Commit Graph

14 Commits

Author SHA1 Message Date
fschneider@google.com b26ff22adf Allocation sinking for contexts.
Improved aliasing computation in presence of Redefinition and AssertAssignable.

Added possibility for inlining annotations via --enable-inlining-annotations flag.

R=vegorov@google.com

Review URL: https://codereview.chromium.org//184523002

git-svn-id: https://dart.googlecode.com/svn/branches/bleeding_edge/dart@41713 260f80e4-7a28-3924-810f-c04153c831b5
2014-11-13 13:53:58 +00:00
fschneider@google.com 04563b8bce Fix crash with --trace-deoptimization-verbose and tweak formatting.
Uninitialized instances (e.g. closures) can't be printed using ToCString.
Print just the class instead.

Print values/addresses consistently with our disassembly output (no leading zeros)
This makes searching in the trace output easier.

Limit printing source lines to max 80 chars. Otherwise the trace output for
minified Dart code gets unreadable.

R=srdjan@google.com

Review URL: https://codereview.chromium.org//658773002

git-svn-id: https://dart.googlecode.com/svn/branches/bleeding_edge/dart@41134 260f80e4-7a28-3924-810f-c04153c831b5
2014-10-15 15:49:40 +00:00
fschneider@google.com bdd6b29fd8 Revert r40306 because of crash.
TBR=vegorov@google.com

Review URL: https://codereview.chromium.org//572913002

git-svn-id: https://dart.googlecode.com/svn/branches/bleeding_edge/dart@40312 260f80e4-7a28-3924-810f-c04153c831b5
2014-09-16 12:14:33 +00:00
fschneider@google.com cd78dbe709 Refactor generating lazy deoptimization descriptors.
Instead of having special handling for closure calls, pass
the number of input operands down so that the correct lazy
deoptimization environment is generated at calls.

There is no functional change here - this is to simplify
having calls with pushed arguments and input operands
(like closure calls).

Runtime calls still do not have a correct lazy deoptimization environment
the environment should only be used for creating stack traces
from optimized code.

Fix a bug with --trace-deoptimization-verbose printing: We
cannot call ToCString on objects when the slots have not
been filled in yet. (e.g. Closures)

R=vegorov@google.com

Review URL: https://codereview.chromium.org//575443002

git-svn-id: https://dart.googlecode.com/svn/branches/bleeding_edge/dart@40306 260f80e4-7a28-3924-810f-c04153c831b5
2014-09-16 10:31:47 +00:00
vegorov@google.com 77c26187f2 Support allocation sinking for compound objects.
- Improve escape/alias analysis:
-- Storing an object into a field of another object does not mean that this object escapes (or has aliases) as long as that object does not have any loads from the same place;
-- Places like X.f and Y.f don't alias if X and Y are two different allocation instructions even if X and Y themselves potentially have aliases;
-- Improve precision of alias analysis for indexed properties;

- Support dematerialization and rematerialization of objects that are referenced by other dematerialized objects.
-- Use fix-point algorithm to collect candidates for allocation sinking;
-- Support aborting unsuccessful allocation sinking.

R=fschneider@google.com, johnmccutchan@google.com

Review URL: https://codereview.chromium.org//395943003

git-svn-id: https://dart.googlecode.com/svn/branches/bleeding_edge/dart@38404 260f80e4-7a28-3924-810f-c04153c831b5
2014-07-18 18:35:08 +00:00
fschneider@google.com bd110db87a Use range information for optimizing integer boxing and fix bug in range analysis.
1. When the input to a BoxInteger operation is in smi-range, we can eliminate
the range check and just perform a smi-tag operation on the input.

2. There was a bug in checking for smi-overflow for range boundaries: Calling Smi::IsValid
with a int64_t argument resulted in silent truncation of the input and therefore a wrong result.

Compiling with -Wconversion would have caught this, but currently we cannot compile with this flag
because of too many broken places.

Instead, I removed Smi::IsValid64 and created one variant Smi::IsValid that is specialized for the
input type with a template parameter. This way calling Smi::IsValid is always safe and will never
 result in silent alteration of the input argument.

R=vegorov@google.com

Review URL: https://codereview.chromium.org//353513002

git-svn-id: https://dart.googlecode.com/svn/branches/bleeding_edge/dart@37657 260f80e4-7a28-3924-810f-c04153c831b5
2014-06-24 13:41:48 +00:00
turnidge@google.com 18a8d7fe4c Fixes bug where we would occasionally materialize a corrupted object.
On the slow path for a stack overflow check we manually spill the live
registers before our call to the runtime function.  We were updating
all uses of these spilled registers in the Environment object but this
was not enough - we also need to update any uses of these spilled
registers in MaterializeObjectInstrs which are hanging off of the
Environment.

This would fail when a deoptimization was triggered during a stack
overflow check and a materialized object happened to refer to
registers.

