Vector ICs: ObjectLiteral refactoring for Oracle feedback

Vector ICs: The Oracle needs to report feedback for the object literals and the count operation.

The refactoring is because it's awkward and error-prone to deterimine which IC slot an
ObjectLiteralProperty uses for feedback. The fix is for each one to know it's own slot. In the
numbering pass, we allocate slots for the ObjectLiteral, then hand out those slots into the
properties.

It adds one word to the ObjectLiteralProperty expression - I'm investigating if thats a
problem.

This changes makes compiling the object literal cleaner across the three compilers. Also, the
slot allocation logic in ObjectLiteral::ComputeFeedbackRequirements() was refactoring to mimic
the style in full-codegen. This is useful since it must remain in sync with
FullCodegen::VisitObjectLiteral().

Committed: https://crrev.com/752b0308df72461bfeb644bf6dd8dd331fcdb722
Cr-Commit-Position: refs/heads/master@{#30686}

[Michael Starzinger, 2015-09-07 08:31:12]

mstarzinger@chromium.org changed reviewers:
+ mstarzinger@chromium.org

[Michael Starzinger, 2015-09-07 08:31:13]

Conceptually I like this change, if memory footprint is acceptable.

https://codereview.chromium.org/1321993004/diff/40001/src/ast.h
File src/ast.h (right):

https://codereview.chromium.org/1321993004/diff/40001/src/ast.h#newcode1429
src/ast.h:1429: int ic_slot_count_;
As discussed offline, please move these fields either to before the "kind" or
after the boolean flags, so that a compiler can put "kind" and booleans into one
word.

Also as discussed offline, lets make sure the memory overhead of these two new
fields are acceptable by measuring/calculating AST memory footprint on common
workloads.

https://codereview.chromium.org/1321993004/diff/40001/src/ast.h#newcode1493
src/ast.h:1493: // Expression* getter;
nit: Looks like a left-over.

https://codereview.chromium.org/1321993004/diff/40001/src/compiler/ast-graph-...
File src/compiler/ast-graph-builder.cc (right):

https://codereview.chromium.org/1321993004/diff/40001/src/compiler/ast-graph-...
src/compiler/ast-graph-builder.cc:1705: void
AstGraphBuilder::BuildAccessor(Node* home_object,
As discussed offline, this is not really a builder, but a visitation method.
Let's call it "VisitObjectLiteralAccessor" instead. The reasoning behind this is
explained in the comments in the ast-graph-builder.h header file.

https://codereview.chromium.org/1321993004/diff/40001/src/compiler/ast-graph-...
File src/compiler/ast-graph-builder.h (right):

https://codereview.chromium.org/1321993004/diff/40001/src/compiler/ast-graph-...
src/compiler/ast-graph-builder.h:256: void BuildAccessor(Node* home_object,
ObjectLiteralProperty* property);
nit: After addressing the comment about naming, this should go down into the
visitation section (preferably with a comment "Dispatched from
VisitObjectLiteral").

[mvstanton, 2015-09-07 09:59:46]

mvstanton@chromium.org changed reviewers:
+ jkummerow@chromium.org

[mvstanton, 2015-09-07 09:59:47]

Hi Michi, hi Jakob, PTAL.

Jakob - General
Michi - Turbofan

Thanks,
--Michael

[Michael Starzinger, 2015-09-07 11:11:51]

LGTM on TurboFan, only nits, didn't look at the rest.

https://codereview.chromium.org/1321993004/diff/60001/src/ast.h
File src/ast.h (right):

https://codereview.chromium.org/1321993004/diff/60001/src/ast.h#newcode1510
src/ast.h:1510: // Expression* setter;
nit: Still look like a left-over.

https://codereview.chromium.org/1321993004/diff/60001/src/compiler/ast-graph-...
File src/compiler/ast-graph-builder.cc (right):

https://codereview.chromium.org/1321993004/diff/60001/src/compiler/ast-graph-...
src/compiler/ast-graph-builder.cc:1900: if (property == NULL) {
nit: s/NULL/nullptr/

https://codereview.chromium.org/1321993004/diff/60001/src/compiler/ast-graph-...
src/compiler/ast-graph-builder.cc:1901: VisitForValueOrNull(NULL);
nit: s/NULL/nullptr/

https://codereview.chromium.org/1321993004/diff/60001/src/compiler/ast-graph-...
src/compiler/ast-graph-builder.cc:1904: VisitForValueOrNull(value);
nit: Just inline "VisitForValueOrNull(property->value())" here.

