[V8][asm.js] Adds support to global const variables.

[V8][asm.js] Adds support to global const variables.

This CL adds support for:
https://discourse.wicg.io/t/allow-const-global-variables/684

It allows global const variables to be used as if they were numeric
literals. For example:

  const f0 = fround(0);
  ...
  function foo() {
    var v = f0;  // no type annotation.
    ...
    return f0;  // no return type annotation.
  }

BUG= https://bugs.chromium.org/p/v8/issues/detail?id=5540

Committed: https://crrev.com/5331e3b61633b5b95b487353b4d08346abf48eee
Cr-Commit-Position: refs/heads/master@{#40477}

[John, 2016-10-19 18:34:41]

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

[commit-bot: I haz the power, 2016-10-19 18:34:47]

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, 2016-10-19 18:54:36]

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

[commit-bot: I haz the power, 2016-10-19 18:54:39]

Dry run: Try jobs failed on following builders:
  v8_mac_rel_ng on master.tryserver.v8 (JOB_FAILED,
http://build.chromium.org/p/tryserver.v8/builders/v8_mac_rel_ng/builds/11160)
  v8_mac_rel_ng_triggered on master.tryserver.v8 (JOB_FAILED,
http://build.chromium.org/p/tryserver.v8/builders/v8_mac_rel_ng_triggered/bui...)

[John, 2016-10-19 19:25:50]

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

[commit-bot: I haz the power, 2016-10-19 19:25:54]

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

[John, 2016-10-19 19:46:37]

jpp@chromium.org changed reviewers:
+ aseemgarg@chromium.org, bradnelson@google.com

[commit-bot: I haz the power, 2016-10-19 20:03:35]

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

[commit-bot: I haz the power, 2016-10-19 20:03:36]

Dry run: This issue passed the CQ dry run.

[bradnelson, 2016-10-19 21:52:06]

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

[bradnelson, 2016-10-19 21:52:07]

lgtm

https://codereview.chromium.org/2435823002/diff/20001/src/asmjs/asm-typer.cc
File src/asmjs/asm-typer.cc (right):

https://codereview.chromium.org/2435823002/diff/20001/src/asmjs/asm-typer.cc#...
src/asmjs/asm-typer.cc:765: VariableInfo::Mutability mutability;
Maybe a comment linking to the errata?

https://codereview.chromium.org/2435823002/diff/20001/src/asmjs/asm-typer.cc#...
src/asmjs/asm-typer.cc:2766: "Undeclared identifier in variable declaration
initializer");
period like the others?

https://codereview.chromium.org/2435823002/diff/20001/test/cctest/asmjs/test-...
File test/cctest/asmjs/test-asm-typer.cc (right):

https://codereview.chromium.org/2435823002/diff/20001/test/cctest/asmjs/test-...
test/cctest/asmjs/test-asm-typer.cc:511: {"var v = not_fround;", "initialize a
global must be a const"},
Add tests that hit:
Undeclared identifier in return statement
Identifier in return statement is not const
Undeclared identifier in variable declaration initializer
Identifier in variable declaration initializer must be const

Add something that covers (since that changed and we didn't catch:
Invalid variable initialization - it should be a literal, const, or
fround(literal).

https://codereview.chromium.org/2435823002/diff/20001/test/mjsunit/asm/asm-va...
File test/mjsunit/asm/asm-validation.js (right):

https://codereview.chromium.org/2435823002/diff/20001/test/mjsunit/asm/asm-va...
test/mjsunit/asm/asm-validation.js:36: // constants dono need type annotations.
dono -> do not ?

[aseemgarg, 2016-10-19 22:24:11]

Didn't realize const was allowed in asmjs. Do we have plumbing to raise an error
if a const is modified later?

[bradn, 2016-10-19 22:45:22]

In the errata (See the bug)... :-) but why?!?!

So this change works with the asm-wasm builder because it does a load from the
global each time.
What we probably want to do (for maximum performance, and correctness) is in the
asm-wasm builder detect the constant and collapse it into a literal. But that
can be a follow up change. (This will at least get all the benchmarks
validating, and will make this box2d_f32 go up)

I don't think assignment is allowed as it will be labeled as immutable. Add a
test though John?

[titzer, 2016-10-20 08:20:53]

On 2016/10/19 22:24:11, aseemgarg wrote:
> Didn't realize const was allowed in asmjs. Do we have plumbing to raise an
error
> if a const is modified later?

