Gmprat logarithms now O(1) not O(N)

... Where the "1" is determined log(3), don't get too excited

So we can finally actually print values outside DBL_MIN->DBL_MAX
Now to go through and remove all the cast to doubles...

diff --git a/src/debugscript.cpp b/src/debugscript.cpp
index 9bb5ea6..3d13245 100644 (file)
--- a/src/debugscript.cpp
+++ b/src/debugscript.cpp
@@ -113,6 +113,12 @@ void DebugScript::ParseAction()
        {
                currentAction.type = AT_RecordPerformance;
        }
+       else if (actionType == "debugfont")
+       {
+               currentAction.type = AT_DebugFont;
+               inp >> currentAction.textargs;
+       }
+
 }
 
 bool DebugScript::Execute(View *view, Screen *scr)
@@ -207,6 +213,10 @@ bool DebugScript::Execute(View *view, Screen *scr)
        case AT_RecordPerformance:
                PrintPerformance(view, scr);
                break;
+       case AT_DebugFont:
+               scr->ShowDebugFont(currentAction.textargs == "1" || currentAction.textargs == "on");
+               currentAction.loops = 1;
+               break;
        default:
                Fatal("Unknown script command in queue.");
        }

diff --git a/src/debugscript.h b/src/debugscript.h
index 9895ddc..d39807d 100644 (file)
--- a/src/debugscript.h
+++ b/src/debugscript.h
@@ -40,6 +40,7 @@ private:
                AT_ClearPerformance,
                AT_PrintPerformance,
                AT_RecordPerformance,
+               AT_DebugFont,
                AT_Quit
        };
 

diff --git a/src/eye_of_the_rabbit.script b/src/eye_of_the_rabbit.script
index af3116d..b4b22c6 100644 (file)
--- a/src/eye_of_the_rabbit.script
+++ b/src/eye_of_the_rabbit.script
@@ -1,6 +1,7 @@
 # Test how well document scales back to original...
 gpu
 lazy
+debugfont off
 clearperf
 label start
 printperf

diff --git a/src/gmprat.h b/src/gmprat.h
index 6782734..3e275d2 100644 (file)
--- a/src/gmprat.h
+++ b/src/gmprat.h
@@ -8,6 +8,7 @@
 
 #include <gmp.h>
 #include <string>
+#include <climits>
 
 class Gmprat
 {
@@ -45,39 +46,36 @@ class Gmprat
                        return s.str();
                }
                
-               double Log10() const
+               double LogE() const
                {
-                       mpz_t num; mpz_init(num); mpq_get_num(num, m_op);
-                       mpz_t den; mpz_init(den); mpq_get_den(den, m_op);
+                       // log(a/b) = log(a) - log(b)
+                       // And if a is represented in base B as:
+                       // a = a_N B^N + a_{N-1} B^{N-1} + ... + a_0
+                       // => log(a) \approx log(a_N B^N)
+                       // = log(a_N) + N log(B)
+                       // where B is the base; ie number of values representable with one digit, ie: ULONG_MAX
                        
-                       double lognum = 0;
-                       double logden = 0;
-                       while (mpz_sizeinbase(num, 10) > 10)
-                       {
-
-                               mpz_div_ui(num, num, 1e10);
-                               lognum += 10;
-                       }
-                       uint64_t n = mpz_get_ui(num);
-                       if (n == 0)
-                       {
-                               return -INFINITY;
-                       }
-                       lognum += log(n)/log(10.0);
-                       //Debug("%lu", mpz_get_ui(den));
-                       while (mpz_sizeinbase(den, 10) > 10)
+                       static double logB = log(ULONG_MAX); // compiler should optimise this anyway?
+                       
+                       // Undefined logs (should probably return NAN if -ve, not -INFINITY, but meh)
+                       if (mpz_get_ui(mpq_numref(m_op)) == 0 || mpz_sgn(mpq_numref(m_op)) < 0)
+                               return -INFINITY;                               
+       
+                       // Log of numerator
+                       double lognum = log(mpq_numref(m_op)->_mp_d[abs(mpq_numref(m_op)->_mp_size) - 1]);
+                       lognum += (abs(mpq_numref(m_op)->_mp_size)-1) * logB;
+                       
+                       // Subtract log of denominator, if it exists
+                       // Note that denominator is not explicitly set to 1, this caused a lot of headache
+                       if (abs(mpq_denref(m_op)->_mp_size) > 0)
                        {
-                               mpz_div_ui(den, den, 1e10);
-                               logden += 10;
+                               lognum -= log(mpq_denref(m_op)->_mp_d[abs(mpq_denref(m_op)->_mp_size)-1]);
+                               lognum -= (abs(mpq_denref(m_op)->_mp_size)-1) * logB;
                        }
-                       uint64_t d = mpz_get_ui(den);
-                       // if d is zero, its been rounded down we hope
-                       if (d != 0)
-                               logden += log(d)/log(10.0);
-                       
-                       return (lognum - logden);
+                       return lognum;
                }
                
+               double Log10() const {return LogE() / log(10.0);}
 
                bool Negative() const {return (mpz_sgn(mpq_numref(m_op)) < 0);}