-
Committer:
John Arbash Meinel
-
Date:
2009-02-24 20:23:07 UTC
-
mto:
This revision was merged to the branch mainline in
revision
4051.
-
Revision ID:
john@arbash-meinel.com-20090224202307-qxv0wgz3cikobn0c
Group records to read by pack file and sort by offset.
This assumes that sort(index_memo) gives the right ordering, but for all
formats so far, this holds true.
This makes a *big* difference when doing 'bzr branch --stacked' or
'bzr checkout --lightweight'.
If you don't group the requests, then the next level splits them up
into many readv() requests, because it splits them by pack/knit file.