current chk version - too many changes on both sides for now ;-(

git-svn-id: svn://ttmath.org/publicrep/ttmath/branches/chk@150 e52654a7-88a9-db11-a3e9-0013d4bc506e

CHANGELOG

@@ -1,4 +1,4 @@
-Version 0.8.5 (2009.05.11):
+Version 0.8.5 prerelease (2009.05.15):
     * fixed:   Big::Mod(x) didn't correctly return a carry
                and the result was sometimes very big (even greater than x)
     * fixed:   global function Mod(x) didn't set an ErrorCode object
@@ -11,6 +11,25 @@ Version 0.8.5 (2009.05.11):
                the same is to Cos() function
     * changed: PrepareSin(x) is using Big::Mod() now when reducing 2PI period
                should be a little accurate especially on a very big 'x'
+    * changed: uint Mul(const UInt<value_size> & ss2, uint algorithm = 100)
+               void MulBig(const UInt<value_size> & ss2, UInt<value_size*2> & result, uint algorithm = 100)
+               those methods by default use MulFastest() and MulFastestBig()
+    * changed: changed a little Mul2Big() to cooperate with Mul3Big()
+    * added:   uint UInt::Mul3(const UInt<value_size> & ss2)
+               void UInt::Mul3Big(const UInt<value_size> & ss2, UInt<value_size*2> & result)
+               a new multiplication algorithm: Karatsuba multiplication,
+               on a vector UInt<100> with all items different from zero this algorithm is faster
+               about 3 times than Mul2Big(), and on a vector UInt<1000> with all items different from
+               zero this algorithm is faster more than 5 times than Mul2Big()
+               (measured on 32bit platform with GCC 4.3.3 with -O3 and -DTTMATH_RELEASE)
+    * added:   uint MulFastest(const UInt<value_size> & ss2)
+               void MulFastestBig(const UInt<value_size> & ss2, UInt<value_size*2> & result)
+               those methods are trying to select the fastest multiplication algorithm
+    * added:   uint AddVector(const uint * ss1, const uint * ss2, uint ss1_size, uint ss2_size, uint * result)
+               uint SubVector(const uint * ss1, const uint * ss2, uint ss1_size, uint ss2_size, uint * result)
+               three forms: asm x86, asm x86_64, no-asm
+               those methods are used by the Karatsuba multiplication algorithm
+               
                
 Version 0.8.4 (2009.05.08):
     * fixed:   UInt::DivInt() didn't check whether the divisor is zero

ttmath/ttmath.h

@@ -96,13 +96,10 @@ namespace ttmath
 			 -2.7 = -3
 	*/
 	template<class ValueType>
-	ValueType Round(const ValueType & x, ErrorCode * err = 0)
+	ValueType Round(const ValueType & x)
 	{
 		ValueType result( x );
-		uint c = result.Round();
+		result.Round();
-
-		if( err )
-			*err = c ? err_overflow : err_ok;
 
 	return result;
 	}
@@ -300,7 +297,7 @@ namespace ttmath
 		(you don't have to call this function) 
 	*/
 	template<class ValueType>
-	uint PrepareSin(ValueType & x, bool & change_sign)
+	void PrepareSin(ValueType & x, bool & change_sign)
 	{
 	ValueType temp;
 
@@ -316,10 +313,12 @@ namespace ttmath
 		// we're reducing the period 2*PI
 		// (for big values there'll always be zero)
 		temp.Set2Pi();
-		
+		if( x > temp )
-		if( x.Mod(temp) )
+		{
-			return 1;
+			x.Div( temp );
-		
+			x.RemainFraction();
+			x.Mul( temp );
+		}
 	
 		// we're setting 'x' as being in the range of <0, 0.5PI>
 
@@ -340,8 +339,6 @@ namespace ttmath
 			x.Sub( temp );
 			x = temp - x;
 		}
-
-	return 0;
 	}
 
 	
@@ -428,7 +425,7 @@ namespace ttmath
 			if( c )
 				// Sin is from <-1,1> and cannot make an overflow
 				// but the carry can be from the Taylor series
-				// (then we only break our calculations)
+				// (then we only breaks our calculations)
 				break;
 
 			if( addition )
@@ -460,28 +457,15 @@ namespace ttmath
 		this function calculates the Sine
 	*/
 	template<class ValueType>
-	ValueType Sin(ValueType x, ErrorCode * err = 0)
+	ValueType Sin(ValueType x)
 	{
 	using namespace auxiliaryfunctions;
 
