| 591ad29eb1aa7c5ac7faadb426f5eb2b4a4ccd02 |
|
01-Mar-2016 |
Aart Bik <ajcbik@google.com> |
Standby list for dyn bce in potentially infinite loops.
Rationale:
The old code relied on "luck" to revisit basic blocks
after dynamic bce and incorporate all bounds checks
in potentially infinite loops that were "made" finite.
Now that revisiting has been removed completely, keeping
a standby list ensures all candidates are considered.
Change-Id: Ida3cf63be1307be6c2b0258d3e64b163f12be235
|
| b6347b7a3dbf9f55cbd638c630e9d044d6d53ac6 |
|
29-Feb-2016 |
Aart Bik <ajcbik@google.com> |
Fixed bug on incorrectly revisiting same block.
Rationale:
Aart's fuzz tester found this particular bug, where
revisiting a block (after dynamic bce) would cause
static array length based bce to feed into itself
and thus incorrectly remove a needed bounds check.
bug=27376274
Change-Id: I9163f283af355d444b4cec707f194fe2b67c2572
|
| e530539656fd7c95c914885b4f67c404daea2ea2 |
|
24-Feb-2016 |
Aart Bik <ajcbik@google.com> |
Make sure constantIndexing2 is optimized with DEOPT.
Rationale:
This particular method is also used in foo() to see if
assertIsManaged() holds after going out of bounds.
During my last rewrite (which added a driver to make
sure more code is actually executed), I changed the
method to show a non-deopt case. However, that would
miss this test. I moved the non-deopt case into a new test.
Change-Id: Iaf3e77e33a82ff6cd300753f9e5211529353401c
|
| bf3f1cf15a021ea1ff8ae860c55e8281da4619b3 |
|
23-Feb-2016 |
Aart Bik <ajcbik@google.com> |
Improved instruction + offset hunting.
Rationale:
This is generally useful for anything using this method
but in particular for deopting something like
bs[ off] = (byte)(n >>> 24);
bs[++off] = (byte)(n >>> 16);
bs[++off] = (byte)(n >>> 8);
bs[++off] = (byte)(n );
where the base + offset is hidden in the increments.
Occurs quite often in real-life code.
Change-Id: I3fa7d285a7368a179a26e590e8eee37f3b64c25d
|
| 1d23982327265e8139f67e37d9a86ba2758f7b4a |
|
09-Feb-2016 |
Aart Bik <ajcbik@google.com> |
Generalized "dom-based" dynamic BCE to symbolic base + offset.
Rationale:
So far, if all others failed, BCE would use a dominator-based
dynamic deoptimization to eliminate bounds checks in e.g. a[0],
a[1], a[2], etc. This CL generalizes this to any symbolic base
with offset in e.g. a[base], a[base+1], etc. The runtime tests
(two for symbolic, one for constant) carefully account for
arithmetic wrap-around.
bug=26680114
Change-Id: I7432a200fd69791914ed776c77fa62567b5863c0
|
| 17fb893c41f2fb2528abbd63e463008d2ddd5d8f |
|
02-Feb-2016 |
Vladimir Marko <vmarko@google.com> |
Optimizing: Do not depend on sharpening in test 449.
The sharpening may or may not remove the ArtMethod*
parameter, so the test must not depend on its absence.
This fixes the test 449 checker test on mips/mips64.
Change-Id: I0c9e7353234365c97b622c7c84708be3dcbbe4ff
|
| 09e8d5ffe52c738c6a74984b1cbc7ad4bc8f5e2c |
|
23-Jan-2016 |
Aart Bik <ajcbik@google.com> |
Some minor simplifications in code and tests.
Background:
This is actually a resubmit of an earlier cl that was
reverted because was test was less robust against
inlining changes (it assumed a virtual call would
never be inlined).
original cl: If8ada79dfd70bea991c11d2b18661b951b6c4cd4
revert cl: I739aaaccd0509d02a62ef01e797a6d45bfe941df
Change-Id: I952680d60ff488874907f066bfdf156a45b409ba
|
| 69fd1b56021ac62c17e188bd0d4dd22fc911558e |
|
22-Jan-2016 |
Nicolas Geoffray <ngeoffray@google.com> |
Revert "Some minor simplifications in code and tests."
