[turbofan] Allow non-speculative operators to consume feedback types.

[turbofan] Allow non-speculative operators to consume feedback types.

Turn the retyping pass of SimplifiedLowering into a proper phase, and
make it possible to propagate feedback types through non-speculative
operators. This defers the output representation selection to the
retyping phase, and checks that we don't mess up.

As a first user, we consume input type feedback for NumberAbs as well.
Long-term we can add all other operators to the mix.

R=jarin@chromium.org

Committed: https://crrev.com/b93cde3744c77cda088ca332bed84b34e6c711ff
Cr-Commit-Position: refs/heads/master@{#37672}

[Benedikt Meurer, 2016-07-12 07:59:33]

The CQ bit was checked by bmeurer@chromium.org to run a CQ dry run

[commit-bot: I haz the power, 2016-07-12 07:59:36]

Dry run: 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, 2016-07-12 08:25:22]

The CQ bit was unchecked by commit-bot@chromium.org

[commit-bot: I haz the power, 2016-07-12 08:25:23]

Dry run: Try jobs failed on following builders:
  v8_linux_dbg_ng on master.tryserver.v8 (JOB_FAILED,
http://build.chromium.org/p/tryserver.v8/builders/v8_linux_dbg_ng/builds/8876)
  v8_linux_dbg_ng_triggered on master.tryserver.v8 (JOB_FAILED,
http://build.chromium.org/p/tryserver.v8/builders/v8_linux_dbg_ng_triggered/b...)

[Jarin, 2016-07-12 08:51:49]

https://codereview.chromium.org/2139203002/diff/1/src/compiler/simplified-low...
File src/compiler/simplified-lowering.cc (right):

https://codereview.chromium.org/2139203002/diff/1/src/compiler/simplified-low...
src/compiler/simplified-lowering.cc:55: //     feedback forwards.
Just say "Propagate types from type feedback forwards."

The traverse from end part is just to get an RPO order, and thus make
convergence faster.

https://codereview.chromium.org/2139203002/diff/1/src/compiler/simplified-low...
src/compiler/simplified-lowering.cc:285: VisitNode(node, info->truncation(),
nullptr);
Hmm, not too happy about uselessly invoking the input truncation propagation
machinery. You only really need to set the representation here.

https://codereview.chromium.org/2139203002/diff/1/src/compiler/simplified-low...
src/compiler/simplified-lowering.cc:670: representation));
I am not sure the subtyping check is really needed here. Convergence does not
really depend on this.

https://codereview.chromium.org/2139203002/diff/1/src/compiler/simplified-low...
src/compiler/simplified-lowering.cc:676: representation));
In this case, the representation should also be the same, not just a subtype.
Otherwise, we could insert a wrong conversion.

https://codereview.chromium.org/2139203002/diff/1/src/compiler/simplified-low...
src/compiler/simplified-lowering.cc:975: if (phase_ == PROPAGATE) {
if (propagate()) {

https://codereview.chromium.org/2139203002/diff/1/src/compiler/simplified-low...
src/compiler/simplified-lowering.cc:979: } else if (phase_ == LOWER) {
} else if (lower()) {

https://codereview.chromium.org/2139203002/diff/1/src/compiler/typer.cc
File src/compiler/typer.cc (right):

https://codereview.chromium.org/2139203002/diff/1/src/compiler/typer.cc#newco...
src/compiler/typer.cc:1044: lhs = ToNumber(lhs, t);
Could you delete this dead code while you are here?