-	ValueType one, result;
+	ValueType one;
 	bool change_sign;	
 	
-		if( err )
+		PrepareSin( x, change_sign );
-			*err = err_ok;
+		ValueType result = Sin0pi05( x );
-
-		if( PrepareSin( x, change_sign ) )
-		{
-			// x is too big, we cannnot reduce the 2*PI period
-			// prior to version 0.8.5 the result was zero
-
-			if( err )
-				*err = err_overflow; // maybe another error code?
-
-		return result; // result we remain as undefined
-		}
-
-		result = Sin0pi05( x );
 	
 		one.SetOne();
 
@@ -506,22 +490,14 @@ namespace ttmath
 		we're using the formula cos(x) = sin(x + PI/2)
 	*/
 	template<class ValueType>
-	ValueType Cos(ValueType x, ErrorCode * err = 0)
+	ValueType Cos(ValueType x)
 	{
 		ValueType pi05;
 		pi05.Set05Pi();
 
-		uint c = x.Add( pi05 );
+		x.Add( pi05 );
 	
-		if( c )
+	return Sin(x);
-		{
-			if( err )
-				*err = err_overflow;
-	
-		return ValueType(); // result is undefined
-		}
-
-	return Sin(x, err);
 	}
 	
 
@@ -538,10 +514,7 @@ namespace ttmath
 	template<class ValueType>
 	ValueType Tan(const ValueType & x, ErrorCode * err = 0)
 	{
-		ValueType result = Cos(x, err);
+		ValueType result = Cos(x);
-		
-		if( err && *err != err_ok )
-			return result;
 
 		if( result.IsZero() )
 		{
@@ -551,7 +524,10 @@ namespace ttmath
 		return result;
 		}
 
-	return Sin(x, err) / result;
+		if( err )
+			*err = err_ok;
+
+	return Sin(x) / result;
 	}
 
 
@@ -578,10 +554,7 @@ namespace ttmath
 	template<class ValueType>
 	ValueType Cot(const ValueType & x, ErrorCode * err = 0)
 	{
-		ValueType result = Sin(x, err);
+		ValueType result = Sin(x);
-
-		if( err && *err != err_ok )
-			return result;
 
 		if( result.IsZero() )
 		{
@@ -591,7 +564,10 @@ namespace ttmath
 		return result;
 		}
 	
-	return Cos(x, err) / result;
+		if( err )
+			*err = err_ok;
+
+	return Cos(x) / result;
 	}
 
 
@@ -2035,18 +2011,15 @@ namespace ttmath
 		the remainder from a division
 
 		e.g.
-		mod( 12.6 ;  3) =  0.6   because 12.6  = 3*4 + 0.6
+		mod( 12.6 ;  3) =  0.6   because 12.6 = 3*4 + 0.6
-		mod(-12.6 ;  3) = -0.6   bacause -12.6 = 3*(-4) + (-0.6)
+		mod(-12.6 ;  3) = -0.6
 		mod( 12.6 ; -3) =  0.6
 		mod(-12.6 ; -3) = -0.6
 	*/
 	template<class ValueType>
-	ValueType Mod(ValueType a, const ValueType & b, ErrorCode * err = 0)
+	ValueType Mod(ValueType a, const ValueType & b)
 	{
-		uint c = a.Mod(b);
+		a.Mod(b);
-
-		if( err )
-			*err = c ? err_overflow : err_ok;
 
 	return a;
 	}

ttmath/ttmathbig.h

@@ -954,7 +954,7 @@ public:
 
 	UInt<man*2> man1;
 	UInt<man*2> man2;
-	uint i,c = 0;
+	uint i,c;
 		
 		if( ss2.IsZero() )
 		{
@@ -978,9 +978,7 @@ public:
 
 		i = man1.CompensationToLeft();
 
