[turbofan] fine grained move operand merge optimization

If all the predecessors of a node have, at the last gap, the exact same
moves, we move those to the node, and remove them from the
predecessors ("merge" them to the common node).

If only some of the moves are common, we don't do anything. This is
what this change addresses.

The bug linked below should be addressed by this change. The only
difference in codegen before/after the change that introduced the bug
was un-merged moves.

BUG=chromium:549262
LOG=N

Committed: https://crrev.com/78b55f2ebc7722c28b7c7f8b65c5402c1062ebac
Cr-Commit-Position: refs/heads/master@{#33481}

[Mircea Trofin, 2015-12-17 23:47:59]

Description was changed from

==========
draft


Can't optimize if predecessor has many successors.

BUG=
==========

to

==========
draft


Can't optimize if predecessor has many successors.

BUG=
==========

[Mircea Trofin, 2016-01-25 03:51:15]

Description was changed from

==========
draft


Can't optimize if predecessor has many successors.

BUG=
==========

to

==========
If all the predecessors of a node have, at the last gap, the exact same
moves, we move those to the node, and remove them from the 
predecessors ("merge" them to the common node). 

If only some of the moves are common, we don't do anything. This is 
what this change addresses.

The bug linked below should be addressed by this change. The only
difference in codegen before/after the change that introduced the bug
was un-merged moves.

BUG=chromium:549262
LOG=N
==========

[Mircea Trofin, 2016-01-25 03:51:15]

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

[Jarin, 2016-01-25 05:23:55]

lgtm

https://codereview.chromium.org/1527203002/diff/120001/src/compiler/move-opti...
File src/compiler/move-optimizer.cc (right):

https://codereview.chromium.org/1527203002/diff/120001/src/compiler/move-opti...
src/compiler/move-optimizer.cc:218: if (pred->SuccessorCount() > 1) return;
Does this really happen? I thought we always have edge-split form CFGs.

https://codereview.chromium.org/1527203002/diff/120001/src/compiler/move-opti...
src/compiler/move-optimizer.cc:271: OperandSet banned_srcs(local_zone());
Perhaps banned -> conflicting?

[Mircea Trofin, 2016-01-25 06:07:33]

The CQ bit was checked by mtrofin@chromium.org

[Mircea Trofin, 2016-01-25 06:07:33]

The patchset sent to the CQ was uploaded after l-g-t-m from jarin@chromium.org
Link to the patchset: https://codereview.chromium.org/1527203002/#ps140001
(title: " ")

[commit-bot: I haz the power, 2016-01-25 06:07:35]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1527203002/140001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1527203002/140001

[Mircea Trofin, 2016-01-25 06:07:38]

https://codereview.chromium.org/1527203002/diff/120001/src/compiler/move-opti...
File src/compiler/move-optimizer.cc (right):

https://codereview.chromium.org/1527203002/diff/120001/src/compiler/move-opti...
src/compiler/move-optimizer.cc:218: if (pred->SuccessorCount() > 1) return;
On 2016/01/25 05:23:55, Jarin wrote:
> Does this really happen? I thought we always have edge-split form CFGs.

I did not know that - I could change this to a DCHECK, on the other hand, the
algorithm works for CFGs "in general" if this stays this way.

https://codereview.chromium.org/1527203002/diff/120001/src/compiler/move-opti...
src/compiler/move-optimizer.cc:271: OperandSet banned_srcs(local_zone());
On 2016/01/25 05:23:55, Jarin wrote:
> Perhaps banned -> conflicting?

Done.

[commit-bot: I haz the power, 2016-01-25 06:32:23]

Description was changed from

==========
If all the predecessors of a node have, at the last gap, the exact same
moves, we move those to the node, and remove them from the 
predecessors ("merge" them to the common node). 

If only some of the moves are common, we don't do anything. This is 
what this change addresses.

The bug linked below should be addressed by this change. The only
difference in codegen before/after the change that introduced the bug
was un-merged moves.

BUG=chromium:549262
LOG=N
==========

to

==========
If all the predecessors of a node have, at the last gap, the exact same
moves, we move those to the node, and remove them from the 
predecessors ("merge" them to the common node). 

If only some of the moves are common, we don't do anything. This is 
what this change addresses.

The bug linked below should be addressed by this change. The only
difference in codegen before/after the change that introduced the bug
was un-merged moves.

BUG=chromium:549262
LOG=N
==========

[commit-bot: I haz the power, 2016-01-25 06:32:23]

Committed patchset #8 (id:140001)

[commit-bot: I haz the power, 2016-01-25 06:33:19]

Description was changed from

==========
If all the predecessors of a node have, at the last gap, the exact same
moves, we move those to the node, and remove them from the 
predecessors ("merge" them to the common node). 

If only some of the moves are common, we don't do anything. This is 
what this change addresses.

The bug linked below should be addressed by this change. The only
difference in codegen before/after the change that introduced the bug
was un-merged moves.

BUG=chromium:549262
LOG=N
==========

to

==========
If all the predecessors of a node have, at the last gap, the exact same
moves, we move those to the node, and remove them from the
predecessors ("merge" them to the common node).

If only some of the moves are common, we don't do anything. This is
what this change addresses.

The bug linked below should be addressed by this change. The only
difference in codegen before/after the change that introduced the bug
was un-merged moves.

BUG=chromium:549262
LOG=N

Committed: https://crrev.com/78b55f2ebc7722c28b7c7f8b65c5402c1062ebac
Cr-Commit-Position: refs/heads/master@{#33481}
==========

[commit-bot: I haz the power, 2016-01-25 06:33:20]

Patchset 8 (id:??) landed as
https://crrev.com/78b55f2ebc7722c28b7c7f8b65c5402c1062ebac
Cr-Commit-Position: refs/heads/master@{#33481}