXH32_intermediateDigest_endian (void* state_in, XXH_endianess endian)
+FORCE_INLINE U32 XXH32_digest_endian (const XXH32_state_t* state_in, XXH_endianess endian)
 {
-    struct XXH_state32_t * state = (struct XXH_state32_t *) state_in;
+    XXH_istate32_t* state = (XXH_istate32_t*) state_in;
     const BYTE * p = (const BYTE*)state->memory;
     BYTE* bEnd = (BYTE*)state->memory + state->memsize;
     U32 h32;
@@ -433,7 +711,7 @@ FORCE_INLINE U32 XXH32_intermediateDigest_endian (void* state_in, XXH_endianess
 
     h32 += (U32) state->total_len;
 
-    while (p<=bEnd-4)
+    while (p+4<=bEnd)
     {
         h32 += XXH_readLE32((const U32*)p, endian) * PRIME32_3;
         h32  = XXH_rotl32(h32, 17) * PRIME32_4;
@@ -457,22 +735,206 @@ FORCE_INLINE U32 XXH32_intermediateDigest_endian (void* state_in, XXH_endianess
 }
 
 
-U32 XXH32_intermediateDigest (void* state_in)
+U32 XXH32_digest (const XXH32_state_t* state_in)
 {
     XXH_endianess endian_detected = (XXH_endianess)XXH_CPU_LITTLE_ENDIAN;
-    
+
     if ((endian_detected==XXH_littleEndian) || XXH_FORCE_NATIVE_FORMAT)
-        return XXH32_intermediateDigest_endian(state_in, XXH_littleEndian);
+        return XXH32_digest_endian(state_in, XXH_littleEndian);
     else
-        return XXH32_intermediateDigest_endian(state_in, XXH_bigEndian);
+        return XXH32_digest_endian(state_in, XXH_bigEndian);
 }
 
 
-U32 XXH32_digest (void* state_in)
+FORCE_INLINE XXH_errorcode XXH64_update_endian (XXH64_state_t* state_in, const void* input, size_t len, XXH_endianess endian)
 {
-    U32 h32 = XXH32_intermediateDigest(state_in);
+    XXH_istate64_t * state = (XXH_istate64_t *) state_in;
+    const BYTE* p = (const BYTE*)input;
+    const BYTE* const bEnd = p + len;
 
-    XXH_free(state_in);
+#ifdef XXH_ACCEPT_NULL_INPUT_POINTER
+    if (input==NULL) return XXH_ERROR;
+#endif
 
