commit 3b05ec333913342258dc67817d6ea6836eeb93aa
author scottb@google.com <scottb@google.com@8db76d5a-ed1c-0410-87a9-c151d255dfc7> Fri May 15 23:23:15 2009 +0000
committer scottb@google.com <scottb@google.com@8db76d5a-ed1c-0410-87a9-c151d255dfc7> Fri May 15 23:23:15 2009 +0000
tree d9986df519faff2bc14f4d9510d83dbd08d5e19c
parent e0704c9324ed7e4586f2291e9c6fb9e2b0acf5e7

Optimize the output production by reusing 16k buffers over and over.

By reusing over and over a buffer small enough to fit into the processor's L2 cache, we realize a huge savings on data transfer.  This replaces the previous method of reading data entirely into huge byte[] and then writing it back in one shot.  Experimentally, this is shown to be much faster.

Review by: spoon, fabbott

git-svn-id: https://google-web-toolkit.googlecode.com/svn/trunk@5404 8db76d5a-ed1c-0410-87a9-c151d255dfc7