ARM64: Faster immediate check and fix corner cases

ARM64: Faster immediate check and fix corner cases

Improve the code used to check for encodable logical immediates, fix some corner
cases associated with moving kWMinInt into W registers, and add tests.

BUG=
R=ulan@chromium.org

Committed: https://code.google.com/p/v8/source/detail?r=22120

[ulan, 2014-06-30 12:50:08]

lgtm

https://codereview.chromium.org/341123003/diff/1/src/arm64/assembler-arm64.cc
File src/arm64/assembler-arm64.cc (right):

https://codereview.chromium.org/341123003/diff/1/src/arm64/assembler-arm64.cc...
src/arm64/assembler-arm64.cc:2565: // Avoid making the assumption that the
most-significant 32 bits are zero by
Yes, this assumption will not hold if negate = true.

https://codereview.chromium.org/341123003/diff/1/src/arm64/assembler-arm64.cc...
src/arm64/assembler-arm64.cc:2634: }
ASSERT(d >= 2) or assert below that CountLeadingZeros(d, kXRegSizeInBits) - 57
is a valid index in multipliers array.

https://codereview.chromium.org/341123003/diff/1/src/arm64/utils-arm64.h
File src/arm64/utils-arm64.h (right):

https://codereview.chromium.org/341123003/diff/1/src/arm64/utils-arm64.h#newc...
src/arm64/utils-arm64.h:60: uint64_t LowestSetBit(uint64_t value);
The name suggests that the function returns a bit position instead of 2^(bit
position). Could you please either add a comment or rename it to something like
LargestPowerOf2Divisor.

https://codereview.chromium.org/341123003/diff/1/test/cctest/test-assembler-a...
File test/cctest/test-assembler-arm64.cc (right):

https://codereview.chromium.org/341123003/diff/1/test/cctest/test-assembler-a...
test/cctest/test-assembler-arm64.cc:599: __ Eor(w13, w0, kWMinInt);
Do we test whether IsImmLogical actually returns true for encodable immediates?

[m.m.capewell, 2014-07-01 09:45:46]

https://codereview.chromium.org/341123003/diff/1/src/arm64/assembler-arm64.cc
File src/arm64/assembler-arm64.cc (right):

https://codereview.chromium.org/341123003/diff/1/src/arm64/assembler-arm64.cc...
src/arm64/assembler-arm64.cc:2565: // Avoid making the assumption that the
most-significant 32 bits are zero by
Altered the comment to make this clear.

https://codereview.chromium.org/341123003/diff/1/src/arm64/assembler-arm64.cc...
src/arm64/assembler-arm64.cc:2634: }
Asserted the index to the multipliers array, as I thought the intent was clearer
this way.

https://codereview.chromium.org/341123003/diff/1/test/cctest/test-assembler-a...
File test/cctest/test-assembler-arm64.cc (right):

https://codereview.chromium.org/341123003/diff/1/test/cctest/test-assembler-a...
test/cctest/test-assembler-arm64.cc:599: __ Eor(w13, w0, kWMinInt);
There are a number of disassembler tests that check that a logical operation
using an immediate that can be encoded in one instruction actually uses the
expected instruction.

[ulan, 2014-07-01 13:05:20]

New changes lgtm.

> There are a number of disassembler tests that check that a logical operation
> using an immediate that can be encoded in one instruction actually uses the
> expected instruction.
OK.

[m.m.capewell, 2014-07-01 13:25:44]

Committed patchset #2 manually as r22120 (presubmit successful).

[m.m.capewell, 2014-07-01 13:47:28]

Fixing the missing UINT64_C definition to green the tree.