-    return h32;
+    state->total_len += len;
+
+    if (state->memsize + len < 32)   // fill in tmp buffer
+    {
+        XXH_memcpy(state->memory + state->memsize, input, len);
+        state->memsize += (U32)len;
+        return XXH_OK;
+    }
+
+    if (state->memsize)   // some data left from previous update
+    {
+        XXH_memcpy(state->memory + state->memsize, input, 32-state->memsize);
+        {
+            const U64* p64 = (const U64*)state->memory;
+            state->v1 += XXH_readLE64(p64, endian) * PRIME64_2;
+            state->v1 = XXH_rotl64(state->v1, 31);
+            state->v1 *= PRIME64_1;
+            p64++;
+            state->v2 += XXH_readLE64(p64, endian) * PRIME64_2;
+            state->v2 = XXH_rotl64(state->v2, 31);
+            state->v2 *= PRIME64_1;
+            p64++;
+            state->v3 += XXH_readLE64(p64, endian) * PRIME64_2;
+            state->v3 = XXH_rotl64(state->v3, 31);
+            state->v3 *= PRIME64_1;
+            p64++;
+            state->v4 += XXH_readLE64(p64, endian) * PRIME64_2;
+            state->v4 = XXH_rotl64(state->v4, 31);
+            state->v4 *= PRIME64_1;
+            p64++;
+        }
+        p += 32-state->memsize;
+        state->memsize = 0;
+    }
+
+    if (p+32 <= bEnd)
+    {
+        const BYTE* const limit = bEnd - 32;
+        U64 v1 = state->v1;
+        U64 v2 = state->v2;
+        U64 v3 = state->v3;
+        U64 v4 = state->v4;
+
+        do
+        {
+            v1 += XXH_readLE64((const U64*)p, endian) * PRIME64_2;
+            v1 = XXH_rotl64(v1, 31);
+            v1 *= PRIME64_1;
+            p+=8;
+            v2 += XXH_readLE64((const U64*)p, endian) * PRIME64_2;
+            v2 = XXH_rotl64(v2, 31);
+            v2 *= PRIME64_1;
+            p+=8;
+            v3 += XXH_readLE64((const U64*)p, endian) * PRIME64_2;
+            v3 = XXH_rotl64(v3, 31);
+            v3 *= PRIME64_1;
+            p+=8;
+            v4 += XXH_readLE64((const U64*)p, endian) * PRIME64_2;
+            v4 = XXH_rotl64(v4, 31);
+            v4 *= PRIME64_1;
+            p+=8;
+        }
+        while (p<=limit);
+
+        state->v1 = v1;
+        state->v2 = v2;
+        state->v3 = v3;
+        state->v4 = v4;
+    }
+
+    if (p < bEnd)
+    {
+        XXH_memcpy(state->memory, p, bEnd-p);
+        state->memsize = (int)(bEnd-p);
+    }
+
+    return XXH_OK;
+}
+
+XXH_errorcode XXH64_update (XXH64_state_t* state_in, const void* input, size_t len)
+{
+    XXH_endianess endian_detected = (XXH_endianess)XXH_CPU_LITTLE_ENDIAN;
+
+    if ((endian_detected==XXH_littleEndian) || XXH_FORCE_NATIVE_FORMAT)
+        return XXH64_update_endian(state_in, input, len, XXH_littleEndian);
+    else
+        return XXH64_update_endian(state_in, input, len, XXH_bigEndian);
+}
+
+
+
+FORCE_INLINE U64 XXH64_digest_endian (const XXH64_state_t* state_in, XXH_endianess endian)
+{
+    XXH_istate64_t * state = (XXH_istate64_t *) state_in;
+    const BYTE * p = (const BYTE*)state->memory;
+    BYTE* bEnd = (BYTE*)state->memory + state->memsize;
+    U64 h64;
+
+    if (state->total_len >= 32)
+    {
+        U64 v1 = state->v1;
+        U64 v2 = state->v2;
+        U64 v3 = state->v3;
+        U64 v4 = state->v4;
+
+        h64 = XXH_rotl64(v1, 1) + XXH_rotl64(v2, 7) + XXH_rotl64(v3, 12) + XXH_rotl64(v4, 18);
+
+        v1 *= PRIME64_2;
+        v1 = XXH_rotl64(v1, 31);
+        v1 *= PRIME64_1;
+        h64 ^= v1;
+        h64 = h64*PRIME64_1 + PRIME64_4;
+
+        v2 *= PRIME64_2;
+        v2 = XXH_rotl64(v2, 31);
+        v2 *= PRIME64_1;
+        h64 ^= v2;
+        h64 = h64*PRIME64_1 + PRIME64_4;
+
+        v3 *= PRIME64_2;
+        v3 = XXH_rotl64(v3, 31);
+        v3 *= PRIME64_1;
+        h64 ^= v3;
+        h64 = h64*PRIME64_1 + PRIME64_4;
+
+        v4 *= PRIME64_2;
+        v4 = XXH_rotl64(v4, 31);
+        v4 *= PRIME64_1;
+        h64 ^= v4;
+        h64 = h64*PRIME64_1 + PRIME64_4;
+    }
+    else
+    {
+        h64  = state->seed + PRIME64_5;
+    }
+
+    h64 += (U64) state->total_len;
+
+    while (p+8<=bEnd)
+    {
+        U64 k1 = XXH_readLE64((const U64*)p, endian);
+        k1 *= PRIME64_2;
+        k1 = XXH_rotl64(k1,31);
+        k1 *= PRIME64_1;
+        h64 ^= k1;
+        h64 = XXH_rotl64(h64,27) * PRIME64_1 + PRIME64_4;
+        p+=8;
+    }
+
+    if (p+4<=bEnd)
+    {
+        h64 ^= (U64)(XXH_readLE32((const U32*)p, endian)) * PRIME64_1;
+        h64 = XXH_rotl64(h64, 23) * PRIME64_2 + PRIME64_3;
+        p+=4;
+    }
+
+    while (p<bEnd)
+    {
+        h64 ^= (*p) * PRIME64_5;
+        h64 = XXH_rotl64(h64, 11) * PRIME64_1;
+        p++;
+    }
+
+    h64 ^= h64 >> 33;
+    h64 *= PRIME64_2;
+    h64 ^= h64 >> 29;
+    h64 *= PRIME64_3;
+    h64 ^= h64 >> 32;
+
+    return h64;
+}
+
+
+unsigned long long XXH64_digest (const XXH64_state_t* state_in)
+{
+    XXH_endianess endian_detected = (XXH_endianess)XXH_CPU_LITTLE_ENDIAN;
+
+    if ((endian_detected==XXH_littleEndian) || XXH_FORCE_NATIVE_FORMAT)
+        return XXH64_digest_endian(state_in, XXH_littleEndian);
+    else
+        return XXH64_digest_endian(state_in, XXH_bigEndian);
 }

