Add AST analysis that flags expressions that will have ToInt32 applied to the...

Add AST analysis that flags expressions that will have ToInt32 applied to them.
Simplify AST analysis of side-effect-free int32 expressions.

Committed: http://code.google.com/p/v8/source/detail?r=4056

[William Hesse, 2010-03-08 14:09:49]

I have two places where a flag is set without using the FLAG_safe_int32_compiler
flag, since the check itself takes time, and these are simple assignments. 
Should I use the guard everywhere?

[Kevin Millikin (Chromium), 2010-03-08 14:43:30]

LGTM with some suggestions for cleanup.

http://codereview.chromium.org/668256/diff/1006/7
File src/ast.h (right):

http://codereview.chromium.org/668256/diff/1006/7#newcode244
src/ast.h:244: // Will ToInt32 be immediately applied to the result of this
function?
Comment should say "Will ToInt32 (ECMA 262-3 9.5) or ToUunt32 (ECMA 262-3 9.6)
be applied to the result of this expression?"

I.e., mention ToUint32, don't say "immediately", "function" ==> "expression".

http://codereview.chromium.org/668256/diff/1006/7#newcode260
src/ast.h:260: class SideEffectFreeField : public BitField<bool, 0, 1> {};
Comment here saying "BitFields have <type, start, size>." or some such.

http://codereview.chromium.org/668256/diff/1006/8
File src/rewriter.cc (right):

http://codereview.chromium.org/668256/diff/1006/8#newcode248
src/rewriter.cc:248: if (FLAG_safe_int32_compiler) {
if (FLAG_...) {
  Slot* slot = var->slot();
  if (slot != NULL) {
    ...

http://codereview.chromium.org/668256/diff/1006/8#newcode430
src/rewriter.cc:430: node->expression()->set_to_int32(true);
You could move this inside the switch inside the if, and put it before ADD and
let it fall through.

http://codereview.chromium.org/668256/diff/1006/8#newcode539
src/rewriter.cc:539: node->right()->side_effect_free());
Indentation is off.

[William Hesse, 2010-03-08 15:20:50]

http://codereview.chromium.org/668256/diff/1006/7
File src/ast.h (right):

http://codereview.chromium.org/668256/diff/1006/7#newcode244
src/ast.h:244: // Will ToInt32 be immediately applied to the result of this
function?
On 2010/03/08 14:43:30, Kevin Millikin wrote:
> Comment should say "Will ToInt32 (ECMA 262-3 9.5) or ToUunt32 (ECMA 262-3 9.6)
> be applied to the result of this expression?"
> 
> I.e., mention ToUint32, don't say "immediately", "function" ==> "expression".

Done.

http://codereview.chromium.org/668256/diff/1006/7#newcode260
src/ast.h:260: class SideEffectFreeField : public BitField<bool, 0, 1> {};
On 2010/03/08 14:43:30, Kevin Millikin wrote:
> Comment here saying "BitFields have <type, start, size>." or some such.

Done.

http://codereview.chromium.org/668256/diff/1006/8
File src/rewriter.cc (right):

http://codereview.chromium.org/668256/diff/1006/8#newcode430
src/rewriter.cc:430: node->expression()->set_to_int32(true);
I want to_int32 to be correct even if --nosafe-int32-compiler.  Moved into the
switch if flag is true.

On 2010/03/08 14:43:30, Kevin Millikin wrote:
> You could move this inside the switch inside the if, and put it before ADD and
> let it fall through.

Done.