[ViewVC] Diff of: cvs/deliantra/server/include/util.h

Comparing deliantra/server/include/util.h (file contents):
Revision 1.71 by root, Sun Apr 20 06:20:38 2008 UTC vs.
Revision 1.97 by root, Fri Mar 26 01:04:44 2010 UTC

 /*
  * This file is part of Deliantra, the Roguelike Realtime MMORPG.
  *
- * Copyright (©) 2005,2006,2007 Marc Alexander Lehmann / Robin Redeker / the Deliantra team
+ * Copyright (©) 2005,2006,2007,2008,2009,2010 Marc Alexander Lehmann / Robin Redeker / the Deliantra team
  *
- * Deliantra is free software: you can redistribute it and/or modify
+ * Deliantra is free software: you can redistribute it and/or modify it under
- * it under the terms of the GNU General Public License as published by
+ * the terms of the Affero GNU General Public License as published by the
- * the Free Software Foundation, either version 3 of the License, or
+ * Free Software Foundation, either version 3 of the License, or (at your
- * (at your option) any later version.
+ * option) any later version.
  *
  * This program is distributed in the hope that it will be useful,
  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
  * GNU General Public License for more details.
  *
- * You should have received a copy of the GNU General Public License
+ * You should have received a copy of the Affero GNU General Public License
- * along with this program. If not, see <http://www.gnu.org/licenses/>.
+ * and the GNU General Public License along with this program. If not, see
+ * <http://www.gnu.org/licenses/>.
  *
  * The authors can be reached via e-mail to <support@deliantra.net>
  */
 #ifndef UTIL_H__
 #define UTIL_H__
+#include <compiler.h>
 #define DEBUG_POISON 0x00 // poison memory before freeing it if != 0
 #define DEBUG_SALLOC  0   // add a debug wrapper around all sallocs
 #define PREFER_MALLOC 0   // use malloc and not the slice allocator
-#if __GNUC__ >= 3
-# define is_constant(c)             __builtin_constant_p (c)
-# define expect(expr,value)         __builtin_expect ((expr),(value))
-# define prefetch(addr,rw,locality) __builtin_prefetch (addr, rw, locality)
-#else
-# define is_constant(c)             0
-# define expect(expr,value)         (expr)
-# define prefetch(addr,rw,locality)
-#endif
-#if __GNUC__ < 4 || (__GNUC__ == 4 || __GNUC_MINOR__ < 4)
-# define decltype(x) typeof(x)
-#endif
-// put into ifs if you are very sure that the expression
-// is mostly true or mosty false. note that these return
-// booleans, not the expression.
-#define expect_false(expr) expect ((expr) != 0, 0)
-#define expect_true(expr)  expect ((expr) != 0, 1)
 #include <pthread.h>
 #include <cstddef>
 #include <cmath>
 #endif
 // use C0X decltype for auto declarations until ISO C++ sanctifies them (if ever)
 #define auto(var,expr) decltype(expr) var = (expr)
-// very ugly macro that basicaly declares and initialises a variable
+// very ugly macro that basically declares and initialises a variable
 // that is in scope for the next statement only
 // works only for stuff that can be assigned 0 and converts to false
 // (note: works great for pointers)
 // most ugly macro I ever wrote
 #define statementvar(type, name, value) if (type name = 0) { } else if (((name) = (value)), 1)
 // as a is often a constant while b is the variable. it is still a bug, though.
 template<typename T, typename U> static inline T min (T a, U b) { return (U)a < b ? (U)a : b; }
 template<typename T, typename U> static inline T max (T a, U b) { return (U)a > b ? (U)a : b; }
 template<typename T, typename U, typename V> static inline T clamp (T v, U a, V b) { return v < (T)a ? (T)a : v >(T)b ? (T)b : v; }
+template<typename T, typename U> static inline void min_it (T &v, U m) { v = min (v, (T)m); }
+template<typename T, typename U> static inline void max_it (T &v, U m) { v = max (v, (T)m); }
+template<typename T, typename U, typename V> static inline void clamp_it (T &v, U a, V b) { v = clamp (v, (T)a, (T)b); }
 template<typename T, typename U> static inline void swap (T& a, U& b) { T t=a; a=(T)b; b=(U)t; }
 template<typename T, typename U, typename V> static inline T min (T a, U b, V c) { return min (a, min (b, c)); }
 template<typename T, typename U, typename V> static inline T max (T a, U b, V c) { return max (a, max (b, c)); }
+// sign returns -1 or +1
+template<typename T>
+static inline T sign (T v) { return v < 0 ? -1 : +1; }
+// relies on 2c representation
+template<>
+inline sint8 sign (sint8 v) { return 1 - (sint8 (uint8 (v) >> 7) * 2); }
+// sign0 returns -1, 0 or +1
+template<typename T>
+static inline T sign0 (T v) { return v ? sign (v) : 0; }
