Switched to MPIR for supporting larger decimal expansion on Windows.

Also, added supporting code in ChudnovskyPiBS class for what I thought would be the most efficient way to handle 10^9 decimal places at the time.

Author: Nomi

ChudnovskyPiBS.cpp

@@ -3,9 +3,13 @@
 #include <barrier>
 #include "ChudnovskyPiBS.h"
 #include <sstream>
+#include "MPIR/gmp.h"
+#include "MPIR/gmpxx.h"
 
 #include <Windows.h> //this should be the last import somehow?
 
+#define HUNDRED_MILLION 100000000
+
 //Global variables for configuration:
 const int NUM_THREADS_IN_CPU = std::thread::hardware_concurrency();
 const int NUM_MAIN_WORKER_THREADS = NUM_THREADS_IN_CPU;
@@ -22,19 +26,51 @@ ChudnovskyPiBS::ChudnovskyPiBS(unsigned long _digits)
 		throw _EXCEPTION_; //precision can't be handled by long double.
 	N = (unsigned long)(digits / DIGITS_PER_TERM + 1);
 
-	mpz_ui_pow_ui(one_squared.get_mpz_t(), 10, 2 * digits);
-
-	//one_squared.get_str();
-
-	mpz_class thousandandfive__times__one_squared = 10005 * one_squared;
-	mpz_sqrt(sqrtC.get_mpz_t(), thousandandfive__times__one_squared.get_mpz_t());
-
-	//sqrtC.get_str();
-
 	mpz_ui_pow_ui(intBigC3_OVER_24.get_mpz_t(), C, 3);
 	mpz_class twentyfour = 24;
 	mpz_fdiv_q(intBigC3_OVER_24.get_mpz_t(), intBigC3_OVER_24.get_mpz_t(), twentyfour.get_mpz_t());
 	//intBigC3_OVER_24.get_str();
+	if (digits <= HUNDRED_MILLION)
+	{
+		mpz_ui_pow_ui(one_squared.get_mpz_t(), 10, 2*_digits);
+		mpz_class thousandandfive__times__one_squared = 10005 * one_squared;
+		mpz_sqrt(sqrtC.get_mpz_t(), thousandandfive__times__one_squared.get_mpz_t());
+		//std::cout << "Needed strlen:" << sqrtC.get_str().length() << std::endl;
+	}
+	else
+	{
+		mp_bitcnt_t prec = digits * log2l(10); //or is it 8 * (digits + 4)??
+		SHIFTER.set_prec(prec);
+		sqrtCF.set_prec(prec);
+
+		mpf_class ten = mpf_class(10);
+		mpf_pow_ui(SHIFTER.get_mpf_t(), ten.get_mpf_t(), digits);
+		//std::cout << "Prec:" << SHIFTER.get_prec() << std::endl;
+		mpf_sqrt_ui(sqrtCF.get_mpf_t(),10005);
+		//sqrtCF += 0.1 / SHIFTER;
+		sqrtCF *= SHIFTER;
+		mpf_trunc(sqrtCF.get_mpf_t(), sqrtCF.get_mpf_t());
+		mp_exp_t exp = 0;
+		std::string tempMiddleWare = sqrtCF.get_str(exp, 10, _digits + 4);
+		//std::cout << "Gotten strlen:" << tempMiddleWare.length() << std::endl;
+		//std::cout << "Decimal pt location:" << exp << std::endl;
+		sqrtC = mpz_class(tempMiddleWare);
+		//sqrtC.get_str();
+		/*
+		////ten = SHIFTER;
+		////for (int i = 1; i < 10; i++) //this is basically doing (10^digits/10)^10 (by power rule, it becomes 10^digits)
+		////	SHIFTER *= ten;
+		////std::cout << one_squared.get_str() << std::endl;
+		////one_squared *= one_squared; //this makes it 10^(digits+digits) = 10^(2*digits) which is what we wanted.
+		////mpz_class lol;
+		////mpz_ui_pow_ui(lol.get_mpz_t(), 10, 2*_digits);
+		////if (lol != one_squared)
+		////{
+		////	std::cout << one_squared.get_str() << std::endl;
+		////	std::cout <<  lol.get_str() << std::endl;
+		////}
+		*/
+	}
 }
 
 bsReturn ChudnovskyPiBS::bs(mpz_class a, mpz_class b) //clearly thread safe because nothing from outside the function is written to.
@@ -233,6 +269,34 @@ bsReturn ChudnovskyPiBS::bs_multithreaded_barrier(mpz_class a, mpz_class b, int
 	return result;
 }
 
