Use constant length for memcpy on A64 simulator.

Use constant length for memcpy on A64 simulator.

Compiler can't optimize away variable length memcpy.
About a 18% boost.

Before - 5:40

Richards: 75.2
DeltaBlue: 105
Crypto: 64.0
RayTrace: 188
EarleyBoyer: 146
RegExp: 23.8
Splay: 96.2
NavierStokes: 91.9
----
Score (version 7): 85.7

After - 4:39

Richards: 90.8
DeltaBlue: 134
Crypto: 81.7
RayTrace: 227
EarleyBoyer: 177
RegExp: 29.7
Splay: 126
NavierStokes: 108
----
Score (version 7): 106

R=rodolph.perfetta@gmail.com, svenpanne@chromium.org

Committed: https://code.google.com/p/v8/source/detail?r=20141

[Sven Panne, 2014-03-20 10:22:37]

https://codereview.chromium.org/203263017/diff/1/src/a64/simulator-a64.h
File src/a64/simulator-a64.h (right):

https://codereview.chromium.org/203263017/diff/1/src/a64/simulator-a64.h#newc...
src/a64/simulator-a64.h:187: memset(&result, 0, sizeof(result));
I think this could be moved into the 2nd case, just like Set.

[Rodolph Perfetta, 2014-03-20 11:33:56]

https://codereview.chromium.org/203263017/diff/1/src/a64/simulator-a64.h
File src/a64/simulator-a64.h (right):

https://codereview.chromium.org/203263017/diff/1/src/a64/simulator-a64.h#newc...
src/a64/simulator-a64.h:171: if (size == kXRegSize) {
STATIC_ASSERT(kXRegSize == kDRegSize);

This is more to document the fact this is also used for S and D registers

same below with W and S

[Fritz, 2014-03-20 19:37:00]

https://codereview.chromium.org/203263017/diff/1/src/a64/simulator-a64.h
File src/a64/simulator-a64.h (right):

https://codereview.chromium.org/203263017/diff/1/src/a64/simulator-a64.h#newc...
src/a64/simulator-a64.h:171: if (size == kXRegSize) {
On 2014/03/20 11:33:57, Rodolph Perfetta wrote:
> STATIC_ASSERT(kXRegSize == kDRegSize);
> 
> This is more to document the fact this is also used for S and D registers
> 
> same below with W and S

Done.

https://codereview.chromium.org/203263017/diff/1/src/a64/simulator-a64.h#newc...
src/a64/simulator-a64.h:187: memset(&result, 0, sizeof(result));
On 2014/03/20 10:22:38, Sven Panne wrote:
> I think this could be moved into the 2nd case, just like Set.

Done.

[Rodolph Perfetta, 2014-03-20 19:58:35]

lgtm

[Fritz, 2014-03-20 20:49:38]

The CQ bit was checked by frkoenig@google.com

[commit-bot: I haz the power, 2014-03-20 20:49:44]

CQ is trying da patch. Follow status at
https://v8-status.appspot.com/cq/frkoenig@google.com/203263017/20001

[commit-bot: I haz the power, 2014-03-20 20:49:44]

The CQ bit was unchecked by commit-bot@chromium.org

[commit-bot: I haz the power, 2014-03-20 20:49:45]

No LGTM from a valid reviewer yet. Only full committers are accepted.
Even if an LGTM may have been provided, it was from a non-committer or
a lowly provisional committer, _not_ a full super star committer.
See http://www.chromium.org/getting-involved/become-a-committer
Note that this has nothing to do with OWNERS files.

[Sven Panne, 2014-03-21 06:56:18]

LGTM, I'll land it for you. Note that the CQ does not work for the v8 project.

[Sven Panne, 2014-03-21 07:08:14]

Committed patchset #2 manually as r20141 (presubmit successful).

[Sven Panne, 2014-03-21 08:05:48]

I had to revert that, some GCCs try to be more clever than they actually are...