-		if( i )
+		c  = exponent.Sub(i);
-			c += exponent.Sub(i);
-
 		c += exponent.Sub(ss2.exponent);
 		
 		for(i=0 ; i<man ; ++i)
@@ -1001,8 +999,8 @@ public:
 		the remainder from a division
 
 		e.g.
-		 12.6 mod  3 =  0.6   because  12.6 = 3*4 + 0.6
+		 12.6 mod  3 =  0.6   because 12.6 = 3*4 + 0.6
-		-12.6 mod  3 = -0.6   bacause -12.6 = 3*(-4) + (-0.6)
+		-12.6 mod  3 = -0.6
 		 12.6 mod -3 =  0.6
 		-12.6 mod -3 = -0.6
 
@@ -1015,25 +1013,18 @@ public:
 
 	uint c = 0;
 
-		if( !SmallerWithoutSignThan(ss2) )
+		Big<exp, man> temp(*this);
-		{
-			Big<exp, man> temp(*this);
 
-			c = temp.Div(ss2);
+		c += temp.Div(ss2);
-			temp.SkipFraction();
+		temp.SkipFraction();
-			c += temp.Mul(ss2);
+		c += temp.Mul(ss2);
-			c += Sub(temp);
+		c += Sub(temp);
-
-			if( !SmallerWithoutSignThan( ss2 ) )
-				c += 1;
-		}
 
 	return (c==0)? 0 : 1;
 	}
 
 
 
-
 	/*!
 		power this = this ^ pow
 		(pow without a sign)
@@ -1885,7 +1876,7 @@ public:
 		// error but I leave it at the moment as is
 		TTMATH_ASSERT( sizeof(double) == 8 )
 
-		// I am not sure what will be on a platform which has 
+		// I am not sure what will be on a plaltform which has 
 		// a different endianness... but we use this library only
 		// on x86 and amd (intel) 64 bits (as there's a lot of assembler code)
 		union 
@@ -3443,7 +3434,7 @@ private:
 	*/
 	int FromString_ReadScientificIfExists(const tchar_t * & source)
 	{
-		uint c = 0;
+	int c = 0;
 
 		bool scientific_read = false;
 		const tchar_t * before_scientific = source;

ttmath/ttmathparser.h

@@ -708,11 +708,7 @@ void Sin(int sindex, int amount_of_args, ValueType & result)
 	if( amount_of_args != 1 )
 		Error( err_improper_amount_of_arguments );
 
-	ErrorCode err;
+	result = ttmath::Sin( ConvertAngleToRad(stack[sindex].value) );
-	result = ttmath::Sin( ConvertAngleToRad(stack[sindex].value), &err );
-
-	if(err != err_ok)
-		Error( err );
 }
 
 void Cos(int sindex, int amount_of_args, ValueType & result)
@@ -720,11 +716,7 @@ void Cos(int sindex, int amount_of_args, ValueType & result)
 	if( amount_of_args != 1 )
 		Error( err_improper_amount_of_arguments );
 
-	ErrorCode err;
+	result = ttmath::Cos( ConvertAngleToRad(stack[sindex].value) );
-	result = ttmath::Cos( ConvertAngleToRad(stack[sindex].value), &err );
-
-	if(err != err_ok)
-		Error( err );
 }
 
 void Tan(int sindex, int amount_of_args, ValueType & result)
@@ -765,10 +757,7 @@ void Round(int sindex, int amount_of_args, ValueType & result)
 	if( amount_of_args != 1 )
 		Error( err_improper_amount_of_arguments );
 
-	result = stack[sindex].value;
+	result = ttmath::Round(stack[sindex].value);
-
-	if( result.Round() )
-		Error( err_overflow );
 }
 
 

ttmath/ttmathtypes.h

@@ -64,7 +64,7 @@
 */
 #define TTMATH_MAJOR_VER		0
 #define TTMATH_MINOR_VER		8
-#define TTMATH_REVISION_VER		5
+#define TTMATH_REVISION_VER		4
 #define TTMATH_PRERELEASE_VER	1
 
 
@@ -120,6 +120,7 @@ namespace ttmath
 	typedef unsigned int uint;
 	typedef signed   int sint;
 
+
 	/*!
 		this type is twice bigger than uint
 		(64bit on a 32bit platforms)
@@ -128,11 +129,8 @@ namespace ttmath
 		but it is defined in C99 and in upcoming C++0x /3.9.1 (2)/ and many compilers support it
 
 		this type is used in UInt::MulTwoWords and UInt::DivTwoWords when macro TTMATH_NOASM is defined
-		but only on a 32bit platform
 	*/
-	#ifdef TTMATH_NOASM
+	typedef unsigned long long int ulint;
-		typedef unsigned long long int ulint;
-	#endif
 
