This re-applies r4220 and r4233, which was reverted in r4254 due to a bug. This bug has now been fixed, with the only change being line 2884 changed from
__ SmiTag(left_side->reg());
to
__ SmiTag(operand->reg());
Added a regression test.
BUG=http://crbug.com/39160
TEST=test/mjsunit/regress/regress-crbug-39160.js
Review URL: http://codereview.chromium.org/1251009
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@4261 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
Instead of testing the value of a constant frame element to determine
the type we compute its type information at construction time.
This speeds up querying the type information during code generation.
This change also adds support for Integer32 constants and sets
the type information accordingly.
Review URL: http://codereview.chromium.org/1277001
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@4256 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
Previously we marked compare operations that occurred as for loop
conditions so as to avoid inlining the floating-point code for them.
Begin marking all expressions that occur as any loop condition
(because they are never dead code).
Review URL: http://codereview.chromium.org/1074009
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@4227 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
Inline floating point compare instead of calling the stub when the following conditions are met:
* Code is in a loop
* Compare is not a for loop condition
* Compare is not an equal comparison
This inlined code handles heap number to heap number and heap number to smi compare. It can also handle smi to smi compare, but whenever there is a chance of comparing two smis the smi compare is inlined before the inlined floating point compare. Support for non SSE2 hardware is included.
A new set of variants of the compare stub without the floating point comparison code is called if the inline comapre fails due to the operands not beeing heap numbers or smis.
The virtual frame has been extended with a branch taking two live results to be carried through to the destination. This makes this change much simpler as the inlined code have two live results in registers and a number of bailouts.
CompareStub::GetName needs to be updated as well. I will do that as a separate change.
Also inlined equality check if both operands can't be NaN. This can only provide positive equals if it is the same object.
Review URL: http://codereview.chromium.org/1117011
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@4220 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
We may encounter an invalid frame after generating code
for the loop body in case the loop body ends in an unconditional
return. Before setting the type information for the loop variable
we need to check for a valid frame.
Review URL: http://codereview.chromium.org/1106002
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@4182 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
The inline runtime functions are now included in the fuzzing of the natives. The chack for the expected number of arguments passed have been moved to the parser which will generate a syntax error if a runtime function (either C++ or inline) is called with a different number of arguments than expected.
Review URL: http://codereview.chromium.org/573056
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@4096 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
This change adds a pass over the AST that computes the
set of assigned variables for locals and parameters for each expression.
The result of this analysis is used to for two purposes:
1. Recognize variables that are trivial subexpressions. A left sub-expression
of a binary operation is trivial if it is a local variable or a parameter
and it is not assigned in the right sub-expression. In the case of a
trivial left sub-expression we evaluate the right first.
Currently only binary operations and compare operations are considered
when finding trivial left sub-expressions.
2. Recogize certain simple for-loops with a constant trip count where the loop
variable is always within smi range. If the loop count variable is not
assigned in the body of the loop (except in the update expression the
for-loop). This allows omitting smi checks on operation using the loop
count variable.
Review URL: http://codereview.chromium.org/669155
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@4087 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
Introducing a virtual-frame-inl.h file containing some platform-independent
virtual frame function which are small enough to be inlined.
Removed unnecessary #include of virtual-frame.h from register-allocator-inl.h
and added the necessary explicit includes in a number of files.
Review URL: http://codereview.chromium.org/660104
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@3962 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
A list of bailout descriptions is kept in the CompilationInfo
structure that is shared between the primary and secondary code
generators. The primary adds a description to the list for each
bailout position.
Responsibility for binding labels is moved from the primary to the
secondary code generator. All the labels still target the start of the
secondary code and the compilation state of the primary is still
ignored.
Move the compilation mode flag to the CompilationInfo.
Review URL: http://codereview.chromium.org/651031
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@3920 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
Add a (currently) syntactic predicate to AST expression nodes telling
whether they are 'trivial'. Trivial expressions have no side effects,
do not require storage to be allocated for them, and can be evaluated
out of order (because their value does not change between when they
are visited by the code generator as expressions in the AST and when
it is consumed).
