Config double register number as 1 for X87 Turbofan.

Currently on X87 platform we use only Double register (stack register) for Turbofan. So we directly
use 1 as allocatable Double register number when setting up the default register configuration..

It does not change the double register configuration of other platforms.

BUG=v8:4135
LOG=N

Committed: https://crrev.com/3fdfebd26bf70778b3bc5a6c61c8063dec7d3dca
Cr-Commit-Position: refs/heads/master@{#29005}

[chunyang.dai, 2015-06-08 14:53:50]

chunyang.dai@intel.com changed reviewers:
+ danno@chromium.org, titzer@chromium.org, weiliang.lin@intel.com

[chunyang.dai, 2015-06-08 14:53:51]

hello, Daniel & Ben Titzer.

As discussed last month, we use one double register for X87 turbofan. Currently
the pass rate of X87 turbo is 100%.
We did minimal change in register-configuration. Two APIs of DoubleRegister
structure will return
the allocatable number for Turbofan but only X87 platform will return different
value. 

Please Review it.
thanks.

[chunyang.dai, 2015-06-08 14:55:16]

On 2015/06/08 14:53:51, chunyang.dai wrote:
> hello, Daniel & Ben Titzer.
> 
> As discussed last month, we use one double register for X87 turbofan.
Currently
> the pass rate of X87 turbo is 100%.
> We did minimal change in register-configuration. Two APIs of DoubleRegister
> structure will return
> the allocatable number for Turbofan but only X87 platform will return
different
> value. 
> 
> Please Review it.
> thanks.

this is the first patch to touch the common code. we will upload the changes for
x87
 platform later.

[titzer, 2015-06-09 14:47:00]

https://codereview.chromium.org/1164813007/diff/1/src/compiler/register-confi...
File src/compiler/register-configuration.cc (right):

https://codereview.chromium.org/1164813007/diff/1/src/compiler/register-confi...
src/compiler/register-configuration.cc:25:
DoubleRegister::NumAllocatableRegistersForTurbo(),
This is the class meant to configure turbofan. Why don't we add the logic here
instead of everywhere else?

[chunyang.dai, 2015-06-09 15:08:34]

Hi, Titzer.
Thank you for your quick review. Do you mean we use the logic like the
following?  


   ArchDefaultRegisterConfiguration()	
       : RegisterConfiguration(Register::kMaxNumAllocatableRegisters,
#if V8_TARGET_ARCH_X87
             DoubleRegister::NumAllocatableRegistersForTurbo(),
             DoubleRegister::NumAllocatableAliasedRegistersForTurbo(),
#else
             DoubleRegister::kMaxNumAllocatableRegisters,	 
             DoubleRegister::NumAllocatableAliasedRegisters()
#endif
             general_register_name_table_,
             double_register_name_table_) {

[chunyang.dai, 2015-06-10 08:21:03]

Hi, Titzer.

 I updated the patch basing on my understanding of your comments. Please review
it. 
 
 Thanks

[Michael Hablich, 2015-06-11 07:44:45]

On 2015/06/10 08:21:03, chunyang.dai wrote:
> Hi, Titzer.
> 
>  I updated the patch basing on my understanding of your comments. Please
review
> it. 
>  
>  Thanks

Hi Chunyang,

could you please reference the bug
https://code.google.com/p/v8/issues/detail?id=4135 by adding

BUG=v8:4135

to the CL description? If you do these your CLs get automatically linked to the
tracking bug.

Thanks,
Michael

[titzer, 2015-06-11 11:17:49]

https://codereview.chromium.org/1164813007/diff/20001/src/compiler/register-c...
File src/compiler/register-configuration.cc (right):

https://codereview.chromium.org/1164813007/diff/20001/src/compiler/register-c...
src/compiler/register-configuration.cc:26:
DoubleRegister::NumAllocatableRegistersForTurbo(),
Why not just 1 here?

[chunyang.dai, 2015-06-11 14:36:02]

Patchset #3 (id:40001) has been deleted

[chunyang.dai, 2015-06-11 14:41:37]

Patchset #3 (id:60001) has been deleted

[chunyang.dai, 2015-06-11 14:42:26]

Patchset #3 (id:80001) has been deleted

[chunyang.dai, 2015-06-11 14:46:44]

hi, Titzer.
The reason why I use one API here is that maybe we will change the available
double register here in the future if we find solution to support more double 
register on X87 platform. Then we do not need to touch the common code. We can
configure it on X87 specific code. 

Anyway, I updated it according to your suggestion. 
Please review it. 

thanks.

[titzer, 2015-06-12 16:29:23]

lgtm

[titzer, 2015-06-12 16:30:08]

The CQ bit was checked by titzer@chromium.org

[commit-bot: I haz the power, 2015-06-12 16:30:28]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1164813007/100001

[commit-bot: I haz the power, 2015-06-12 17:22:53]

Committed patchset #3 (id:100001)

[commit-bot: I haz the power, 2015-06-12 17:23:07]

Patchset 3 (id:??) landed as
https://crrev.com/3fdfebd26bf70778b3bc5a6c61c8063dec7d3dca
Cr-Commit-Position: refs/heads/master@{#29005}