+void ChudnovskyPiBS::getSqrtC(unsigned long digits)
+{
+	if (digits <= HUNDRED_MILLION)
+	{
+		mpz_ui_pow_ui(one_squared.get_mpz_t(), 10, 2 * digits);
+	}
+	else
+	{
+		mpz_class ten = 10;
+		mpz_pow_ui(one_squared.get_mpz_t(), ten.get_mpz_t(), digits / 10); //10^(digits/10)
+		/*mpz_pow_ui(one_squared.get_mpz_t(), one_squared.get_mpz_t(), 10);*/
+		ten = one_squared;
+		for (int i = 1; i < 10; i++) //this is basically doing (10^digits/10)^10 (by power rule, it becomes 10^digits)
+			one_squared *= ten;
+		//std::cout << one_squared.get_str() << std::endl;
+		one_squared *= one_squared; //this makes it 10^(digits+digits) = 10^(2*digits) which is what we wanted.
+		//mpz_class lol;
+		//mpz_ui_pow_ui(lol.get_mpz_t(), 10, 2*_digits);
+		//if (lol != one_squared)
+		//{
+		//	std::cout << one_squared.get_str() << std::endl;
+		//	std::cout <<  lol.get_str() << std::endl;
+		//}
+	}
+	mpz_class thousandandfive__times__one_squared = 10005 * one_squared;
+	mpz_sqrt(sqrtC.get_mpz_t(), thousandandfive__times__one_squared.get_mpz_t());
+}
+
 int ChudnovskyPiBS::getTotalNumThreadsFromUsefulNumThreads(int usefulThreadCountWanted)
 {
 	if (usefulThreadCountWanted == 2)
@@ -257,7 +321,8 @@ mpz_class ChudnovskyPiBS::calculatePi()
 	bsReturn BSResult = bs_multithreaded(0, N, TOTAL_THREAD_COUNT); //bs_multithreaded(0, N, TOTAL_THREAD_COUNT); //bs_multithreaded_barrier(0, N, TOTAL_THREAD_COUNT, 0); //bs(0, N); //apparently Q and T gotten are wrong.
 	//BSResult.Q.get_str();
 	//BSResult.T.get_str();
-	//return mpz_fdiv_q((Q * 426880 * sqrtC) / T
+
+	/*getSqrtC(digits);*/
 	mpz_class result = (BSResult.Q * 426880 * sqrtC);
 	mpz_fdiv_q(result.get_mpz_t(), result.get_mpz_t(), BSResult.T.get_mpz_t());
 
@@ -266,6 +331,8 @@ mpz_class ChudnovskyPiBS::calculatePi()
 }
 
 
+
+
 /* DEPRECATED/OLD MULTITHREADING SOLUTION
 mpz_class ChudnovskyPiBS::calculatePi()
 {

ChudnovskyPiBS.h

@@ -3,7 +3,8 @@
 #include <math.h>
 #include <stdexcept>
 #include <barrier>
-#include "GMP/gmpxx.h"
+#include "MPIR/gmp.h"
+#include "MPIR/gmpxx.h"
 
 //Helper structs:
 struct bsReturn
@@ -19,16 +20,19 @@ class ChudnovskyPiBS //Calculates Pi using Chudnovsky algorithm with Binary Spli
 private:
 	const long double C = 640320;
 	const long double C3_OVER_24 = powl(C,3)/24;
-	mpz_class intBigC3_OVER_24 = 0;
+	mpz_class intBigC3_OVER_24 = mpz_class(0);
 	const long double DIGITS_PER_TERM = log10l(C3_OVER_24 / 6 / 2 / 6);
 
 	unsigned long N;
 	unsigned long digits;
-	mpz_class one_squared;
-	mpz_class sqrtC;
+	mpz_class one_squared = mpz_class(-1);
+	mpf_class SHIFTER = mpf_class(-1);
 
-	mpz_class oneClass = 1;
-	mpz_ptr one = oneClass.get_mpz_t();
+	mpz_class sqrtC = mpz_class(-1);
+	mpf_class sqrtCF = mpf_class(-1);
+
+	mpz_class digitOneClass = mpz_class(1);
+	mpz_ptr one = digitOneClass.get_mpz_t();
 
 	/// <summary>
 	/// Computes the terms for binary splitting the Chudnovsky infinite series.
@@ -48,7 +52,7 @@ class ChudnovskyPiBS //Calculates Pi using Chudnovsky algorithm with Binary Spli
 	void directlyCompute__P_Q_T__from_A_to_AplusOne(mpz_class& a, mpz_class& Pab, mpz_class& Qab, mpz_class& Tab);
 	bsReturn bs_multithreaded(mpz_class a, mpz_class b, int threadCount);
 	bsReturn bs_multithreaded_barrier(mpz_class a, mpz_class b, int threadCount, int depth); //uses a barrier to wait for all main worker threads to spawn.
-
+	void getSqrtC(unsigned long digits);
 public:
 	/// <summary>
 	/// Constructor for ChudnovskyPiBS Class.