[mvstanton, 2015-09-07 13:36:10]

Hi guys,
Addressed Michi's comments, did ports, fixed a bug.
Thanks much,
--Michael

https://codereview.chromium.org/1321993004/diff/40001/src/ast.h
File src/ast.h (right):

https://codereview.chromium.org/1321993004/diff/40001/src/ast.h#newcode1429
src/ast.h:1429: int ic_slot_count_;
On 2015/09/07 08:31:13, Michael Starzinger wrote:
> As discussed offline, please move these fields either to before the "kind" or
> after the boolean flags, so that a compiler can put "kind" and booleans into
one
> word.
> 
> Also as discussed offline, lets make sure the memory overhead of these two new
> fields are acceptable by measuring/calculating AST memory footprint on common
> workloads.

Excellent. I've reduced the two new words to one, and am trying to get a picture
of the overall size increase.

https://codereview.chromium.org/1321993004/diff/40001/src/compiler/ast-graph-...
File src/compiler/ast-graph-builder.cc (right):

https://codereview.chromium.org/1321993004/diff/40001/src/compiler/ast-graph-...
src/compiler/ast-graph-builder.cc:1705: void
AstGraphBuilder::BuildAccessor(Node* home_object,
On 2015/09/07 08:31:13, Michael Starzinger wrote:
> As discussed offline, this is not really a builder, but a visitation method.
> Let's call it "VisitObjectLiteralAccessor" instead. The reasoning behind this
is
> explained in the comments in the ast-graph-builder.h header file.

Right on, done.

https://codereview.chromium.org/1321993004/diff/40001/src/compiler/ast-graph-...
File src/compiler/ast-graph-builder.h (right):

https://codereview.chromium.org/1321993004/diff/40001/src/compiler/ast-graph-...
src/compiler/ast-graph-builder.h:256: void BuildAccessor(Node* home_object,
ObjectLiteralProperty* property);
On 2015/09/07 08:31:13, Michael Starzinger wrote:
> nit: After addressing the comment about naming, this should go down into the
> visitation section (preferably with a comment "Dispatched from
> VisitObjectLiteral").

Done.

https://codereview.chromium.org/1321993004/diff/60001/src/ast.h
File src/ast.h (right):

https://codereview.chromium.org/1321993004/diff/60001/src/ast.h#newcode1510
src/ast.h:1510: // Expression* setter;
On 2015/09/07 11:11:50, Michael Starzinger wrote:
> nit: Still look like a left-over.

Done.

https://codereview.chromium.org/1321993004/diff/60001/src/compiler/ast-graph-...
File src/compiler/ast-graph-builder.cc (right):

https://codereview.chromium.org/1321993004/diff/60001/src/compiler/ast-graph-...
src/compiler/ast-graph-builder.cc:1900: if (property == NULL) {
On 2015/09/07 11:11:50, Michael Starzinger wrote:
> nit: s/NULL/nullptr/

Done.

https://codereview.chromium.org/1321993004/diff/60001/src/compiler/ast-graph-...
src/compiler/ast-graph-builder.cc:1901: VisitForValueOrNull(NULL);
On 2015/09/07 11:11:50, Michael Starzinger wrote:
> nit: s/NULL/nullptr/

Done.

https://codereview.chromium.org/1321993004/diff/60001/src/compiler/ast-graph-...
src/compiler/ast-graph-builder.cc:1904: VisitForValueOrNull(value);
On 2015/09/07 11:11:50, Michael Starzinger wrote:
> nit: Just inline "VisitForValueOrNull(property->value())" here.

Heh, done. :)

[Jakob Kummerow, 2015-09-07 14:09:02]

lgtm

[mvstanton, 2015-09-10 14:36:44]

The CQ bit was checked by mvstanton@chromium.org

[mvstanton, 2015-09-10 14:36:44]

The patchset sent to the CQ was uploaded after l-g-t-m from
mstarzinger@chromium.org, jkummerow@chromium.org
Link to the patchset: https://codereview.chromium.org/1321993004/#ps140001
(title: "REBASE, turned off flag.")

[commit-bot: I haz the power, 2015-09-10 14:36:54]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1321993004/140001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1321993004/140001

[commit-bot: I haz the power, 2015-09-10 15:02:27]

Committed patchset #8 (id:140001)

[commit-bot: I haz the power, 2015-09-10 15:02:46]

Patchset 8 (id:??) landed as
https://crrev.com/752b0308df72461bfeb644bf6dd8dd331fcdb722
Cr-Commit-Position: refs/heads/master@{#30686}