[Benedikt Meurer, 2016-07-12 09:43:49]

https://codereview.chromium.org/2139203002/diff/1/src/compiler/simplified-low...
File src/compiler/simplified-lowering.cc (right):

https://codereview.chromium.org/2139203002/diff/1/src/compiler/simplified-low...
src/compiler/simplified-lowering.cc:55: //     feedback forwards.
On 2016/07/12 08:51:49, Jarin wrote:
> Just say "Propagate types from type feedback forwards."
> 
> The traverse from end part is just to get an RPO order, and thus make
> convergence faster.
> 

Done.

https://codereview.chromium.org/2139203002/diff/1/src/compiler/simplified-low...
src/compiler/simplified-lowering.cc:285: VisitNode(node, info->truncation(),
nullptr);
On 2016/07/12 08:51:49, Jarin wrote:
> Hmm, not too happy about uselessly invoking the input truncation propagation
> machinery. You only really need to set the representation here.

Acknowledged.

https://codereview.chromium.org/2139203002/diff/1/src/compiler/simplified-low...
src/compiler/simplified-lowering.cc:670: representation));
On 2016/07/12 08:51:49, Jarin wrote:
> I am not sure the subtyping check is really needed here. Convergence does not
> really depend on this.

Done.

https://codereview.chromium.org/2139203002/diff/1/src/compiler/simplified-low...
src/compiler/simplified-lowering.cc:676: representation));
On 2016/07/12 08:51:49, Jarin wrote:
> In this case, the representation should also be the same, not just a subtype.
> Otherwise, we could insert a wrong conversion.

Done.

https://codereview.chromium.org/2139203002/diff/1/src/compiler/simplified-low...
src/compiler/simplified-lowering.cc:975: if (phase_ == PROPAGATE) {
On 2016/07/12 08:51:49, Jarin wrote:
> if (propagate()) {

Done.

https://codereview.chromium.org/2139203002/diff/1/src/compiler/simplified-low...
src/compiler/simplified-lowering.cc:979: } else if (phase_ == LOWER) {
On 2016/07/12 08:51:49, Jarin wrote:
> } else if (lower()) {

Done.

https://codereview.chromium.org/2139203002/diff/1/src/compiler/typer.cc
File src/compiler/typer.cc (right):

https://codereview.chromium.org/2139203002/diff/1/src/compiler/typer.cc#newco...
src/compiler/typer.cc:1044: lhs = ToNumber(lhs, t);
On 2016/07/12 08:51:49, Jarin wrote:
> Could you delete this dead code while you are here?

Done.

[Jarin, 2016-07-12 09:46:57]

lgtm

[Benedikt Meurer, 2016-07-12 09:47:11]

The CQ bit was checked by bmeurer@chromium.org

[commit-bot: I haz the power, 2016-07-12 09:47:13]

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, 2016-07-12 10:08:51]

Committed patchset #2 (id:20001)

[commit-bot: I haz the power, 2016-07-12 10:09:08]

CQ bit was unchecked.

[commit-bot: I haz the power, 2016-07-12 10:09:15]

Description was changed from

==========
[turbofan] Allow non-speculative operators to consume feedback types.

Turn the retyping pass of SimplifiedLowering into a proper phase, and
make it possible to propagate feedback types through non-speculative
operators. This defers the output representation selection to the
retyping phase, and checks that we don't mess up.

As a first user, we consume input type feedback for NumberAbs as well.
Long-term we can add all other operators to the mix.

R=jarin@chromium.org
==========

to

==========
[turbofan] Allow non-speculative operators to consume feedback types.

Turn the retyping pass of SimplifiedLowering into a proper phase, and
make it possible to propagate feedback types through non-speculative
operators. This defers the output representation selection to the
retyping phase, and checks that we don't mess up.

As a first user, we consume input type feedback for NumberAbs as well.
Long-term we can add all other operators to the mix.

R=jarin@chromium.org

Committed: https://crrev.com/b93cde3744c77cda088ca332bed84b34e6c711ff
Cr-Commit-Position: refs/heads/master@{#37672}
==========

[commit-bot: I haz the power, 2016-07-12 10:09:16]

Patchset 2 (id:??) landed as
https://crrev.com/b93cde3744c77cda088ca332bed84b34e6c711ff
Cr-Commit-Position: refs/heads/master@{#37672}