Add asm.js typer / validator.

Add asm.js typer / validator.

Walk asm.js module ASTs, attach concrete type information
in preparation for generating a WASM module.

cctest test coverage (mjsunit coming in later CL).
Expressions, function tables, and foreign functions have coverage.
Statement coverage to be expanded in a later CL.

BUG= https://code.google.com/p/v8/issues/detail?id=4203
TEST=test-asm-validator
R=rossberg@chromium.org,titzer@chromium.org
LOG=N

Committed: https://crrev.com/546d9a70ace27ccecbf4adaad44bdeee6de135b7
Cr-Commit-Position: refs/heads/master@{#30520}

[bradn, 2015-08-30 06:38:30]

Patchset #1 (id:1) has been deleted

[titzer, 2015-08-31 09:13:36]

https://codereview.chromium.org/1322773002/diff/40001/src/typing-asm.cc
File src/typing-asm.cc (right):

https://codereview.chromium.org/1322773002/diff/40001/src/typing-asm.cc#newco...
src/typing-asm.cc:59: #define RECURSE_EXPR(expr, expected_type, msg)            
      \
Can these four macros be made into helper functions? AFAICT the only early
return is from RETURN() but that doesn't seem necessary.

https://codereview.chromium.org/1322773002/diff/40001/src/typing-asm.cc#newco...
src/typing-asm.cc:143: if (decl != NULL) SetType(decl->proxy()->var(),
Type::Any(zone()));
Shouldn't globals start at type none?

https://codereview.chromium.org/1322773002/diff/40001/src/typing-asm.cc#newco...
src/typing-asm.cc:664: if (computed_type_->IsArray()) {
Can we factor this to a helper function for array (heap) accesses?

https://codereview.chromium.org/1322773002/diff/40001/src/typing-asm.cc#newco...
src/typing-asm.cc:869: expr->op() == Token::MUL ? Type::Any() : Type::Number();
I don't understand why MUL takes Type::Any, despite the comment. Shouldn't it
always be Number?

https://codereview.chromium.org/1322773002/diff/40001/src/typing-asm.cc#newco...
src/typing-asm.cc:889: if (intish_ > (1 << 20)) {
kMaxIntishness (or better name)

[bradn, 2015-08-31 17:34:44]

PTAL

https://codereview.chromium.org/1322773002/diff/40001/src/typing-asm.cc
File src/typing-asm.cc (right):

https://codereview.chromium.org/1322773002/diff/40001/src/typing-asm.cc#newco...
src/typing-asm.cc:59: #define RECURSE_EXPR(expr, expected_type, msg)            
      \
On 2015/08/31 09:13:36, titzer wrote:
> Can these four macros be made into helper functions? AFAICT the only early
> return is from RETURN() but that doesn't seem necessary.

Done.

https://codereview.chromium.org/1322773002/diff/40001/src/typing-asm.cc#newco...
src/typing-asm.cc:143: if (decl != NULL) SetType(decl->proxy()->var(),
Type::Any(zone()));
On 2015/08/31 09:13:36, titzer wrote:
> Shouldn't globals start at type none?

Oops, yep.
Done.

https://codereview.chromium.org/1322773002/diff/40001/src/typing-asm.cc#newco...
src/typing-asm.cc:664: if (computed_type_->IsArray()) {
On 2015/08/31 09:13:36, titzer wrote:
> Can we factor this to a helper function for array (heap) accesses?

Done.

https://codereview.chromium.org/1322773002/diff/40001/src/typing-asm.cc#newco...
src/typing-asm.cc:869: expr->op() == Token::MUL ? Type::Any() : Type::Number();
On 2015/08/31 09:13:36, titzer wrote:
> I don't understand why MUL takes Type::Any, despite the comment. Shouldn't it
> always be Number?

Ah, yeah. This was a carryover from Andreas' protype, Number should be fine.
Changed.

https://codereview.chromium.org/1322773002/diff/40001/src/typing-asm.cc#newco...
src/typing-asm.cc:889: if (intish_ > (1 << 20)) {
On 2015/08/31 09:13:36, titzer wrote:
> kMaxIntishness (or better name)

Done.

