minor code refactoring, mostly around __builtin_expect

git-svn-id: https://lz4.googlecode.com/svn/trunk@59 650e7d94-2a16-8b24-b05c-7c0b3f6821cd

1 files changed, 23 insertions, 20 deletions

diff --git a/lz4.c b/lz4.c
index 28a892e..df62be6 100644
--- a/lz4.c
+++ b/lz4.c
@@ -98,6 +98,8 @@
 #define restrict // Disable restrict

 #endif

 

+#define GCC_VERSION (__GNUC__ * 100 + __GNUC_MINOR__)

+

 #ifdef _MSC_VER  // Visual Studio

 #define inline __forceinline // Visual is not C99, but supports some kind of inline

 #include <intrin.h>          // _BitScanForward

@@ -109,6 +111,15 @@
 #define bswap16(x) ((unsigned short int) ((((x) >> 8) & 0xffu) | (((x) & 0xffu) << 8)))

 #endif

 

+#if (GCC_VERSION >= 302) || (__INTEL_COMPILER >= 800) || defined(__clang__)

+# define expect(expr,value)    (__builtin_expect ((expr),(value)) )

+#else

+# define expect(expr,value)    (expr)

+#endif

+

+#define likely(expr)     expect((expr) != 0, 1)

+#define unlikely(expr)   expect((expr) != 0, 0)

+

 

 //**************************************

 // Includes

@@ -210,14 +221,6 @@ typedef struct _U64_S { U64 v; } U64_S;
 #define LZ4_WRITE_LITTLEENDIAN_16(p,v) { A16(p) = v; p+=2; }

 #endif

 

-#if __GNUC__ >= 3

-# define expect(expr,value)         __builtin_expect ((expr),(value))

-#else

-# define expect(expr,value)         (expr)

-#endif

-

-#define expect_true(expr)  expect ((expr) != 0, 1)

-#define expect_false(expr) expect ((expr) != 0, 0)

 

 //**************************************

 // Local structures

@@ -249,7 +252,7 @@ inline static int LZ4_NbCommonBytes (register U64 val)
     unsigned long r = 0;

     _BitScanReverse64( &r, val );

     return (int)(r>>3);

-    #elif defined(__GNUC__) && ((__GNUC__ * 100 + __GNUC_MINOR__) >= 304) && !defined(LZ4_FORCE_SW_BITCOUNT)

+    #elif defined(__GNUC__) && (GCC_VERSION >= 304) && !defined(LZ4_FORCE_SW_BITCOUNT)

     return (__builtin_clzll(val) >> 3); 

     #else

 	int r;

@@ -263,7 +266,7 @@ inline static int LZ4_NbCommonBytes (register U64 val)
     unsigned long r = 0;

     _BitScanForward64( &r, val );

     return (int)(r>>3);

-    #elif defined(__GNUC__) && ((__GNUC__ * 100 + __GNUC_MINOR__) >= 304) && !defined(LZ4_FORCE_SW_BITCOUNT)

+    #elif defined(__GNUC__) && (GCC_VERSION >= 304) && !defined(LZ4_FORCE_SW_BITCOUNT)

     return (__builtin_ctzll(val) >> 3); 

     #else

 	static const int DeBruijnBytePos[64] = { 0, 0, 0, 0, 0, 1, 1, 2, 0, 3, 1, 3, 1, 4, 2, 7, 0, 2, 3, 6, 1, 5, 3, 5, 1, 3, 4, 4, 2, 5, 6, 7, 7, 0, 1, 2, 3, 3, 4, 6, 2, 6, 5, 5, 3, 4, 5, 6, 7, 1, 2, 4, 6, 4, 4, 5, 7, 2, 6, 5, 7, 6, 7, 7 };

@@ -281,7 +284,7 @@ inline static int LZ4_NbCommonBytes (register U32 val)
     unsigned long r = 0;

     _BitScanReverse( &r, val );

     return (int)(r>>3);

-    #elif defined(__GNUC__) && ((__GNUC__ * 100 + __GNUC_MINOR__) >= 304) && !defined(LZ4_FORCE_SW_BITCOUNT)

+    #elif defined(__GNUC__) && (GCC_VERSION >= 304) && !defined(LZ4_FORCE_SW_BITCOUNT)

     return (__builtin_clz(val) >> 3); 

     #else

 	int r;

@@ -294,7 +297,7 @@ inline static int LZ4_NbCommonBytes (register U32 val)
     unsigned long r = 0;

     _BitScanForward( &r, val );

     return (int)(r>>3);

