It was only used for Math.log, and even then only in full code and in %_MathLog. For crankshafted code, Intel already used the FP operations directly, while the ARM/MIPS ports were a bit lazy and simply called the stub. The latter directly call the C library now without any cache. It would be possible to directly generate machine code if somebody has the time, from what I've seen out in the wild it should be only about a dozen instructions.
LOG=y
R=yangguo@chromium.org
Review URL: https://codereview.chromium.org/113343003
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@18344 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
This removes tons of architecture-specific code and makes it easy to
experiment with other pseudo-RNG algorithms. The crankshafted code is
extremely good, keeping all things unboxed and doing only minimal
checks, so it is basically equivalent to the handwritten code.
When benchmarks are run without parallel recompilation, we get a few
percent regression on SunSpider's string-validate-input and
string-base64, but these benchmarks run so fast that the overall
SunSpider score is hardly affected and within the usual jitter. Note
that these benchmarks actually run even faster when we don't
crankshaft at all on the main thread (the regression is not caused by
bad code, it is caused by Crankshaft needing a few hundred microsecond
for compilation of a trivial function). Luckily, when parallel
recompilation is enabled, i.e. in the browser, we see no regression at
all!
R=mstarzinger@chromium.org
Review URL: https://codereview.chromium.org/68723002
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@17955 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
To keep the structure of the serializer more or less untouched, we use
some ingenious Corry-approved(TM) 3-step technology (a.k.a. "hack"):
* Create copies of code objects.
* Wipe out all absolute addresses in these copies.
* Write out the cleaned copies instead of the originals.
In conjunction with --random-seed, our snapshots are reproducible now.
BUG=v8:2885
R=bmeurer@chromium.org, erik.corry@gmail.com
Review URL: https://codereview.chromium.org/54823002
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@17473 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
Previously, the result of target_reference_address() could only be
read, writing to it would have had an architecture-dependent effect,
e.g. writing into the code on ia32, a no-op on arm, etc.
This refactoring-only CL turns this into a simple getter, making it
impossible to use incorrectly.
More to come...
R=bmeurer@chromium.org
Review URL: https://codereview.chromium.org/46583006
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@17467 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
This change means that code which is never executed is garbage collected immediately, and code which is only executed once is collected more quickly (limiting heap growth), however, code which is re-executed is reset to the young age, thus being kept around for the same number of GC generations as currently.
BUG=280984
R=danno@chromium.org, hpayer@chromium.org
Review URL: https://codereview.chromium.org/23480031
Patch from Ross McIlroy <rmcilroy@chromium.org>.
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@17343 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
New abstract class CodeEventListener was created.
CodeEventLogger which is the base class for Jit, LowLevel
and CodeAddressMap loggers was inherited from CodeEventListener.
CodeAddressMap class was moved to serializer.cc because serializer is the only user for it. Actually it collects code names and pushes them to the standard log as SnapshotCodeNameEvent. So I extracted this code into separate function CodeNameEvent. It happens that this method works only when Serializer serializes an object. So I added direct log call there.
CodeEventLogger class declaration was moved to the header
because CodeAddressMap needs it.
The code for the nested class CodeEventLogger::NameBuffer was left in the cc file.
CpuProfiler now is inherit CodeEventListener but not used
the loggers infrastructure yet due to the complex initialization schema. I'd like to fix that in a separate cl.
BUG=none
TEST=current test set.
R=yangguo@chromium.org, yurys@chromium.org
Review URL: https://codereview.chromium.org/19724007
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@15911 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
* Allocations of AllocationSites occur in generated code, so generated code needs to be able to add to the list. For now I have a special hydrogen instruction, though it would be nice to use general purpose instructions.
* The snapshot contains AllocationSites, and these need to be re-threaded into the list on deserialization.
Something nice is that the AllocationSites are only created in old space, so a special new space visitor isn't required.
BUG=
R=mstarzinger@chromium.org
Review URL: https://codereview.chromium.org/18173013
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@15715 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
in preparation of the introduction of ES6 'symbols' (aka private/unique names).
The SymbolTable became the StringTable. I also made sure to adapt all comments. The only remaining use of the term "symbol" (other than unrelated uses in the parser and such) is now 'NewSymbol' in the API and the 'V8.KeyedLoadGenericSymbol' counter, changing which might break embedders.
The one functional change in this CL is that I removed the former 'empty_string' constant, since it is redundant given the 'empty_symbol' constant that we also had (and both were used inconsistently).
R=yangguo@chromium.org
BUG=
Review URL: https://codereview.chromium.org/12210083
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@13781 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
Unified parameter order of CreateHandle with the rest of v8 on the way. A few
Isolate::Current()s had to be introduced, which is not nice, and not every place
will win a beauty contest, but we can clean this up later easily in smaller steps.
Review URL: https://codereview.chromium.org/12300018
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@13717 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
This includes:
* Adding support for saving callee-clobbered double registers in Crankshaft code.
* Adding a new "HTrapAllocationMemento" hydrogen instruction to handle AllocationSiteInfo data in crankshafted stubs.
* Adding a new "HAllocate" hydrogen instruction that can allocate raw memory from the GC in crankshafted code.
* Support for manipulation of the hole in HChange instructions for Crankshafted stubs.
* Utility routines to manually build loops and if statements containing hydrogen code.
Review URL: https://codereview.chromium.org/11659022
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@13585 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
When code objects in the heap for FUNCTIONs and OPTIMIZED_FUNCTIONs are marked by the GC, their prologue is patched with a call to a stub that removes the patch. This allows the collector to quickly identify code objects that haven't been executed since the last full collection (they are the ones that sill contain the patch). The functionality is currently disabled, but can be activated by specifying the "--age-code".
R=mstarzinger@chromium.org
Review URL: https://codereview.chromium.org/10837037
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@12898 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
the speed of deserializing code. The current startup
time improvement for V8 is around 6%, but code deserialization
is speeded up disproportionately, and we will soon have more
code in the snapshot.
* Removed support for deserializing into large object space.
The regular pages are 1Mbyte now and that is plenty. This
is a big simplification.
* Instead of reserving space for the snapshot we actually
allocate it now. This removes some special casing from
the memory management and simplifies deserialization since
we are just bumping a pointer rather than calling the
normal allocation routines during deserialization.
* Record in the snapshot how much we need to boot up and
allocate it instead of just assuming that allocations in
a new VM will always be linear.
* In the snapshot we always address an object as a negative
offset from the current allocation point. We used to
sometimes address from the start of the deserialized data,
but this is less useful now that we have good support for
roots and repetitions in the deserialization data.
* Code objects were previously deserialized (like other
objects) by alternating raw data (deserialized with memcpy)
and pointers (to external references, other objects, etc.).
Now we deserialize code objects with a single memcpy,
followed by a series of skips and pointers that partially
overwrite the code we memcopied out of the snapshot.
The skips are sometimes merged into the following
instruction in the deserialization data to reduce dispatch
time.
* Integers in the snapshot were stored in a variable length
format that gives a compact representation for small positive
integers. This is still the case, but the new encoding can
be decoded without branches or conditional instructions,
which is faster on a modern CPU.
Review URL: https://chromiumcodereview.appspot.com/10918067
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@12505 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
