motiejus
/
turbonss
1
Fork 0
Code Releases Activity

Fixed first_edge initialization bug.

This commit is contained in:
davi 2010-11-08 22:02:18 -02:00
parent cde9f72c9e
commit 676d34073c
8 changed files with 84 additions and 57 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
cxxmph/cmph_hash_map.h
View File

@ -110,8 +110,10 @@ CMPH_METHOD_DECL(void_type, rehash)() {
<< slack_.size() << " keys in slack "
<< values_.size() << " key in total" << std::endl;
slack_type().swap(slack_);
table_.Reset(make_iterator_first(values_.begin()),
make_iterator_first(values_.end()));
bool success = table_.Reset(
make_iterator_first(values_.begin()),
make_iterator_first(values_.end()));
assert(success);
std::vector<value_type> new_values(values_.size());
for (const_iterator it = values_.begin(), end = values_.end();
it != end; ++it) {

1
cxxmph/mphtable.cc
View File

@ -69,6 +69,7 @@ bool MPHTable::GenerateQueue(
// At this point queue head is the number of edges touching at least one
// vertex of degree 1.
cerr << "Queue head " << queue_head << " Queue tail " << queue_tail << endl;
graph->DebugGraph();
while (queue_tail != queue_head) {
cmph_uint32 current_edge = queue[queue_tail++];
graph->RemoveEdge(current_edge);

6
cxxmph/mphtable.h
View File

@ -3,6 +3,7 @@
// Minimal perfect hash abstraction implementing the BDZ algorithm
#include <cassert>
#include <cmath>
#include <unordered_map> // for std::hash
#include <vector>
@ -129,6 +130,11 @@ cmph_uint32 MPHTable::index(const Key& key) const {
h[0] = h[0] % r_;
h[1] = h[1] % r_ + r_;
h[2] = h[2] % r_ + (r_ << 1);
assert(g_.size());
cerr << "g_.size() " << g_.size() << " h0 >> 2 " << (h[0] >> 2) << endl;
assert((h[0] >> 2) <g_.size());
assert((h[1] >> 2) <g_.size());
assert((h[2] >> 2) <g_.size());
cmph_uint32 vertex = h[(get_2bit_value(g_, h[0]) + get_2bit_value(g_, h[1]) + get_2bit_value(g_, h[2])) % 3];
cerr << "Search found vertex " << vertex << endl;
return Rank(vertex);

14
cxxmph/trigraph.cc
View File

@ -9,7 +9,7 @@ using std::endl;
using std::vector;
namespace {
static const cmph_uint8 kInvalidEdge = std::numeric_limits<cmph_uint8>::max();
static const cmph_uint32 kInvalidEdge = std::numeric_limits<cmph_uint32>::max();
}
namespace cxxmph {
@ -65,5 +65,17 @@ void TriGraph::RemoveEdge(cmph_uint32 current_edge) {
--vertex_degree_[vertex];
}
}
void TriGraph::DebugGraph() const {
int i;
for(i = 0; i < edges_.size(); i++){
cerr << i << " " << edges_[i][0] << " " << edges_[i][1] << " " << edges_[i][2]
<< " nexts " << next_edge_[i][0] << " " << next_edge_[i][1] << " " << next_edge_[i][2] << endl;
}
for(i = 0; i < first_edge_.size();i++){
cerr << "first for vertice " <<i << " " << first_edge_[i] << endl;
}
}
} // namespace cxxmph

1
cxxmph/trigraph.h
View File

@ -30,6 +30,7 @@ class TriGraph {
void AddEdge(const Edge& edge);
void RemoveEdge(cmph_uint32 edge_id);
void ExtractEdgesAndClear(std::vector<Edge>* edges);
void DebugGraph() const;
const std::vector<Edge>& edges() const { return edges_; }
const std::vector<cmph_uint8>& vertex_degree() const { return vertex_degree_; }

2
src/Makefile.am
View File

@ -2,7 +2,7 @@ bin_PROGRAMS = cmph
lib_LTLIBRARIES = libcmph.la
include_HEADERS = cmph.h cmph_types.h cmph_time.h chd_ph.h
libcmph_la_SOURCES = hash.h hash.c \
jenkins_hash.h jenkins_hash.c\
jenkins_hash.h jenkins_hash.c MurmurHash2.h\
hash_state.h debug.h \
vstack.h vstack.c vqueue.h vqueue.c\
graph.h graph.c bitbool.h \

1
src/bdz.c
View File

@ -178,6 +178,7 @@ static int bdz_generate_queue(cmph_uint32 nedges, cmph_uint32 nvertices, bdz_que
};
};
DEBUGP("Queue head %d Queue tail %d\n", queue_head, queue_tail);
bdz_dump_graph(graph3,graph3->nedges,graph3->nedges+graph3->nedges/4);
while(queue_tail!=queue_head){
curr_edge=queue[queue_tail++];
bdz_remove_edge(graph3,curr_edge);