Mark 'this' and literals as trivial. Allow them to be pushed on the
virtual frame. Make use of them to push 'this' more lazily in this
property assignments.
Review URL: http://codereview.chromium.org/647018
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@3906 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
Fix for issue 603.
Revision r3484 removed the property name from the call stack for
call ICs. When a non-function was called via a call IC and
Function.prototype.call, an extra value was left on the stack that the
caller could not know to clean up.
Fix is to change the JS builtin used for calling non-functions. It
now gets the callee as receiver, rather than iterating stack frames
and finding it on the expression stack of its JS caller.
Review URL: http://codereview.chromium.org/604064
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@3882 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
Currently we only pass a boolean parameter indicating whether
the input operands to the GenericBinaryOpStub are guaranteed
to be numbers or not.
Instead we pass the complete number type as a parameters. This
allows to use more precise type information for code generation
in the stub.
Also make the computation of the result type more precise and correct on both ia32 and x64.
Review URL: http://codereview.chromium.org/593110
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@3873 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
Each frame element gets a new attribute with number type information. A frame element can be:
- smi
- heap number
- number (i.e. either of the above)
- or something else.
The type information is propagated along with all virtual frame operations.
Results popped from the frame carry the number information with them.
Two optimizations in the code generator make use of the new
information:
- GenericBinaryOpSyub omits map checks if input operands are numbers.
- Boolean conversion for numbers: Emit inline code for converting a number (smi or heap number) to boolean. Do not emit call to ToBoolean stub in this case.
Review URL: http://codereview.chromium.org/545007
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@3861 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
All two character string results from adding two strings and from sub string used to be handled in the runtime system as a lookup in the symbol table was done before allocating a new string. The native code for string add and sub string now probes the symbol cache for two character strings to avoid the runtime call. If the result string is not found in the symbol table within a fixed number of probes a new string is just allocated. Newly allocated two character strings are not added to the symbol table immediately.
Review URL: http://codereview.chromium.org/598062
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@3842 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
bailouts. For now the virtual-frame state at entry of a function is
hard-coded when using the virtual-frame based compiler as the
secondary compiler.
Setup frame pointer correctly on function entry on ARM in
fast-codegen-arm.
Review URL: http://codereview.chromium.org/569010
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@3776 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
on ia32.
1. Operate on the values in edx,eax when possible (all operations
except DIV and MOD). This saves moving them on entry and when falling
out to the non-smi code.
2. Do not perform ADD and SUB before the smi check of their inputs.
This saves undoing the operation in the case that we fall through to
the non-smi case due to non-smi inputs (probably common?), and we can
avoid emitting the smi check code twice (code size reduction).
3. Don't perform OR twice (once to smi check the inputs and once to
smi check the result).
Review URL: http://codereview.chromium.org/556019
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@3712 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
As the start index is already passed it is easy to calculate the "at start" boolean in generated code. Also as direct entry has been implemented this needs to be done in generated code anyway, and therefore might as well be moved to the generated code for RegExp. The "at start" value is now calcualted as a local variable on the native RegExp frame based on the value of the start index argument.
The x64 version have been tested on both Linux and 64-bit Windows Vista.
For ARM I have tested cctest/test-regexp on ARM hardware, but the rest of the tests have only been run on the ARM simulator.
Review URL: http://codereview.chromium.org/554078
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@3709 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
Currently arguments are never passed on registers (due to the way ArgsInRegistersSupported is written) and
if they were, the stub would break in several places because registers are not preserved properly in the
course of execution. This CL makes use of registers more often (than never) and makes sure that registers are
handler properly.
A peformance gain is small (0.2-0.3%) but stable.
This CL was extracted from the one sent out earlier (http://codereview.chromium.org/551093).
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@3692 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
This is a preparation step for including number type information
in the virtual frame. We need a common place where we can update
the number type information of the result of a binary operation
since we should not modify the state of the virtual frame elements
directly.
Review URL: http://codereview.chromium.org/551080
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@3661 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
When a function is called with a value type as the receiver this is now boxed as an object.
