Now we remember new space's top pointer after the last GC to find out if there was a new space allocation since the last GC.
Unfortunately, this not completely safe - the debugger has a callback hook (that can call to JS) at the end of the GC epilogue that can in theory allocate and possibly make the heap non-iterable. We can only hope this does not happen.
BUG=373283
R=hpayer@chromium.org
LOG=N
Review URL: https://codereview.chromium.org/291193005
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@21431 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
The Android build system support for gyp has been fixed to handle
target-dependent host binaries correctly without requiring them to
include the target architecture in the name. Remove the suffixes to make
referring to these targets simpler again.
This reverts r14209.
BUG=
R=jkummerow@chromium.org
Review URL: https://codereview.chromium.org/236833004
Patch from Richard Coles <torne@chromium.org>.
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@21428 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
The change relative to the previous CL is a logic change in
DropActivationsInActiveThreadImpl. The previous CL skipped the matcher
unless the frame was a JS frame; this was correct for
MultipleFunctionTarget but not for SingleFrameTarget.
I have not been able to reproduce the original failures on either
architecture (ia32 or x64; stack frame dropping is unsupported on other
architectures).
R=yangguo@chromium.org
LOG=N
TEST=mjsunit/harmony/generators-debug-liveedit.js
BUG=
Review URL: https://codereview.chromium.org/270283002
Patch from Andy Wingo <wingo@igalia.com>.
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@21419 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
The reuse of CodeCreateEvent for deopt events caused a CodeCreateEvent
fired twice for a code object. When the event was processed for the first
time it seized the no-fp-ranges from code object, so the second event
had no ranges info leaving code entry without them.
As a result when a cpu profile sample falls into the region it missed the
2nd stack frame.
LOG=N
BUG=
R=bmeurer@chromium.org, loislo@chromium.org
Review URL: https://codereview.chromium.org/290093005
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@21418 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
Port r21380 (ef20a0a)
Original commit message:
This CL updates RelocInfo update operations and set_target_address_at to enable
skipping of the icache flush if it going to be batched up later.
Code::CopyFrom and Code::Relocate are modified to avoid individual icache
flushes since the whole code area will be flushed after the reloc info is
updated.
These changes reduce a regression when enabling the OOL constant pool on Arm,
since this change can cause MovT/MovW instructions for relocatable targets
if the constant pool is full.
Scores for Mandreel latency on a Nexus 5:
- OOL CP disabled: 3533
- OOL CP enabled, without this CL: 1825
- OOL CP enabled, with change: 3015
BUG=
R=plind44@gmail.com
Review URL: https://codereview.chromium.org/296723003
Patch from Balazs Kilvady <kilvadyb@homejinni.com>.
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@21390 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
This changes how Map/Set interacts with its iterators. When the
underlying table is rehashed or cleared, we create a new table (like
before) but we add a reference from the old table to the new table. We
also add an array describing how to transition the iterator from the
old table to the new table.
When Next is called on the iterator it checks if there is a newer table
that it should transition to. If there is, it updates the index based
on the previously recorded changes and finally changes itself to point
at the new table.
With these changes Map/Set no longer keeps the iterators alive. Also,
as before, the iterators keep the underlying table(s) alive but not the
actual Map/Set.
BUG=v8:1793
LOG=Y
R=mstarzinger@chromium.org, rossberg@chromium.org
Review URL: https://codereview.chromium.org/289503002
Patch from Erik Arvidsson <arv@chromium.org>.
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@21389 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
There's actually no need to have the transition as part of the HStoreNamedField instruction. In fact, it is cleaner and faster to generate a separate HStoreNamedField for the transition map. This will also help to eliminate map stores with store elimination, as well as reduce register pressure for transitioning stores on ia32.
R=hpayer@chromium.org
Review URL: https://codereview.chromium.org/295743002
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@21383 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
This CL updates RelocInfo update operations and set_target_address_at to enable
skipping of the icache flush if it going to be batched up later.
Code::CopyFrom and Code::Relocate are modified to avoid individual icache
flushes since the whole code area will be flushed after the reloc info is
updated.
These changes reduce a regression when enabling the OOL constant pool on Arm,
since this change can cause MovT/MovW instructions for relocatable targets
if the constant pool is full.
Scores for Mandreel latency on a Nexus 5:
- OOL CP disabled: 3533
- OOL CP enabled, without this CL: 1825
- OOL CP enabled, with change: 3015
R=rodolph.perfetta@arm.com, ulan@chromium.org
Review URL: https://codereview.chromium.org/284153004
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@21380 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
