[Assorted-commits] SF.net SVN: assorted: [328] hash-join/trunk/src/hashjoin.cc

[<yan...@us...>]

Revision: 328
          http://assorted.svn.sourceforge.net/assorted/?rev=328&view=rev
Author:   yangzhang
Date:     2008-02-06 09:09:53 -0800 (Wed, 06 Feb 2008)

Log Message:
-----------
fixed some bugs

Modified Paths:
--------------
    hash-join/trunk/src/hashjoin.cc

Modified: hash-join/trunk/src/hashjoin.cc
===================================================================
--- hash-join/trunk/src/hashjoin.cc	2008-02-06 16:19:42 UTC (rev 327)
+++ hash-join/trunk/src/hashjoin.cc	2008-02-06 17:09:53 UTC (rev 328)
@@ -338,7 +338,7 @@
   //size_t bucket_size = max(1000000,buflen / ncpus * 2); //2 * buflen / ncpus;
   int bucket_size = max(1000000UL,buflen / ncpus * 3);
   if (heads[bucket] + nbytes < bs[bucket].bufs.back() + bucket_size) {
-    memcpy(heads[bucket], s, nbytes);
+    memcpy(heads[bucket], p, nbytes);
     heads[bucket] += nbytes;
     return -1;
   } else {
@@ -390,7 +390,7 @@
   for (unsigned int i = 0; i < ncpus; i++) {
     bs[i].sz.back() = heads[i] - bs[i].bufs.back();
   }
-  cout << "movie count " << counter << " vs " << bs[0].sz.back()<< endl;
+  cout << "movie count " << counter << " nbytes " << bs[0].sz.back()<< endl;
 }
 
 void
@@ -415,7 +415,7 @@
   // Statistics (TODO dynamic allocation)
   int counter = 0, mincount = INT_MAX;
   char *p = partstart, *end = partend;
-  while (p < end - 999) {
+  while (p < end) {
     char *name = p;
     p = strchr(p, '\0') + 1;
     strcpy(tmp, name);
@@ -426,12 +426,14 @@
       p = strchr(p, '\0') + 1;
 
       strcpy(subtmp, title);
-      size_t strl = strlen(subtmp);
+      size_t tmplen = subtmp + strlen(subtmp) + 2 - tmp;
+      check(tmplen < 1024);
+      tmp[tmplen-1] = '\0';
 
       // Copy this line into the correct local bucket.
       //cout << "hashing " << title << endl;
       unsigned int bbb;
-      if (-1 != (bbb = push_bucket(heads, bs, title, tmp, subtmp + strl + 1 - tmp))) {
+      if (-1 != (bbb = push_bucket(heads, bs, title, tmp, tmplen))) {
         //size_t bucket_size = max(1000000,buflen / ncpus * 2); //2 * buflen / ncpus;
         //int bucket_size = max(1000000UL,buflen / ncpus * 3);
         //cout << "FUCK " << heads[0] - bs[0].buf << " " << bucket_size << " " << heads[1] - bs[1].buf << " " << p - title << endl;
@@ -453,7 +455,7 @@
   for (unsigned int i = 0; i < ncpus; i++) {
     bs[i].sz.back() = heads[i] - bs[i].bufs.back();
   }
-  cout << "actress count " << counter << " vs " << bs[0].sz.back()<< endl;
+  cout << "actress count " << counter << " nbytes " << bs[0].sz.back()<< endl;
 }
 
 const hmap *
@@ -514,23 +516,24 @@
 actdb::probe1(unsigned int pid, const hmap *hh, const bucket **actbucs)
 {
   const hmap &h = *hh;
+  int hits = 0, misses = 0;
   for (unsigned int i = 0; i < ncpus; i++) {
     char *p = actbucs[i][pid].bufs[0],
          *end = actbucs[i][pid].bufs[0] + actbucs[i][pid].sz[0];
-    int hits = 0, misses = 0;
     while (p < end) {
       char *name = p;
       p = strchr(p, '\0') + 1;
       while (true) {
         char *title = p;
         p = strchr(p, '\0') + 1;
-        // cout << "name " << name << "title: " << title << p - title << endl;
+        //cout << "name " << name << " title: " << title << p - title << endl;
         // Emit any joined tuple.
         if (h.find(title) != h.end()) {
           //cout << "  HIT" << endl;
           hits++;
           join(title, name);
         } else {
+          //cout << "  MISS" << endl;
           misses++;
         }
         // End of tuple?
@@ -540,9 +543,8 @@
         }
       }
     }
-    //cout << "cpu " << pid << " src " << i << " hits " << hits << " misses " <<
-      //misses << endl;
   }
+  cout << "cpu " << pid << " hits " << hits << " misses " << misses << endl;
 }
 
 // vim:et:sw=2:ts=2


This was sent by the SourceForge.net collaborative development platform, the world's largest Open Source development site.