Fails 530-checker-loops on arm
This reverts commit bf03fcd10a3ffa15468d335f26697b0473e45b36.
Change-Id: I739aaaccd0509d02a62ef01e797a6d45bfe941df
|
| bf03fcd10a3ffa15468d335f26697b0473e45b36 |
|
04-Jan-2016 |
Aart Bik <ajcbik@google.com> |
Some minor simplifications in code and tests.
Rationale: fell through the cracks of previous "intrinsics" CL.
Change-Id: If8ada79dfd70bea991c11d2b18661b951b6c4cd4
|
| 5949fa0cb9a8d26ac20b9b02065a63b4b20443be |
|
18-Dec-2015 |
Nicolas Geoffray <ngeoffray@google.com> |
Revert "Revert "Tweak inlining heuristics.""
This reverts commit b17d1ccff0ac26fc22df671907ba2b4f4c656ce4.
Change-Id: I26f6f8702a448c3da12662cbc6bc0f6e562bc40b
|
| 5d75afe333f57546786686d9bee16b52f1bbe971 |
|
14-Dec-2015 |
Aart Bik <ajcbik@google.com> |
Improved side-effects/can-throw information on intrinsics.
Rationale: improved side effect and exception analysis gives
many more opportunities for GVN/LICM/BCE.
Change-Id: I8aa9b757d77c7bd9d58271204a657c2c525195b5
|
| 4a34277c55279ba57ab361f7580db846a201d9b1 |
|
30-Nov-2015 |
Aart Bik <ajcbik@google.com> |
Dynamic BCE (based on induction range analysis)
Rationale:
A rewritten dynamic BCE that uses induction variable analysis
to generate the run-time tests before a loop in order to
eliminate bounds-checks from its body. This CL removes now
obsoleted induction related code inside the BCE module.
Also, the dynamic test generation is placed more strategically,
since we missed a few cases where static analysis does better.
Most significant performance improvements (filtering noise) is about:
Linpack +20%
LU > +10%
Change-Id: I03d7631857154b6a131b132f26a2dc568af1b3a1
|
| d59c70627cc42878cc30b46bd29ff497b4483b22 |
|
21-Nov-2015 |
Aart Bik <ajcbik@google.com> |
Revert "Dynamic BCE (based on induction range analysis)"
This reverts commit 0b5849be045c5683d4a6b6b6c306abadba5f0fcc.
Change-Id: Id33f5da42bbdfb1aff7e2281417c8a7aa492df05
Rationale: so close :-( but bullhead-userdebug (linux) build in git_mnc-dr-dev-plus-aosp reported a breakage with a type inconsistency (long vs int in probably the codegen of dynamic bce); no time to investigate and fix this fully before my trip, so rolling back for now
|
| 0b5849be045c5683d4a6b6b6c306abadba5f0fcc |
|
19-Oct-2015 |
Aart Bik <ajcbik@google.com> |
Dynamic BCE (based on induction range analysis)
Rationale: A rewritten dynamic BCE that uses induction variable analysis
to generate the run-time tests before a loop in order to
eliminate bounds-checks from its body. This CL removes now
obsoleted induction related code inside the BCE module.
Also, the dynamic test generation is placed more strategically,
since we missed a few cases where static analysis does better.
Most significant performance improvements (after filtering noise) is about:
Linpack +20%
LU > +10%
Change-Id: I4e7b8bab0288beff6f98a14856e3536103d32742
|
| 951779839f0d35ed5336f399c8f521fd9a6b7c27 |
|
30-Oct-2015 |
David Brazdil <dbrazdil@google.com> |
ART: Enable inlining under try/catch
This patch updates the inliner to set try/catch information
when inlining into a method with try/catch. It does not yet
allow inlining of methods with try/catch because that will
require generating catch stack maps with inline info.
Change-Id: I7d57e1454e7da537d75c5c7eda60b22f3a30fa60
|
| 0dfc9bc0cca3ae15dc75bbea3ca9b6d7508ce893 |
|
01-Oct-2015 |
Andreas Gampe <agampe@google.com> |
ART: Add more deopt tests
Extend libarttest with the ability to turn off asserts for method
state. Use this to put asserts into test functions, but turn them
off if the test wouldn't react as we want, e.g., when we're in
interpreter mode.
