Introduce PushInt64AsTwoSmis and PopInt64AsTwoSmis macro instructions for X64

Introduce PushInt64AsTwoSmis and PopInt64AsTwoSmis macro instructions for X64

R=danno@chromium.org

Committed: https://code.google.com/p/v8/source/detail?r=16376

[haitao.feng, 2013-08-09 06:53:20]

Addressed danno's comment at
https://codereview.chromium.org/21014003/patch/17001/18007. I also gather all
the SMI functions together in the macro-assembler-x64.[h|cc] files.

[danno, 2013-08-13 09:46:55]

I think this is actually two separate changes, please split them up. There is
consolidation of smi operations into the same area and then there is an addition
of PushInt64AsTwoSmis. By the way, as discussed in another CL, for the x32 port,
PushInt64AsTwoSmis doesn't always work because you can only encode 62 bits in
two smis. How are you planning on handling this? Asserting if the top to bits
are different than bit 61?

[haitao.feng, 2013-08-13 11:02:04]

danno, thanks for the review. I will split them up. I might need to choose a
better name for PushInt64AsTwoSmis. It is a general SMI (all the tagged bits are
zero). The logic in PushInt64AsTwoSmis works for both kSmiShift=1 and
kSmiShift=32, but the value might not be in the range of 31 bit SMI. Do you
think PushInt64AsTwoZeroTaggedValues are OK?

[danno, 2013-08-13 12:12:26]

I think PushInt64AsTwoSmis is probably OK, but it's just a little misleading. In
the 31-bit case, you shift the 32nd bit up to the 33rd bit of the 64-bit
register, so it's no longer part of the smi. However, for the purpose of just
saving registers that don't look like pointers, I think it will be fine. Just go
ahead with your current approach fo PushInt64AsTwoSmis.

[haitao.feng, 2013-08-13 13:13:23]

Thanks for the comment. I have made this CL to include the PushInt64AsTwoSmis
only. I added a comment in the header file, the comment is a little misleading
also before 31-bit SMI is not introduced as currently they are exactly SMIs.

[danno, 2013-08-19 22:00:40]

lgtm if you improve the comment as suggested, please wait to land until the V8
tree is no longer throttled

https://chromiumcodereview.appspot.com/22348005/diff/7001/src/x64/macro-assem...
File src/x64/macro-assembler-x64.h (right):

https://chromiumcodereview.appspot.com/22348005/diff/7001/src/x64/macro-assem...
src/x64/macro-assembler-x64.h:724: // Push a 64-bit integer as two "general"
Smis on stack so that the garbage
I find the comments a bit confusing. I think the following is probably clearer:

Save away a 64-bit integer on the stack as two 32-bits integers masquerading as
smis so that the garbage collector skips visiting them.

https://chromiumcodereview.appspot.com/22348005/diff/7001/src/x64/macro-assem...
src/x64/macro-assembler-x64.h:728: // Reconstruct a 64-bit integer from two
"general" Smis from stack.
// Reconstruct a 64-bit integer from two 32-bits integers masquerading as smis
on the top of stack.

[haitao.feng, 2013-08-20 13:20:14]

danno, thanks for refining the comments.

https://codereview.chromium.org/22348005/diff/7001/src/x64/macro-assembler-x64.h
File src/x64/macro-assembler-x64.h (right):

https://codereview.chromium.org/22348005/diff/7001/src/x64/macro-assembler-x6...
src/x64/macro-assembler-x64.h:724: // Push a 64-bit integer as two "general"
Smis on stack so that the garbage
On 2013/08/19 22:00:40, danno wrote:
> I find the comments a bit confusing. I think the following is probably
clearer:
> 
> Save away a 64-bit integer on the stack as two 32-bits integers masquerading
as
> smis so that the garbage collector skips visiting them.

Done.

https://codereview.chromium.org/22348005/diff/7001/src/x64/macro-assembler-x6...
src/x64/macro-assembler-x64.h:728: // Reconstruct a 64-bit integer from two
"general" Smis from stack.
On 2013/08/19 22:00:40, danno wrote:
> // Reconstruct a 64-bit integer from two 32-bits integers masquerading as smis
> on the top of stack.

Done.

[haitao.feng, 2013-08-28 01:07:45]

Committed patchset #3 manually as r16376 (presubmit successful).