[llvm] r181753 - This is the first of three patches which creates stubs used for

[Eric Christopher]

Ping on these?


>
>> This pass is a self contained module pass which solves a very messy problem
>> in mips16 in a very clear way. There are many cases to this and pic and
>> static have some very different issues and there are endian issues too. In
>> addition I have to change instruction sets because the code is being
>> compiled for mips16 but the stubs are mips32 code.
>
> I'm not sure that this is an issue. Why is this a problem for you?
>
>>
>> I want to just finish checking in the solution.
>>
>> After I do, if someone wants to look at the totality of the problem and
>> offer a better solution, I'm open to making changes. I think it's unfair to
>> insist there is a better way without looking
>> at the whole problem.
>>
>
> Why not submit the whole patch for review first then?
>
>> I did a very simple part of this problem using selection dag earlier and it
>> was messy. I've said before that maybe someone that is more comfortable with
>> DAG transformation would offer a better solution.
>>
>> This was in  r173320
>
> I saw that. There's also MI lowering, which could be easier as I said.
>
> It may be an aesthetic choice, but given how many people are saying
> that your aesthetics are wrong you may wish to reconsider your
> position.
>
> Ultimately if it is really going to touch nothing in any other pass or
> backend and I don't have to maintain it then I'm not going to put up
> too much of a fuss.
>
> One specific about your testing though, if this is going to be a
> module pass you should run your test just the input/output of the pass
> and not the entire backend. Please provide tests for this.
>


-eric