110
src/jenkins_hash.c
View File

@ -9,6 +9,7 @@
#define DEBUG
#include "debug.h"
#include "MurmurHash2.h"
#define hashsize(n) ((cmph_uint32)1<<(n))
#define hashmask(n) (hashsize(n)-1)
@ -99,63 +100,67 @@ void jenkins_state_destroy(jenkins_state_t *state)
inline void __jenkins_hash_vector(cmph_uint32 seed, const char *k, cmph_uint32 keylen, cmph_uint32 * hashes)
{
register cmph_uint32 len, length;
/* Set up the internal state */
length = keylen;
len = length;
hashes[0] = hashes[1] = 0x9e3779b9; /* the golden ratio; an arbitrary value */
hashes[2] = seed; /* the previous hash value - seed in our case */
/*---------------------------------------- handle most of the key */
while (len >= 12)
{
hashes[0] += ((cmph_uint32)k[0] +((cmph_uint32)k[1]<<8) +((cmph_uint32)k[2]<<16) +((cmph_uint32)k[3]<<24));
hashes[1] += ((cmph_uint32)k[4] +((cmph_uint32)k[5]<<8) +((cmph_uint32)k[6]<<16) +((cmph_uint32)k[7]<<24));
hashes[2] += ((cmph_uint32)k[8] +((cmph_uint32)k[9]<<8) +((cmph_uint32)k[10]<<16)+((cmph_uint32)k[11]<<24));
mix(hashes[0],hashes[1],hashes[2]);
k += 12; len -= 12;
int i;
for (i = 0; i < 3; ++i) {
hashes[i] = MurmurHash2(k, keylen, seed + i);
}
/*------------------------------------- handle the last 11 bytes */
hashes[2] += length;
switch(len) /* all the case statements fall through */
{
case 11:
hashes[2] +=((cmph_uint32)k[10]<<24);
case 10:
hashes[2] +=((cmph_uint32)k[9]<<16);
case 9 :
hashes[2] +=((cmph_uint32)k[8]<<8);
/* the first byte of hashes[2] is reserved for the length */
case 8 :
hashes[1] +=((cmph_uint32)k[7]<<24);
case 7 :
hashes[1] +=((cmph_uint32)k[6]<<16);
case 6 :
hashes[1] +=((cmph_uint32)k[5]<<8);
case 5 :
hashes[1] +=(cmph_uint8) k[4];
case 4 :
hashes[0] +=((cmph_uint32)k[3]<<24);
case 3 :
hashes[0] +=((cmph_uint32)k[2]<<16);
case 2 :
hashes[0] +=((cmph_uint32)k[1]<<8);
case 1 :
hashes[0] +=(cmph_uint8)k[0];
/* case 0: nothing left to add */
}
mix(hashes[0],hashes[1],hashes[2]);
// register cmph_uint32 len, length;
//
// /* Set up the internal state */
// length = keylen;
// len = length;
// hashes[0] = hashes[1] = 0x9e3779b9; /* the golden ratio; an arbitrary value */
// hashes[2] = seed; /* the previous hash value - seed in our case */
//
// /*---------------------------------------- handle most of the key */
// while (len >= 12)
// {
// hashes[0] += ((cmph_uint32)k[0] +((cmph_uint32)k[1]<<8) +((cmph_uint32)k[2]<<16) +((cmph_uint32)k[3]<<24));
// hashes[1] += ((cmph_uint32)k[4] +((cmph_uint32)k[5]<<8) +((cmph_uint32)k[6]<<16) +((cmph_uint32)k[7]<<24));
// hashes[2] += ((cmph_uint32)k[8] +((cmph_uint32)k[9]<<8) +((cmph_uint32)k[10]<<16)+((cmph_uint32)k[11]<<24));
// mix(hashes[0],hashes[1],hashes[2]);
// k += 12; len -= 12;
// }
//
// /*------------------------------------- handle the last 11 bytes */
// hashes[2] += length;
// switch(len) /* all the case statements fall through */
// {
// case 11:
// hashes[2] +=((cmph_uint32)k[10]<<24);
// case 10:
// hashes[2] +=((cmph_uint32)k[9]<<16);
// case 9 :
// hashes[2] +=((cmph_uint32)k[8]<<8);
// /* the first byte of hashes[2] is reserved for the length */
// case 8 :
// hashes[1] +=((cmph_uint32)k[7]<<24);
// case 7 :
// hashes[1] +=((cmph_uint32)k[6]<<16);
// case 6 :
// hashes[1] +=((cmph_uint32)k[5]<<8);
// case 5 :
// hashes[1] +=(cmph_uint8) k[4];
// case 4 :
// hashes[0] +=((cmph_uint32)k[3]<<24);
// case 3 :
// hashes[0] +=((cmph_uint32)k[2]<<16);
// case 2 :
// hashes[0] +=((cmph_uint32)k[1]<<8);
// case 1 :
// hashes[0] +=(cmph_uint8)k[0];
// /* case 0: nothing left to add */
// }
//
// mix(hashes[0],hashes[1],hashes[2]);
}
cmph_uint32 jenkins_hash(jenkins_state_t *state, const char *k, cmph_uint32 keylen)
{
cmph_uint32 hashes[3];
__jenkins_hash_vector(state->seed, k, keylen, hashes);
return hashes[2];
/* cmph_uint32 a, b, c;
// cmph_uint32 hashes[3];
// __jenkins_hash_vector(state->seed, k, keylen, hashes);
// return hashes[2];
cmph_uint32 a, b, c;
cmph_uint32 len, length;
// Set up the internal state
@ -209,7 +214,6 @@ cmph_uint32 jenkins_hash(jenkins_state_t *state, const char *k, cmph_uint32 keyl
/// report the result
return c;
*/
}
void jenkins_hash_vector_(jenkins_state_t *state, const char *k, cmph_uint32 keylen, cmph_uint32 * hashes)