Extend run-test 449 by adding asserts for the method state for the
expected deopt cases. This tests both standard behavior, as well
as single-frame deoptimization.
Bug: 21611912
Change-Id: I27fa5e3e44d8c5eab57962d144c6aec96587bf45
|
| 639bdd13993644a267f177f8f5936496bda65e2b |
|
03-Jun-2015 |
Andreas Gampe <agampe@google.com> |
ART: Single-frame deopt
Add deoptimization of a single frame. Works by removing the managed
code frame and jumping into the quick-to-interpreter bridge, and
the bridge understanding a stored ShadowFrame.
We need a separate fixup pass. For x86, we leave the return address
on the stack so we don't need to push it there.
Bug: 21611912
Change-Id: I06625685ced8b054244f8685ab50b238a705b9d2
|
| 681652d8e8a33bc07c5c082a71aea13d0f15e0a0 |
|
23-Jul-2015 |
Mingyao Yang <mingyao@google.com> |
HDeoptimize should hold values live in env.
Values that are not live in compiled code anymore may still be needed in
interpreter, due to code motion, etc.
(cherry-picked from commit 718493c6c3c8e380663cb8a94e57ce160a6c473f)
Bug: 22665511
Change-Id: I8b85833c5c462f8fe36f86d6026a51b07563995a
|
| 718493c6c3c8e380663cb8a94e57ce160a6c473f |
|
23-Jul-2015 |
Mingyao Yang <mingyao@google.com> |
HDeoptimize should hold values live in env.
Values that are not live in compiled code anymore may still be needed in
interpreter, due to code motion, etc.
Bug: 22665511
Change-Id: I8b85833c5c462f8fe36f86d6026a51b07563995a
|
| 6a92a033c33e383541d77607fbe8cd982875d13d |
|
23-Jul-2015 |
Roland Levillain <rpl@google.com> |
Cosmetic changes in ART run-tests.
Perform a copyright headers clean up mostly.
Change-Id: I26c855b2f54d0887676d9b9b8281b14b7978ab29
|
| bca381a12965a98e3727e93986dd0a195db500a0 |
|
20-May-2015 |
Mingyao Yang <mingyao@google.com> |
Fix premature deoptimization if the loop body isn't entered.
Add a test between initial_ and end_ to see if the loop body is entered.
If the loop body isn't entered at all, we jump to the loop header. Loop header is
still executed and is going to test the condition again and loop body won't be
entered. This makes sure no deoptimization is triggered if the loop body isn't
even entered.
Bug: 21034044
(cherry picked from commit 3584bce5b1f45e5741d3a6ca24884a36320ecb6b)
Change-Id: I2b6de1f22fbc4568ca419f76382ebd87806d9694
|
| 3584bce5b1f45e5741d3a6ca24884a36320ecb6b |
|
20-May-2015 |
Mingyao Yang <mingyao@google.com> |
Fix premature deoptimization if the loop body isn't entered.
Add a test between initial_ and end_ to see if the loop body is entered.
If the loop body isn't entered at all, we jump to the loop header. Loop header is
still executed and is going to test the condition again and loop body won't be
entered. This makes sure no deoptimization is triggered if the loop body isn't
even entered.
Bug: 21034044
Change-Id: I2b6de1f22fbc4568ca419f76382ebd87806d9694
|
| a06d66a4ee60926127b9498b7ff0b3e37a24fccf |
|
28-May-2015 |
David Brazdil <dbrazdil@google.com> |
ART: Distinguish Checker lines from comments
In order to prevent tests passing due to lines with hard-to-spot
formatting errors begin ignored, e.g. by forgetting the colon after
"//CHECK", Checker will now require its assertions to start with "///"
or "##", respectivelly for Java and Smali. Such lines will never be
ignored and will fail the test unless successfully parsed.
Change-Id: I0da9a8f13eb96d950af8c85df17d1899a853a299
|
| 9d750efd66ae7f4b790af3c1ff8de972bbe826d9 |
|
27-Apr-2015 |
Mingyao Yang <mingyao@google.com> |
BCE: don't add deoptimization if the loop has early exit.
Also make the way to detect loop_body_successor to be
more accurate.
Change-Id: I29680f93396383c478a8f40ad28735e4f3f07c1b
|
| 206d6fd6cae5ba8c4d5f0e230111fe77b9d5c0a5 |
|
14-Apr-2015 |
Mingyao Yang <mingyao@google.com> |
Deoptimization-based BCE for unknown loop bounds.