 	/*!
 		the mask for the highest bit in the unsigned 32bit word (2^31)

ttmath/ttmathuint.h

@@ -143,7 +143,10 @@ public:
 	*/
 	void SetZero()
 	{
-		memset(table,0,sizeof(table));
+		// in the future here can be 'memset'
+
+		for(uint i=0 ; i<value_size ; ++i)
+			table[i] = 0;
 
 		TTMATH_LOG("UInt::SetZero")
 	}
@@ -2069,7 +2072,8 @@ public:
 	*/
 	void FromUInt(uint value)
 	{
-		memset(table,0,sizeof(table));
+		for(uint i=1 ; i<value_size ; ++i)
+			table[i] = 0;
 
 		table[0] = value;
 
@@ -2901,6 +2905,7 @@ public:
 
 
 private:
+public: // !!! chwilowo public
 	uint Rcl2_one(uint c);
 	uint Rcr2_one(uint c);
 	uint Rcl2(uint bits, uint c);

ttmath/ttmathuint_noasm.h

@@ -116,7 +116,7 @@ namespace ttmath
 			table[1] = 30 + 2;
 			table[2] = 5;
 
-		of course if there was a carry from table[2] it would be returned
+		of course if there was a carry from table[3] it would be returned
 	*/
 	template<uint value_size>
 	uint UInt<value_size>::AddInt(uint value, uint index)
@@ -175,7 +175,7 @@ namespace ttmath
 	{
 	uint i, c;
 
-		TTMATH_ASSERT( index < value_size - 1 )
+		TTMATH_ASSERT( index < value_size )
 
 
 		c = AddTwoWords(table[index],   x1, 0, &table[index]);
@@ -255,7 +255,7 @@ namespace ttmath
 			table[1] = 30 - 2;
 			table[2] = 5;
 
-		of course if there was a carry from table[2] it would be returned
+		of course if there was a carry from table[3] it would be returned
 	*/
 	template<uint value_size>
 	uint UInt<value_size>::SubInt(uint value, uint index)
@@ -473,8 +473,8 @@ namespace ttmath
 
 		uint mask = 1;
 
-		if( bit > 1 )
+		while( bit-- > 0 )
-			mask = mask << bit;
+			mask = mask << 1;
 
 		uint last = value & mask;
 		value     = value | mask;
@@ -601,6 +601,7 @@ namespace ttmath
 	*/
 	
 
+	// !! maybe returns something? a carry? or when c is zero?
 	/*!
 		this method calculates 64bits word a:b / 32bits c (a higher, b lower word)
 		r = a:b / c and rest - remainder
@@ -647,6 +648,10 @@ namespace ttmath
 		{
 			*r    = b / c;
 			*rest = b % c;
+
+#ifdef TTMATH_WARTOWNIK
+			++tester_wartownik1; // !!!!! skasowac
+#endif
 		}
 		else
 		if( c_.u_.high == 0 )
@@ -669,6 +674,10 @@ namespace ttmath
 			*rest        = temp2.u % c;
 
 			*r = res_.u;
+#ifdef TTMATH_WARTOWNIK
+			++tester_wartownik2; // !!!!! skasowac
+#endif
+
 		}
 		else
 		{
@@ -681,13 +690,6 @@ namespace ttmath
 
 #ifdef TTMATH_PLATFORM64
 
-
-	/*!
-		this method is available only on 64bit platforms
-		
-		the same algorithm like the third division algorithm in ttmathuint.h
-		but now with the radix=2^32
-	*/
 	template<uint value_size>
 	void UInt<value_size>::DivTwoWords2(uint a, uint b, uint c, uint * r, uint * rest)
 	{
@@ -702,6 +704,7 @@ namespace ttmath
 		c_.u  = c;
 
 		// normalizing
+		// a0 will actually not be used
 		uint d = DivTwoWordsNormalize(a_, b_, c_);
 
