| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
DCE_GVN does not take into account the following case:
mov a, b
...
mov c, b
when optimization tries to replace a with c it must ensure that
for all uses of a there is no new definition of c before use.
Otherwise that use will incorrectly substituted with new c instead
of original b.
Bug: 23102860
Signed-off-by: Serguei Katkov <serguei.i.katkov@intel.com>
(cherry picked from commit 2f2f17399f6bdfc5ec94a875152c31ef79620520)
Change-Id: I1f08c99cedbe4fd1b96cad11f17d60ab551c7cf7
|
|
|
|
|
|
|
|
|
|
| |
Squashed cherry-picks:
* 067f1ed7816cf4eb5d6258ca31b387ddb2073ab7
* 750f7c2827318f6d07620f2ef0321218ea4d8670
* 2f90b3415aadc2587d26c767c6bfb235797119a8
* 2ea7b70b2347969f3735bd0ec1b462bd6d2ff1bd
Bug: 22881413
|
|
|
|
|
|
|
|
|
|
|
| |
array-length can throw NPE so it should be taking into account.
Bug: 22521944
Signed-off-by: Serguei Katkov <serguei.i.katkov@intel.com>
(cherry picked from commit b016c6dd3c30b04104a0a43dc294ce93e5f63874)
Change-Id: I6232430a02f9e6d156ad2aae0eb5a630118c0e79
|
|
|
|
|
|
|
|
| |
Bug: 21702651
(cherry picked from commit 511d408909b7fe9be3c95e032cc7f426f4a01e6e)
Change-Id: I059e0b841543334cd08a27431e0ea853b42e4f35
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This CL properly fixes the high word marking and reverts
https://android-review.googlesource.com/150352
which was just covering up the underlying issue. A unit test
for the encountered issue is provided, though it does not
expose the deficiency in the cover-up CL.
Bug: 20640451
(cherry picked from commit 9cacef6e811940c2f21e7e54055379a2c43f0d06)
Change-Id: I4c4ca82fe4b2e34feb38090d88a5d5c754914f89
|
|
|
|
|
|
|
|
| |
Bug: 20640451
(cherry picked from commit ade58e14582c6ab81978e9cfcbbf5e54cc23fe18)
Change-Id: I8ac7fea3d9d2a434ee79fb917e68a4dc96e5482f
|
|
|
|
|
|
|
|
|
|
|
| |
Previously we missed some cases of overlap with registers
coming from previous blocks.
Bug: 20640451
(cherry picked from commit 83d46ef1eaa8fdecadfdb9564d80e50b42646c37)
Change-Id: I1be879edfbc900b70cee411d9e31e5a4b524530a
|
|
|
|
|
|
|
|
|
|
|
| |
If we're inserting a new Phi and all its inputs are the same
SSA reg (for example, we allow a single-input Phi), some
subsequent insn may actually refer to that reg, so we need
to rename it to keep the graph in a consistent state.
(cherry picked from commit f60715c2c0a2398822a143ffb6b75507bffca703)
Change-Id: I46822b0db21c187601b79d2ff1a650494558772a
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Fix LVN::GetEndingVregValueNumberImpl() to check whether
the requested wideness matches the SSA register type as
recorded in MIRGraph::reg_location_.
Add DCHECKs that the wideness matches when getting/setting
sreg values, update Phi handling in LVN/DCE to use the type
from MIRGraph::reg_location_ instead of determining it from
the sreg value maps which would now trigger the DCHECKs.
Update tests to initialize MIRGraph::reg_location_.
Reenable DCE.
Bug: 20572509
(cherry picked from commit a5e69e87c630c08c0de1740427e60d531ce851b9)
Change-Id: Ieb97ac9e3672b977e36fd7f369a975bae7d5271e
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
When eliminating a move by renaming its source register,
check that it doesn't conflict with vreg usage by insns
between the defining insn and the move.
Improve the GVN/DCE cleanup so that it can handle cases
where GVN or DCE is individually disabled in the pass driver
but not in the disable_opt flags.
Bug: 19419671
Change-Id: I49bb67b81509f51fbaf90c6016c509962be43736
|
|
|
|
|
|
|
|
|
| |
Fix the type of the ArtMethod* SSA register.
Bug: 19419671
This reverts commit 1b717f63847de8762e7f7bdd6708fdfae9d24a67.
Change-Id: Ie4da3c03a0e0334a39a24718f6dc31f9255cfb53
|
|
|
|
|
|
|
|
| |
Breaks arm64, as the method register is not correctly flagged
as ref and thus 32bit.
Bug: 19419671
This reverts commit e490b01c12d33f3bd5c247b55b47e507cc9c8fab.
|
|
|
|
|
|
|
|
|
|
|
| |
Use method signatures, field types and types embedded in dex
insns for type inference. Perform the type inference in two
phases, first a simple pass that records all types implied
by individual insns, and then an iterative pass to propagate
those types further via phi, move, if-cc and aget/aput insns.
Bug: 19419671
Change-Id: Id38579d48a44fc5eadd13780afb6d370093056f9
|
|
|
|
| |
Change-Id: Ie6aba02f4223b1de02530e1515c63505f37e184c
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Eliminate check-cast if the result of an instance-of with
the very same type on the same value is used to branch to
the check-cast's block or a dominator of it.
Note that there already exists a verifier-based elimination
of check-cast but it excludes check-cast on interfaces. This
new optimization works for interface types and, since it's
GVN-based, it can better recognize when the same reference
is used for instance-of and check-cast.
Change-Id: Ib315199805099d1cb0534bb4a90dc51baa409685
|
|
Change-Id: I5b77411a8f088f0b561da14b123cf6b0501c9db5
|