ext/xxhash.h

@@ -1,5 +1,5 @@
 /*
-   xxHash - Fast Hash algorithm
+   xxHash - Extremely Fast Hash algorithm
    Header File
    Copyright (C) 2012-2014, Yann Collet.
    BSD 2-Clause License (http://www.opensource.org/licenses/bsd-license.php)
@@ -7,14 +7,14 @@
    Redistribution and use in source and binary forms, with or without
    modification, are permitted provided that the following conditions are
    met:
-  
+
        * Redistributions of source code must retain the above copyright
    notice, this list of conditions and the following disclaimer.
        * Redistributions in binary form must reproduce the above
    copyright notice, this list of conditions and the following disclaimer
    in the documentation and/or other materials provided with the
    distribution.
-  
+
    THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
    "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
    LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
@@ -52,8 +52,8 @@ CRC32           0.43 GB/s     9
 MD5-32          0.33 GB/s    10       Ronald L. Rivest
 SHA1-32         0.28 GB/s    10
 
-Q.Score is a measure of quality of the hash function. 
-It depends on successfully passing SMHasher test set. 
+Q.Score is a measure of quality of the hash function.
+It depends on successfully passing SMHasher test set.
 10 is a perfect score.
 */
 
@@ -64,101 +64,93 @@ extern "C" {
 #endif
 
 
-//****************************
-// Type
-//****************************
+/*****************************
+   Includes
+*****************************/
+#include <stddef.h>   /* size_t */
+
+
+/*****************************
+   Type
+*****************************/
 typedef enum { XXH_OK=0, XXH_ERROR } XXH_errorcode;
 
 
 
-//****************************
-// Simple Hash Functions
-//****************************
+/*****************************
+   Simple Hash Functions
+*****************************/
 
-unsigned int XXH32 (const void* input, int len, unsigned int seed);
+unsigned int       XXH32 (const void* input, size_t length, unsigned seed);
+unsigned long long XXH64 (const void* input, size_t length, unsigned long long seed);
 
 /*
 XXH32() :
-    Calculate the 32-bits hash of sequence of length "len" stored at memory address "input".
-    The memory between input & input+len must be valid (allocated and read-accessible).
+    Calculate the 32-bits hash of sequence "length" bytes stored at memory address "input".
+    The memory between input & input+length must be valid (allocated and read-accessible).
     "seed" can be used to alter the result predictably.
     This function successfully passes all SMHasher tests.
     Speed on Core 2 Duo @ 3 GHz (single thread, SMHasher benchmark) : 5.4 GB/s
-    Note that "len" is type "int", which means it is limited to 2^31-1.
-    If your data is larger, use the advanced functions below.
+XXH64() :
+    Calculate the 64-bits hash of sequence of length "len" stored at memory address "input".
 */
 
 
 
-//****************************
-// Advanced Hash Functions
-//****************************
-
-void*         XXH32_init   (unsigned int seed);
-XXH_errorcode XXH32_update (void* state, const void* input, int len);
-unsigned int  XXH32_digest (void* state);
+/*****************************
+   Advanced Hash Functions
+*****************************/
+typedef struct { long long ll[ 6]; } XXH32_state_t;
+typedef struct { long long ll[11]; } XXH64_state_t;
 
