Adds intrinsics for arm64.

Adds intrinsics for arm64.

Also enables tests that no longer time-out.

R=regis@google.com

Committed: https://code.google.com/p/dart/source/detail?r=36330

[regis, 2014-05-19 20:14:02]

LGTM

https://codereview.chromium.org/285403004/diff/1/runtime/vm/assembler_arm64.h
File runtime/vm/assembler_arm64.h (right):

https://codereview.chromium.org/285403004/diff/1/runtime/vm/assembler_arm64.h...
runtime/vm/assembler_arm64.h:796: addw(rd, ZR, Operand(rn, LSR, shift));
Is a shift amount > 31 allowed? Are all 64 bits of rn taken into consideration?

https://codereview.chromium.org/285403004/diff/1/runtime/vm/intermediate_lang...
File runtime/vm/intermediate_language_arm64.cc (right):

https://codereview.chromium.org/285403004/diff/1/runtime/vm/intermediate_lang...
runtime/vm/intermediate_language_arm64.cc:2460: __ b(overflow, HI);
Thanks for implementing my suggestion!

https://codereview.chromium.org/285403004/diff/1/runtime/vm/intermediate_lang...
File runtime/vm/intermediate_language_x64.cc (right):

https://codereview.chromium.org/285403004/diff/1/runtime/vm/intermediate_lang...
runtime/vm/intermediate_language_x64.cc:2598: // overflow = (result + 2^52) >
2^53.
the comparison must be unsigned

https://codereview.chromium.org/285403004/diff/1/runtime/vm/intrinsifier_arm.cc
File runtime/vm/intrinsifier_arm.cc (right):

https://codereview.chromium.org/285403004/diff/1/runtime/vm/intrinsifier_arm....
runtime/vm/intrinsifier_arm.cc:521: __ SmiUntag(R0);  // untags R6. We only want
result shifted by one.
Untags

https://codereview.chromium.org/285403004/diff/1/runtime/vm/intrinsifier_arm6...
File runtime/vm/intrinsifier_arm64.cc (right):

https://codereview.chromium.org/285403004/diff/1/runtime/vm/intrinsifier_arm6...
runtime/vm/intrinsifier_arm64.cc:42: __ b(&fall_through, NE);  // Index is not
an smi, fall through
period

https://codereview.chromium.org/285403004/diff/1/runtime/vm/intrinsifier_arm6...
runtime/vm/intrinsifier_arm64.cc:229: __ b(&fall_through, NE);  // Index is not
an smi, fall through
period

https://codereview.chromium.org/285403004/diff/1/runtime/vm/intrinsifier_arm6...
runtime/vm/intrinsifier_arm64.cc:238: __ ldr(R6, FieldAddress(R1,
GrowableObjectArray::data_offset()));  // data
Data.

https://codereview.chromium.org/285403004/diff/1/runtime/vm/intrinsifier_arm6...
runtime/vm/intrinsifier_arm64.cc:668: TestBothArgumentsSmis(assembler,
&fall_through);  // checks two smis
Check two smis.

https://codereview.chromium.org/285403004/diff/1/runtime/vm/intrinsifier_arm6...
runtime/vm/intrinsifier_arm64.cc:682: TestBothArgumentsSmis(assembler,
&fall_through);  // checks two smis
Check two smis.

https://codereview.chromium.org/285403004/diff/1/runtime/vm/intrinsifier_arm6...
runtime/vm/intrinsifier_arm64.cc:697: TestBothArgumentsSmis(assembler,
&fall_through);  // checks two smis
Check two smis.

https://codereview.chromium.org/285403004/diff/1/runtime/vm/intrinsifier_arm6...
runtime/vm/intrinsifier_arm64.cc:1278: __ csinc(R0, R0, ZR, NE);  // R0 <- (R0
!= 0) ? R0 : (ZR + 1).
nice!

[zra, 2014-05-19 21:13:18]

Thanks!

https://codereview.chromium.org/285403004/diff/1/runtime/vm/assembler_arm64.h
File runtime/vm/assembler_arm64.h (right):

https://codereview.chromium.org/285403004/diff/1/runtime/vm/assembler_arm64.h...
runtime/vm/assembler_arm64.h:796: addw(rd, ZR, Operand(rn, LSR, shift));
On 2014/05/19 20:14:02, regis wrote:
> Is a shift amount > 31 allowed? Are all 64 bits of rn taken into
consideration?

Ah, no. A shift > 31 is not allowed. I've added an ASSERT. Instructions in
32-bit mode only operate on the bottom 32 bits, so this instruction will shift
in zeros.

https://codereview.chromium.org/285403004/diff/1/runtime/vm/intermediate_lang...
File runtime/vm/intermediate_language_x64.cc (right):

https://codereview.chromium.org/285403004/diff/1/runtime/vm/intermediate_lang...
runtime/vm/intermediate_language_x64.cc:2598: // overflow = (result + 2^52) >
2^53.
On 2014/05/19 20:14:02, regis wrote:
> the comparison must be unsigned

Amended TODO.

https://codereview.chromium.org/285403004/diff/1/runtime/vm/intrinsifier_arm.cc
File runtime/vm/intrinsifier_arm.cc (right):

https://codereview.chromium.org/285403004/diff/1/runtime/vm/intrinsifier_arm....
runtime/vm/intrinsifier_arm.cc:521: __ SmiUntag(R0);  // untags R6. We only want
result shifted by one.
On 2014/05/19 20:14:02, regis wrote:
> Untags

Done.

https://codereview.chromium.org/285403004/diff/1/runtime/vm/intrinsifier_arm6...
File runtime/vm/intrinsifier_arm64.cc (right):

https://codereview.chromium.org/285403004/diff/1/runtime/vm/intrinsifier_arm6...
runtime/vm/intrinsifier_arm64.cc:42: __ b(&fall_through, NE);  // Index is not
an smi, fall through
On 2014/05/19 20:14:02, regis wrote:
> period

Done.

https://codereview.chromium.org/285403004/diff/1/runtime/vm/intrinsifier_arm6...
runtime/vm/intrinsifier_arm64.cc:229: __ b(&fall_through, NE);  // Index is not
an smi, fall through
On 2014/05/19 20:14:02, regis wrote:
> period

Done.

https://codereview.chromium.org/285403004/diff/1/runtime/vm/intrinsifier_arm6...
runtime/vm/intrinsifier_arm64.cc:238: __ ldr(R6, FieldAddress(R1,
GrowableObjectArray::data_offset()));  // data
On 2014/05/19 20:14:02, regis wrote:
> Data.

Done.

https://codereview.chromium.org/285403004/diff/1/runtime/vm/intrinsifier_arm6...
runtime/vm/intrinsifier_arm64.cc:668: TestBothArgumentsSmis(assembler,
&fall_through);  // checks two smis
On 2014/05/19 20:14:02, regis wrote:
> Check two smis.

Done.

https://codereview.chromium.org/285403004/diff/1/runtime/vm/intrinsifier_arm6...
runtime/vm/intrinsifier_arm64.cc:682: TestBothArgumentsSmis(assembler,
&fall_through);  // checks two smis
On 2014/05/19 20:14:02, regis wrote:
> Check two smis.

Done.

https://codereview.chromium.org/285403004/diff/1/runtime/vm/intrinsifier_arm6...
runtime/vm/intrinsifier_arm64.cc:697: TestBothArgumentsSmis(assembler,
&fall_through);  // checks two smis
On 2014/05/19 20:14:02, regis wrote:
> Check two smis.

Done.

[zra, 2014-05-19 21:35:39]

Committed patchset #2 manually as r36330 (presubmit successful).