review (S) for 6792161: assert("No dead instructions after post-alloc")

[Vladimir Kozlov]

You are right, lcm moves Precedent edge to prec().
Your code is good.

Vladimir

Tom Rodriguez wrote:
> You're right that the code isn't quite right but it works as written and your code isn't quite right either and wouldn't work.  The Precedent input is really a prec edge so req() == Precedent but len() == Precedent + 1.  So it would be more clearly written as n->len() >= MemBarNode::Precedent + 1.  Is that ok?
> 
> tom
> 
> On Jan 21, 2010, at 3:22 PM, Vladimir Kozlov wrote:
> 
>> !   if (n->req() == MemBarNode::Precedent &&
>>                                        ^ + 1
>>
>> or
>>
>> !   if (n->req() > MemBarNode::Precedent
>>
>> Vladimir
>>
>> Tom Rodriguez wrote:
>>> http://cr.openjdk.java.net/~never/6792161
>