[titzer, 2015-08-31 19:16:11]

https://codereview.chromium.org/1322773002/diff/60001/src/typing-asm.cc
File src/typing-asm.cc (right):

https://codereview.chromium.org/1322773002/diff/60001/src/typing-asm.cc#newco...
src/typing-asm.cc:343: RECURSE(VisitWithExpectation(stmt->cond(), cache_.kInt32,
Is this also true? I thought asm allowed naked conditions like if(a < 3) and do
{x } while (a < 4). It certainly allows do .. while(false)

https://codereview.chromium.org/1322773002/diff/60001/src/typing-asm.cc#newco...
src/typing-asm.cc:366: RECURSE(VisitWithExpectation(stmt->cond(), cache_.kInt32,
Same.

https://codereview.chromium.org/1322773002/diff/60001/src/typing-asm.cc#newco...
src/typing-asm.cc:435: "conditional then branch expected to be integer"));
You have number in the expectation and then number in the comment.

Shouldn't conditions use the expectation of the outer expression passed down?

[bradn, 2015-08-31 19:40:13]

PTAL

https://codereview.chromium.org/1322773002/diff/60001/src/typing-asm.cc
File src/typing-asm.cc (right):

https://codereview.chromium.org/1322773002/diff/60001/src/typing-asm.cc#newco...
src/typing-asm.cc:343: RECURSE(VisitWithExpectation(stmt->cond(), cache_.kInt32,
On 2015/08/31 19:16:11, titzer wrote:
> Is this also true? I thought asm allowed naked conditions like if(a < 3) and
do
> {x } while (a < 4). It certainly allows do .. while(false)

At least not in the text of the spec:

do stmt:Statement while ( expr:Expression ) ;
succeeds if ValidateStatement succeeds for stmt and ValidateExpression succeeds
for expr with a subtype of int.

Same with the others.

do..while(0) would be ok.

https://codereview.chromium.org/1322773002/diff/60001/src/typing-asm.cc#newco...
src/typing-asm.cc:366: RECURSE(VisitWithExpectation(stmt->cond(), cache_.kInt32,
On 2015/08/31 19:16:11, titzer wrote:
> Same.

See above.

https://codereview.chromium.org/1322773002/diff/60001/src/typing-asm.cc#newco...
src/typing-asm.cc:435: "conditional then branch expected to be integer"));
On 2015/08/31 19:16:11, titzer wrote:
> You have number in the expectation and then number in the comment.
> 
> Shouldn't conditions use the expectation of the outer expression passed down?

The comment is wrong, fixed.
Also, good point about the expected type, using that and updating the message.

[titzer, 2015-09-01 14:28:37]

lgtm with comment

https://codereview.chromium.org/1322773002/diff/110001/src/typing-asm.cc
File src/typing-asm.cc (right):

https://codereview.chromium.org/1322773002/diff/110001/src/typing-asm.cc#newc...
src/typing-asm.cc:44: AsmTyper::AsmTyper(CompilationInfo* info)
Can we make this take a ParseInfo instead, or better yet, just the literal,
script, and zone?

[bradn, 2015-09-01 16:28:41]

Will do a similar switch on the other AstVisitors in another CL.

https://codereview.chromium.org/1322773002/diff/110001/src/typing-asm.cc
File src/typing-asm.cc (right):

https://codereview.chromium.org/1322773002/diff/110001/src/typing-asm.cc#newc...
src/typing-asm.cc:44: AsmTyper::AsmTyper(CompilationInfo* info)
On 2015/09/01 14:28:37, titzer wrote:
> Can we make this take a ParseInfo instead, or better yet, just the literal,
> script, and zone?

Done.

[bradn, 2015-09-01 16:28:53]

The CQ bit was checked by bradnelson@google.com

[bradn, 2015-09-01 16:28:54]

The patchset sent to the CQ was uploaded after l-g-t-m from titzer@chromium.org
Link to the patchset: https://codereview.chromium.org/1322773002/#ps130001
(title: "fix")

[commit-bot: I haz the power, 2015-09-01 16:29:03]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1322773002/130001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1322773002/130001

[commit-bot: I haz the power, 2015-09-01 18:30:39]

Committed patchset #7 (id:130001)

[commit-bot: I haz the power, 2015-09-01 18:30:47]

Patchset 7 (id:??) landed as
https://crrev.com/546d9a70ace27ccecbf4adaad44bdeee6de135b7
Cr-Commit-Position: refs/heads/master@{#30520}