+// div* only work correctly for div > 0
+// div, with correct rounding (< 0.5 downwards, >=0.5 upwards)
+template<typename T> static inline T div    (T val, T div)
+{
+  return expect_false (val < 0) ? - ((-val + (div - 1) / 2) / div) : (val + div / 2) / div;
+}
+// div, round-up
+template<typename T> static inline T div_ru (T val, T div)
+{
+  return expect_false (val < 0) ? - ((-val                ) / div) : (val + div - 1) / div;
+}
+// div, round-down
+template<typename T> static inline T div_rd (T val, T div)
+{
+  return expect_false (val < 0) ? - ((-val + (div - 1)    ) / div) : (val          ) / div;
+}
+// lerp* only work correctly for min_in < max_in
+// Linear intERPolate, scales val from min_in..max_in to min_out..max_out
 template<typename T>
 static inline T
 lerp (T val, T min_in, T max_in, T min_out, T max_out)
 {
-  return (val - min_in) * (max_out - min_out) / (max_in - min_in) + min_out;
+  return min_out + div   <T> ((val - min_in) * (max_out - min_out), max_in  - min_in);
+}
+// lerp, round-down
+template<typename T>
+static inline T
+lerp_rd (T val, T min_in, T max_in, T min_out, T max_out)
+{
+  return min_out + div_rd<T> ((val - min_in) * (max_out - min_out), max_in  - min_in);
+}
+// lerp, round-up
+template<typename T>
+static inline T
+lerp_ru (T val, T min_in, T max_in, T min_out, T max_out)
+{
+  return min_out + div_ru<T> ((val - min_in) * (max_out - min_out), max_in  - min_in);
 }
 // lots of stuff taken from FXT
 /* Rotate right. This is used in various places for checksumming */
   int32_t d = b - a;
   d &= d >> 31;
   return b - d;
 }
-// this is much faster than crossfires original algorithm
+// this is much faster than crossfire's original algorithm
 // on modern cpus
 inline int
 isqrt (int n)
 {
   return (int)sqrtf ((float)n);
+}
+// this is kind of like the ^^ operator, if it would exist, without sequence point.
+// more handy than it looks like, due to the implicit !! done on its arguments
+inline bool
+logical_xor (bool a, bool b)
+{
+  return a != b;
+}
+inline bool
+logical_implies (bool a, bool b)
+{
+  return a <= b;
 }
 // this is only twice as fast as naive sqrtf (dx*dy+dy*dy)
 #if 0
 // and has a max. error of 6 in the range -100..+100.
 absdir (int d)
 {
   return ((d - 1) & 7) + 1;
 }
+// avoid ctz name because netbsd or freebsd spams it's namespace with it
+#if GCC_VERSION(3,4)
+static inline int least_significant_bit (uint32_t x)
+{
+  return __builtin_ctz (x);
+}
+#else
+int least_significant_bit (uint32_t x);
+#endif
+#define for_all_bits_sparse_32(mask, idxvar)      \
+  for (uint32_t idxvar, mask_ = mask;   \
+       mask_ && ((idxvar = least_significant_bit (mask_)), mask_ &= ~(1 << idxvar), 1);)
 extern ssize_t slice_alloc; // statistics
 void *salloc_ (int n)            throw (std::bad_alloc);
 void *salloc_ (int n, void *src) throw (std::bad_alloc);
       g_slice_free1 (n * sizeof (T), (void *)ptr);
       assert (slice_alloc >= 0);//D
     }
 }
+// nulls the pointer
+template<typename T>
+inline void sfree0 (T *&ptr, int n = 1) throw ()
+{
+  sfree<T> (ptr, n);
+  ptr = 0;
+}
 // makes dynamically allocated objects zero-initialised
 struct zero_initialised
 {
   void *operator new (size_t s, void *p)
   {
   }
   void *operator new[] (size_t s)
   {
     return salloc0<char> (s);
+  }
+  void operator delete (void *p, size_t s)
+  {
+    sfree ((char *)p, s);
+  }
+  void operator delete[] (void *p, size_t s)
+  {
+    sfree ((char *)p, s);
+  }
+};
+// makes dynamically allocated objects zero-initialised
+struct slice_allocated
+{
+  void *operator new (size_t s, void *p)
+  {
+    return p;
+  }
+  void *operator new (size_t s)
+  {
+    return salloc<char> (s);
+  }
+  void *operator new[] (size_t s)
+  {
+    return salloc<char> (s);
   }
   void operator delete (void *p, size_t s)
   {
     sfree ((char *)p, s);
 // P. L'Ecuyer, “Maximally Equidistributed Combined Tausworthe Generators”, Mathematics of Computation, 65, 213 (1996), 203–213.
 // http://www.iro.umontreal.ca/~lecuyer/myftp/papers/tausme.ps
 // http://www.iro.umontreal.ca/~lecuyer/myftp/papers/tausme2.ps
 struct tausworthe_random_generator
 {
-  // generator
   uint32_t state [4];
   void operator =(const tausworthe_random_generator &src)
   {
     state [0] = src.state [0];
     state [3] = src.state [3];
   }
   void seed (uint32_t seed);
   uint32_t next ();
+};
-  // uniform distribution
+// Xorshift RNGs, George Marsaglia
+// http://www.jstatsoft.org/v08/i14/paper
+// this one is about 40% faster than the tausworthe one above (i.e. not much),
