Rev 24: Tweak batch sizes to have roughly the same memory consumption for both in file:///net/bigmamac/Volumes/home/vila/.bazaar/plugins/fix277537/

[Vincent Ladeuil]

At file:///net/bigmamac/Volumes/home/vila/.bazaar/plugins/fix277537/

------------------------------------------------------------
revno: 24
revision-id: v.ladeuil+lp at free.fr-20090206172203-z6w3rgos1eccdpln
parent: v.ladeuil+lp at free.fr-20090206144757-998yz1pc6entod17
committer: Vincent Ladeuil <v.ladeuil+lp at free.fr>
branch nick: fix277537
timestamp: Fri 2009-02-06 18:22:03 +0100
message:
  Tweak batch sizes to have roughly the same memory consumption for both
  storage kinds.
  
  * reconcile.py:
  (InventoryAncestryReconcilePacker._copy_fixed_nodes): Final tweak
  for fulltext batch sizes.
-------------- next part --------------
=== modified file 'reconcile.py'
--- a/reconcile.py	2009-02-06 14:47:57 +0000
+++ b/reconcile.py	2009-02-06 17:22:03 +0000
@@ -178,8 +178,8 @@
         if getattr(osutils, 'chunks_to_lines', None) is None:
             storage_kind = 'fulltext'
             convert_bytes = osutils.split_lines
-            min_batch_size = 16
-            max_batch_size = 128
+            min_batch_size = 4
+            max_batch_size = 64
         else:
             storage_kind = 'chunked'
             convert_bytes = osutils.chunks_to_lines
@@ -250,7 +250,9 @@
                         # wrong.
                         import pdb; pdb.set_trace()
                     if storage_kind == 'fulltext':
-                        # No need for more parents there
+                        # Experiments showed that the performance gain wasn't
+                        # worth the memory consumption here, a single parent
+                        # was the best trade-off.
                         parent_texts = {key : content}
                     else:
                         parent_texts[key] = content