r5147 wrongly assumed that a code cache for a slow case map is always empty.
This patch solves this: whenever we attempt to add a stub to a map's code cache
we check that this map is cached. If it is we give the object its own copy
of the map and only then modify the map.
Review URL: http://codereview.chromium.org/3134027
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5342 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
Object model changes
----------------------------------------
New fixed_cow_array_map is used for the elements array of a JSObject
to mark it as COW. The JSObject's map and other fields are not
affected. The JSObject's map still has the "fast elements" bit set. It
means we can do only the receiver map check in keyed loads and the
receiver and the elements map checks in keyed stores. So introducing
COW arrays doesn't hurt performance of these operations. But note that
the elements map check is necessary in all mutating operations because
the "has fast elements" bit now means "has fast elements for reading".
EnsureWritableFastElements can be used in runtime functions to perform
the necessary lazy copying.
Generated code changes
----------------------------------------
Generic keyed load is updated to only do the receiver map check (this
could have been done earlier). FastCloneShallowArrayStub now has two
modes: clone elements and use COW elements. AssertFastElements macro
is added to check the elements when necessary. The custom call IC
generators for Array.prototype.{push,pop} are updated to avoid going
to the slow case (and patching the IC) when calling the builtin should
work.
COW enablement
----------------------------------------
Currently we only put shallow and simple literal arrays in the COW
mode. This is done by the parser.
Review URL: http://codereview.chromium.org/3144002
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5275 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
The IC stub is completely generic, so there will only be one such stub
in the system.
Added a new overloaded version of the macro assembler RecordWrite
method for cases where we have the address we store to computed up
front.
Review URL: http://codereview.chromium.org/2804029
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@4991 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
Similar or duplicate checks are scattered around the code before doing the dictionary load.
Also the entire branch in GenerateCallNormal that handles global/builtin receiver is
guaranteed to bail out from GenerateDictionaryLoad, so there is no point in generating it at all.
The purpose of the patch is:
- making C++ code more compact and transparent,
- not generating dead code.
There is a tiny performance gain. The patch is ia32 only for now.
Please tell me if I am missing anything.
Review URL: http://codereview.chromium.org/2801007
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@4926 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
* Faster hashing for sequential strings.
* When adding short external two-byte strings try to convert them
back to ascii. Chances are high the embedder uses two-byte
representation even for ascii strings. This optimization saves
memory and makes hashing faster.
Review URL: http://codereview.chromium.org/1444001
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@4300 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
NumberToString in runtime JavaScript is inlined through a call to a stub. Currently the stub only checks the number string cache and only if the number is a smi. Code is shared with the inlining of number string cache lookup when adding a smi to a string.
Review URL: http://codereview.chromium.org/604062
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@3865 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
Added a stub to allocate and fill a string object with a substring from another string.
Use the rep movs instruction to copy the string data as it turned out to be the fastest way.
While preparing this I experimented with some SSE2 instructions, so the instructions movdqa and movdqu are still in the IA-32 assembler even though they are not used.
Review URL: http://codereview.chromium.org/525085
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@3554 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
Calls to RegExp no longer have to be via a call to the runtime system. A new stub have been added which can handle this call in generated code. The stub checks all the parameters and creates RegExp entry frame in the same way as it is created by the runtime system. Bailout to the runtime system is done whenever an uncommon situation is encountered or when the static data used is not initialized. After running the native RegExp code the last match info is updated like in the runtime system.
Currently only ASCII strings are handled.
Added another argument to the RegExp entry frame. It indicated whether the call is direct from JavaScript code or through the runtime system. This information is used when RegExp execution is interrupted. If an interruption happens when RegExp code is called directly a retry is issued causing the interruption to be handled via the runtime system. The reason for this is that the direct call to RegExp code does not support garbage collection.
Review URL: http://codereview.chromium.org/521028
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@3542 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
Instead of weak handles external strings use a separate table. This
table uses 5 times less memory than weak handles. Moreover, since we
don't have to follow the weak handle callback protocol we can collect
the strings faster and even on scavenge collections.
Review URL: http://codereview.chromium.org/467037
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@3439 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
This adds a code stub which can do most of what Heap::AllocateConsString can do. It bails out if the result cannot fit in new space or if the result is a short (flat) string and one argument is an ascii string and the other a two byte string. It also bails out if adding two one character strings as Heap::AllocateConsString has special handling of this utilizing the symbol table. The stub is used both for the binary add operation and for StringAdd calls from runtime JavaScript files. Extended the string add test to cover all sizes of flat result stings.
Review URL: http://codereview.chromium.org/442024
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@3400 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
specification under development. The optimizations are patterned after
those previously done for CanvasPixelArray. This CL adds all of the
necessary framework but continues to use the generic KeyedLoadIC and
KeyedStoreIC code, to create a baseline for benchmarking purposes. The
next CL will add the optimized ICs to ic-ia32.cc and ic-x64.cc.
These new CanvasArray types have different semantics than
CanvasPixelArray; out-of-range values are clamped via C cast
semantics, which is cheaper than the clamping behavior specified by
CanvasPixelArray. Out-of-range indices raise exceptions instead of
being silently ignored.
As part of this work, pulled FloatingPointHelper::AllocateHeapNumber
up to MacroAssembler on ia32 and x64 platforms. Slightly refactored
KeyedLoadIC and KeyedStoreIC. Fixed encoding for fistp_d on x64 and
added a few more instructions that are needed for the new ICs. The
test cases in test-api.cc have been verified by hand to exercise all
of the generated code paths in the forthcoming specialized ICs.
Review URL: http://codereview.chromium.org/293023
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@3096 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
fast code generator is optimized for compilation time and code size.
