OSDN Git Service

ARM Baker's read barrier fast path implementation.
authorRoland Levillain <rpl@google.com>
Fri, 18 Dec 2015 10:38:42 +0000 (10:38 +0000)
committerRoland Levillain <rpl@google.com>
Fri, 8 Jan 2016 12:55:31 +0000 (12:55 +0000)
commitc928591f5b2c544751bb3fb26dc614d3c2e67bef
treeb6c8a5e08c4d4c7a66a70f4d91e209ededf22334
parent5ee288c9dd99614e3a238f5efceeec6456e3499d
ARM Baker's read barrier fast path implementation.

Introduce an ARM fast path implementation in Optimizing for
Baker's read barriers (for both heap reference loads and GC
root loads).  The marking phase of the read barrier is
performed by a slow path, invoking the runtime entry point
artReadBarrierMark.

Other read barrier algorithms continue to use the original
slow path based implementation, which has been renamed as
GenerateReadBarrierSlow/GenerateReadBarrierForRootSlow.

Bug: 12687968
Change-Id: Ie7ee85b1b4c0564148270cebdd3cbd4c3da51b3a
compiler/optimizing/code_generator_arm.cc
compiler/optimizing/code_generator_arm.h
compiler/optimizing/intrinsics_arm.cc
compiler/optimizing/nodes.cc
compiler/optimizing/nodes.h
compiler/optimizing/parallel_move_resolver.cc