Add vstr and vldr floating point load and store to ARM assembler, disassemble...

Add vstr and vldr floating point load and store to ARM assembler, disassembler, and simulator.

Committed: http://code.google.com/p/v8/source/detail?r=3687

[William Hesse, 2010-01-21 16:20:02]

Simulator not done yet.  How do you unit test these?  Do you?

[Erik Corry, 2010-01-22 09:11:01]

Perhaps we can use the opportunity to clean up the disassembler and simulator to
use more meaningful names for the bit fields.  Page A7-24 and A7-26 have
overviews of the encodings of the VFP instructions.  Perhaps we could use that
to get rid of most of the bit(x, y) = z stuff.

http://codereview.chromium.org/545155/diff/1/4
File src/arm/disasm-arm.cc (right):

http://codereview.chromium.org/545155/diff/1/4#newcode1006
src/arm/disasm-arm.cc:1006: if (instr->Bit(23) == 1) {
This looks wrong to me.  Bit 23 can be either 1 or 0 for a vldr instruction
according to the manual.

http://codereview.chromium.org/545155/diff/1/4#newcode1018
src/arm/disasm-arm.cc:1018: if ((instr->Bits(27, 24) == 0xC) &&
This seems silly.  Most of these bits are already determined by the fact that
it's a type6 instruction.

[William Hesse, 2010-01-22 14:07:12]

Simulator and unit test added.  It looks done to me.

http://codereview.chromium.org/545155/diff/6001/6007
File test/cctest/test-assembler-arm.cc (right):

http://codereview.chromium.org/545155/diff/6001/6007#newcode283
test/cctest/test-assembler-arm.cc:283: CHECK_EQ(true, (dummy == dummy));
Changed to USE(dummy);

[Erik Corry, 2010-01-22 14:29:13]

LGTM

http://codereview.chromium.org/545155/diff/6001/6003
File src/arm/assembler-arm.cc (right):

http://codereview.chromium.org/545155/diff/6001/6003#newcode1376
src/arm/assembler-arm.cc:1376: int offset,
I feel strongly that the offset should be byte oriented as it is for other load
instructions.  With an assert that says it is divisble by 4.

http://codereview.chromium.org/545155/diff/6001/6005
File src/arm/simulator-arm.cc (right):

http://codereview.chromium.org/545155/diff/6001/6005#newcode2079
src/arm/simulator-arm.cc:2079: #ifdef BIG_ENDIAN_FLOATING_POINT
I don't think vfp and big endian floating point exist together except perhaps on
big endian ARM.  But V8 doesn't support big endian ARM.

[William Hesse, 2010-01-25 11:49:00]

http://codereview.chromium.org/545155/diff/6001/6003
File src/arm/assembler-arm.cc (right):

http://codereview.chromium.org/545155/diff/6001/6003#newcode1376
src/arm/assembler-arm.cc:1376: int offset,
On 2010/01/22 14:29:13, Erik Corry wrote:
> I feel strongly that the offset should be byte oriented as it is for other
load
> instructions.  With an assert that says it is divisble by 4.

Done.

http://codereview.chromium.org/545155/diff/6001/6005
File src/arm/simulator-arm.cc (right):

http://codereview.chromium.org/545155/diff/6001/6005#newcode2079
src/arm/simulator-arm.cc:2079: #ifdef BIG_ENDIAN_FLOATING_POINT
On 2010/01/22 14:29:13, Erik Corry wrote:
> I don't think vfp and big endian floating point exist together except perhaps
on
> big endian ARM.  But V8 doesn't support big endian ARM.

Done.