[turbofan] Optimize tagged conversion based on type.

[turbofan] Optimize tagged conversion based on type.

If we have to convert a float64 value to tagged representation and we
already know that the value is either in Signed31/Signed32 or
Unsigned32 range, then we can just convert the float64 to word32 and
use the fast word32 to tagged conversion. Doing this in
ChangeLowering (or the effect linearization pass) would be unsound, as
the types on the nodes are no longer usable.

This removes all Type uses from effect linearization. There's still some
work to be done for ChangeLowering tho.

[email protected]

Committed: https://crrev.com/861295bf167f045a3323fac086261b8bb264f8b3
Cr-Commit-Position: refs/heads/master@{#35713}

[Benedikt Meurer, 2016-04-21 18:51:33]

Hey Jaro,

Start moving access to types out of ChangeLowering.
Please take a look.

Thanks,
Benedikt

[Benedikt Meurer, 2016-04-22 04:50:27]

Description was changed from

==========
[turbofan] Optimize float64->tagged conversion based on type.

If we have to convert a float64 value to tagged representation and we
already know that the value is either in Signed32 or Unsigned32 range,
then we can just convert the float64 to word32 and use the fast word32
to tagged conversion. Doing this in ChangeLowering (or the effect
linearization pass would be unsound, as the types on the nodes are no
longer usable).

[email protected]
==========

to

==========
[turbofan] Optimize tagged conversion based on type.

If we have to convert a float64 value to tagged representation and we
already know that the value is either in Signed31/Signed32 or
Unsigned32 range, then we can just convert the float64 to word32 and
use the fast word32 to tagged conversion. Doing this in
ChangeLowering (or the effect linearization pass) would be unsound, as
the types on the nodes are no longer usable.

This removes all Type uses from effect linearization. There's still some
work to be done for ChangeLowering tho.

[email protected]
==========

[Benedikt Meurer, 2016-04-22 04:52:42]

Patchset #1 (id:1) has been deleted

[Benedikt Meurer, 2016-04-22 06:44:37]

Patchset #1 (id:20001) has been deleted

[Jarin, 2016-04-22 07:08:20]

Nice. lgtm.

[Benedikt Meurer, 2016-04-22 07:08:35]

The CQ bit was checked by [email protected]

[commit-bot: I haz the power, 2016-04-22 07:08:45]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1908093002/40001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1908093002/40001

[Benedikt Meurer, 2016-04-22 07:26:56]

The CQ bit was unchecked by [email protected]

[Benedikt Meurer, 2016-04-22 07:29:43]

The CQ bit was checked by [email protected]

[Benedikt Meurer, 2016-04-22 07:29:43]

The patchset sent to the CQ was uploaded after l-g-t-m from [email protected]
Link to the patchset: https://codereview.chromium.org/1908093002/#ps60001
(title: "Fix ChangeLowering::ChangeUint32ToTagged too.")

[commit-bot: I haz the power, 2016-04-22 07:29:55]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1908093002/60001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1908093002/60001

[commit-bot: I haz the power, 2016-04-22 07:42:17]

The CQ bit was unchecked by [email protected]

[commit-bot: I haz the power, 2016-04-22 07:42:18]

Try jobs failed on following builders:
  v8_win_nosnap_shared_rel_ng on tryserver.v8 (JOB_FAILED,
http://build.chromium.org/p/tryserver.v8/builders/v8_win_nosnap_shared_rel_ng...)
  v8_win_nosnap_shared_rel_ng_triggered on tryserver.v8 (JOB_FAILED,
http://build.chromium.org/p/tryserver.v8/builders/v8_win_nosnap_shared_rel_ng...)

[Benedikt Meurer, 2016-04-22 08:17:12]

The CQ bit was checked by [email protected]

[Benedikt Meurer, 2016-04-22 08:17:12]

The patchset sent to the CQ was uploaded after l-g-t-m from [email protected]
Link to the patchset: https://codereview.chromium.org/1908093002/#ps80001
(title: "Fixes")

[commit-bot: I haz the power, 2016-04-22 08:17:24]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1908093002/80001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1908093002/80001

[commit-bot: I haz the power, 2016-04-22 08:39:36]

Description was changed from

==========
[turbofan] Optimize tagged conversion based on type.

If we have to convert a float64 value to tagged representation and we
already know that the value is either in Signed31/Signed32 or
Unsigned32 range, then we can just convert the float64 to word32 and
use the fast word32 to tagged conversion. Doing this in
ChangeLowering (or the effect linearization pass) would be unsound, as
the types on the nodes are no longer usable.

This removes all Type uses from effect linearization. There's still some
work to be done for ChangeLowering tho.

[email protected]
==========

to

==========
[turbofan] Optimize tagged conversion based on type.

If we have to convert a float64 value to tagged representation and we
already know that the value is either in Signed31/Signed32 or
Unsigned32 range, then we can just convert the float64 to word32 and
use the fast word32 to tagged conversion. Doing this in
ChangeLowering (or the effect linearization pass) would be unsound, as
the types on the nodes are no longer usable.

This removes all Type uses from effect linearization. There's still some
work to be done for ChangeLowering tho.

[email protected]
==========

[commit-bot: I haz the power, 2016-04-22 08:39:37]

Committed patchset #3 (id:80001)

[commit-bot: I haz the power, 2016-04-22 19:15:02]

Description was changed from

==========
[turbofan] Optimize tagged conversion based on type.

If we have to convert a float64 value to tagged representation and we
already know that the value is either in Signed31/Signed32 or
Unsigned32 range, then we can just convert the float64 to word32 and
use the fast word32 to tagged conversion. Doing this in
ChangeLowering (or the effect linearization pass) would be unsound, as
the types on the nodes are no longer usable.

This removes all Type uses from effect linearization. There's still some
work to be done for ChangeLowering tho.

[email protected]
==========

to

==========
[turbofan] Optimize tagged conversion based on type.

If we have to convert a float64 value to tagged representation and we
already know that the value is either in Signed31/Signed32 or
Unsigned32 range, then we can just convert the float64 to word32 and
use the fast word32 to tagged conversion. Doing this in
ChangeLowering (or the effect linearization pass) would be unsound, as
the types on the nodes are no longer usable.

This removes all Type uses from effect linearization. There's still some
work to be done for ChangeLowering tho.

[email protected]

Committed: https://crrev.com/861295bf167f045a3323fac086261b8bb264f8b3
Cr-Commit-Position: refs/heads/master@{#35713}
==========

[commit-bot: I haz the power, 2016-04-22 19:15:03]

Patchset 3 (id:??) landed as
https://crrev.com/861295bf167f045a3323fac086261b8bb264f8b3
Cr-Commit-Position: refs/heads/master@{#35713}