various issues related to encoding larger gaps than anticipated, mostly ...

cpp/Makefile

@@ -0,0 +1,14 @@
+CXXFLAGS=-Wall -MMD -O3 
+CFLAGS=-Wall -MMD -O3
+all: gcs
+
+-include *.d
+
+GCS_OBJECTS = main.o gcs.o md5.o
+
+gcs: $(GCS_OBJECTS)
+	g++ $(LDFLAGS) $(GCS_OBJECTS) -o $@
+clean:
+	rm -f *~ *.o *.d gcs
+
+

cpp/gcs.cpp

@@ -31,10 +31,14 @@
 #include <algorithm>
 #include <assert.h>
 
-#define BITMASK(n)    ((1 << (n)) - 1)
+inline uint64_t BITMASK(unsigned int n) 
+{
+  assert(n < 63);
+  return     ((1ULL << (n)) - 1ULL);
+}
 
 
-static uint32_t gcs_hash(const void *data, int size, int N, int P)
+static uint32_t gcs_hash(const void *data, int size, unsigned int N, unsigned int P)
 {
 	unsigned char digest[16];
 	MD5Context ctx;
@@ -72,7 +76,7 @@ class BitWriter
 	BitWriter(std::ostream &_f)
 		: f(_f), accum(0), n(0) {}
 
-	void write(int nbits, unsigned value)
+	void write(int nbits, uint64_t value)
 	{
 		assert(nbits >= 0);
 		while (nbits)
@@ -136,7 +140,7 @@ class GolombEncoder
 };
 
 
-GCSBuilder::GCSBuilder(int _n, int _p)
+GCSBuilder::GCSBuilder(unsigned int _n, unsigned int _p)
 	: N(_n), P(_p)
 {
 	assert(N <= ~(hash_t)0 / P);

cpp/gcs.h

@@ -34,18 +34,18 @@ typedef uint32_t hash_t;
 
 class GCSBuilder
 {
-	int N, P;
+	unsigned int N, P;
 	std::vector<hash_t> values;
 
 public:
-	GCSBuilder(int N, int P);
+	GCSBuilder(unsigned int N, unsigned int P);
 	void add(const void *data, int size);
 	void finalize(std::ostream &f);
 };	
 
 class GCSQuery
 {
-	int N, P;
+	unsigned int N, P;
 	std::istream &f;
 	uint8_t *gcs;
 	int gcs_len;
@@ -56,6 +56,5 @@ class GCSQuery
 	bool query(const void *data, int size);
 };
 
-
 #endif /* GCS_H */
 

cpp/md5.c

@@ -24,7 +24,6 @@
 
 static void byteReverse(unsigned char *buf, unsigned longs)
 {
-    uint32 t;
     do {
         *(uint32*)buf = O32_HOST_TO_LE(*(uint32*)buf);
         buf += 4;