Adds ldr and str to arm64.

Adds ldr and str to arm64.

R=regis@google.com

Committed: https://code.google.com/p/dart/source/detail?r=34831

[regis, 2014-04-07 23:38:18]

LGTM

https://codereview.chromium.org/227593012/diff/1/runtime/vm/assembler_arm64.h
File runtime/vm/assembler_arm64.h (right):

https://codereview.chromium.org/227593012/diff/1/runtime/vm/assembler_arm64.h...
runtime/vm/assembler_arm64.h:94: // imm9 case, the offset is the number of
bytes, but is unscaled.
I do not understand "We translate to number of operands".

https://codereview.chromium.org/227593012/diff/1/runtime/vm/assembler_arm64_t...
File runtime/vm/assembler_arm64_test.cc (right):

https://codereview.chromium.org/227593012/diff/1/runtime/vm/assembler_arm64_t...
runtime/vm/assembler_arm64_test.cc:300: }
It would be good to add a test with an offset of -1, with a base register
unaligned by 1 byte, corresponding to our heap tag.

https://codereview.chromium.org/227593012/diff/1/runtime/vm/constants_arm64.h
File runtime/vm/constants_arm64.h (right):

https://codereview.chromium.org/227593012/diff/1/runtime/vm/constants_arm64.h...
runtime/vm/constants_arm64.h:222: case kWord: case kUnsignedWord: return 2;
It would be more readable to use one case per line.

[zra, 2014-04-08 15:14:13]

Thanks!

https://codereview.chromium.org/227593012/diff/1/runtime/vm/assembler_arm64.h
File runtime/vm/assembler_arm64.h (right):

https://codereview.chromium.org/227593012/diff/1/runtime/vm/assembler_arm64.h...
runtime/vm/assembler_arm64.h:94: // imm9 case, the offset is the number of
bytes, but is unscaled.
On 2014/04/07 23:38:19, regis wrote:
> I do not understand "We translate to number of operands".

Tried to clarify. Also added an assert.

https://codereview.chromium.org/227593012/diff/1/runtime/vm/assembler_arm64_t...
File runtime/vm/assembler_arm64_test.cc (right):

https://codereview.chromium.org/227593012/diff/1/runtime/vm/assembler_arm64_t...
runtime/vm/assembler_arm64_test.cc:300: }
On 2014/04/07 23:38:19, regis wrote:
> It would be good to add a test with an offset of -1, with a base register
> unaligned by 1 byte, corresponding to our heap tag.

Yes. I notice that this will likely be pretty annoying because it isn't possible
to have a negative offset without also pre- or post-indexing with writeback.

https://codereview.chromium.org/227593012/diff/1/runtime/vm/constants_arm64.h
File runtime/vm/constants_arm64.h (right):

https://codereview.chromium.org/227593012/diff/1/runtime/vm/constants_arm64.h...
runtime/vm/constants_arm64.h:222: case kWord: case kUnsignedWord: return 2;
On 2014/04/07 23:38:19, regis wrote:
> It would be more readable to use one case per line.

Done.

[zra, 2014-04-08 15:43:20]

Committed patchset #3 manually as r34831 (presubmit successful).