In JavaScript, this is a dynamic error, so there is no static checking for this
situation.

[John, 2016-10-20 15:08:31]

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

[commit-bot: I haz the power, 2016-10-20 15:08:35]

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

[John, 2016-10-20 15:13:22]

ptal

https://codereview.chromium.org/2435823002/diff/20001/src/asmjs/asm-typer.cc
File src/asmjs/asm-typer.cc (right):

https://codereview.chromium.org/2435823002/diff/20001/src/asmjs/asm-typer.cc#...
src/asmjs/asm-typer.cc:765: VariableInfo::Mutability mutability;
On 2016/10/19 21:52:07, bradnelson wrote:
> Maybe a comment linking to the errata?

Done (in asm-typer.h)

https://codereview.chromium.org/2435823002/diff/20001/src/asmjs/asm-typer.cc#...
src/asmjs/asm-typer.cc:2766: "Undeclared identifier in variable declaration
initializer");
On 2016/10/19 21:52:07, bradnelson wrote:
> period like the others?

Done.

https://codereview.chromium.org/2435823002/diff/20001/test/cctest/asmjs/test-...
File test/cctest/asmjs/test-asm-typer.cc (right):

https://codereview.chromium.org/2435823002/diff/20001/test/cctest/asmjs/test-...
test/cctest/asmjs/test-asm-typer.cc:511: {"var v = not_fround;", "initialize a
global must be a const"},
On 2016/10/19 21:52:07, bradnelson wrote:
> Add tests that hit:
> Undeclared identifier in return statement
> Identifier in return statement is not const
> Undeclared identifier in variable declaration initializer
> Identifier in variable declaration initializer must be const
> 
> Add something that covers (since that changed and we didn't catch:
> Invalid variable initialization - it should be a literal, const, or
> fround(literal).

Done.

https://codereview.chromium.org/2435823002/diff/20001/test/mjsunit/asm/asm-va...
File test/mjsunit/asm/asm-validation.js (right):

https://codereview.chromium.org/2435823002/diff/20001/test/mjsunit/asm/asm-va...
test/mjsunit/asm/asm-validation.js:36: // constants dono need type annotations.
On 2016/10/19 21:52:07, bradnelson wrote:
> dono -> do not ?

Done.

[bradn, 2016-10-20 15:20:34]

lgtm

[bradn, 2016-10-20 15:20:39]

The CQ bit was unchecked by bradnelson@google.com

[bradn, 2016-10-20 15:20:39]

The CQ bit was checked by bradnelson@google.com

[bradn, 2016-10-20 15:20:40]

The patchset sent to the CQ was uploaded after l-g-t-m from
bradnelson@chromium.org
Link to the patchset: https://codereview.chromium.org/2435823002/#ps40001
(title: "Addresses comments.")

[commit-bot: I haz the power, 2016-10-20 15:20:45]

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, 2016-10-20 15:48:33]

Committed patchset #3 (id:40001)

[commit-bot: I haz the power, 2016-11-17 22:08:30]

Description was changed from

==========
[V8][asm.js] Adds support to global const variables.

This CL adds support for:
https://discourse.wicg.io/t/allow-const-global-variables/684

It allows global const variables to be used as if they were numeric
literals. For example:

  const f0 = fround(0);
  ...
  function foo() {
    var v = f0;  // no type annotation.
    ...
    return f0;  // no return type annotation.
  }

BUG= https://bugs.chromium.org/p/v8/issues/detail?id=5540
==========

to

==========
[V8][asm.js] Adds support to global const variables.

This CL adds support for:
https://discourse.wicg.io/t/allow-const-global-variables/684

It allows global const variables to be used as if they were numeric
literals. For example:

  const f0 = fround(0);
  ...
  function foo() {
    var v = f0;  // no type annotation.
    ...
    return f0;  // no return type annotation.
  }

BUG= https://bugs.chromium.org/p/v8/issues/detail?id=5540

Committed: https://crrev.com/5331e3b61633b5b95b487353b4d08346abf48eee
Cr-Commit-Position: refs/heads/master@{#40477}
==========

[commit-bot: I haz the power, 2016-11-17 22:08:32]

Patchset 3 (id:??) landed as
https://crrev.com/5331e3b61633b5b95b487353b4d08346abf48eee
Cr-Commit-Position: refs/heads/master@{#40477}