[Turbofan] Add Simd128 registers to RegisterConfiguration.

[Turbofan] Add Simd128 registers to RegisterConfiguration.
-Defines SIMD128_REGISTERS for all platforms.
-Adds Simd128 register information to RegisterConfiguration, and implements
aliasing calculations.

LOG=N
BUG=v8:4124
Committed: https://crrev.com/aca3716a504c262ff8e5cd8b5ae7ba45d63b2e89
Cr-Commit-Position: refs/heads/master@{#37437}

[bbudge, 2016-06-25 21:18:54]

bbudge@chromium.org changed reviewers:
+ bmeurer@chromium.org, mtrofin@chromium.org

[Benedikt Meurer, 2016-06-27 03:59:20]

LGTM from my side (I wish we could simplify some of the register allocation
logic though, maybe some followup CLs?)

[bbudge, 2016-06-27 10:58:49]

On 2016/06/27 03:59:20, Benedikt Meurer wrote:
> LGTM from my side (I wish we could simplify some of the register allocation
> logic though, maybe some followup CLs?)

Yes, I plan to use the kSimpleFPAliasing flag to avoid all the extra fixed live
range handling. I'll keep looking for ways to speed up ARM.

[Mircea Trofin, 2016-06-27 20:32:17]

On 2016/06/27 10:58:49, bbudge wrote:
> On 2016/06/27 03:59:20, Benedikt Meurer wrote:
> > LGTM from my side (I wish we could simplify some of the register allocation
> > logic though, maybe some followup CLs?)
> 
> Yes, I plan to use the kSimpleFPAliasing flag to avoid all the extra fixed
live
> range handling. I'll keep looking for ways to speed up ARM.

lgtm

[Mircea Trofin, 2016-06-27 20:32:26]

https://codereview.chromium.org/2092103004/diff/100001/src/compiler/move-opti...
File src/compiler/move-optimizer.cc (right):

https://codereview.chromium.org/2092103004/diff/100001/src/compiler/move-opti...
src/compiler/move-optimizer.cc:70: int aliases = config->GetAliases(rep,
loc.register_code(), other_rep1, &base);
Is base expected to be positive after this?

[bbudge, 2016-06-27 20:43:21]

https://codereview.chromium.org/2092103004/diff/100001/src/compiler/move-opti...
File src/compiler/move-optimizer.cc (right):

https://codereview.chromium.org/2092103004/diff/100001/src/compiler/move-opti...
src/compiler/move-optimizer.cc:70: int aliases = config->GetAliases(rep,
loc.register_code(), other_rep1, &base);
On 2016/06/27 20:32:26, Mircea Trofin wrote:
> Is base expected to be positive after this?

It is left unchanged if the return value is 0 (no aliases.) I initialize it to
-1 to fail quicker if someone tries to use it when the return value is 0.

[Mircea Trofin, 2016-06-27 20:49:20]

https://codereview.chromium.org/2092103004/diff/100001/src/compiler/move-opti...
File src/compiler/move-optimizer.cc (right):

https://codereview.chromium.org/2092103004/diff/100001/src/compiler/move-opti...
src/compiler/move-optimizer.cc:70: int aliases = config->GetAliases(rep,
loc.register_code(), other_rep1, &base);
On 2016/06/27 20:43:21, bbudge wrote:
> On 2016/06/27 20:32:26, Mircea Trofin wrote:
> > Is base expected to be positive after this?
> 
> It is left unchanged if the return value is 0 (no aliases.) I initialize it to
> -1 to fail quicker if someone tries to use it when the return value is 0.

Should we have the DCHECK right after the call to GetAliases?

[bbudge, 2016-06-27 21:08:02]

https://codereview.chromium.org/2092103004/diff/100001/src/compiler/move-opti...
File src/compiler/move-optimizer.cc (right):

https://codereview.chromium.org/2092103004/diff/100001/src/compiler/move-opti...
src/compiler/move-optimizer.cc:70: int aliases = config->GetAliases(rep,
loc.register_code(), other_rep1, &base);
On 2016/06/27 20:49:20, Mircea Trofin wrote:
> On 2016/06/27 20:43:21, bbudge wrote:
> > On 2016/06/27 20:32:26, Mircea Trofin wrote:
> > > Is base expected to be positive after this?
> > 
> > It is left unchanged if the return value is 0 (no aliases.) I initialize it
to
> > -1 to fail quicker if someone tries to use it when the return value is 0.
> 
> Should we have the DCHECK right after the call to GetAliases?

Good idea, I've added a DCHECK at all GetAliases call sites.

[bbudge, 2016-06-29 23:05:09]

Patchset #10 (id:180001) has been deleted

[bbudge, 2016-06-29 23:05:20]

Patchset #9 (id:160001) has been deleted

[bbudge, 2016-06-29 23:34:08]

Despite all my thrashing around with patchsets, I don't think this has changed
significantly. However, I'll hold off committing until tomorrow in case anyone
wants to PTAL.

[bbudge, 2016-06-30 11:26:33]

The CQ bit was checked by bbudge@chromium.org

[bbudge, 2016-06-30 11:26:34]

The patchset sent to the CQ was uploaded after l-g-t-m from
bmeurer@chromium.org, mtrofin@chromium.org
Link to the patchset: https://codereview.chromium.org/2092103004/#ps220001
(title: "Eliminate some dead code, simplify.")

[commit-bot: I haz the power, 2016-06-30 11:26:40]

CQ is trying da patch. Follow status at
 
https://chromium-cq-status.appspot.com/v2/patch-status/codereview.chromium.or...

[commit-bot: I haz the power, 2016-06-30 11:29:27]

Committed patchset #10 (id:220001)

[commit-bot: I haz the power, 2016-06-30 11:29:51]

Description was changed from

==========
[Turbofan] Add Simd128 registers to RegisterConfiguration.
-Defines SIMD128_REGISTERS for all platforms.
-Adds Simd128 register information to RegisterConfiguration, and implements
aliasing calculations.

LOG=N
BUG=v8:4124
==========

to

==========
[Turbofan] Add Simd128 registers to RegisterConfiguration.
-Defines SIMD128_REGISTERS for all platforms.
-Adds Simd128 register information to RegisterConfiguration, and implements
aliasing calculations.

LOG=N
BUG=v8:4124

Committed: https://crrev.com/aca3716a504c262ff8e5cd8b5ae7ba45d63b2e89
Cr-Commit-Position: refs/heads/master@{#37437}
==========

[commit-bot: I haz the power, 2016-06-30 11:29:53]

Patchset 10 (id:??) landed as
https://crrev.com/aca3716a504c262ff8e5cd8b5ae7ba45d63b2e89
Cr-Commit-Position: refs/heads/master@{#37437}