-    #elif defined(__GNUC__) && ((__GNUC__ * 100 + __GNUC_MINOR__) >= 304) && !defined(LZ4_FORCE_SW_BITCOUNT)

+    #elif defined(__GNUC__) && (GCC_VERSION >= 304) && !defined(LZ4_FORCE_SW_BITCOUNT)

     return (__builtin_ctz(val) >> 3); 

     #else

 	static const int DeBruijnBytePos[32] = { 0, 0, 3, 0, 3, 1, 3, 0, 3, 2, 2, 1, 3, 2, 0, 1, 3, 3, 1, 2, 2, 2, 2, 0, 3, 1, 2, 0, 1, 0, 1, 1 };

@@ -381,7 +384,7 @@ int LZ4_compressCtx(void** ctx,
 			ip = forwardIp;

 			forwardIp = ip + step;

 

-			if (expect_false(forwardIp > mflimit)) { goto _last_literals; }

+			if unlikely(forwardIp > mflimit) { goto _last_literals; }

 

 			forwardH = LZ4_HASH_VALUE(forwardIp);

 			ref = base + HashTable[h];

@@ -390,7 +393,7 @@ int LZ4_compressCtx(void** ctx,
 		} while ((ref < ip - MAX_DISTANCE) || (A32(ref) != A32(ip)));

 

 		// Catch up

-		while ((expect_false(ip>anchor) && expect_false(ref>(BYTE*)source) && (ip[-1]==ref[-1]))) { ip--; ref--; }

+		while ((ip>anchor) && (ref>(BYTE*)source) && unlikely(ip[-1]==ref[-1])) { ip--; ref--; }

 

 		// Encode Literal length

 		length = ip - anchor;

@@ -408,7 +411,7 @@ _next_match:
 		// Start Counting

 		ip+=MINMATCH; ref+=MINMATCH;   // MinMatch verified

 		anchor = ip;

-		while (expect_true(ip<matchlimit-(STEPSIZE-1)))

+		while likely(ip<matchlimit-(STEPSIZE-1))

 		{

 			UARCH diff = AARCH(ref) ^ AARCH(ip);

 			if (!diff) { ip+=STEPSIZE; ref+=STEPSIZE; continue; }

@@ -532,7 +535,7 @@ int LZ4_compress64kCtx(void** ctx,
 		} while (A32(ref) != A32(ip));

 

 		// Catch up

-		while (((ip>anchor) && expect_false(ref>(BYTE*)source) && (ip[-1]==ref[-1]))) { ip--; ref--; }

+		while ((ip>anchor) && (ref>(BYTE*)source) && (ip[-1]==ref[-1])) { ip--; ref--; }

 

 		// Encode Literal length

 		length = ip - anchor;

@@ -576,7 +579,7 @@ _endCount:
 		// Test next position

 		ref = base + HashTable[LZ4_HASH64K_VALUE(ip)];

 		HashTable[LZ4_HASH64K_VALUE(ip)] = ip - base;

-		if (expect_true(A32(ref) == A32(ip))) { token = op++; *token=0; goto _next_match; }

+		if (A32(ref) == A32(ip)) { token = op++; *token=0; goto _next_match; }

 

 		// Prepare next loop

 		anchor = ip++;

@@ -658,7 +661,7 @@ int LZ4_uncompress(const char* source,
 

 		// copy literals

 		cpy = op+length;

-		if (expect_false(cpy>oend-COPYLENGTH))

+		if unlikely(cpy>oend-COPYLENGTH)

 		{

 			if (cpy > oend) goto _output_error;

 			memcpy(op, ip, length);

@@ -675,7 +678,7 @@ int LZ4_uncompress(const char* source,
 		if ((length=(token&ML_MASK)) == ML_MASK) { for (;*ip==255;length+=255) {ip++;} length += *ip++; }

 

 		// copy repeated sequence

-		if (expect_false(op-ref<STEPSIZE))

+		if unlikely(op-ref<STEPSIZE)

 		{

 #if LZ4_ARCH64

 			size_t dec2table[]={0, 0, 0, -1, 0, 1, 2, 3};

@@ -764,7 +767,7 @@ int LZ4_uncompress_unknownOutputSize(
 		if ((length=(token&ML_MASK)) == ML_MASK) { while (ip<iend) { int s = *ip++; length +=s; if (s==255) continue; break; } }

 

 		// copy repeated sequence

-		if (expect_false(op-ref<STEPSIZE))

+		if unlikely(op-ref<STEPSIZE)

 		{

 #if LZ4_ARCH64

 			size_t dec2table[]={0, 0, 0, -1, 0, 1, 2, 3};