- Return the correct NaN when an invalid operation generates a NaN.
- When one or more operands are NaN, handle them as the processor
would, prioritising signalling NaNs and making them quiet.
- Fix fmadd and related instructions:
- Fnmadd is fma(-n, m, -a), not -fma(n, m, a).
- Some common libc implementations incorrectly implement fma for
zero results, so work around these cases.
- Replace some unreliable tests.
This patch also adds support for Default-NaN mode, since once all the
other work was done, it only required a couple of lines of code.
Default-NaN mode was used for an optimisation in ARM, and it should now
be possible to apply the same optimisation to A64.
BUG=
R=jochen@chromium.org
Review URL: https://codereview.chromium.org/199083005
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@19927 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
- Distinguish between context bound scripts (Script) and context unbound scripts
(UnboundScript).
- Add ScriptCompiler (which will later contain functions for async compilation).
This is a breaking change, in particular, Script::New no longer exists (it is
replaced by ScriptCompiler::CompileUnbound). Script::Compile remains as a
backwards-compatible shorthand for ScriptCompiler::Compile.
Passing CompilerOptions with produce_data_to_cache = true doesn't do anything
yet; the only way to generate the data to cache is the old preparsing API. (To
be fixed in the next version.)
This is a fixed version of https://codereview.chromium.org/186723005/
BUG=
R=dcarney@chromium.org
Review URL: https://codereview.chromium.org/199063003
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@19925 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
Attempting to re-use the type feedback vector stored in the
SharedFunctionInfo turns out to be difficult among the various cases.
It will be much easier to do this when deferred type feedback processing
is removed, as is in the works.
Created bug v8:3212 to track re-introducing the optimization of reusing
the type vector on recompile before optimization.
The CL also brings back the type vector on the SharedFunctionInfo.
BUG=351257
LOG=Y
R=bmeurer@chromium.org, bmeuer@chromium.org
Review URL: https://codereview.chromium.org/199973004
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@19919 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
- Distinguish between context bound scripts (Script) and context unbound scripts
(UnboundScript).
- Add ScriptCompiler (which will later contain functions for async compilation).
This is a breaking change, in particular, Script::New no longer exists (it is
replaced by ScriptCompiler::CompileUnbound). Script::Compile remains as a
backwards-compatible shorthand for ScriptCompiler::Compile.
Passing CompilerOptions with produce_data_to_cache = true doesn't do anything
yet; the only way to generate the data to cache is the old preparsing API. (To
be fixed in the next version.)
BUG=
R=dcarney@chromium.org
Review URL: https://codereview.chromium.org/186723005
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@19881 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
The escape analysis calculates the number of slots in an object as
no-of-slots = object-size / pointer-size. This gives 3 slots for
heap numbers on 32-bit architectures (one slot for the map, two for
the double value); however, my argument materialization code assumed
just two slots (map + value). Since Hydrogen allocates heap numbers
quite rarely, it is hard to produce a more meaningful repro than the
one provided by Clusterfuzz. Any suggestions are welcome.
The fix is simple - we just read out all extra slots (beyond the map
and the double) for heap numbers.
R=mstarzinger@chromium.org
BUG=351315
LOG=N
Review URL: https://codereview.chromium.org/196283004
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@19874 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
This is preparatory work to get rid of UnsafePersistent in blink.
The previous version had to be reverted due to timeouts in win32/Debug: https://codereview.chromium.org/197173002/
The timeouts happened because the STL version on that platform contains sanity checking code which opens a 'debug window' in the GUI, patiently waiting for the user to click ok/cancel/somethirdoption. It turns out, the cause for that debug window was totally valid and the test had a use-after-free issue.
The 1st patch set is the code as before. The 2nd patch set contains the fix.
Related blink changes are here: https://codereview.chromium.org/180363004/
This patch is largely based on https://codereview.chromium.org/175503003/, with some methods added to support the blink change mentioned above.
BUG=
R=dcarney@chromium.org
Review URL: https://codereview.chromium.org/197263002
Patch from Daniel Vogelheim <vogelheim@chromium.org>.
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@19873 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
It's possible to get a transitioned map with no links to the origin
map if it's a shared map. Code in KeyedStoreIC::StoreElementStub
assumes it can check if two maps are in the same family by
traversing the transition array. Long term, the "family" relationship
should be recognized with the Normalized Map Cache. For now, allow
the IC to remain monomorphic in this case if the receiver map and
the previous receiver map are the same.
Filed V8 issue 3210 (https://code.google.com/p/v8/issues/detail?id=3210)
to track the issue with the Normalized Map Cache.
BUG=350884
LOG=N
R=verwaest@chromium.org
Review URL: https://codereview.chromium.org/194623005
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@19847 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