 /*
-These functions calculate the xxhash of an input provided in several small packets,
+These structures allow static allocation of XXH states.
+States must then be initialized using XXHnn_reset() before first use.
+
+If you prefer dynamic allocation, please refer to functions below.
+*/
+
+XXH32_state_t* XXH32_createState(void);
+XXH_errorcode  XXH32_freeState(XXH32_state_t* statePtr);
+
+XXH64_state_t* XXH64_createState(void);
+XXH_errorcode  XXH64_freeState(XXH64_state_t* statePtr);
+
+/*
+These functions create and release memory for XXH state.
+States must then be initialized using XXHnn_reset() before first use.
+*/
+
+
+XXH_errorcode XXH32_reset  (XXH32_state_t* statePtr, unsigned seed);
+XXH_errorcode XXH32_update (XXH32_state_t* statePtr, const void* input, size_t length);
+unsigned int  XXH32_digest (const XXH32_state_t* statePtr);
+
+XXH_errorcode      XXH64_reset  (XXH64_state_t* statePtr, unsigned long long seed);
+XXH_errorcode      XXH64_update (XXH64_state_t* statePtr, const void* input, size_t length);
+unsigned long long XXH64_digest (const XXH64_state_t* statePtr);
+
+/*
+These functions calculate the xxHash of an input provided in multiple smaller packets,
 as opposed to an input provided as a single block.
 
-It must be started with :
-void* XXH32_init()
-The function returns a pointer which holds the state of calculation.
+XXH state space must first be allocated, using either static or dynamic method provided above.
 
-This pointer must be provided as "void* state" parameter for XXH32_update().
-XXH32_update() can be called as many times as necessary.
-The user must provide a valid (allocated) input.
+Start a new hash by initializing state with a seed, using XXHnn_reset().
+
+Then, feed the hash state by calling XXHnn_update() as many times as necessary.
+Obviously, input must be valid, meaning allocated and read accessible.
 The function returns an error code, with 0 meaning OK, and any other value meaning there is an error.
-Note that "len" is type "int", which means it is limited to 2^31-1. 
-If your data is larger, it is recommended to chunk your data into blocks 
-of size for example 2^30 (1GB) to avoid any "int" overflow issue.
 
-Finally, you can end the calculation anytime, by using XXH32_digest().
-This function returns the final 32-bits hash.
-You must provide the same "void* state" parameter created by XXH32_init().
-Memory will be freed by XXH32_digest().
+Finally, you can produce a hash anytime, by using XXHnn_digest().
+This function returns the final nn-bits hash.
+You can nonetheless continue feeding the hash state with more input,
+and therefore get some new hashes, by calling again XXHnn_digest().
+
+When you are done, don't forget to free XXH state space, using typically XXHnn_freeState().
 */
 
 
-int           XXH32_sizeofState();
-XXH_errorcode XXH32_resetState(void* state, unsigned int seed);
-
-#define       XXH32_SIZEOFSTATE 48
-typedef struct { long long ll[(XXH32_SIZEOFSTATE+(sizeof(long long)-1))/sizeof(long long)]; } XXH32_stateSpace_t;
-/*
-These functions allow user application to make its own allocation for state.
-
-XXH32_sizeofState() is used to know how much space must be allocated for the xxHash 32-bits state.
-Note that the state must be aligned to access 'long long' fields. Memory must be allocated and referenced by a pointer.
-This pointer must then be provided as 'state' into XXH32_resetState(), which initializes the state.
-
-For static allocation purposes (such as allocation on stack, or freestanding systems without malloc()),
-use the structure XXH32_stateSpace_t, which will ensure that memory space is large enough and correctly aligned to access 'long long' fields.
-*/
-
-
-unsigned int XXH32_intermediateDigest (void* state);
-/*
-This function does the same as XXH32_digest(), generating a 32-bit hash,
-but preserve memory context.
-This way, it becomes possible to generate intermediate hashes, and then continue feeding data with XXH32_update().
-To free memory context, use XXH32_digest(), or free().
-*/
-
-
-
-//****************************
-// Deprecated function names
-//****************************
-// The following translations are provided to ease code transition
-// You are encouraged to no longer this function names
-#define XXH32_feed   XXH32_update
-#define XXH32_result XXH32_digest
-#define XXH32_getIntermediateResult XXH32_intermediateDigest
-
-
-
 #if defined (__cplusplus)
 }
 #endif