For loop like:
for (int i = start; i < end; i++) {
array[i] = 1;
}
We add the following to the loop pre-header:
if (start < 0) deoptimize();
if (end > array.length) deoptimize();
Then we can eliminate bounds-check of array[i] inside the loop.
We also take care of indexing with induction variable plus some offsets,
like array[i - 1]/array[i + 1] inside the loop, and adjust the condition
for deoptimization accordingly.
Change-Id: I9e24c6b5e134ff95eff5b5605ff8f95d6546616f
|
| d43b3ac88cd46b8815890188c9c2b9a3f1564648 |
|
01-Apr-2015 |
Mingyao Yang <mingyao@google.com> |
Revert "Revert "Deoptimization-based bce.""
This reverts commit 0ba627337274ccfb8c9cb9bf23fffb1e1b9d1430.
Change-Id: I1ca10d15bbb49897a0cf541ab160431ec180a006
|
| 0ba627337274ccfb8c9cb9bf23fffb1e1b9d1430 |
|
24-Mar-2015 |
Andreas Gampe <agampe@google.com> |
Revert "Deoptimization-based bce."
This breaks compiling the core image:
Error after BCE: art::SSAChecker: Instruction 219 in block 1 does not dominate use 221 in block 1.
This reverts commit e295e6ec5beaea31be5d7d3c996cd8cfa2053129.
Change-Id: Ieeb48797d451836ed506ccb940872f1443942e4e
|
| e295e6ec5beaea31be5d7d3c996cd8cfa2053129 |
|
07-Mar-2015 |
Mingyao Yang <mingyao@google.com> |
Deoptimization-based bce.
A mechanism is introduced that a runtime method can be called
from code compiled with optimizing compiler to deoptimize into
interpreter. This can be used to establish invariants in the managed code
If the invariant does not hold at runtime, we will deoptimize and continue
execution in the interpreter. This allows to optimize the managed code as
if the invariant was proven during compile time. However, the exception
will be thrown according to the semantics demanded by the spec.
The invariant and optimization included in this patch are based on the
length of an array. Given a set of array accesses with constant indices
{c1, ..., cn}, we can optimize away all bounds checks iff all 0 <= min(ci) and
max(ci) < array-length. The first can be proven statically. The second can be
established with a deoptimization-based invariant. This replaces n bounds
checks with one invariant check (plus slow-path code).
Change-Id: I8c6e34b56c85d25b91074832d13dba1db0a81569
|
| 4559f000b323b64e4bd179b72cfb788b30b25b23 |
|
27-Feb-2015 |
Mingyao Yang <mingyao@google.com> |
bce: handle a pattern for circular buffer
Change-Id: Ie54bdd7c044af58deea0d0addaaa8186cabf3532
|
| 57e04754d5b7fb3cc99d6b9f70da73cf4c65b416 |
|
10-Feb-2015 |
Mingyao Yang <mingyao@google.com> |
bce: add support to narrow two MonotonicValueRange's at the same time.
Change-Id: I545da4f375619ce47e01bb5aa5c8b1a4a9d1df41
|
| 8c8bad89ff367712a6f5bbf3a5836cd398391067 |
|
10-Feb-2015 |
Mingyao Yang <mingyao@google.com> |
More checker tests for BCE.
Also make sure the check on MonotonicValueRange narrow is more strict.
Plus some handling on array length division such as array.length/2.
Added checker tests for each case.
Change-Id: I9f32fc5f6ca1f3da8edec576de66b44d85a50bc6
|
| 0304e182adee81be32c744fd3c0d28add29974ff |
|
31-Jan-2015 |
Mingyao Yang <mingyao@google.com> |
Improve bce so that more bounds checks can be eliminated.
For pattern like "int[] array = new int[size+1]", we record this range
for size:
[-1, array.length-1]
This can eliminate more bounds checks.
Also simplify overflow/underflow handling and make it more solid.
Enhance instruction simplifier such that if array is a result of
NewArray with a constant size, replace array.length with that constant.
Plan to move all bce gtests to checker in another change.
Change-Id: Ibe7cc7940b68fb6465dc3e0ff3ebdb0fd6487aa9
|