Added some reading material, floating point support.

git-svn-id: http://www.cdkersey.com/harp/harptool@9 0246edb2-e076-4747-b392-db732a341fa2

src/unit_test/float_test.cpp

@@ -0,0 +1,49 @@
+#include "include/harpfloat.h"
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <time.h>
+
+using namespace Harp;
+using namespace std;
+
+static void try_val(double d, Size sz) {
+  Float f(d, sz);
+  Word w(f);
+  double e(Float(w, sz));
+  printf("%uB: %g -> 0x%llx -> %g\n", sz, d, (unsigned long long)w, e);
+}
+
+int main() {
+  /* Loop over available integer sizes. */
+  unsigned randseed(time(NULL));
+  for (unsigned sz = 1; sz <= 8; sz++) {
+    srand(randseed);
+
+    /* First test some random ordinary numbers and their conversions. */
+    for (unsigned i = 0; i < 2; i++) {
+      int n = rand() - RAND_MAX/2;
+      double d = n * 0.0000001;
+      try_val(d, sz);
+    }
+
+    /* Next, let's try +/- infinity. */
+    for (unsigned i = 0; i < 2; i++) {
+      double d(i?HUGE_VAL:-HUGE_VAL);
+      try_val(d, sz);
+    }
+
+    /* Last, let's try some random subnormal numbers and their conversions. */
+    double mote;
+    if (sz < 4) mote = pow(2, int(-14 - (sz*8 - 6)));
+    else if (sz < 8) mote = pow(2, int(-126 - (sz*8 - 9)));
+    else mote = pow(2, int(-1022 - (sz*8 - 12)));
+    for (unsigned i = 0; i < 2; i++) {
+      int n = rand()%256;
+      double d = n * mote;
+      try_val(d, sz);
+    }
+  }
+
+  return 0;
+}