Currently it is only implemented on IA32. It is potentially triggered
for any code in the global scope (including code eval'd in the global
scope). It performs a syntactic check and chooses to compile in fast
mode if the AST contains only supported constructs and matches some
other constraints.
Initially supported constructs are
* ExpressionStatement,
* ReturnStatement,
* VariableProxy (variable references) to parameters and
stack-allocated locals,
* Assignment with lhs a parameter or stack-allocated local, and
* Literal
This allows compilation of literals at the top level and not much
else.
All intermediate values are allocated to temporaries and the stack is
used for all temporaries. The extra memory traffic is a known issue.
The code generated for 'true' is:
0 push ebp
1 mov ebp,esp
3 push esi
4 push edi
5 push 0xf5cca135 ;; object: 0xf5cca135 <undefined>
10 cmp esp,[0x8277efc]
16 jnc 27 (0xf5cbbb1b)
22 call 0xf5cac960 ;; code: STUB, StackCheck, minor: 0
27 push 0xf5cca161 ;; object: 0xf5cca161 <true>
32 mov eax,[esp]
35 mov [ebp+0xf4],eax
38 pop eax
39 mov eax,[ebp+0xf4]
42 mov esp,ebp ;; js return
44 pop ebp
45 ret 0x4
48 mov eax,0xf5cca135 ;; object: 0xf5cca135 <undefined>
53 mov esp,ebp ;; js return
55 pop ebp
56 ret 0x4
Review URL: http://codereview.chromium.org/273050
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@3067 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
The calls to GenericBinaryStub can now pass the arguments in registers instead of on the stack. It is supported for ADD, SUB, MUL and DIV. The convention in GenericBinaryStub is not changed so the left operand is passed in edx and the right one in eax. When the stub contains smi code arguments are always passed on the stack as the smi code has to have left and right operands on eax and ebx, so moving from edx,eax to eax,ebx is not worth it and the smi code also trashes the registers so if arguments where passed in registers they would have to be saved on the stack anyway.
Added flags to disable the use of certain Intel CPU features to make it easier to test different code paths.
Review URL: http://codereview.chromium.org/246075
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@3041 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
The construction of arrays when using the the Array function either as a constructor or a normal function is now handled fully in generated code in most cases. Only when Array is called with one argument which is either negative or abowe JSObject::kInitialMaxFastElementArray (which is currently 1000) or if the allocated object cannot fit in the room left in new space is the runtime system entered.
Two new native code built-in functions are added one for normal invocation and one for the construct call. The existing C++ builtin is renamed, but kept. When the normal invocation cannot be handled in generated code the C++ builtin is called. When the construct invocation cannot be handled in native code the generic construct stub is called (which will end up in the C++ builtin through a construct trampoline).
One thing that might be changed is preserving esi (constructor function) during the handling of a construct call. We know precisily what function we where calling anyway and can just reload it. This could remove the parameter construct_call to ArrayNativeCode and remove the handling of this from that function.
The X64 and ARM implementations are not part of this changelist.
Review URL: http://codereview.chromium.org/193125
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@2899 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
For objects which only have simple assignments of the form this.x = ...; a
specialized constructor stub is now generated. This generated code allocates the
object and fills in the initial properties directly. If this fails for some
reason code continues in the generic constructor stub which in turn might pass
control to the runtime system.
Added counter to see how many objects are constructed using a specialized stub.
The specialized stub is only implemented for ia32 architecture in this change.
For x64 and ARM the generic construct stub is used.
This is change is identical to http://codereview.chromium.org/174392 (committed in r2753 and reverted in r2754) except that a few parts have already been committed from http://codereview.chromium.org/173469 (committed in r2762).
Review URL: http://codereview.chromium.org/173470
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@2764 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
For objects which only have simple assignments of the form this.x = ...; a specialized constructor stub is now generated. This generated code allocates the object and fills in the initial properties directly. If this fails for some reason code continues in the generic constructor stub which in turn might pass control to the runtime system.
Added counter to see how many objects are constructed using a specialized stub.
The specialized stub is only implemented for ia32 architecture in this change. For x64 and ARM the generic construct stub is used.
Review URL: http://codereview.chromium.org/174392
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@2753 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
During parsing functions are analyzed for statements of the form this.x = ...;. These assignments are categorized in two types: simple and non simple. The simple ones are where the right hand side is known to be either a constant or an argument to the function. If a function only contains statements of this type the property names are collected and for the simple assignments the index of the argument or the constant value assigned are stored as well.
When the initial map for a function is created and the function consists of only this type of assignemnts the initial map is created with a descriptor array describing these properties which will be known to always exist in an object created from the function.
The information on this property assignments is not collected during pre-parsing so if compiling using pre-parse data these optimization hints are not available.
Next step will be to use the information collected for the simple assignments to generate constructor code which will create and initialize the object from this information without calling the code for the function.
Review URL: http://codereview.chromium.org/172088
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@2710 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
This issue was raised by Brett Wilson while reviewing my changelist for readability. Craig Silverstein (one of C++ SG maintainers) confirmed that we should declare one namespace per line. Our way of namespaces closing seems not violating style guides (there is no clear agreement on it), so I left it intact.
Review URL: http://codereview.chromium.org/115756
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@2038 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
an object that holds a setter. If there are no store ics then no
flushing is done. The implementation has been tweaked so that no ICs
are cleared during normal context creation.
This may cost us some performance but I'm submitting it as it is and
if there are problems we can either decide to be smarter about when,
what and/or how we clear, or back this change out altogether.
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@1509 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
calling a stub. The map to check against is unknown when generating
the code, so we patch the map check in the IC initialization code.
Loop nesting is currently not tracked on ARM. I'll file feature
request bug reports for implementing this on ARM and add the number to
the TODOs before I commit.
Review URL: http://codereview.chromium.org/16409
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@1015 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
