This adds the basics necessary to support float32 operations in TurboFan.
The actual functionality required to detect safe float32 operations will
be added based on this later. Therefore this does not affect production
code except for some cleanup/refactoring.
In detail, this patchset contains the following features:
- Add support for float32 operations to arm, arm64, ia32 and x64
backends.
- Add float32 machine operators.
- Add support for float32 constants to simplified lowering.
- Handle float32 representation for phis in simplified lowering.
In addition, contains the following (related) cleanups:
- Fix/unify naming of backend instructions.
- Use AVX comparisons when available.
- Extend ArchOpcodeField to 9 bits (required for arm64).
- Refactor some code duplication in instruction selectors.
BUG=v8:3589
LOG=n
R=dcarney@chromium.org
Review URL: https://codereview.chromium.org/1044793002
Cr-Commit-Position: refs/heads/master@{#27509}
The (32 - fraction_bits) value should be encoded so that the least
significant bit is set to bit 5 and the four next bits to bits 0-3. Fix
the previously incorrect encoding. This bug did not cause behavioral
issues before, since in existing uses of the function the order of the
bits in the immediate value does not matter, as they are all 1.
BUG=3256
LOG=N
R=ulan@chromium.org
Review URL: https://codereview.chromium.org/223623003
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@20508 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
Refactoring:
* consistent use of SmiTag/Untag
* added a few Smi macros and helpers
Improvements
* small optimisations (e.g. merging untag and cmp #0)
* added fixed point to double conversion instructions for simpler conversions
More on the last point: a Smi can be seen as a fixed point number with the
a one bit fractional part. Fixed to double instructions allow us to convert
a Smi to a double without untagging.
BUG=none
TEST=none
Review URL: https://chromiumcodereview.appspot.com/15085026
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@14724 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
I tried to limit the use of v8::Isolate::GetCurrent() and v8::internal::Isolate::Current() as much as possible, but sometimes this would have involved restructuring tests quite a bit, which is better left for a separate CL.
BUG=v8:2487
Review URL: https://codereview.chromium.org/12716010
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@13953 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
First of all, it has nothing to do with Isolates, it is related to the assembler
at hand. Furthermore, the saving/restoring is platform-independent. Cleaned up
some platform-specific stuff on the way.
Note that there are some things which still need some cleanup, like e.g. using
EnumSet instead of uint64_t, making Probe() more uniform across platforms etc.,
but the CL is already big enough.
BUG=v8:2487
Review URL: https://codereview.chromium.org/12391055
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@13823 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
Previously, we would disassemble some VFP instructions like this:
vmla.f64eq d16, d17, d18
This patch moves the condition to the right place:
vmlaeq.f64 d16, d17, d18
Spotted by Rodolph Perfetta!
BUG=none
Review URL: https://chromiumcodereview.appspot.com/12335129
Patch from Hans Wennborg <hans@chromium.org>.
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@13752 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
The assembler has 8 different vmov variants. The one for vmov.32 and for moving
an immediate into a double reg only differs in the type of the second
paremeter: vmov.32 takes an int, the other takes a double.
The situation is dangerous because C++ will happily implicitly convert between
int and double.
This patch changes the signature of the vmov.32 assembler function so that it
cannot be confused with the other vmovs.
BUG=none
Review URL: https://chromiumcodereview.appspot.com/12255031
Patch from Hans Wennborg <hans@chromium.org>.
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@13668 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
When calculating Math.pow where the exponent is a smi use a simple loop to calculate the result.
Added support for the vmov instruction moving from one doubleword extension register to another.
Added some Math.pow tests which partially covers what is in the Sputnik tests.
Review URL: http://codereview.chromium.org/2804033
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@4990 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
Zhang Kun. For now we only emit movw and movt in places where no
relocation is needed. Small performance boost (around 0.5%).
Also adds support for turning ALU operations (eor etc.) with
large immediates into mvn or movw followed by a register-based
ALU operation.
Review URL: http://codereview.chromium.org/2821014
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@4913 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
The bkpt instruction is now supported by both the disassembler and the simulator. In the simulator it breaks to the native debugger if any like int3 on Intel.
Moved the handling of the miscellaneous instructions to a separate part.
Review URL: http://codereview.chromium.org/1576026
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@4360 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
