Eliminate the use of the MASKMOVDQU instruction, to speed up decompression performanc...

git-svn-id: svn+ssh://svn.code.sf.net/p/libjpeg-turbo/code/trunk@836 632fc199-4ca6-4c93-a231-07263d6284db

diff --cc ChangeLog.txt
index d4808e8e5d9bd5dffa5ad6069879c26219719228,e80ac6c3a8accf455c9265c9b0a45717b8e2c87b..5453869e7d86a550e5e6dacc4cc52c90a2feee5c
--- 1/ChangeLog.txt
--- 2/ChangeLog.txt
+++ b/ChangeLog.txt
@@@ -34,9 -19,13 +34,16 @@@ calling conventions
  images (specifically, images in which the component count was erroneously set
  to a large value) would cause libjpeg-turbo to segfault.
  
 -[5] Worked around a severe performance issue with "Bobcat" (AMD Embedded APU)
 +[8] Extended the TurboJPEG Java API so that it can be used to decompress a
 +JPEG image into an arbitrary position in a large output buffer.
 +
++[9] Worked around a severe performance issue with "Bobcat" (AMD Embedded APU)
+ processors.  The MASKMOVDQU instruction, which was used by the libjpeg-turbo
+ SSE2 SIMD code, is apparently implemented in microcode on AMD processors, and
+ it is painfully slow on Bobcat processors in particular.  Eliminating the use
+ of this instruction improved performance by an order of magnitude on Bobcat
+ processors and by a small amount (typically 5%) on AMD desktop processors.
+ 
  
  1.2.0
  =====