I've been chasing this for a while and I was able to get it to
reliably reproduce by using the --stacktrace_every=N and the
--stacktrace_filter flags, which were added for this purpose.

ALSO:

Set the top context to null on exceptions/errors.  This is needed for
the --verify-incoming-context flag.

Minor printing improvements.

R=fschneider@google.com

Review URL: https://codereview.chromium.org//246303004

git-svn-id: https://dart.googlecode.com/svn/branches/bleeding_edge/dart@35443 260f80e4-7a28-3924-810f-c04153c831b5
2014-04-25 20:24:27 +00:00
fschneider@google.com 1fbec40a7f Change deoptimization description for materialized objects.
Use field offset instead of fields as arguments in the deoptimization
instructions for materialization of objects. This is to prepare for
materializing contexts and fixed-size arrays when doing allocation sinking.

R=johnmccutchan@google.com

Review URL: https://codereview.chromium.org//183683013

git-svn-id: https://dart.googlecode.com/svn/branches/bleeding_edge/dart@33267 260f80e4-7a28-3924-810f-c04153c831b5
2014-03-04 11:53:16 +00:00
johnmccutchan@google.com ed5894d0cf Unbox/Box Float64x2 and inline typed array loads and stores
R=fschneider@google.com

Review URL: https://codereview.chromium.org//172653002

git-svn-id: https://dart.googlecode.com/svn/branches/bleeding_edge/dart@32869 260f80e4-7a28-3924-810f-c04153c831b5
2014-02-20 20:41:54 +00:00
fschneider@google.com 15dfa97ab6 Fix bug and refactor optimized try-catch.
When untagged values flow into a exception-throwing instruction
 they need to be converted eagerly. Catch block entries assume all
 values are tagged at the entry.

 I changed the generation of moves at throwing
 instructions in try-catch: Instead of manuallu emitting moves,
 construct a parallel move and use the resolver to emit the native code.
 This eliminates a lot of duplicated code from all platforms.

 It will also allow to re-use stack space that is currently allocated
 separately for each individual try-catch in a function.

 I added a few more unrelated minor changes in various parts of the VM
 * Simpilify guard code generation
 * Resolve refactoring TODO in deoptimization
 * Improve names

R=johnmccutchan@google.com

Review URL: https://codereview.chromium.org//125103002

git-svn-id: https://dart.googlecode.com/svn/branches/bleeding_edge/dart@31536 260f80e4-7a28-3924-810f-c04153c831b5
2014-01-07 09:58:45 +00:00
fschneider@google.com e994d8b697 Revert r31326.
TBR=johnmccutchan@google.com,

Review URL: https://codereview.chromium.org//118133004

git-svn-id: https://dart.googlecode.com/svn/branches/bleeding_edge/dart@31332 260f80e4-7a28-3924-810f-c04153c831b5
2013-12-20 12:21:09 +00:00
fschneider@google.com fe726e1971 Use parallel move resolver for optimized try-catch.
This CL has no change in functionality and is purely clean up and
refactoring. Instead of manually generating the moves at throwing
instructions in try-catch, construct a parallel move and use the resolver
to emit the native code. This eliminates a lot of duplicated code
from all platforms.

It will also allow to re-use stack space that is currently allocated
separately for each individual try-catch in a function.

I added a few more unrelated minor changes in various parts of the VM
* Simpilify guard code generation
* Resolve refactoring TODO in deoptimization
* Improve names

R=johnmccutchan@google.com

Review URL: https://codereview.chromium.org//119213002

git-svn-id: https://dart.googlecode.com/svn/branches/bleeding_edge/dart@31326 260f80e4-7a28-3924-810f-c04153c831b5
2013-12-20 09:41:42 +00:00
johnmccutchan@google.com 29c01044bc This is the final breaking change in dart:typed_data needed for Dart 1.0. We need this change because the ECMAScript SIMD specification only includes Int32x4 and Int32x4List and our types must match.
R=srdjan@google.com

Review URL: https://codereview.chromium.org//56023004

git-svn-id: https://dart.googlecode.com/svn/branches/bleeding_edge/dart@29849 260f80e4-7a28-3924-810f-c04153c831b5
2013-11-04 21:56:59 +00:00
turnidge@google.com baeb23f9a5 Refactor some deoptimization code.
Primarily this change moves deoptimization state/code out of Isolate
and into DeoptContext (formerly DeoptimizationContext).  The lifetime
of DeoptizationContext changes to survive through the entire
deoptimization process.

Some minor renaming.  DeoptizationContext -> DeoptContext to make it
consistent with DeoptInstr and to save my wrists.

New files deferred_object.{cc,h} contain a bunch of the stuff lifted
from isolate.{cc,h}.

R=fschneider@google.com, srdjan@google.com

Review URL: https://codereview.chromium.org//24834002

git-svn-id: https://dart.googlecode.com/svn/branches/bleeding_edge/dart@28112 260f80e4-7a28-3924-810f-c04153c831b5
2013-10-01 19:34:12 +00:00