svn commit: r1519272 - /subversion/trunk/subversion/libsvn_subr/cache-membuffer.c

[st...@apache.org]

Author: stefan2
Date: Sun Sep  1 14:04:40 2013
New Revision: 1519272

URL: http://svn.apache.org/r1519272
Log:
In membuffer caching, simplify the L1->L2 promotion logic.  No longer
prevent cache index entries pertaining to L1 from being moved to another
entry in the same group.  Check for that special condition in the
caller now where it is more obvious.

* subversion/libsvn_subr/cache-membuffer.c
  (promote_entry): assert that we actually call this for L1 entries only
  (ensure_data_insertable_l2): don't care whether we move L1 entries
  (ensure_data_insertable_l1): check for moved L1 entries here

Modified:
    subversion/trunk/subversion/libsvn_subr/cache-membuffer.c

Modified: subversion/trunk/subversion/libsvn_subr/cache-membuffer.c
URL: http://svn.apache.org/viewvc/subversion/trunk/subversion/libsvn_subr/cache-membuffer.c?rev=1519272&r1=1519271&r2=1519272&view=diff
==============================================================================
--- subversion/trunk/subversion/libsvn_subr/cache-membuffer.c (original)
+++ subversion/trunk/subversion/libsvn_subr/cache-membuffer.c Sun Sep  1 14:04:40 2013
@@ -1120,6 +1120,7 @@ promote_entry(svn_membuffer_t *cache, en
   apr_uint32_t idx = get_index(cache, entry);
   apr_size_t size = ALIGN_VALUE(entry->size);
   assert(get_cache_level(cache, entry) == &cache->l1);
+  assert(idx == cache->l1.next);
 
   /* copy item from the current location in L1 to the start of L2's
    * insertion window */
@@ -1142,16 +1143,14 @@ promote_entry(svn_membuffer_t *cache, en
  * 
  * If necessary, enlarge the insertion window of CACHE->L2 until it is at
  * least TO_FIT_IN->SIZE bytes long. TO_FIT_IN->SIZE must not exceed the
- * data buffer size allocated to CACHE->L2.  IDX is the item index of
- * TO_FIT_IN and is given for performance reasons.
+ * data buffer size allocated to CACHE->L2.
  * 
  * Return TRUE if enough room could be found or made.  A FALSE result
  * indicates that the respective item shall not be added.
  */
 static svn_boolean_t
 ensure_data_insertable_l2(svn_membuffer_t *cache,
-                          entry_t *to_fit_in,
-                          apr_uint32_t idx)
+                          entry_t *to_fit_in)
 {
   entry_t *entry;
   apr_uint64_t average_hit_value;
@@ -1170,9 +1169,6 @@ ensure_data_insertable_l2(svn_membuffer_
   /* accumulated "worth" of items dropped so far */
   apr_size_t drop_hits = 0;
 
-  /* verify parameters */
-  assert(idx == get_index(cache, to_fit_in));
-
   /* This loop will eventually terminate because every cache entry
    * would get dropped eventually:
    * - hit counts become 0 after the got kept for 32 full scans
@@ -1244,16 +1240,6 @@ ensure_data_insertable_l2(svn_membuffer_
             {
               keep = TRUE;
             }
-          else if (cache->l2.next / GROUP_SIZE == idx / GROUP_SIZE)
-            {
-              /* Special case: we cannot drop entries that are in the same
-               * group as TO_FIT_IN because that might the latter to become
-               * invalidated it it happens to be the highest used entry in
-               * the group.  So, we must keep ENTRY unconditionally.
-               * (this is a very rare condition)
-               */
-              keep = TRUE;
-            }
           else if (   entry->priority < SVN_CACHE__MEMBUFFER_DEFAULT_PRIORITY
                    && to_fit_in->priority > entry->priority)
             {
@@ -1331,8 +1317,6 @@ ensure_data_insertable_l2(svn_membuffer_
 static svn_boolean_t
 ensure_data_insertable_l1(svn_membuffer_t *cache, apr_size_t size)
 {
-  entry_t *entry;
-
   /* Guarantees that the while loop will terminate. */
   if (size > cache->l1.size)
     return FALSE;
@@ -1344,9 +1328,11 @@ ensure_data_insertable_l1(svn_membuffer_
     {
       /* first offset behind the insertion window
        */
+      apr_uint32_t entry_index = cache->l1.next;
+      entry_t *entry = get_entry(cache, entry_index);
       apr_uint64_t end = cache->l1.next == NO_INDEX
                        ? cache->l1.start_offset + cache->l1.size
-                       : get_entry(cache, cache->l1.next)->offset;
+                       : entry->offset;
 
       /* leave function as soon as the insertion window is large enough
        */
@@ -1370,12 +1356,14 @@ ensure_data_insertable_l1(svn_membuffer_
           /* Remove the entry from the end of insertion window and promote
            * it to L2, if it is important enough.
            */
-          entry = get_entry(cache, cache->l1.next);
+          svn_boolean_t keep = ensure_data_insertable_l2(cache, entry);
 
-          if (ensure_data_insertable_l2(cache, entry, cache->l1.next))
-            promote_entry(cache, entry);
-          else
-            drop_entry(cache, entry);
+          /* We might have touched the group that contains ENTRY. Recheck. */
+          if (entry_index == cache->l1.next)
+            if (keep)
+              promote_entry(cache, entry);
+            else
+              drop_entry(cache, entry);
         }
     }