This is a low-impact solution where the receiver is only boxed when required. For IC calls to the V8 builtins values are not boxed and as most of the functions on String.prototype, Number.prototype and Boolean.prototype are sitting there most IC calls on values will not need any boxing of the receiver.
For calls which are not IC calls but calls through the CallFunctionStub a flag is used to determine whether the receiver might be a value and only when that is the case will the receiver be boxed.
No changtes to Function.call and Function.apply - they already boxed values. According to the ES5 spec the receiver should not be boxed for these functions, but current browsers have not adopted that change yet.
BUG=223
TEST=test/mjsunit/value-wrapper.js
TEST=test/mjsunit/regress/regress-crbug-3184.js
Review URL: http://codereview.chromium.org/542087
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@3617 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
When initializing the special local variable containing the reference to the enclosing
function in named functions we now (correctly) emit an INIT_CONST instead of INIT_VAR,
and we correctly bail out in the top-level code generator.
Also part of this change is adding missing statement position information
for some statements in the top-level code generator.
Review URL: http://codereview.chromium.org/536029
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@3587 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
* Rename the operation of filling a context with a value. Formerly it
was 'Move', now it's 'Apply' so as to avoid confusion with various
other Moves (eg, in the toplevel codegen, in the macro assemblers).
* Use the abstraction Drop rather than math on the stack pointer.
* Add a predicate on AST expression nodes to abstract a repeated test
whether we should us a regular (named) IC or a keyed IC.
Review URL: http://codereview.chromium.org/546006
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@3580 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
Added newly added instructions to test-disasem-ia32.cc and implemented the missi
ng ones in the disasembler.
Added some asserts to 8-bit instructions which only work with eax, ebx, ecx and
edx (al, bl, cl and dl).
Removed the loope instruction.
Review URL: http://codereview.chromium.org/548002
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@3577 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
to floating point and then converting back we convert directly to a
32 bit integer. In addition the bit twiddling implementation of float-
to-integer conversion has been ported from ARM. Testing has shown that
this runs faster than the x87 or SSE3 rounding instructions. This change
is IA32 only. There may be a smaller benefit from doing the same on x64.
Review URL: http://codereview.chromium.org/506052
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@3492 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
The different length string types was used to encode the string length and the hash in one field. This is now split into two fields one for length and one for hash. The hash field still encodes the array index of the string if it has one. If an array index is encoded in the hash field the string length is added to the top bits of the hash field to avoid a hash value of zero.
On 32-bit this causes an additional 4 bytes to be used for all string objects. On 64-bit this will be half on average dur to pointer alignment.
Review URL: http://codereview.chromium.org/436001
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@3350 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
Summary:
This change fixes a performance regression introduced by the special
handling of regular expressions in typeof expressions.
As a result we regain ~8% speedup on 3d-raytrace and ~13% on boyer
(vs bleeding edge)
Description:
The macros IS_OBJECT and IS_FUNCTION are frequently used in the
JS runtime functions.
By introducing new inlined runtime functions %_IsFunction and %_IsObject
we avoid invoking the more expensive %_ClassOf function plus comparing
its result to a string.
Review URL: http://codereview.chromium.org/399111
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@3335 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
Change name of shifts picking the shift count from cl to sal_cl, shl_cl and shr_cl.
Add special encoding of shift by one for shr which was missing it.
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@3314 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
This change fixes the problem with the original version of this approach
(r3032) that may lead to a corrupted stack if we would invoke spilling during
syncing a large SMI constant (unsafe SMIs) in the virtual frame.
The new code for storing unsafe SMI constants does not use an extra temporary
register. This prevents the compiler from ever having to spill during a
virutal frame sync operation.
For storing a large SMI constant we previously generated:
mov ecx, (large_smi & 0x0000ffff)
xor ecx, (large_smi & 0xffff0000)
push ecx
we now generate:
push (large_smi & 0x0000ffff)
or [esp], (large_smi & 0xffff0000)
Not using a temporary register avoids spilling within an nvocation
of VirtualFrame::SyncRange.
Review URL: http://codereview.chromium.org/391079
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@3313 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