 		// loop from j=1 to j=0
@@ -745,7 +748,12 @@ namespace ttmath
 			a_.u = a_.u << 1; // carry bits from 'a' are simply skipped 
 
 			if( bc )
+			{
 				a_.u = a_.u | 1;
+	#ifdef TTMATH_WARTOWNIK
+				++tester_wartownik3; // !!!!! skasowac
+	#endif
+			}
 		}
 
 	return d;
@@ -794,11 +802,23 @@ namespace ttmath
 
 			if( decrease )
 			{
+				#ifdef TTMATH_WARTOWNIK
+				++tester_wartownik4; // !!!!! skasowac
+				#endif
+
 				--qp_.u;
 				rp_.u += v_.u_.high;
 
 				if( rp_.u_.high == 0 ) 
+				{
 					next_test = true;
+
+					#ifdef TTMATH_WARTOWNIK
+					++tester_wartownik5; // !!!!! skasowac
+					#endif
+				}
+
+				
 			}
 		}
 		while( next_test );
@@ -829,12 +849,20 @@ namespace ttmath
 		temp_.u_.low  = u_.u_.high;
 		c = SubTwoWords(temp_.u, res_high, c, &sub_res_high_.u);
 
+#ifdef TTMATH_WARTOWNIK
+		++tester_wartownik6; // !!!!! skasowac
+#endif
+
 		if( c )
 		{
 			--q;
 
 			c = AddTwoWords(sub_res_low_.u, v_.u, 0, &sub_res_low_.u);
 			AddTwoWords(sub_res_high_.u, 0, c, &sub_res_high_.u);
+
+			#ifdef TTMATH_WARTOWNIK
+			++tester_wartownik7; // !!!!! skasowac
+			#endif
 		}
 
 		u_.u_.high = sub_res_high_.u_.low;

ttmath/ttmathuint_x86_64.h

@@ -158,7 +158,7 @@ namespace ttmath
 			table[1] = 30 + 2;
 			table[2] = 5;
 
-		of course if there was a carry from table[2] it would be returned
+		of course if there was a carry from table[3] it would be returned
 	*/
 	template<uint value_size>
 	uint UInt<value_size>::AddInt(uint value, uint index)
@@ -374,7 +374,7 @@ namespace ttmath
 			table[1] = 30 - 2;
 			table[2] = 5;
 
-		of course if there was a carry from table[2] it would be returned
+		of course if there was a carry from table[3] it would be returned
 	*/
 	template<uint value_size>
 	uint UInt<value_size>::SubInt(uint value, uint index)
@@ -695,7 +695,7 @@ namespace ttmath
 
 
 	/*
-		this method returns the number of the highest set bit in one 64-bit word
+		this method returns the number of the highest set bit in one 32-bit word
 		if the 'x' is zero this method returns '-1'
 
 		***this method is created only on a 64bit platform***
@@ -800,17 +800,18 @@ namespace ttmath
 
 
 	/*!
-		multiplication: result_high:result_low = a * b
+		multiplication: result2:result1 = a * b
-		result_high - higher word of the result
+		result2 - higher word
-		result_low  - lower word of the result
+		result1 - lower word of the result
 	
 		this methos never returns a carry
-		this method is used in the second version of the multiplication algorithms
 
 		***this method is created only on a 64bit platform***
+
+		it is an auxiliary method for version two of the multiplication algorithm
 	*/
 	template<uint value_size>
-	void UInt<value_size>::MulTwoWords(uint a, uint b, uint * result_high, uint * result_low)
+	void UInt<value_size>::MulTwoWords(uint a, uint b, uint * result2, uint * result1)
 	{
 	/*
 		we must use these temporary variables in order to inform the compilator
@@ -843,8 +844,8 @@ namespace ttmath
 		#endif
 
 
-		*result_low  = result1_;
+		*result1 = result1_;
-		*result_high = result2_;
+		*result2 = result2_;
 	}
 
 

ttmath/ttmathuint_x86_amd64_msvc.asm

@@ -114,9 +114,9 @@ loop1:
 		mov		r9, 0				; set to 0 -> cy still set!
 		dec		rdx
 		jnz		loop1
+		jc		return_1			; most of the times, there will be NO carry (I hope)
 
 done:
-		jc		return_1	; most of the times, there will be NO carry (I hope)
 		xor		rax, rax
 		ret
 	
@@ -184,9 +184,9 @@ loop1:
 		mov		r9, 1
 		dec		rdx
 		jnz		loop1
+		jc		return_1	; most of the times, there will be NO carry (I hope)
 
 done:
-		jc		return_1	; most of the times, there will be NO carry (I hope)
 		xor		rax, rax
 		ret