+// despite the inlining, and has the issue of only creating 2**32-1 numbers.
+// see also http://www.iro.umontreal.ca/~lecuyer/myftp/papers/xorshift.pdf
+struct xorshift_random_generator
+{
+  uint32_t x, y;
+  void operator =(const xorshift_random_generator &src)
+  {
+    x = src.x;
+    y = src.y;
+  }
+  void seed (uint32_t seed)
+  {
+    x = seed;
+    y = seed * 69069U;
+  }
+  uint32_t next ()
+  {
+    uint32_t t = x ^ (x << 10);
+    x = y;
+    y = y ^ (y >> 13) ^ t ^ (t >> 10);
+    return y;
+  }
+};
+template<class generator>
+struct random_number_generator : generator
+{
+  // uniform distribution, 0 .. max (0, num - 1)
   uint32_t operator ()(uint32_t num)
   {
-    return is_constant (num)
+    return !is_constant (num) ? get_range (num)                        // non-constant
-             ? (next () * (uint64_t)num) >> 32U
+         : num & (num - 1)    ? (this->next () * (uint64_t)num) >> 32U // constant, non-power-of-two
-             : get_range (num);
+         :                      this->next () & (num - 1);             // constant, power-of-two
   }
   // return a number within (min .. max)
   int operator () (int r_min, int r_max)
   {
 protected:
   uint32_t get_range (uint32_t r_max);
   int get_range (int r_min, int r_max);
 };
-typedef tausworthe_random_generator rand_gen;
+typedef random_number_generator<tausworthe_random_generator> rand_gen;
-extern rand_gen rndm;
+extern rand_gen rndm, rmg_rndm;
 INTERFACE_CLASS (attachable)
 struct refcnt_base
 {
   typedef int refcnt_t;
 typedef refptr<object>    object_ptr;
 typedef refptr<archetype> arch_ptr;
 typedef refptr<client>    client_ptr;
 typedef refptr<player>    player_ptr;
+#define STRHSH_NULL 2166136261
+static inline uint32_t
+strhsh (const char *s)
+{
+  // use FNV-1a hash (http://isthe.com/chongo/tech/comp/fnv/)
+  // it is about twice as fast as the one-at-a-time one,
+  // with good distribution.
+  // FNV-1a is faster on many cpus because the multiplication
+  // runs concurrently with the looping logic.
+  uint32_t hash = STRHSH_NULL;
+  while (*s)
+    hash = (hash ^ *s++) * 16777619;
+  return hash;
+}
+static inline uint32_t
+memhsh (const char *s, size_t len)
+{
+  uint32_t hash = STRHSH_NULL;
+  while (len--)
+    hash = (hash ^ *s++) * 16777619;
+  return hash;
+}
 struct str_hash
 {
   std::size_t operator ()(const char *s) const
   {
-    unsigned long hash = 0;
-    /* use the one-at-a-time hash function, which supposedly is
-     * better than the djb2-like one used by perl5.005, but
-     * certainly is better then the bug used here before.
-     * see http://burtleburtle.net/bob/hash/doobs.html
-     */
-    while (*s)
-      {
-        hash += *s++;
-        hash += hash << 10;
-        hash ^= hash >>  6;
-      }
-    hash += hash <<  3;
-    hash ^= hash >> 11;
-    hash += hash << 15;
-    return hash;
+    return strhsh (s);
+  }
+  std::size_t operator ()(const shstr &s) const
+  {
+    return strhsh (s);
   }
 };
 struct str_equal
 {
     erase (&obj);
   }
 };
 // basically does what strncpy should do, but appends "..." to strings exceeding length
+// returns the number of bytes actually used (including \0)
-void assign (char *dst, const char *src, int maxlen);
+int assign (char *dst, const char *src, int maxsize);
 // type-safe version of assign
 template<int N>
-inline void assign (char (&dst)[N], const char *src)
+inline int assign (char (&dst)[N], const char *src)
 {
-  assign ((char *)&dst, src, N);
+  return assign ((char *)&dst, src, N);
 }
 typedef double tstamp;
 // return current time as timestamp
 tstamp now ();
 int similar_direction (int a, int b);
-// like sprintf, but returns a "static" buffer
+// like v?sprintf, but returns a "static" buffer
-const char *format (const char *format, ...);
+char *vformat (const char *format, va_list ap);
+char *format (const char *format, ...) attribute ((format (printf, 1, 2)));
+// safety-check player input which will become object->msg
+bool msg_is_safe (const char *msg);
 /////////////////////////////////////////////////////////////////////////////
 // threads, very very thin wrappers around pthreads
 struct thread

Diff Legend

-–
+Removed lines
-+
+Added lines
-<
+Changed lines
->
+Changed lines

Comparing deliantra/server/include/util.h (file contents): Revision 1.71 by root, Sun Apr 20 06:20:38 2008 UTC vs. Revision 1.97 by root, Fri Mar 26 01:04:44 2010 UTC

Diff Legend

Comparing deliantra/server/include/util.h (file contents):
Revision 1.71 by root, Sun Apr 20 06:20:38 2008 UTC vs.
Revision 1.97 by root, Fri Mar 26 01:04:44 2010 UTC